Sampling and feature selection in a genetic algorithm for document clustering

In this paper we describe a Genetic Algorithm for document clustering that includes a sampling technique to reduce computation time. This algorithm calculates an approximation of the optimum k value, and solves the best grouping of the documents into these k clusters. We evaluate this algorithm with sets of documents that are the output of a query in a search engine. Two types of experiment are carried out to determine: (1) how the genetic algorithm works with a sample of documents, (2) which document features lead to the best clustering according to an external evaluation. On the one hand, our CA with sampling performs the clustering in a time that makes interaction with a search engine viable. On the other hand, our CA approach with the representation of the documents by means of entities leads to better results than representation by lemmas only.