The inherent complexity of microservices architectures represents a significant obstacle to
the realization of the right to be forgotten, as required by the General Data Protection Law
(LGPD). The fragmentation of data across independently managed services, each with
its own business logic and data storage, makes consistent and reliable deletion a complex
challenge. This research addresses this problem by proposing an automated middleware
solution designed to orchestrate the deletion process across distributed microservices, using
the well-known two-phase commit (2PC) pattern to ensure data consistency, in addition to
offering a library to be imported into existing code and to facilitate adoption, requiring only
the implementation of the service’s own business rules. By evaluating its architecture and
performance in simulated environments, this study aims to demonstrate the effectiveness
and compliance of this automated approach to a problem currently fraught with manual
complexity and potential legal ramifications. Ultimately, this research strives to deliver
a robust and scalable solution that empowers organizations to effectively manage data
deletion in microservices environments, promoting greater trust and adherence to data
protection regulations.

The Brazilian Judiciary, made up of ninety-four Courts and their respective Information and
Communication Technology departments, faces the crucial challenge of unifying its judicial
systems. Despite the efforts of the National Justice Council (CNJ), a significant disparity
between administrative and support systems still persists. This diversity of solutions,
combined with different personnel structures, technical capabilities and infrastructures,
makes it difficult to control the applications in use, meet deadlines and, especially, software
security. To address these challenges, it is proposed to implement software that supports
IT areas in adopting the DevSecOps methodology, integrating development, security and
operations. This approach aims to break down knowledge silos, distribute responsibilities
and information more efficiently, increase transparency in the IT area, improve quality
and reduce development time, in addition to optimizing software maintenance throughout
its life cycle. . The development of the solution was based on a literature review to identify
best practices and tools for vulnerability analysis, elicitation of minimum functional and
non-functional requirements, choice of appropriate technology for programming language,
development tools and database. Use cases and tools for proof of concept were defined,
focused on application security assessments, using calls to the application programming
interface (API) and presenting results through a web interface. Key benefits identified
include centralized visualization of information about applications and their dependencies,
effective vulnerability analysis, flexible integration of security tools, and expanded visibility
into application security for the entire IT team.

The increasing demand for food security and the scarcity of natural resources pose global challenges, particularly in agriculture, which accounts for a significant share of water consumption. In Brazil, agriculture holds great economic importance; however, its intensive water use necessitates the adoption of more sustainable practices. Managing irrigation adaptively and efficiently is challenging due to the complexity of multiple environmental variables. This reduces water use efficiency and negatively impacts agricultural productivity. IrrigaFlow is a modular architecture that automates irrigation through IoT, fuzzy logic, and distributed processing. It is composed of three layers: IoT Module, Network Edge, and Cloud, enabling real-time monitoring and adjustments based on local environmental data. This approach optimizes water use and enhances responsiveness to climatic conditions. Simulations conducted in a controlled environment allowed for evaluating the architecture’s efficiency across varied scenarios, highlighting its capacity to adjust water management to the specific needs of each crop and environment. IrrigaFlow emerges as a promising alternative to address food security challenges, reduce water waste, and promote sustainable agricultural practices.

The unprecedented growth in the creation and persistence of unsctructured textual documents in public institutions poses challenges for efficient information retrieval and data analysis. This research addresses these challenges, by proposing a prototype of a search assistant using the Retrieval-Augmented Generation (RAG) approach, specifically applied to documents produced by Assembleia Legislativa do Estado do Rio Grande do Norte (Alern). The proposed system leverages Natural Language Processing (NLP) techniques, vector databases, and Large Language Models (LLMs) to enable semantic search and the generation of relevant content as answers to query inputs. The research introduces an architecture capable of retrieving document fragments based on semantic similarity. User-provided queries are processed and used to search content with contextual relevance, which is then synthesized into coherent and contextually appropriate responses through an LLM. Results from automated evaluations using BERTScore demonstrate the system's effectiveness in retrieving information based on user input data -- with precision and recall achieving values of 79% and 69% respectively, which are satisfactory values in text generation scenarios. Being powered by the RAG approach, the proposed assistant not only reduces the cognitive load associated with the manual analysis of large document collections but also provides a scalable and adaptable solution for continuously evolving datasets. This research contributes to bridging the gap between the availability of public data and the generation of searchable information, aligning with goals of transparency and accessibility in the legislative environment.

Software architecture consists of fundamental decisions that guide the construction and
evolution of a system. Over time, accumulated modifications throughout the software
lifecycle can violate previously defined criteria, characterizing architectural degradation.
The absence of a formal continuous evaluation process causes these problems to be
detected reactively, often only after causing significant impacts. To mitigate these risks,
static analysis tools can be adopted to validate architecture proactively, enabling constant
monitoring. One example of a tool that follows this approach is Clj-Depend, which validates
dependencies between modules in systems implemented using the Clojure programming
language. In this context, this work aims to investigate the occurrence and evolution
of these violations in systems implemented in Clojure. For this purpose, a quantitative
longitudinal study based on software repository mining was conducted, involving five
services from a multinational technology company, where they process on average more than
eight hundred requests per minute. The results showed that all services presented violation
accumulation correlated with code base growth, with concentration in few critical rules.
Statistical analyses confirmed significant association between lines of code and architectural
degradation when adequate governance is absent. Despite scope limitations, the study
demonstrated that architectural degradation is monitorable and can be integrated into
the team’s development workflow through automated processes. The developed protocol
offers a systematic approach for continuous architectural auditing, contributing to more
sustainable software evolution practices.

