Hybridization of CBR and numeric soft computing techniques for mining of scarce construction databases

Due to the nature of construction projects, collection of sufficient data is usually challenging for knowledge discovery in construction databases. Previous researchers had explored the capabilities of various artificial intelligence (AI) techniques for the mining of construction databases including symbolic reasoning techniques (e.g., case based reasoning, CBR), and numeric reasoning techniques (e.g., artificial neural networks, ANN, and neuro-fuzzy system, NFS). Both of the above paradigms own their merits and drawbacks on data mining. This paper proposes a hybridization of both symbolic and numeric reasoning techniques, to form a new data mining technique that can achieve a higher mining accuracy and overcome the restrictions of traditional numeric reasoning techniques on data scarcity problems. The testing results show the proposed hybridization can improve relative system accuracy by 44% (ANN) and 68% (NFS) compared with traditional CBR, or by 33.62% (ANN) and 72.88% (NFS) compared with traditional ANN and NFS, respectively. It provides profound potential for improving performance of traditional AI techniques in data mining for scarce construction databases.