Enhancing Legal Document Analysis and Judgment Prediction with Machine Learning and Deep Learning Techniques

Abstract

The primary objective of this study is to create a judicial judgement prediction system that achieves a high level of accuracy by employing sophisticated machine learning methods. The goal is to improve the predicting skills of models in legal scenarios by utilising the Texas Wolf Optimisation (TWO) algorithm alongside a Deep Bidirectional Long Short- Term Memory (BiLSTM) network. The approach entails enhancing the BiLSTM model's hyperparameters by the utilisation of the TWO algorithm in order to enhance its capacity to identify intricate patterns within legal documents. The collection consists of previous legal cases from the Supreme Court, which include comprehensive annotations on legal references, arguments, and judgements. Multiple models, such as LR, SVM, CNN, and LSTM, are evaluated for their performance, and the TWO-BiLSTM model demonstrates improved outcomes. Models are evaluated using performance criteria such as accuracy, F-score, precision, and recall. The findings demonstrate that the TWO-BiLSTM model has superior performance compared to current models, with a 97% accuracy and a 97.29% F-score in scenarios with a True Positive (TP) rate of 90. Furthermore, it consistently demonstrates robust performance in K-fold cross-validation, with an impressive accuracy rate of 96%. The study showcases the efficacy of the suggested TWO- BiLSTM model as a robust tool for forecasting judicial outcomes, presenting significant enhancements compared to conventional methods.