This dissertation investigates the limited coverage of air-quality monitoring in Brazil and
proposes an alternative solution: an IoT-based platform capable of expanding environmen-
tal observability and operational automation. The object of study is a layered, software-
and hardware-agnostic architecture designed to integrate acquisition (IoT devices), trans-
mission, processing and assessment (calculation of the IQAr), time-series persistence, and
interactive visualization, with real-time notifications. The overall objective was to design
and validate, in a reproducible testbed, such an architecture; the specific objectives compri-
sed: (i) modeling and specifying a layered architecture for IoT-based air-quality monitoring;
(ii) implementing the experimental environment (testbed) with devices and sensors simula-
ted in Node-RED; (iii) demonstrating Zabbix’s potential as a monitoring platform applied
both to air quality and to IoT devices; and (iv) analyzing, from a technical-operational
standpoint, the results produced by the testbed and the feasibility of the proposed solution.
All objectives were achieved. The results show automated configuration of monitoring
via Zabbix templates, dashboards that display a map and per-station time series (with
pollutant-level detail and meteorological correlation), device and sensor supervision, and
real-time alerts delivered via instant messenger. The architecture proved applicable to
the technological arrangement used. Zabbix exhibited very efficient performance and full
compatibility with the application context, standing out as a tool with high potential for
monitoring IoT devices and air-quality systems; furthermore, the scarcity of academic
studies analyzing its use in this domain underscores this research’s contribution. As an
overall contribution, the work delivers a vendor-agnostic architectural guide, a replicable
testbed, and a robust proof of concept of the Node-RED–MQTT–Zabbix–Grafana inte-
gration for environmental monitoring, pointing to a concrete, scalable, and economically
viable path to expand air-quality monitoring coverage in the country.

The management of services offered in a computer network demands a manual effort to maintain the infrastructure. Typically, network administrators need to access multiple servers, install multiple software, and configure numerous files. This manual approach often results in errors and impacts the availability of services, as well as increasing the response time for change requests. The approach of infrastructure as code emerges as a solution to minimize the errors associated with traditional service management, but it requires network administrators to acquire programming skills, which are generally not part of their curriculum.
In this context, this work presents the development of an integrated platform that uses infrastructure as code to automate service management. This platform allows network administrators to use a graphical interface to access servers and perform settings that were previously done manually. As a result, it is possible to reduce errors caused by configuration deviations, speed up the delivery of changes and promote greater consistency in the services offered.

Software testing plays a crucial role in ensuring the quality and reliability of software systems. In recent years, medium and large-scale systems have been developed based on REST APIs. REST APIs stand out in this scenario due to their simplicity, scalability, and platform independence. In this context, a tool called EvoMaster stands out for its ability to effectively generate automated test cases. This dissertation evaluates this tool's effectiveness in generating test cases for REST applications, particularly in large-scale web systems. The evaluation was conducted on the SigSaude web system, and the results were evaluated in collaboration with one of its developers. Our results show that EvoMaster uncovered a significant number of defects. However, it revealed challenges related to its adoption, such as difficulty configuring the tool and regression testing with low code coverage.

Cancer is one of the main causes of death in the world, with chemotherapy being one of the most frequently used treatments. Despite its effectiveness in eliminating cancer cells, this approach negatively impacts healthy cells, triggering considerable side effects, both physical and psychological. 

Therefore, it is common to develop some degree of depression and abandon treatment, making the development of support strategies for patients crucial. 

Coping with these symptoms at home is challenging for patients, especially considering the scarcity of tools in Brazilian Portuguese that offer support. In this context, the proposal to develop the AMIA tool arises, a chatbot that uses a structured dialogue based on a flowchart prepared by a medical team from the League Against Cancer. Patients can choose pre-defined symptoms and receive corresponding guidance. 

Additionally, the tool allows patients to enter free text, using natural language processing to identify symptoms and offer appropriate guidance based on a previous list of symptoms with their respective management strategies. The proposal aims to improve the quality of life of patients undergoing chemotherapy treatment, 

enabling more effective self-management of adverse symptoms in their daily environment.

Real-Time Strategy Games are environments that generally simulate real military situations and present a series of challenges for the field of Artificial Intelligence, such as high complexity, large action and state spaces, partially observable maps, dealing with multiple units at the same time, and also the fact that tasks can be performed within the scope of micromanagement or macromanagement. In particular, Reinforcement Learning has stood out in the application and evolution of techniques capable of dealing with these challenges. A systematic review of the literature was then carried out to understand the state-of-the-art use of Deep Reinforcement Learning in Real-Time Strategy games. Some of the most relevant information raised were the use of StarCraft II as the main simulation environment, the need for more studies addressing macromanagement, the good performance of hierarchical architectures, and the importance of using action masking and transfer learning techniques, which can reduce the complexity of the problem and the computational cost. Given this and the challenges in this field, this work proposes Jaguar, a hierarchical architecture capable of handling both macro and micromanagement, while requiring few resources during training. Jaguar implements techniques for action and state shaping, invalid action masking through the state and the communication between hierarchical levels, different types of rewards, and uses the DQN model for micro decisions and DDQN for strategic decisions. The agent was trained in a base scenario and transfer learning was applied to two other scenarios. The results show that the agent was able to perform both micromanagement and macromanagement, presenting good learning in the base scenario and promising results in subsequent scenarios, demonstrating that the proposed approach was able to deal significantly well with the complexity of StarCraft II scenarios using few computational resources, leaving room for future improvement.

In recent decades, technological evolution has brought numerous advances allowing intensive collection, processing, and storage of personal data. There is much evidence, mainly revelations, about the operations and data breaches of large companies with data as their most significant asset, such as Facebook, Google, Amazon, and Uber. Due to this finding, there is a growing concern about using these data, evidenced by the profusion of laws worldwide that aim to protect individuals' privacy. The various legislations point to the need to implement processes and techniques that guarantee data privacy, among which is the (pseudo)anonymization of data. It is in this context and seeking to contribute to the protection of privacy that, in this work, an architecture is proposed for the multilevel (pseudo)anonymization of health data. Multilevel, as data is pseudonymized at two levels, one local and one global, thus ensuring that data from multiple providers can be related yet (pseudo)anonymized. The focus on the health area is, on the one hand, a challenging application, given the sensitivity of the data.
The architecture proposed in this work was implemented as a proof of concept and evaluated from a set of tests. Test results suggest that the architecture enables correct anonymization at the source, secure linking of (pseudo)anonymized data across multiple sources, and even allows reidentification for cases involving the security of the individuals involved.

