Data Mining 4 Dummies: A Web application for automatic selection of data mining algorithms for new problems

Abstract

The interest in the area of classification and prediction is growing rapidly in industry and commerce. A large number of data mining tools are already available. However, such tools are still of limited use to end-users who are not experts. This is due to the fact that machine learning systems are non-trivial. As a result, users of machine learning/data mining systems are faced with two major problems: selecting the most suitable algorithm to use on a given dataset, and combining this algorithm with useful and effective transformations of the data. Traditionally, these problems are solved by trial-and-error or consulting experts. The first solution is time consuming and unreliable, while the second is expensive and based on preferences of the experts. Clearly, neither solution is completely satisfactory for the non-expert end-users. Therefore automatic and systematic guidance is required.By analysing the state of the art we can see how different attempts have been made to address this problem, and although some of them have shown very interesting results, they are still tool restricted and present a lack of satisfactory user guidance, simplicity and process transparency. The focus of this dissertation is to improve support to machine learning/data mining end-users, by creating a new system that will allow the recommendation and use of the most promising algorithms in a distributed and collaborative way.