Han Jianguo, Marifel Grace C. Kummer
ABSTRACT
This study aimed to develop an asset management system with predictive analytics for Lianxiang Company. It employed a multi-layer feedforward artificial neural network with backpropagation learning to predict consultees’ enrollment status, alongside time series analysis to forecast the company’s future demand for office supplies and devices. The neural network model was trained, tested, and validated using 21,690 student samples who inquired at G College from school years 2021 to 2024, with data compiled from admissions records and institutional research databases. Seven input variables were used: Applied Program, Graduated Students’ Average Employment Rate of the Program (SER), Total Enrollment Fee-to-Salary Ratio (TSR), Program Rating (PR), whether the consultee’s NCEE score met or exceeded the least admission score (IsAbove), Total Enrollment Fee-to-Average Monthly Income of the Consultee’s City (TIR), and willingness to accept program adjustment if the applied program was not met (AcpAdj). A random search tuner was applied to determine the optimal combination of hyperparameters for improved model performance. For demand forecasting, historical consumption data from January 2022 to December 2024 were analyzed using scatter plots, autocorrelation functions to identify seasonality, moving average methods to compute seasonal indices, and linear regression to model trends. Predictions generated by the developed models achieved less than 6% Mean Absolute Percentage Error (MAPE) in estimating major supply and device consumption, indicating strong predictive accuracy. These results demonstrate that the system can support administrators and procurement staff in planning purchases more effectively, minimizing over-procurement, strengthening internal control, and improving overall operational performance.
Keywords: Artificial neural network, asset management system, enrolment prediction, predictive analytics, time-series forecasting
https://doi.org/10.57180/burn4695