The Javascript language is one of the most famous development tools today, gaining visibility in several areas such as web games, three-dimensional renderings, artificial intelligence and, mainly, the development of web applications, with its major role in the construction of interfaces through front end development. In this ecosystem, several libraries and frameworks were built, the most famous being the React library, developed and published by Meta (Facebook). Applications built on React, like any other system, need to remain usable and relevant over time. As empirical evidence shows, the presence of bad smells in the code might compromise the software evolvability. Based on this context, this work presents, based on mapping studies of the academic and gray literature, a proposal of bad smells oriented to the React library, integrating them to a code detection tool called ReactLint, which will flag code flaws and will indicate possible solutions to developers who use it. This work aims to validate the proposed smells, as well as the built tool, in order to identify whether they can affect the performance of a React application in the short and long term

Fostered by hardware and software advances, Machine Learning (ML) started to ramp up exponentially in the last few decades, and has become instrumental for advancing the work in the most varied areas of knowledge. Though generally restricted to controlledspace experiments, over previously obtained and curated data samples, results have been outstanding, which gave rise to such levels of popularity for ML applications that it’s hard to find an area of human knowledge left untouched by Machine Learning. In such context, establishing minimum performance guarantees over unknown, real-world data, becomes paramount, especially in Healthcare applications, where errors can lead to life-threatening situations. There’s an ML discipline, called Machine Learning Operations (or MLOps, for short), which concerns itself with ML Models’ lifecycle management, from conception to deployment in production (real-world) environments, including monitoring its real-world behavior. Once deployed, models are subject to performance decay issues, such as drift, which has motivated recent studies on continual learning and Continuous Monitoring of ML models. The present work focuses on identifying state-of-the-art techniques for evaluating model fitness in real-world usage scenarios, and on how to establish a feedback-loop to include continuous monitoring and change management in the models’ lifecycle. Finally, the present work aims to apply evaluation techniques in a case study of ML models applied to Healthcare, and establish a process for evaluating models. The target models were developed as part of the Remote Assistance Platform (PAR), and is currently in effective use in an oncologic ICU.

The internet is becoming increasingly accessible and available to a larger number of people. However, internet users nowadays are no longer enthusiasts or academics as they were in its origin; they are ordinary people. Consequently, users who need to manage a local network under their responsibility are not technically skilled individuals with knowledge of computer network Niche language. In the context of a residential local network, the growing presence of devices and the need to control network access demand solutions that are accessible and intuitive for common users who want greater control over their network.
However, home router configuration interfaces are currently limited, unintuitive, and unfriendly to the general public. Consequently, they hinder access control and customization of advanced settings as they require specialized technical knowledge. Moreover, the lack of standardization in these interfaces creates accessibility barriers for users. As a result, users face difficulties in effectively configuring their residential routers, limiting their ability to control network access and customize settings. The use of natural language, whether through voice or text, becomes an interface that presents an advantage in this context due to its simplicity and accessibility for this audience.
This work addresses this issue and proposes a layered architecture that utilizes natural language and automation resources to facilitate the configuration of residential networks, particularly focusing on device access control and device class control. This approach aims to eliminate the need for specialized technical knowledge from the user. To achieve this goal, this work reviews the state of the art regarding the use of natural language in the context of intent-based network configuration and contributes with a systematic mapping of the use of natural language for network configuration in the domestic context, along with existing strategies for device identification and classification in a network with heterogeneous devices. By using natural language processing tools facilitated by artificial intelligence and features such as automatic device identification and classification, it is expected to offer common users a more intuitive and efficient experience in configuring their residential networks.
To validate the proposed solution, a prototype in the Python language will be developed, establishing a hierarchy of device types and a hierarchy of configuration intentions, both compatible with the residential environment. The prototype will utilize communication between APIs and the SSH protocol to provide interoperability between the proposed layers.
The research results indicate that the use of natural language for network configuration is still underexplored in the residential context, particularly concerning automatic detection of device types present in the network. At this point, the proposed hierarchies and the necessary descriptors for the prototype's functioning have been established.

The Internet arising and the Information and Communication Technologies development expanded the volume and diversification of data sources, thus opening up new opportunities in the industry and academic fields for Machine Learning Techiniques and Big Data related applications. In the same perspective is the extensive amount of data generated by mobile networks infrastructures worldwide. The Radio Access Networks (RAN), crucial for the telecommunications infrastructure, work as a really important layer for the wireless communications and produce a significant data volume due to the network counters measurements which stand as the enablers for the monitoring and visibility on network performance indicators and service quality. The present work consists of applying clustering algorithms to create profiles on datasets related to 5G Radio Access Network Indicators regarding traffic, volume and channel quality so the labeled data can get used on classification problems and as a support tool for identifying improvements, performance management and operational efficiency of Radio Access Networks.

The number of IoT devices in the world grows exponentially and is expected to exceed 75 billion in the coming years. Its use in home automation processes has grown and, as a result, an increasing number of users without technical knowledge (or with reduced technical knowledge) have made use of these devices. The wide variety of manufacturers and the lack of standardization in the features available can generate doubts among users and lead them to make wrong settings, harming the management of these equipment. Performing these settings can be even more challenging for people with disabilities. According to data presented by IBGE, more than 45 million people have some type of disability, which corresponds to almost 24% of the population. Faced with these challenges, this work proposes a layered architecture, based on independent modules, where this architecture will allow, through voice commands or text and using natural language, a user can perform complex configurations, without the need for advanced technical knowledge regarding infrastructure or specifics of the network device used. For this purpose, among other technologies, natural language processing, machine learning and text-to-speech conversion will be used. A survey of the state of the art regarding solutions aimed at configuring, activating and monitoring devices is presented, as well as a prototype for validation and evaluation of the functioning of the architecture proposed in this research. For future improvements can be implemented in the chatbot, its design was submitted to an evaluation to identify possible existing problems. Finally, in the quantitative tests carried out to verify the assertiveness of the prototype developed, its accuracy was greater than 83%.

