Development of a decision support tool for analytic customer relationship management integrating data mining and multi criteria decision making methods

In recent years, there is an enormous interest in sharing a wide range of experiences and opinions regarding various products and services in blogs, social media platforms and websites specialized on customer reviews. The customer reviews derived from those platforms involve valuable information for potential customers, who might read and evaluate the comments of experienced customers before making a purchase decision on a specific product or service. Furthermore, the companies may measure customer satisfaction regarding to their products or services through online customer reviews and use this information in customer relationship management applications. In this thesis, a decision support tool, which can be used by potential customers and companies, is developed to measure and evaluate the customer satisfaction level on a specific product or service and to enable potential customers ranking the products/services according to this evaluation. The decision support system mainly consists of two phases. In the first phase, sentiment analysis is employed to convert the online customer reviews into customer satisfaction scores. In the second phase, the alternatives are ranked by using a novel multi criteria decision making (MCDM) methodology according to the performance scores obtained in the first level. The MCDM methodology developed in the second phase integrates intuitionistic fuzzy (IF) ELECTRE and VIKOR methods in a novel way. The system utilizes intuitionistic fuzzy sets (IFSs) to effectively represent the customer reviews including hesitant expressions in decision matrix. The applicability of the developed decision support system is explored by a case study, in which customer reviews about hotel experiences are evaluated using lexicon based sentiment analysis and alternative hotels are ranked according to the findings from the sentiment analysis by the IF ELECTRE integrated with VIKOR methodology.