In most situations, we would like to recognize specific patterns or subgroups in a set of observations about
individuals. Therefore, the general goal would be to recognize in a massive dataset some common features of subgroups of individuals that could be relevant for understanding collective behaviour and predictive possible future behaviour. One of the main methodology dealing with such a problem is ¿cluster analysis¿. Inside the literature on cluster, a special role is played by those techniques that provide a "soft" (i.e. fuzzy) representation of the information. Fuzzy clustering algorithms are, concisely, models for assigning to each individual his/her membership degree to a particular subgroup of the population. Apart from possible computational advantages , their main convenience is grounded on the fact that they provide: a natural way of encoding and represents relevant information (which can also be expressed with vagueness and imprecision); a realistic representation of individual (e.g. consumer) behaviour that cannot be described by single numbers, but requires a more general specification (the so-called fuzzy set or fuzzy number representation).In this research project, we aim at developing specific theoretical, computational and applied investigations about the use of fuzzy clustering methods in order to deal with several
problems arising in marketing and consumer behaviour. Our final aim is to contribute to the development of practical tools that may help the decision makers (e.g., managers) to understand the behaviour of consumers.
The expected output of the project is the preparation of scientific papers to be submitted to international journals and the dissemination of the results to the international scientific community. Moreover, the participants will analyse the possibility of exploitation of the developed models in terms of novel knowledge and possible policy implications.
This project entails a general theoretical part of novel methods for cluster analysis, and applies its finding to market segmentation, with particular emphasis on tourism. Over the years in both the marketing and the more specific tourism literature, a great debate, still open, has been generated over the clustering types and techniques to use in segmentation. Since the introduction of market segmentation in the late 1950s, the number and types of approaches to market segmentation have grown enormously (Liao et al. 2012). In marketing and tourism literature, cluster analysis remains the most favoured method (Wedel and Kamakura, 2000) even if it has been criticized for its overestimation of the validity of the segmentation results (Dolnicar and Lazarevski, 200 9). Generally, clustering procedures are divided into two groups: a hard algorithm allows each observation to belong to a single segment and/or group only, a soft (fuzzy) algorithm assigns each observation to each cluster with a certain degree of membership (Tuma et al., 2011). The main topic of this project is related to soft methods for cluster analysis. In fact, we believe that a soft approach for cluster analysis presents many advantages over the hard clustering methods . From a theoretical perspective 1) fuzzy clustering methods are computationally more stable because dramatic changes in the value of cluster membership are less likely to occur in estimation procedures (Coppi et al., 2012; D'Urso, 2014; D¿Urso and De Giovanni, 2014); 2) fuzzy clustering has been shown to be less affected by local optima problems in the estimation procedures (D'Urso, 2007); 3) fuzzy clustering provides the best performance in stability criterion when compared to hard methods (Wang et al., 2008).
Our final aim is to contribute to the development of innovative management tools for consumer segmentation and market analysis in tourism. In fact, we believe that the use of quantitative (yet intuitive, easy-to-handle) methodology will provide fruitful hints for the knowledge of consumer behaviour.