CLASSIFICATION OF THE FEATURE-LEVEL RATING SENTIMENTS FOR TELUGU LANGUAGE REVIEWS USING WEIGHTED XGBOOST CLASSIFIER

Sentiment analysis (SA) is the opinion extraction that studies the attitude, sentiments, opinions and emotion of people. The huge number of active users will provide informations about their opinions in E-Commerce websites which gives the effective review about products. Various sentiment analysis approaches were presented to classify the sentiments as positive, negative, and neutral. However, the existing methods are not effective and ignores the subtle sentiment classification among various text. But, the supervised learning methods was achieved some satisfactory performance on dimensional sentiment analysis, although they needed multiple labels to train the system, that are cost effective and consumes time for annotation of data.In order to overcome such an issue,proposed an Weighted XGBoost Classifier method for the sentimental analysis of the amazon telugu reviews. Initially, the amazon telugu reviews are collected and required features are selected. Then, preprocessing is carried to convert the unstructured data into structured data. Further, relevant sentences are extracted from the structured data. Then, weighted XGBoost classifier is used to classify the product reviews and ratings are generated. When the ratings are generated the reviews are categorized as Terrible (1star), Poor (2stars), Average (3stars), Very Good (4stars) and Excellent (5stars). The results obtained from the predictive analysis are computed in terms of performance measures such as Accuracy, precision, Recall and F-Measure.