Validation of a Model Developed for Value Engineering Approach Performance on Gravel Roads Maintenance Projects in Tanzania

Abstract

A validated model offers a consistent framework for decision-makers to understand the critical factors influencing performance across different data sets. Gravel roads (GR) are vital in Tanzania, comprising over 75% of the road network, with 65% in poor condition. Value engineering (VE) has emerged as a promising tool to enhance GR maintenance, accounting up to 83.3% of the variance, as demonstrated by a model developed using partial least squares structural equation modeling (PLS-SEM). This paper therefore evaluates the validation of a Model Developed for Value Engineering Approach Performance on Gravel Roads Maintenance Projects in Tanzania using split data methodology and the PLSpredict tool in SmartPLS, which assesses the out-of-sample predictive power of PLS-SEM. The results revealed that the model exhibits medium predictive relevance for the corresponding constructs, with 65.38% and 61.22% of indicators in the PLS-SEM yielding smaller prediction errors compared to the naïve linear regression model (LM) benchmark for training and validation data sets, respectively. These findings validate the model’s ability to predict future data effectively, supporting its use for decision-making and strategic planning. The study concludes that adopting a VE approach to enhance GR maintenance projects in Tanzania and other regions is crucial, given the model’s predictive relevance across different data sets.