Perbandingan Algoritma Support Vector Machine dan Naïve Bayes dalam Menganalisis Sentimen Pinjaman Online di Twitter

Unemployment is one of the poverty factors in society, the large economic needs make it difficult for people to meet their daily needs, thus triggering high demand for loans in society. With the advancement of technology, online loans are now available to help people meet their economic needs. However, over time, many irresponsible parties have taken advantage of this. Marked by the emergence of many illegal online loans, which have triggered negative impacts such as the spread of customer personal data, terror on social media, to debt collection using debt collectors. So that it raises a lot of sentiment in society regarding online loans. For this reason, it is necessary to conduct a sentiment analysis with the aim of public response to online loans, which can be positive, negative or neutral responses. There are two datasets used, namely legal online loans and illegal online loans. This study uses two algorithms, namely SVM and Naive Bayes, the two algorithms will be compared to find out which algorithm is better at analyzing online loan sentiment. In addition, in its use, the two algorithms will also use the SMOTE technique to stabilize the data. The results obtained on legal loan data classification using SVM are quite better than Naive Bayes, with an accuracy rate of 69% with sentiment that often appears is positive sentiment. For illegal loan data, classification using the Naive Bayes algorithm is better than SVM with an accuracy of 75% and sentiment that often appears is neutral sentiment. Based on these results, it can be concluded that in analyzing sentiment using legal loan data, the best algorithm is the SVM algorithm, and for illegal loan data, the best algorithm is the Naive Bayes algorithm.