Proposing a Hybrid Approach to Predict, Schedule and Select the Most Robust Project Portfolio under Uncertainty

Abstract

Suitable project portfolio selection in inconsistent economy that can reduce the portfolio risks and increasing utilities for investors has gained significant research attentions.   This article addresses the project portfolio selection in which conventional certain (1) prediction, (2) optimization and (3) clustering approaches cannot be used to face uncertainty. To predict the real value of affecting project risk parameters, neural network has been used; Then to determine the optimized sequences and procedures, the proposed model have been evaluated using the multi-objective shuffle frog leaping algorithm (SFLA) by robust optimization approach; To suggest different risk criteria, K-means algorithm utilized to categorize the candidate projects and differentiating the clusters.  As the proposed hybrid methodology is studied on 420 different construction projects in an Iranian construction company in two economic stable years and an instable year in Iran real estate market. The results show 96 percent prediction-optimization capability due to different desired criteria.