Improved Hybrid Similarity for Clustering of Text Documents Using GA

Similarity is an important parameter for making clusters of text documents in the age of technology. There are various similarities through which clustering can be done by access the documents. These are named as Cosine, Link and Jaccard similarities. Authors proposed improved hybrid similarity by doing the integration of improved rank similarity and hybrid similarity for generating the clusters of the text documents. It employs the data-set of text files taken from amazon musical instrument review and theses reviews are taken from Kaggle repository. Authors have taken two parameters i.e., accuracy of generating the clusters and memory consumed per document during clustering of documents. Results are taken based on improved hybrid similarity and it gives better results on the basis of the chosen parameters of accuracy and memory consumption per document. Further enhancement is done to the results by applying genetic algorithm to the similarity measure and comparison is done on various techniques in the presence and absence of GA. The average accuracy with alone Improved Hybrid similarity is 72.34% and with the application of GA on Improved Hybrid similarity the average accuracy is 84.14%, means an increase of 16.31%. Memory consumed per document is also improved by 19.4%.