Text Clustering Quality Improvement using a hybrid Social spider optimization

Text document clustering is one of the most widely studied data mining problems. It organizes text documents into groups such that each group has similar text documents. While grouping text documents, several issues have been observed. Accuracy and Efficiency are the main issues in text document clustering. Recently, as clustering problem can be mapped to optimization problem, evolutionary optimization techniques have been used by researchers to improve accuracy and efficiency. Evolutionary techniques are stochastic general purpose methods for solving optimization problems. Swarm Intelligence is one such technique that deals with aggregative behavior of swarms and their complex interactions without any supervision. In this paper, we proposed a novel swarm intelligence algorithm called Social Spider Optimization SSO for textual document clustering. We compared it with K-means clustering and other state-of-art clustering algorithms such as PSO, ACO and Improved Bee Colony Optimization and found it to give better accuracy. Then we proposed two hybrid clustering techniques namely SSO + K-means and K-means + SSO and found SSO + K-means clustering outperformed Kmeans + SSO, KPSO(K-means + PSO) , KGA (K-means + GA), KABC(K-means + Artificial Bee colony) and Interleaved K-means + IBCO clustering techniques. We used Sum of intra cluster distances, average cosine similarity , accuracy and Inter cluster distance to measure the performance of clustering techniques.