An Intelligent Learning Approach to Case Analysis for Predicting Legal Outcomes Using Random Forest and K-Nearest Neighbors
Puneet Bafna, Hardeo Kumar Thakur, Jagendra Pratap Singh, Jannam Sadana, Motaz Hassan, M. Sree Vani · 2024
Machine learning is an increasingly popular area of research that holds great potential for enhancing the effectiveness and precision of legal analysis and decision-making. More specifically, machine learning can be used to predict the most probable outcomes of a legal case, and the purpose of this project is to utilize machine learning with case law data from the Chennai District Court to predict the outcomes of cases that involve criminal offences and the charge of murder. In other words, the goal of this project is to predict whether a case will result in the conviction of a victim from a list of case factors. For this research, several approaches to machine learning, including K-Nearest Neighbors, Support Vector Machines, Naive Bayes, Random Forests, and Decision Trees, are implemented to identify the best solutions for predicting the case outcomes. The case factors used in the research provided in the court judgments and applied various preprocessing tools to eliminate random factors that could lead to unreliable predictions. Finally, the various performance scores of the model are assessed, which includes the accuracy of each trained model in terms of precision, recall, Fl-score, and the Area Under the ROC Curve. As a result, about 95% of models displayed high accuracy rates, while KNN performed the best - 96.76% precision and 0.985 AUC-ROC, with SVM at 94.5% and 0.975 extremely close to perfection. In addition, NB, RF, and DT also showed good levels of accuracy at 92.1%, 88.9%, and 85.6%, respectively. This research confirms that machine learning can be highly effective in predicting legal outcomes and increasing legal practice accuracy.