Academic journal article Library Philosophy and Practice

Improving Efficacy of Library Services: ARIMA Modelling for Predicting Book Borrowing for Optimizing Resource Utilization

Academic journal article Library Philosophy and Practice

Improving Efficacy of Library Services: ARIMA Modelling for Predicting Book Borrowing for Optimizing Resource Utilization

Article excerpt


Academic libraries are an essential component of colleges and universities throughout the world and play a central role in academic life (1). Regular library usage by students improves critical thinking (2,3), student learning and engagement, collaborative learning, student-faculty interactions, and academic challenges (4,5). Even the academic performance of students improves with library usage (6,7). Academic librarians face immense pressure to meet the needs of ever-increasing users and commitments for modernization and improvements (8). Further, with the advent of information-communication revolution, academic libraries are exploring service developments to support a series of new scenarios such as publication and scholarly communications; extensive use of digital resources; increasing heterogeneous student population; ICT-based learning and distance learning. Hence, libraries with limited resources, particularly in terms of staffing face enormous challenges in management (9). Librarians need to predict library usage over time, as this helps in planning the resources required to meet the usage. The future usage might not be stationary and there could be periods of huge demand, interleaved with long periods. A model for predicting the library usage should take into account this variability to be accurate in its prediction. Hence, seasonal prediction models are likely to improve the accuracy of prediction. The study proposes a prediction model that forecasts library usage for a period of 12 months.

Time series analysis allows examining temporal dependency of data and generating predictions. Time series consists of two components: trend and seasonality. Trend depicts a linear or non-linear component that is non-repetitive within the time range, whereas seasonality repeats at systematic intervals over time (10). Time series analysis using Autoregressive Integrated Moving Average (ARIMA) models are very effective in modelling the time-dependent data as they structure time dependence between the observations (11). The predictions made using ARIMA models are more accurate than those obtained by other statistical methods (12-15). Earlier, ARIMA models are successfully forecasted household electric consumption (16); software evolution (17); incidence of several infectious diseases (12,18-25) and use of health facilities (26,27). However, such predictions are not available in the context of library usage. Hence, in this study, the authors used Box-Jenkings ARIMA modelling to monitor and predict book borrowing in Sri Venkateswara College of Engineering and Technology, Chittoor. The data pertaining to book borrowing from 1998 to 2013 fit into an ARIMA model using Box-Jenkings approach (28). The fitted model has been used to predict book borrowing for the year 2014 with 12-steps ahead and 1-step ahead approach. Root mean squared error (RMSE) was used to compare the predictive power of the two approaches. Further, both approaches have been tested using Wilcoxon signed rank test (29).

Objectives of the study

* To detect any seasonal pattern in the book borrowing services and identify the underlying factors associated with it;

* To identify trends in the distribution of book borrowing over time; and

* To validate the usage of ARIMA model with seasonal information to accurate forecasting of book borrowing.


The study was conducted at the college library of Sri Venkateswara College of Engineering and Technology, Chittoor established in 1998. For the analyses, monthly data on book borrowing for the period 1998-2013 was used to fit an ARIMA model, which later used to predict out-of-sample book borrowing for the year 2014. Since estimation procedures are available only for stationary series, Augmented Dickey-Fuller (ADF) unit root test and Box-Jenkins approach on ARIMA modelling of time series, used to check the stationary series of the data, which was ascertained by observing the plot of book borrowing that involves a four-step process (11,15,24,28-33). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.