Neural Network-Based Prediction of the Performance Budget for Road Defects and Highway Maintenance

Abstract

The goal of this study is to create a model that can forecast the allocation for maintenance of roads and bridges in DPWH in Region XII. The study comes up with a new way to improve road maintenance, which is important for a country's social and economic growth. In this research, artificial neural networks (ANN) were employed to make a prediction model that could accurately predict how much it would cost to maintain roads and bridges in the future. At eight (8) District Engineering Offices in the DPWH Region XII between 2019 and 2022, the researcher used the maintenance section's prior allocation. In this study, final allocation served as the output variable, and the manipulated factors utilized to run the model were road length, road equivalent maintenance kilometer (EMK), bridge length, bridge EMK, and road condition. Levenberg-Marquardt (LM) based on ANN was utilized as the training algorithm (TA), while the Hyperbolic Tangent Sigmoid (HTS) function was used as the transfer function (TF). The outcomes of the analysis demonstrate that the model attained R values of 0.99197, which are approaching the optimal value of 1.00. It shows that neural network models perform 118 times better than multiple linear regression (MLR) models regarding prediction accuracy. This study provides a model for government agencies to use in predicting allocation costs for road and bridge maintenance more precisely.