capstone_project

1. Project Statement The project aims to employ Legal Analytics in order to predict the winning party given a case. The primary objective of a Law Firm is to fight cases where chances of success is high. Appropriate analysis and Machine Learning models aid a firm or a lawyer to make an informed choice based on historical data and sound statistical analysis.

2. Dataset Details  The database used dates back to the year 1946 during the beginning of Vinston Court and Warren Court. There have been many versions since then. The data available can be used with various statistical packages to perform various analyses with respect to what the business/ Law firm considers important. The advanced modelling and EDA techniques used in the project will enable the business to calculate and view relationships among the variables which shall aid the management in making informed decisions. The database consists of information from 1946. Thus with increase in information, the needed responsibility to alter/ transform the data is carried out

3. Complexity Involved a. Find the key area, gaps identified in the topic survey where the project can add value to the customers and business. b. Since datasets involving court cases have multiple variables, the variables that aid in the prediction of "partyWinning" have to be selected. Certain predictor variables involve information that a law firm may not have access to, before the case begins in the supreme court. Hence these variables are removed c. Merging of different columns so that it makes sense which further reduces processing and execution time and then building an accurate model with decent respective model scores using required libraries and algorithms.To obtain duration of the case , various columns were merged through appropriate datetime techniques. d. Since each column is unique, the null value imputation was done diligently using hot deck imputation. e. Finding out the best target variable as the dataset contained multiple potential target variables.

4. Project Outcome

• Commercial: Given the details of a case, the law firm will be able to predict the outcome which shall enable the agency to maintain a better track record and select cases where the chances of winning are high.

• Social:Should the petitioner reconsider filing a case based on model results. This will help the petitioner in saving time and money. Hence the law firm will be able to provide alternate solutions to the petitioner and gain reputation for the same.

5. METHODOLOGY TO BE FOLLOWED [a] The Primary step of this project is to understand the relationship between data and the field of Law. The US supreme court database is used for this project. A dataset which focuses on the case details and final judgement is chosen for analysis . [b] As Legal Analytics is a broad field, the plausible target variables and their respective predictor variables are studied. The Target variable has been identified as the winning party and the case details as the predictor variables. [c] The necessary data is prepared only when the data has been studied for its important characteristics and the potential anomalies it may possess. Exploratory Data Analysis is performed on the data using Univariate-Bivariate-Multivariate analysis through visual inspection or statistical tests based on the type of data. These steps build the necessary foundation to perform feature engineering, data reduction and transformation. [d] The next phase is to create a model which is capable of capturing the characteristics of the target variable. This involves selecting an optimal Hypothesis function (ML model ) which captures the true relationship between the explanatory and response variables (Low Bias). At the same time the model must also be able to make good predictions for data that it has not seen before (Low Variance). Basic models such as Logistic Regression, Linear Regression to advanced models such as AdaBoost and XGBoost could be tested for the same. [e] Various metrics can be used to evaluate the efficiency of a model. Based on what the business/ research considers important, specific metrics and thresholds can be used to make our model more appropriate for the Business problem at hand. On achieving the necessary sufficiency , the model and the methodology is further evaluated in the presence of an industry mentor.