Clustering mixed datasets using multi objective genetic algorithm

Collecting and extracting the useful information for users from the datasets becomes very popular and important among the research areas of computer sciences. For using the extracted information people can easily create links between the different data and make clustering or classification operations with them. In order to do that information extraction process, there are remarkable number of algorithms are developed and they are used in areas like banking, bioinformatics and medicine.There are lot of algorithms which are do clustering operations for datasets which are included only numerical attributes or only categorical attributes. However the number of the algortihms convenient for the mixed datasets, which are included both numerical and categorical attributes, are very low.In this thesis, it has been stutied on developing a new clustering algorithm for all the three types (numerical, categorical and mixed) of datasets. The algorithm which is proposed is seperating the types of the attributes as numerical and categorical, calculating the distances between the data and returning a clustering result. For calculating the distance between two datum, there are fitness functions. Fitness functions are also seperated for numerical and categorical attributes and they are use in the same way as the fitness functions in the k ? modes and k ? means algorithm. Finally the clustering results, which are returned from the algorithm, are evaluated and the optimal clustering numbers are detected. The success of the results are evaluated with purity index and they are compared with the results of the other algorithms.