Optimizing the Efficiency of Machine Learning Techniques

The prediction of judicial decisions based on historical datasets in the legal domain is a challenging task. To answer the question about how the court will render a decision in a particular case has remained an important issue. Prior studies conducted on the prediction of judicial case decisions have datasets with limited size by experimenting less efficient set of predictors variables applied to different machine learning classifiers. In this work, we investigate and apply more efficient sets of predictors variables with a machine learning classifier over a large size legal dataset for court judgment prediction. Experimental results are encouraging and depict that incorporation of feature selection technique has significantly improved the performance of predictive classifier.