Scribd Logo
Upload

Welcome to Scribd!

  • Finance and Risk Analytics GRP Assgn

    Uploaded by

    psyish
    140 views7 pages
    The document describes the creation of a logistic regression model to predict credit risk (default) using Indian company financial data. It outlines the following steps: [1] Data preparation including identifying defaulters, removing small companies, and transforming variables; [2] Shortlisting 14 important variables using ratios of mean values for default and non-default companies; [3] Building the logistic regression model in Knime and selecting important variables; [4] Evaluating the model's accuracy on test and validation data, achieving over 93% accuracy.

    Original Description:

    analytics

    Original Title

    Finance and Risk Analytics Grp Assgn.docx

    Copyright

    © © All Rights Reserved

    Available Formats

    DOCX, PDF, TXT or read online from Scribd

    Did you find this document useful?

    Is this content inappropriate?

    Report this Document
    The document describes the creation of a logistic regression model to predict credit risk (default) using Indian company financial data. It outlines the following steps: [1] Data preparation including identifying defaulters, removing small companies, and transforming variables; [2] Shortlisting 14 important variables using ratios of mean values for default and non-default companies; [3] Building the logistic regression model in Knime and selecting important variables; [4] Evaluating the model's accuracy on test and validation data, achieving over 93% accuracy.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as docx, pdf, or txt
    140 views7 pages

    Finance and Risk Analytics GRP Assgn

    Uploaded by

    psyish
    The document describes the creation of a logistic regression model to predict credit risk (default) using Indian company financial data. It outlines the following steps: [1] Data preparation including identifying defaulters, removing small companies, and transforming variables; [2] Shortlisting 14 important variables using ratios of mean values for default and non-default companies; [3] Building the logistic regression model in Knime and selecting important variables; [4] Evaluating the model's accuracy on test and validation data, achieving over 93% accuracy.

    Copyright:

    © All Rights Reserved
    Download as DOCX, PDF, TXT or read online from Scribd
    Flag for inappropriate content
    Download as docx, pdf, or txt
    of 7

    Group III

    FRA Group
    Assignment
    Logistic regression model

    Priyank, Sakshi, Varun, Vinay


    Finance and Risk Analytics – Group Assignment

    Problem Statement

    Create India credit risk(default) model, using the data provided in the spreadsheet raw-data.xlsx, and
    validate it on validation_data.xlsx. Please use the logistic regression framework to develop the credit
    default model.

    Data Insights

    The data provided in raw-data comprises of financial data.

    Major data points or variables are

    Net worth next year, Total assets, Net worth, Total income, Total expenses, Profit after tax, PBDITA,
    PBT (Profit Before Tax), Cash profit, PBDITA as % of total income, PBT as % of total income, Cash
    profit as % of total income, PAT as % of net worth, Sales, Total capital, Reserves and funds,
    Borrowings, Current liabilities & provisions, Capital employed, Net fixed assets, Investments, Net
    working capital, Debt to equity ratio (times), Cash to current liabilities (times), Total liabilities.

    In addition to the above variables there are other financial parameters which define the financial
    strength of the organization taking the total tally of variables to 51.

    Data Preparation

    Step 1 (Identifying Defaulters)

    Companies were classified into probable Defaulter and Non Defaulters based on the Net worth next
    year. Ones with negative worth were classified as Defaulters (marked as 1) and the rest as Non
    Defaulters (marked s 0).

    Step 2 (Remove very small companies)

    Removed companies with Total Net asset of 0 to 3

    Step 3 (Basic Statistics)

    Calculated Min, Max, Mean, Standard Deviation, Median and Percentiles (1 st to 4th and 99th to 97th )

    Step 4 (Identified Floor and Cap)

    Based on data understanding, identified floor and cap (Min / Max or based on Percentile)

    Step 4 (Outlier and missing value Treatment)

    Data imputation - Transformed the data with replacing the values with floor and cap.

    Missing values were replaced with median.

    Step 5 (Creation of new variables)

    Created new variables to be used for model building. New variables are Critical variables divided by
    Total assets.
    New variables created

    Net worth/ Total assets,Total income / Total assets, Total expenses / Total assets, Profit after tax /
    Total assets, PBT / Total assets, Sales / Total assets. Current liabilities & provisions / Total assets,
    Capital employed / Total assets, Net fixed assets / Total assets, Investments / Total assets, Total
    liabilities / Total assets

    Step 6 (Variable shortlisting)

    Calculated Mean for Non Default and Default companies and then derived the ratio. Variables with
    ratio > 3 and ratio < 1/3 were shortlisted for making the model.

    Microsoft xls in which calculations were made and new transformed variables were created is
    attached below

    Following 14 variables were shortlisted

    1-Net worth (Transfrmd)

    2-Profit after tax (Transfrmd)

    3-PBDITA (Transfrmd)

    4-PBT (Transfrmd)

    5-Cash profit (Transfrmd)

    6-PBT as % of total income (Transfrmd)

    7-Cash profit as % of total income (Transfrmd)

    8-PAT as % of net worth (Transfrmd)

    9-Sales (Transfrmd)

    10-Reserves and funds (Transfrmd)

    11-Net working capital (Transfrmd)

    12-Profit after tax / Total assets

    13-PBT / Total assets

    14. Default

    Logistic regression model was built in Knime.

    Data was partitioned to test the accuracy before validating the accuracy with validation_data
    Model building flow

    Data Partitioning
    Model Coefficients and Statistics

    Following variables highlighted in yellow are selected for model

    1-Net worth (Transfrmd)

    2-Profit after tax (Transfrmd)

    3-PBDITA (Transfrmd)

    4-PBT (Transfrmd)

    5-Cash profit (Transfrmd)

    6-PBT as % of total income (Transfrmd)

    7-Cash profit as % of total income (Transfrmd)

    8-PAT as % of net worth (Transfrmd)

    9-Sales (Transfrmd)

    10-Reserves and funds (Transfrmd)

    11-Net working capital (Transfrmd)

    12-Profit after tax / Total assets

    13-PBT / Total assets

    14. Default

    Confusion Matrix
    Model Stats

    Measure Value
    Sensitivity 0.956
    Specificity 0.691
    Precision 0.986
    Negative Predictive Value 0.397
    False Positive Rate 0.31
    False Discovery Rate 0.014
    False Negative Rate 0.045
    Accuracy 0.945

    Now model ll be tested against validation_data

    Measure Value

    Sensitivity 0.9649

    Specificity 0.6667

    Precision 0.9693
    Negative Predictive Value 0.6349

    False Positive Rate 0.3333

    False Discovery Rate 0.0307

    False Negative Rate 0.0351

    Accuracy 0.9399

    You might also like