Prediction; University Libraries; Time series; Random Forest; ARIMA. 

Providing books to users is one of the most important aspects of a university library. Consequently, forecasting the use of the collection is a vital activity for planning your actions. This allows the library to manage its resources more effectively and anticipate the needs of its users, proactively meeting their demands and expectations. Given the above, this objective work forecasts the loan of books from the current collection of the Zila Mamede Central Library (BCZM) through historical data on book loans. To carry out a modeling, Random Forest and ARIMA were used with data from 2011 to 2019. The study used an adjusted model to predict aid for the year 2019 through two approaches: weekly and monthly. These models were evaluated using the error metrics: MAE, MSE and RMSE. The results indicate that, in grouping by week, both models performed similarly, with Random Forest performing slightly better in predicting books grouped by week than ARIMA. For the monthly grouping of books, the models showed unsatisfactory performance, as they obtained high error rates. Thus, it can be concluded that both the ARIMA and Random Forest models are promising in terms of predicting book-to-book borrowing, when grouped by week. In addition, both models were able to identify credit trend patterns, making their application plausible for the Zila Mamede Central Library.