Using data mining techniques to predict and detect important features for book borrowing rate in academic libraries

Library usage and book borrowing are important factors in student's academic performance. It is highly essential for academic institutions to estimate how frequently their libraries are used by students and how often the materials in these facilities will be needed by students during the academic session. In this paper, we present a classification method to predict hook borrowing rate in academic libraries by students, based on their library usage behaviors. We conducted a survey of 200 university students on their usage of the library and used this data to establish a correlation between features and outcome. We tested several types of tree classification with different parameters and used the elimination method on features to identify the best possible parameters for prediction. We reached a %71.9 accuracy rate during training and %72 on test data. We identified that some of the features from the survey questionnaire may he irrelevant to classification. We used Python libraries during the building and testing of the classification methods.