ANALYZING AND PREDICTING SENTIMENT RESULT FOR AMAZON PRODUCTS USING TEXTBLOB AND SUPPORT VECTOR MACHINE

Nowadays, numerous of reviews from buyers spreaded widely on online shopping platform, especially Amazon which it give an important information for knowing how good the condition and service of this item from various available categories. The bulk of reviews from buyers complicate to indentify all of sentiments of many comments, so that program was made to identify and analyze all of sentiments of many comments using sentiment analysis method. This program identifies sentiment from each of reviews based of average valuation from automatic polarity of reviews using TextBlob library from Python programming language and the overall of each comments which changed into certain numbers such as -1, -0,5, 0, 0,5, 1 for 1, 2, 3, 4, 5 overalls. If there are a comment which have helpful votes, so there are additional changed overall values to count the average of this sentiment valuation. The average valuation of sentiment will determine whether sentiment are positive or negative. Then the sentiment result of reviews are processed into Support Vector Machine (SVM) algorithm which later measuring accuracy level and predict other reviews from another prediction dataset to prove exactness of it. From the project result that performed, the Polynomial Kernel from SVM algorithm is the highest accuracy of 92,85% on prediction result in Without Stemming experiment and for highest examination of sentiment identification prediction accuracy is Radial Basis Function (RBF) from SVM algorithm with 92,64% which 667 of 720 from another dataset are correctly predicted in Without Undersampling experiment