The provision of health services is increasingly being supported by different technological solutions, such as Health Information Systems (HIS), Electronic Health Records (EHR), or solutions based on eHealth and mHealth. These solutions aim to support decisions and optimize processes in the provision of health services through the collection, processing, and availability of health data. Therefore, the use of digital solutions in the health context brings benefits both from the point of view of the management of health facilities and the clinical team that deals directly or indirectly with patients. Digital solutions for the health context are used in different environments, such as hospitals, clinics, laboratories, among others. Thus, it is common that different solutions are created and used to solve specific problems, with different characteristics and for different environments, but still in the health context. For example, a hospital may use different systems in its day-to-day activities, such as an Electronic Patient Record (PEP) to manage information regarding patient consultations and treatments; a system for monitoring patients admitted to the Intensive Care Unit (ICU); and a system for managing outpatient examinations. These different systems deal with different types of data, resulting in an ecosystem of heterogeneous solutions that are not easily integrated, making interoperability between these solutions difficult. Interoperability consists of the ability of an application to transfer data to another application (syntactic interoperability) and the ability to interpret the data that is transferred (semantic interoperability). Considering the diversity of types of solutions that can be used in the health context and the potential gain that can be obtained by integrating data from different solutions in this context, the interoperability requirement becomes important to be contemplated by Health Information Systems solutions. Health and that is why Middleware-based solutions emerge as an option for solutions that support interoperability between health systems. A Middleware consists of the layer of software that is between the technological and application levels, being above the operating system and below the application level, providing a common programming abstraction and aiming to facilitate the integration between systems through strategies of abstraction of the schedule. By promoting high-level abstractions for programmers, middlewares facilitate integration steps between heterogeneous systems, making code more portable. In the health context, there are a series of middlewares that propose to support the integration between already used applications. However, the review of the state of the art carried out in this work shows that the solutions are focused on different areas (such as image exams, sensing, and mobile computing solutions) and do not use open interoperability standards that facilitate the exchange of information from different clinical data, such as HL7 FHIR (Fast Healthcare Interoperability Resources - HL7 FHIR). It is based on this context that this work describes a Middleware Based on HL7 FHIR for Health Information Systems. The proposed solution aims to promote interoperability between different applications and health information systems, making the details of their implementations clearer become invisible to developers, focusing only on the communication layer and enabling integration between these different solutions. In this sense, throughout the work, the requirements of the solution will be described, as well as details of its architecture and implementation, resulting in the realization of a Proof of Concept carried out with 5 applications integrated through Middleware, namely: the Remote Assistance Platform (PAIR); the Clinical Data Platform (SigSaúde); the Information System for the Management of ICU Beds (Beds); the System for the Management of Notifications of COVID Cases in RN (SUVEPI); and the Testing Management System for Suspected COVID-19 Patients in RN (Exams).

Chatbots are software that make use of Artificial Intelligence (IA) to provide interaction between humans and machines through Natural Language (NL), resembling a conversation between two people. It is a current topic that has received a lot of attention in recent years. However, this type of application presents a behavior that is sometimes unpredictable, as different users can request the same set of actions in different ways and the application can return something different to each of the users. This makes the software testing activity somewhat complex, since it is not simple to define the test oracle. The test oracle is the process that verifies the test result, judging the response of the application, program or function as correct or incorrect through specification and planning. In this scenario, an appropriate testing technique to deal with the oracle problem is Metamorphic Testing (MT). This technique was proposed as a solution to minimize the oracle definition problem, since the application response verification process is performed by the definition of Metamorphic Relations (MRs), elements that assume the role of the oracle. Metamorphic relationships are properties of the test object application that define the behavior that the application must maintain when submitted to a given test case. However, despite Metamorphic Testing being a well-established technique to deal with the problem of the lack of a test oracle, it is still little explored in the context of chatbots. Thus, the objective of this work was to investigate the applicability and effectiveness of the MT technique in a chatbot application developed by a public company. Therefore, it was necessary to develop a tool to automate the test process by applying the MT concepts. For the evaluation of the technique we selected 22 Source Test Cases (STCs) divided into 3 Intents and input this information into the automation tool to run the tests of the 20 Metamorphic Relations. Through the execution of the tool, it was possible to generate more than 4100 Follow-up Test Cases (FTCs), derived from the STCs. The verification took place in two test batteries using two versions of the same application with 3 months of development difference. With the application of the MT technique, it was observed that 25% of the FTCs revealed defects, representing more than 900 disapproved FTCs, per version tested.

The purchasing and contracting processes are essential to meet the operational needs of any organization, thus enabling it to reach its goals. This reality is also common to all the agencies that make up the Public Administration. Unlike the private sector, public purchasing and contracting must be carried out through competitive biddings, ensuring wide disclosure and transparency of information. Over the years, the Brazilian legislation dealing with this issue has been improved several times. The electronic auction and the Price Registration System (SRP), for example, were important instruments that contributed to the modernization of this important process. The supply contracts (Acts of Price Registry) that have their origin in the electronic bidding conducted by SRP require an efficient information management process from public managers, so that the public budget is executed optimally. However, some information related to this domain, contained in the System for the Administration of General Services (SIASG), are dispersed in different modules, favoring the use of parallel controls, prone to problems such as redundancy and inconsistency in the data. Without a trusted place to consult this information, managers' decision-making can be compromised. Although the literature on the themes and methods covered in this work is extensive, no solution to the problem was found. In this master dissertation, we present the specification, design, and evaluation of a software architecture that aims to structure the essential information for the management of price registration minutes, which aims to contribute to the decision-making process related to the application of budget resources. The execution of these steps was based on a process model widely experienced by the industry, with some adaptations. Design Science Research (DSR) was the research method that guided the conception of the proposed artifact. The results of the evaluations show that future instances of the proposed architecture have the potential to solve the problems identified.

Datacenters are critical environments for the availability of technology-based services. Aiming at the high availability of these services, performance metrics of nodes such as Virtual Machines (VM) or VM clusters are widely monitored. These metrics, such as CPU and memory utilization, can show anomalous patterns associated with failures and performance degradation, culminating in resource exhaustion and total node failure. The field of early detection of anomalies, that is, patterns in data with different behavior than expected, can enable remediation measures, such as VM migration and resource reallocation before losses occur. However, traditional monitoring tools often use fixed thresholds for detecting problems on nodes and lack automatic ways to detect anomalies at runtime. In this sense, machine learning techniques have been reported to detect anomalies in computer systems with online and offline approaches. This work proposes the application called DC Health, as an approach to anticipate the online detection of anomalies in datacenter nodes. The purpose of DC Health is to detect anomalies in the behavior of hosts and alert datacenter operators so that investigation and remediation measures can be taken. For this, this research was conducted from i) Systematic Literature Mapping, ii) problem modeling from real VM data and iii) DC Health evaluation using the prequential method in 6 real-world datasets. The results showed that DC Health excelled in constant memory usage and detection accuracy above 75%. As future work, it is mainly expected to evaluate the detection tool in cloud computing scenarios and develop automated mechanisms for diagnosis and remediation.

