Data mining on multi-criteria rating values

The development of information and communication technologies offers the possibility of sharing on customer views, comments and ratings about products and services over the Internet. Customers evaluate services or products by taking into account multiple criteria and in this context there are datasets collected from customers. Customer expectations and profiles can be effectively analyzed using data mining techniques over multi criteria customer reviews. In this study, we focus on how multi criteria rating values will be investigated using data mining techniques. Using in-flight experience reviews of airline passenger, passenger trends are tried to be identified and how passenger profiles can be formed is discussed. Data are examined with feature-based and similarity-based clustering approaches. While feature-based approach grouped customers according to selected features, in the second approach novel similarity-based clustering algorithms are proposed in the view of research problem. In the proposed clustering methods, similarity score is defined to be computed for each users and according to this score characteristic users are determined. Passengers are clustered based on similarity between passenger and characteristic users. Then, ReliefF algorithm is applied for each obtained cluster, features are ranked according to importance in the view of passengers.