An Endogenous Project Performance Evaluation Approach Based on Random Forests and IN-PROMETHEE II Methods

In order to identify the best or poorest alternative project by an overall ranking result in the scenario of assessing multiple infrastructure projects, multicriteria decision aid methods need to be incorporated into evaluating project performance. Most previous methods for assessing infrastructure project performance may not be applicable to frequent cases with numerous evaluation criteria but inadequate observation data. This paper proposed an objective performance evaluation approach from annual field-survey data through Random Forests and IN-PROMETHEE II methods together. Random Forests method is employed to predict performance values under selected criteria as the single-valued performance scores. IN-PROMETHEE II method is further developed to quantify the preference index among different projects under each criterion. By calculating a weighted average of single-criterion preference index, the multicriteria preference index can be obtained to determine the ultimate ranking of alternative projects. A comprehensive empirical study reveals that this approach is able to successfully avoid subjective bias. It is helpful in tracing decisive factors of project performance for practical projects in multicriteria cases. The analysis results have proved that the proposed method can be widely used in performance evaluation of complicated infrastructure projects.