In recent years, one of the major challenges in education has been to predict students’ performance in a certain and reliable way in order to apply different strategies toimprove their academic deficiencies. In this way, there are many works focused on finding Machine Learning (ML) models individually, but few use the knowledge acquired from student data of a course degree to predict the students’ outcome of other curses’s degree. Therefore, the main goal of this work is to analise the portability of ML models by using Transfer Learning (TL) in students’ logs extracted form the Moodle system, aiming to predict undergraduate students performance in different courses. Through the applied experimental methodology, each of the two types of groupings formed by the disciplines
will be evaluated: those formed according to the undergraduate course and those according to the activities used in Moodle. The data extraction of each group will be performed from the Moodle logs, using the following evaluation methods: cross-validation and hold-out. With this, it will be possible to know whether these evaluations, all performed on the predictive models with the J48 algorithm, tend to show different results in relation to the portability of forecasting models. For evaluation, two scenarios were developed for the execution of experiments, so that each experiment consists of two parts: the choice of models, using the AUC ROC index for Experiment 1 and the F-Measure for Experiment 2; and the validation of the models, using the Precision index for Experiment 1 and the Recall for Experiment 2. The results, even in the evaluation phase, allow us to affirm that it is possible to apply transfer learning between models of the same group in some cases.

This work was conceived from experiences in the Public Ministry of Rio Grande do Norte (MPRN) regarding the creation of a platform to organize products open and closed MPRN data. Thus, firstly, a systematic mapping of the literature, in order to seek the best practices related to data publication, as well as existing processes or  methodologies, such as other initiatives developed by other Public Ministries. The results were then analyzed.
After doing the analysis, it was found that there is a need to improvement of data sharing processes by the Public Ministry. Therefore, we are developing a new process to support sharing open and closed data through the MPRN platform in Data. After that, some case studies were carried out in order to verify the usefulness of the process proposed, as well as to identify the difficulties faced, evaluating its general efficiency and if it can really contribute to improving productivity within the Ministry Public of RN.

Automated machine learning (AutoML) is a field of great interest to both industry and academia. AutoML has allowed developers working on machine learning (ML) applications to achieve satisfactory results with little to no ML expertise. More recently, AutoML tools focused on deep learning (DL) models have proven especially useful to applications where domain-specific algorithms are predominant, as in computer vision (CV) tasks. Still, AutoML tools focused on simpler ML pipelines remain an important alternative, since DL models have high computational resource requirements and offer reduced model interpretability.
Among the main AutoML approaches, AutoML based on an algorithm configurator (AC) is commonly used to produce simpler pipelines, whereas neural architecture search (NAS) is used to produce deep learning models. These two approaches also intersect, since an AC can be used as a NAS algorithm.

In this work, we study the application of the irace AC to both these AutoML methods. The irace configurator has been successfully applied to design effective algorithms for optimization problems, but it has not yet been applied to AutoML. Our assessment is performed in two stages. First, we propose an irace-based AutoML tool to produce simple and effective ML pipelines. The tool is dubbed iSklearn, for which we define a configuration space and setup. We demonstrate that iSklearn is able to produce effective ML pipelines using irace as its AC, with comparable performance to more complex ensembles produced by AutoSklearn, an established configuration-based AutoML tool. Moreover, we show the benefits of the configuration space and setup proposed for iSklearn, even when coupled with another AC. In the second part of our work, we assess irace as a NAS algorithm. To do so, we evaluate irace on NAS-Bench-101, a recent NAS benchmark for the CIFAR-10 CV dataset. We benchmark irace not only as to final-quality assessment, but also as to anytime performance through a bi-objective formulation. Results demonstrate that  irace can be used a NAS algorithm, obtaining comparable results to the best NAS algorithms included in NAS-Bench-101 in terms of final quality. However, further work is required to improve its anytime performance in this context. Finally, we discuss other design choices made in the original benchmark, showing how they affect the relative performance of NAS algorithms, and provide guidelines for improving the assessment of NAS algorithms through the use of NAS-Bench-101.

In order to obtain commercial advantages over competitors, companies in all segments are improving their relationship with customers. The supermarket segment is no different and investments in customer relationship management (CRM) are increasing over the last years. The first step towards a successful CRM strategy is to know customers better, for which customer segmentation plays an important role.
In this work, we segment customers from Nordestão, the third largest supermarket chain in the Northeast of Brazil. To do so, we adapt the recency-frequency-monetary model, enrich it with new features, and use Gaussian mixture models to clusterize the data. Furthermore, we employ a well-established a priori segmentation from the Brazilian supermarket literature. For each a priori segment, customer groups were obtained for each retail store, with each group representing a different customer profile. Among the most interesting are prime and opportunity customers, who respectively focus on high-end and on sale products. Importantly, most of the behaviours are consistent across the different stores, varying only as to store-specific parameters. We conclude our work with a further algorithmic validation and interpretabilityanalysis of our findings.

Domain-Driven Design (DDD) is in creasingly adopted as a particular way of modeling systems. Among DDD’s strategic design patterns is the context map, a powerful tool that helps development teams understand the business domain. Context maps are usually created from the analysis of existing systems, initial requirements or based on the experience of the team, which can lead to the construction of domain models that do not reflect the reality of the organization. In this sense, the analysis of business process models is presented as an alternative to extract domain information and create context maps, since these models capture the logic of organizations’ operations, providing a comprehensive and high-level view of the business. However, approaches for creating context maps from domain knowledge extracted from business process models are lacking. Thus, this research aims to propose the BPM2DDD technique, an approach to support the identification of information domains and the creation of context maps, which can be used in the stages of analysis and modeling of systems, even serving as an alternative for identifying possible software services. The technique allows the identification of domains, subdomains and limited contexts, based on the analysis of BPMN (Business Process Model and Notation) models. The technique was developed and applied in real business processes of the Urban Mobility Secretariat of the city of Natal (STTU), using the action research methodology, during the years 2019 and 2020, in the context of the development of new urban mobility systems. To evaluate the technique, multiple applications were carried out by business and systems analysts, using one of the STTU business processes, in addition to validations by the domain experts of the generated context map. The results indicate that the techniqueis applicable to software processes adopted by the industry, allowing the discovery of information that enables the creation of a context map more aligned to the organization’s domain.

