An ontology for supporting data mining process

Data mining has attracted increasing interests in recent years. Although there are several data mining software suits available, it is not easy for an end user to apply data mining techniques without the help of the data mining expert. The difficult is that with huge amount of data mining algorithms, how to choose a set of algorithms appropriate to their data that can satisfy their requirement. In other words, the users need the knowledge of the character of the data mining algorithms. In addition, we believe even a data mining expert also lacks this type of knowledge. The no free lunch theorem has shown that no algorithm is universally better than other algorithms for any datasets. Therefore an algorithm relatively better than other algorithms for some type of datasets in some measure criteria might perform worse in other cases. To circumvent this problem, we propose a method to extract and represent the knowledge of mining algorithms. The knowledge is represented by ontology. Users or agents could select mining algorithms easily with the data mining ontology.