Predictive Modeling of Library Circulation Trends Based on Time Series Analysis

Abstract

Time series analysis forecasts future patterns based on past data, but it might not take into consideration abrupt changes or outside influences that could have a big impact on circulation patterns. Unexpected occurrences could cause trends to diverge, modifications in user behavior, shifts in the economy, or the arrival of new technology. In this manuscript, Predictive Modeling of Library Circulation Trends Based on Time Series Analysis (PMLCT-TSA-DTGMGCN)is proposed. Initially, the data collected from the Fine Free Public Library Circ Totals by Month (FFPLCTM) are given as input. Afterward, the collected data are fed to pre-processing. In the preprocessing stage, input data is normalized using Privacy-Preserving Distributed Kalman Filtering (PP-DKF). The preprocessed data is then fed into a DTGMGCN to predict library literature circulation. In general, the DTGMGCN predictor does not indicate how to modify optimization algorithms to find the ideal parameters for precise library literature circulation prediction. Hence, the Pufferfish Optimization Algorithm (POA) is to optimize the weight parameter of DTGMGCN which accurately predict the library literature circulation. The effectiveness of proposed PMLCT-TSA-DTGMGCN approach is implemented in python and evaluated through performance metrics, like accuracy, precision, recall, F1score and error rate is analyzed. The performance of the proposed PMLCT-TSA-DTGMGCN approach contain 27.36%, 26.42% and 28.17%high accuracy; 28.26%, 25.42% and 29.27% high precision when analysed to the existing methods like methods A Seasonal Autoregressive Integrated Moving Average with Exogenous Factors Forecasting Model-Based Time Series Approach (SARIMAX-FMTSA-LSTM), Deep Transformer Networks for Anomaly Detection in Multivariate Time Series Data(DTNAD-MTSD-BNN), and Short-Term Traffic Flow Prediction for Urban Road Sections Based on Time Series Analysis(STTP-TSA -BILSTM) respectively.