Implementation of Machine Learning in Predicting Length of Punishment at Bandung Court

When undergoing trial in court, what the defendant is waiting for is the decision that will be given by the judge because the result of the decision will determine the length of the punishment to be given to the defendant. Therefore, it is very important for the community and lawyer to be able to predict the length of time the sentence will be given to the defendant. This needs to be predicted so that lawyer can determine what strategy will be made during the trial. Therefor we need a method to predict the length of the punishment will be given by the judge. In this study, we use two machine learning methods i.e., K-Nearest Neighbor (KNN) and Naive Bayes Classifier methods in predicting the length of the punishment will give the defendant. The data used are 100 data on the results of judges’ decisions obtained from the Bandung district court. The data was divided into two, 75 data as training data and 25 data as test data which were selected randomly. Before making predictions, we perform text preprocessing by using the TF-IDF method. Based on the test results using the KNN method, an accuracy rate of 84% is obtained, the weighted average precision value is 90% and the weighted average recall value is 84%, while the test results using the Naive Bayes Classifier method obtained an accuracy rate of 80%, the weighted average precision value of 72.16844% and the weighted average recall value is 80 %.