RISK-BASED PREDICTIVE MODELLING FOR HIGHWAY PROJECT DURATIONS IN NIGERIA USING MULTIPLE LINEAR REGRESSION AND ARTIFICIAL NEURAL NETWORK

Abstract

Across the globe, highway infrastructure plays a pivotal role in national development by supporting the movement of people, goods, and services, making the timely delivery of such projects essential for economic growth and public welfare. Highway infrastructure projects attract huge budgetary allocations; hence, their timely completion is of significant importance. In Nigeria, schedule overruns on highway projects remain a persistent challenge, and previous studies on forecasting highway project duration relied mainly on conventional methods, which have yielded limited accuracy. This study presents a more robust approach to predicting highway construction durations by integrating both contemporary and traditional modelling techniques “Artificial Neural Networks (ANN) and Multiple Linear Regression (MLR).” A dataset comprising 103 completed federal highway projects executed between 2002 and 2022 was compiled using a snowball sampling strategy, drawing from the 2017 Federal Ministry of Works and Housing publication as well as additional inputs from highway engineers and quantity surveyors across Nigeria. For each project, a key professional was selected using the purposive sampling method to solicit detailed information on schedule-related risk factors via a structured questionnaire. These identified risks, combined with historical schedule data for each identified project, were employed as predictor variables for developing the schedule estimation models. Comparative analysis of model performance indicated that the ANN technique produced significantly more accurate duration forecasts than the MLR model. The study contributes to a practical and data-driven predictive tool that can assist government agencies, consulting firms, and contractors in enhancing the reliability of schedule planning and mitigating delays in future highway infrastructure delivery.