Classification of Public Complaint Data in SMS Complaint Using Naive Bayes Multinomial Method

SMS Complaint is an electronic public complaint tool for reporting issues on government performance. Text mining classification utilized to determine the value of each complaint category. The SMS data in this study sourced from the SMS Complaint Service of Ambon City Government. There were 6 categories of classification, namely Public Service, Infrastructure, Bureaucracy, Health, Education, and Social. The classification performed to measure levels of accuracy of the Stemming process and non-Stemming process represented in Matrix with values of recall, precision, and f1 score. The methods used in the measurement were Naıve Bayes Multinomial. With the naıve Bayes method, an accuracy level with stemming of 91.38% obtained, and while the accuracy level without stemming was 90.73%. The result showed that the naıve Bayes method could be used effectively to predict complaint data through stemming.