An Extended Hybrid Clustering Method Utilzing Svm As Cross Validator

Abstract

The massive data accumulation from the internet creates attention for the researchers. The data collected in the form of structured and unstructured data. The structured data consists of messages, transactions, conversations, etc. while unstructured represents video and audio clips. This essentially manages the raw data problem in which unreferenced clustering is used. A hybrid approach is proposed using Cosine Similarity and soft cosine. A novel clustering technique is designed which is cross-validated using the Support Vector Machine (SVM). The validated approach is further verified by using K- means clustering. The clustering results have been further evaluated using parameters precision, recall, and F-measure. The evaluated results show the improvement in precision and recall accuracy due to hybridization of cosine similarity and soft cosine techniques.