Predicting Completion Risk in PPP Projects Using Big Data Analytics

Abstract

Accurate prediction of potential delays in public private partnerships (PPP) projects could provide valuable information relevant for planning and mitigating completion risk in future PPP projects. However, existing techniques for evaluating completion risk remain incapable of identifying hidden patterns in risk behavior within large samples of projects, which are increasingly relevant for accurate prediction. To effectively tackle this problem in PPP projects, this study proposes a Big Data Analytics predictive modeling technique for completion risk prediction. With data from 4294 PPP project samples delivered across Europe between 1992 and 2015, a series of predictive models have been devised and evaluated using linear regression, regression trees, random forest, support vector machine, and deep neural network for completion risk prediction. Results and findings from this study reveal that random forest is an effective technique for predicting delays in PPP projects, with lower average test predicting error than other legacy regression techniques. Research issues relating to model selection, training, and validation are also presented in the study.