Evolutionary fuzzy neural inference system for prediction of project success

Project performances are affected by various factors in different construction stages. Due to the impact of the factors varying with delivery, the success of project is hard to predict. In addition, those impact factors containing uncertain, vague, and incomplete information increase the predictive difficulty. The objective of this research is to adopt the Evolutionary Fuzzy Neural Inference Model (EFNIM) to develop a dynamic project success prediction model. To achieve the goal, the major factors affecting the project success in the construction time frame were identified. The proposed method can assist project mangers to make proper decisions enforcing the management and control of the influencing factors. This study firstly developed a dynamic project success prediction database based on the research results of the CAPP (Continuous Assessment of Project Performance) system. CAPP system was used to identify the significant factors influencing the project success. Combining the Data Mining technique, 52 historical construction projects were clustered into groups using the K-means method. Cases with higher similarity were categorized within each cluster as training sets for EFNIM. Furthermore, the predictive results before and after clustering were compared to prove that training cases preprocessed through Data Mining treatment can improve the prediction accuracy of the EFNIM.