Analisis Sentimen Ulasan Produk Kecantikan Menggunakan Metode BM25 dan Improved K-Nearest Neighbor dengan Seleksi Fitur Chi-Square

Abstrak

The impact of beauty product which makes women interest. With the convenience provided at this time there is a special platform based on the web and mobile phone that is Female Daily. Female Daily is an information media site that contains the content of beauty products about facial care to the body to be given reviews by consumers who have tried or using the product. Review Data can be used as a reference before consumers want to try out beauty products. A lot of reviews resulted makes consumers has difficult to choose the product they want. In this research, it helps consumers to know the review data is entered into positive or negative sentiments. The process in sentiment analysis requires the BM25 method used as a weighted word, Improved K-Nearest Neighbor as a determination in sorting the sentiments and Chi-Square feature selection to reduce the number of words in the classification on the text. The test used is 5-fold cross validation with the best results on  k-rank=15 resulting on the average value of precision= 0.9, recall= 0.8, accuracy = 0.7806 and f-measure= 0.8428. Then, testing used selection of Chi-square features selection based on percentages with k-rank= 15 parameters, the highest percentage value of 40% and 50% with the value of precision = 0.888, recall = 0.8, accuracy = 0.7818 and f-measure = 0.842.