One of the main research fields under Artificial Intelligence, on the context of Digital Games, is the study of Real-Time Strategy Games (RTS), which are commonly considered the successors of classic strategy games such as Checkers, Chess, Backgammon and Go, and impose great challenges to this area’s researchers due to the great complexity involved. Currently, the field aims to study the RTS using StarCraft I and II as stage of experimentation. The main feature sought in artificial agents developed to this kind of game is high performance, having as its main objective to defeat specialist human players. On this context it is inserted the generalization problematic, that is the capacity of an artificial agent of reusing previous experiences, from different contexts, to a new environment. Generalization is a very studied field by the scientific community, but still poorly explored on the context of RTS. By this reason, this work proposes the Beyond Star model, which consists in an architecture to generically represent the state-space of Real-Time Strategy Games, using as base deep reinforcement learning techniques aiming to learn effective strategies to be applied in several RTS environments. As a basis to the architecture, it was developed a platform titled URNAI, a tool that integrates several learning algorithms and environments, such as StarCraft II and DeepRTS. To analyse if the solution is capable of allowing agent generalization, trainings were done in DeepRTS and tests were carried out in StarCraft II. It was verified that the trained agents were capable of generalizing their knowledge from one environment to the other, showing a promising result that allowed to validate this work’s proposal.

The Internet of Things is booming and more and more devices are connected, generating massive amounts of data. Usually this large amount of data is not used to discover additional information. This is also a reality in healthcare, where large amounts of data can already be generated from devices that use open protocols or connected open source hardware sensors, such as environment sensors or human body sensors. This dissertation aims to propose an intelligence module that can, obtaining data from IoT sensors in health, return categorization and prognostic information of medical complications, thus bringing direct benefits to the patient's health as well as streamlining the patient's health. hospital care and cost reduction.

The food and restaurant sector are segments that are facing challenges to adapt to a new reality that requires less food waste, quality food and efficient management. Consumers in rich countries are estimated to waste around 222 million tonnes of food products and most of them are from restaurants. In Brazil, ANVISA requires restaurants to be monitored by nutritionists, one of whose tasks is to control the temperature of food preservation to verify compliance with its resolutions. In order for restaurants to avoid both food waste and working within these regulatory standards, efficient management is needed to optimize internal processes and make more accurate decisions. Therefore, the objective of this work is to create a smart platform for monitoring the quantity and temperature of a self-service restaurant buffet in order to minimize food waste, improve the handling of nutritional information and optimize its management. The system will be able to check the temperature consumption of food exposed in tanks of a buffet and that the data will be sent to restaurant sectors like cooking, nutrition and management in real time. The system will be created using the concepts and features of the Internet of Things (IoT), Big Data and Web Programming in JavaScript. Finally, software like Node-red, ThingSpeak, Arduino IDE among other tools will compose the framework of technologies used to develop this platform.

Software quality has become a key factor in the software development process today. To be able to measure the quality of different types of tests is today an important challenge to be met. The research conducted in this study considered the evaluation of the effectiveness of systems tests performed by MOPA Embedded Systems in the use of Ginga middleware for Digital TV. To perform the system test suite evaluation performed by the company, it was proposed a methodology to evaluate it through the code coverage criterion, followed by an exploratory study to investigate possible reasons for the results obtained from the coverage. Finally, exploratory tests were conducted to confirm the findings from the exploratory study phase. Through these steps, it was possible to establish an opinion on the quality of the tests performed, as well as the automation performed by the company.

This work will present an interface for non-invasive adaptation of Virtual Reality input
devices, for use with desktop platform games incompatible with such devices, with a focus
on games for rehabilitation, since the Universidade Federal do Rio Grande do Norte has a
ongoing research on application and development of new controller devices for therapeutic
games. In addition to being a tool for adapting joysticks, it will also contain a module for
therapy, this module besides collecting information can help the patient to correct the
movements performed, artifice useful when therapy is performed at home, away from the
therapist’s supervision.

For a variety of reasons, when there is a monitoring system that allows analyzing and correlating data so that relevant information can be extracted, even the smallest problem can be detected and corrected efficiently, reducing downtime.Thus, effective monitoring helps applications and services be available as long as possible. This work specifies the design and development of a monitoring system for software and hardware components for IoT based platforms.The hypothesis describes the proposed monitoring system will help to maintain the guarantee of the availability and reliability of IoT based platform components.

Increasingly, companies have cherished well-being, motivation and satisfaction of their employees since they are the most valuable resource of organization. Motivated employees tend to produce with more quality and productivity, helping to solve problems with better performance. When companies acts successfully to positively influence motivation and better work environment, receives in exchange not just greater engagement but also makes a good external marketing to attract new talents, being seen by external community as a company where “many others want to work”. To make this possible, it is necessary recognize and provide a feedback to employees on their performance. Depending on hierarchical complexity and quantity of employees and diversity of company positions, evaluate their performance may represent a big challenge, being necessary the use of many resources to make it happens in acceptable time. This work presents software that provide a performance feedback to the teams of a company by use of performance indicators, in addition to implement a public recognition model of employees in a view of good attitudes. To measure the impact of the software utilization, a study will be conducted before and after use this software, to measure their influence on the motivation of members of development teams and what the variation of satisfaction of them within the work environment.

Information systems have become increasingly robust, working with high volumes of data, objects and data processes. In the context of the Federal Institute of Rio Grande do Norte (IFRN), a system was developed to attend all the administrative and academic activities of the institution: SUAP, Unified System of Public Administration. Over the course of 11 years, the application has evolved from just one module to over 40, in addition to the dizzying growth of users -- which today permeates the number of 35 thousand. These aspects often impact directly and negatively on systems performance, leading software architects to seek alternatives to better scale their applications. In the case of the IFRN, even after high investments in infrastructure, SUAP performance problems persist. In order to make better use of the available hardware structure, it is necessary to understand the reasons for these performance problems, so that these computational resources are used dynamically and incrementally, managed more efficiently and able to withstand surplus loads without prejudice to application stability. Thus, the need to evaluate more comprehensive aspects of the current architecture of the application is presented, so that elements that impair its performance can be known. This work aims to perform a systematized analysis of the SUAP performance through the application of the Architecture Trade-off Analysis Method (ATAM). Based on the analysis performed, a set of load tests were identified that allow the identification of its main points of slowness. With this, it is hoped to contribute to the definition of a strategy for tests of performance in the SUAP that can be incorporated by the development team. 

