Machine Learning Approach for Sentiment Analysis in Crime Information Retrieval

Abstract

Crime information is usually announced periodically. To get the information in real-time, the information should be retrieved through the information retrieval system automatically. The system should choose some appropriate keywords for retrieving the crime data. Eight keywords have been chosen which represented the most viral topic. The keywords in this study were analyzed regarding their sentiment from the hashtags in twitter posts. The Machine Learning algorithms were utilised such as Multinomal Naive Bayes, Random Forest Classifier, Linear SVM, and Nearestneighborhood (kNN) finding a better classifier. Sentiments, both positive and negative, are usually have been used by the website content designer to attract the reader. Most keywords showed negative sentiment which showed the negative reaction from the people. The sentiment analysis in Bahasa Indonesia also useful for understanding the people’s view on the types of crime as well as for keyword selection in the crime information retrieval system. As the result, near-repeat crime effect as a condition where criminal activity tends to repeat in the near place and time, can be predicted.