Predicting patron satisfaction on library service using machine learning approaches: A case study in Indonesia

Abstract

The advancement of information technology has diminished the traditional role of libraries, necessitating modernization to enhance patron satisfaction. This study focuses on predicting patron satisfaction with library services in Indonesia using machine learning models. By examining key factors such as collection, service, infrastructure, building, and staff, the research aims to identify the most significant contributors to patron satisfaction. Various machine learning algorithms, including KNN, Logistic Regression, CART, Random Forest, LightGBM, CatBoost, and XGBoost, were compared to determine their predictive performance. Data from the National Library of Indonesia’s Reading Habit Survey in 2023, encompassing 10,200 respondents, was utilized. The findings reveal that LightGBM achieved the highest accuracy and lowest RMSE, highlighting its efficacy in predicting library patron satisfaction. Key predictors of satisfaction identified include staff, service, and infrastructure. These insights can guide librarians and policymakers in enhancing library services to meet patron needs better. Future work will explore advanced machine learning techniques and expand the study to other public service domains.