The software architecture represents a set of design decisions, often taken before implementing the system, with the goal of achieving desired levels of software quality attributes. An architectural pattern provides a set of specific design decisions applicable to recurrent design problems. Violating these decisions may not only hinder quality attributes, but it may also lead to architectural degradation. The Model-Template-View (MTV) pattern implemented by the Django framework has a set of decisions that promote low coupling and rigorous separation between the parts of an application. However, during the evoution of an applicatoin, design decision may be violated. This way, we investiigate the detection of design problems related to the MTV pattern with the goal of aiding the prevention of architectural degradation in MTV applications. The main contributions of this work are the elaboration of a catalogue of MTV-specific design problems e the implementation of tool for statically detecting these violations. The catalogue and the tool were validated in the context of the Sistema Unificado de Administração Pública (SUAP) developed and maintained by the Instituto Federal de Educação, Ciência e Tecnologia do Rio Grande do Norte – IFRN.

Since the dawn of civilization, humans beings had always been fascinated by games, because they are fun and a great way to pass time. In recent decades, digital games became popular among young people in school age. Among the digital game styles available on the market today, the RPG (Role-playing game) has attracted attention, since its titles are several years in the bestseller list. This is due various features available, such as character customization and progression, interaction with the new and fictional world and problem solving in this world. In this scenario, we find the concept of quest, which are missions given to the player that must be accomplished for the conclusion of the game or to evolute the narrative. In an effort to facilitate the quest produce process, this work propose a methodology called Quest Design Canvas, which allows for faster production, with well defined creative process steps, with the purpose of, finally, obtain a  narrative line inserted within a set of missions within a mental map. So, it is intended to reach all the goals to be achieved with a digital game, even being educational, introspective or promotional. As validation, This methodology will be used to produce quests for the massive and multiplayer online RPG game (MMORPG) UFRN - The Video game.

The methodology based on competence aims at training of workers with a focus on skills to perform tasks related to their occupation in the labor market. But this new proposal brought challenges pedagogical and teaching teams, the act of planning has been an arduous task, considering so many processes and information that teachers need to understand and develop. Because of this, this paper proposes a solution through a software tool that minimizes the problems faced and provide resources to facilitate the understanding of the methodology. For it was developed an exploratory research with the goal of finding difficulties patterns through online forms and focus group were collected, with teachers from schools of SENAI-DR / RN, relevant data to formulate a proposed solution. The proposed application is currently under development.

The Learning Objects (LO) are importants modular and reusable resources to Distance Education thatt composes the Learning Management Systems (LMS), building a tool kit that can be used for teachers in their learning project and a course’s activity fluxs.The LOs can ber hosted directly into LMS or stored into Learning Objects Repositories (LOR), that are suitable to host them in centered form, describing their details using metadata and with eficient search engine for find the requested object. The stored objects in the LOR can be accessed and used in severals LMSs that include them, increasing the resource available variety to mount a course whith their evaluative activities. This dissertation aproachs the proposal of architecture for integration of the Distance Education Secretary’s (DES) LOR of Federal University of Rio Grande do Norte with LMS Moodle using tha LTI especification developed by IMS Global Leaning Consortion.

The border firewall exercises a fundamental role for corporate network security, acting as the first defense barrier delimiting a secure perimeter. However, the culture of Bring Your Own Device weakens the notion of secure domain, contributing to the creation of attack vectors inside the secure perimeter. Distributed firewall systems emerged with the proposal of protecting individual hosts against attacks originating from inside the network. In such systems, firewall rules are created centrally, then distributed and enforced on all servers that compose the firewall, restricting which services should be available. Meanwhile, software vulnerabilities can make network services vulnerable to attacks, since firewalls usually do not scan application protocols. In this sense, from the discovery of a vulnerability until the publication and application of patches there is an exposure window that must be reduced. Given this scenario, the present work presents a proposal of a self-adaptive distributed firewall system. This proposal is based on monitoring hosts and using a vulnerability assessment system to detect vulnerable services, integrated with components capable of deciding and applying firewall rules on affected hosts. In this way, our approach is able to respond to vulnerabilities discovered in these hosts, helping to mitigate the risk of exploiting the vulnerability. Our system was evaluated in the context of a case study that simulates a corporate network environment, where the results achieved demonstrate its viability.

A software product line (SPL) represents a family of related systems that share commonalities and variabilities to meet the needs of a specific market or mission. The adoption of SPL has been applied in several areas in the software industry due to the benefits achieved, such as reduction of development costs, quality improvement and reduction of time to market. However, distinct scenarios can be found when developing a SPL, which lead to 3 approaches for adopting a SPL: (1) proactive approach: there areno software being developed and a SPLis developed from scratch; (2) reactive approach: there is a SPL which is evolved to address new features and products; (3) extractive approach: SPL is developed from the assets of a system or a set of related systems that are already in production. In the context of the extractive and reactive approaches, this dissertation aims to define a method to develop a SPL from existing systems implemented in the Java language. A practical study that extracts and evolves a SPL of the domain of space allocation management is used as a background to define the method.  

The necessity of software evolution for those which solve daily problems became even more frequent and the software engineering had to be adapted in order to be able to delivery products with good quality in tight dead lines. In order to the software continues being useful during its life cycle, to the main purpose whose was developed, its necessary to apply changes or include new features due to changes which happens in the business.

Rewrite the same block of code is not a recommended approach on software development, because it spreads code in a distributed and disordered way. Applying the code reuse, the same block of code already wrote is referenced by different use cases through the dependency chain and method calls, where different parts of the code, which are being relate to differents funcionalitys, going to depend to the same component. Changes applyed to a block of code which has direct or indirect relation with differents use cases may lead to umpredictable fails, depending on the number of different artifacts related and the extension of dependency chain related to the artifact which was modified, this change may cause a impact on another use case which, by a first look, does not have any relation which the modified block of code. Predict impacts from in a specific artifact is a task which demands time to analysis, deep knowledge of the source-code and test effort.

