Intelligent decision support system for industrial site classification: a GIS-based hierarchical neuro-fuzzy approach

Industrial site selection involves a large number of criteria and location alternatives; consequently, the selection process leads to extended decision-making periods and requires complex knowledge management, classification and analysis skills. The selection criteria are generally described by a number of different features expressed as both quantitative and qualitative measures that can involve some uncertainty. Moreover, the goals considered in the selection process are frequently nonlinearly related to the criteria; thus, they give rise to an optimization problem that is nonlinear with respect to each goal. Consequently, decision making requires appropriate support to enable efficient data optimization and classification under uncertainty before the final selection of an industrial site is made. This paper presents a novel intelligent decision support system for classifying industrial sites according to quality criteria estimated by exploiting a geographic information system, expert knowledge and machine learning techniques. The proposed system is based on a geographic information system for generating location alternatives and a hierarchical neuro-fuzzy approach for site classification. The neuro-fuzzy method is based on a knowledge base designed by experts in the field and uses a neural approach to tune the parameters of the membership functions. Experimental results on real-world problems show that the proposed system provides accurate results for industrial site classification at the local level (micro locations).