A NAIVE BAYES MODEL USING SEMI-SUPERVISED PARAMETERS FOR ENHANCING THE PERFORMANCE OF TEXT ANALYTICS

Sentiment mining is an emerging area that utilizes the feedback of the users to make intelligent decisions in marketing. Consumer reviews helps the user to convert oral conversations to digital versions to be used in marketing analytics. Since e-commerce takes intrinsic forms in present day business, the need to get positive reviews for specific products is vital on an organization perspective. Both sentiment mining and TF-IDF (Term Frequency-Inverse Document Frequency) are text mining tasks but have their unique applications and characteristics. Sentiment mining classifies the documents into positive, negative or neutral opinions that are used to derive intelligent decisions. TF-IDF classifies the documents into sub-categories within the documents itself. Utilizing the TF-IDF as a feature for sentiment mining improves the performance of classification. In this research work a strategy is applied for feature preprocessing to evaluate the consumer��s sentiments accurately. This strategy makes use of the semi-supervised parameter estimation Na誰ve Bayes Model. The experimental results demonstrate that TF-IDF tuning approach results in better optimized results.