This paper presents an approach to automatize the identification of possible indirect conflicts using the developed tool, whose can determinate which use cases are more defect prone by source-code changes, providing a more effective direction to the test's efforts. A Study Case was elaborated, assessing a real project of extensive dimension whose doesn't have a automatized test case suite, and the developed tool was able to identify the indirect conflicts on differents cenarios and besides, the tool was able to proof in a empiric experiment which the major failures, caused by indirect conflicts could be avoided if the tool were be used during the development fase.

Para alunos que ingressam no primeiro ano de um curso técnico integrado, a matemática é uma das disciplinas de difícil entendimento. Fatores como falta de atenção, problemas com interpretação de texto e problemas no aprendizado no ensino fundamental, colaboram com as reprovações na disciplina, gerando um aumento no desestímulo do aluno no curso. Com base em relatos dos professores da disciplina de matemática feitos ao setor pedagógico da escola sobre as dificuldades desses alunos em resolver as atividades trabalhadas em sala de aula e após a realização de entrevistas com um grupo de alunos, os referidos fatores de dificuldade foram identificados e possíveis soluções foram traçadas. Uma delas é a implementação de um jogo digital, a ser trabalhado no ambiente escolar, que proporcione o exercício da matemática fundamental, como forma de trabalhar aspectos cognitivos e o conhecimento dos conteúdos da disciplina, gerando dados sobre o desempenho do jogador, proporcionando um feedback ao professor sobre dificuldades e/ou facilidades na resolução das questões. Este trabalho apresenta a proposta de jogo concebida.

In general, laboratories or labs are workplaces that provide controlled conditions for experiments and measurements to be performed. For this reason, controlling temperature and humidity is an important requirement that needs to be achieved in order to guarantee the reproducibility of processes carried out in labs.

Aiming to propose efficient environmental controlling mechanisms, specifically for chemical analysis laboratories, we present in this work an intelligent environmental control system based on sensors network and pattern recognition. Our prototype uses its own data generated by sensors distributed in the environment to identify a pattern of behavior. Through the use of machine learning algorithms, the system identifies the classes within the data (clustering), does the training and testing procedures (classification), so that it can generalize what was learned. Finally, the rules are created in association with previously identified classes in order to control air conditioners, both the main and the spare ones, plus the dehumidifier. In this sense, the prototype keeps temperature and humidity stable and in an effective way.

Quality and time are important factors in software development process, because  we  expect  to  deliver  a  high  quality  product  with  acceptable  cost  and deadline. These factors are even more relevant to innovative environments and small businesses, like a startup. Many companies still don’t carry out activities to improve quality of use of software they had developed. Lack of knowledge of HCI evaluation  methods  and  expectation  of  large  increase  in  coast  and  deadlines have been arguments to maintain the current culture in some companies. This work compares three HCI  evaluation methods well known in literature: heuristic evaluation, communicability evaluation and usability testing. We investigate how they  contribute  to  designer’s  work  during  conception  of  their  interaction  and user interface solution. We compared the activities, time of execution, used materials, profile and number of involved evaluators and users, results and difficulties of evaluators during the execution of methods. This comparison was made by a designer as part of SoftUrbano’s juridical system for smartphone. The results can better inform professionals about the HCI evaluation practice and contribute to changing the software development culture in many companies, especially small businesses and startups.

Regardless of the area of operation and due to business needs, given
the pace of innovation, public and private organizations are
constantly adapting to technological changes, either with the best
market practices or upgrading their technologies. These frequent
changes often end up creating in the company a very heterogeneous
environment, where the systems involved in the business were developed
using various platforms (for example, programming language and / or
database) operate in an isolated way. In situations like this, the
company needs to adopt strategies to provide interoperability between
their systems and thus ensure a good communication, aiming to promote
the exchange of information between departments and the integration of
business rules. The main objective of this work was to propose and
develop strategies enabling the integration of systems in these
heterogeneous environments, regardless of platform and development
language of the systems. In order to solve the current communication
problems and performance between different information systems in the
Department of Distance Education (SEDIS) of the Federal University of
Rio Grande do Norte (UFRN), it was developed a so-called middleware
INTEGRA designed on a service-oriented architecture. The proposal It
was created for the SEDIS, but can be extended to other public
institutions of higher education. One of the advantages of the
architecture proposed is to provide the addition of new systems
without causing a severe impacts of adjustments and changes, ensuring
general maintainability of the software. The middleware creates
communication between different help desk systems and issue tracking
offering a range of services through a communication protocol that
standardizes the exchange and routing of messages. Through the
proposed strategy as showed in the poc, it was possible to get a
balance between the principle of decoupling and the principle of
uniformly distributing responsibility, thus guaranteeing a low
coupling and high cohesion, desirable principles in Software
Engineering.

Usually, the software architect responsible for the software architecture definition and evaluation cannot rank which non-functional requirements must be prioritized during the development of their systems. Because of that, failures may happen during the system execution requiring more time and resources to fix it. In many cases, due to the inexperience of architects or the need for rapid deployment of a system, the non-functional requirements are not considered in the software architecture definition phase and the its documentation is absent or incomplete, making the software architecture difficult to be accessed and understood by others team members and also more difficult the maintenance of its components/modules and relationships. This work raises with public and private software companies which are their main strategies used to define software architecture in terms of its documentation and to ensure it is suitable for non-functional requirements of interest. Our study had the participation of 17 software architects to answer the survey. With the completion of the survey we identified that the non-functional requirements for performance and reliability are the most important to be addressed by the architecture and even with the existence of some approaches to evaluation architectures, they do not seem to be well disseminated and / or used among architects. Besides that, we propose a guide to help software architects in the task of achieving such non-functional requirements during the evolution of software systems. The proposed guide prioritizes the non-functional requirement of performance.

This work presents a tool composed by a domain-specific language used to describe textually performance tests, and by a component that uses those described tests to generate projects in different performance test execution platforms. The purpose is to use concepts defined in the language to abstract the concepts of each platform, which are often modeled differently (regarding nomenclature and/or structure) and even dependent on the tool rather than just the domain. The proposed tool supports SOAP, REST and web applications performance tests to JMeter and Silk Performer. It also allows extension to new test types and target platforms. Studies has been made to evaluate the tool’s usage: tests from a company were rewritten with the DSL and then projects were generated on the target platforms, to be executed. This study also served to help in the development of the tool. It was possible to observe the feasibility of using the DSL to different application types, customizing it according to concepts and characterists of platforms and organizations.