Judgement Prediction of Government Employee’s Retirement Benefits Cases using TensorFlow Decision Forests

Abstract

Listen

Translate article

A judgement prediction model with TensorFlow Decision Forests (DF) is a machine learning model that uses decision trees as the building blocks for classification and forecasting of writ petition. The TensorFlow library provides tools for implementing and training decision forests, which are collections of decision trees, in order to make predictions. In this paper we analyze the basic description of the retirement benefits matters like grant of pension on superannuation, family pension, commutation of pension, gratuity, group insurance saving fund, leave encashment and superannuation, provident fund which are filed in appropriate Courts/Tribunals. The outcome is either “Allowed” or “Dismissed”. Rarely “Partly Allowed” petitions are treated as “Allowed”. TensorFlow DF looks a better choice which applied on labeled dataset based on factors affecting the outcomes of cases. The objective of this paper is to classify and forecast the outcome whether a filed petition will be ‘allowed, or ‘dismissed using TensorFlow DF, a supervised learning approach. Classification and forecasting can help the delinquent, learned counsel, and presiding officers finish the petition and understand it. The model for judgement prediction is implemented on retirement benefits cases which are filed in concerned court against different departments in U.P. State in India. The accuracy of the proposed model varies between 94% to 98%. As a result, the effectiveness of models developed in the TensorFlow DF is highly useful in forecasting the outcome of petitions pertaining to the employment related cases of civil servants, as well as employees of local governments and government-owned enterprises.