Shruti Tiware Patil
Excel & Power Query Automation | MIS Reporting | Power BI Dashboards ·
Hire
Recently, I completed an academic project on Loan Default Prediction using Machine Learning as part of my M.Sc. Statistics studies.
The objective was to analyse borrower-related factors and develop predictive models to identify the likelihood of loan default.
Key Highlights:
- Performed Exploratory Data Analysis (EDA).
- Conducted statistical hypothesis te… Recently, I completed an academic project on Loan Default Prediction using Machine Learning as part of my M.Sc. Statistics studies.
The objective was to analyse borrower-related factors and develop predictive models to identify the likelihood of loan default.
Key Highlights:
- Performed Exploratory Data Analysis (EDA).
- Conducted statistical hypothesis testing using Chi-Square tests.
- Analysed correlations and multicollinearity using VIF.
- Identified important predictors using Random Forest.
- Applied SMOTE and SMOTEENN to handle class imbalance.
- Built and compared Logistic Regression, Random Forest and LightGBM models.
- Evaluated model performance using Accuracy, Recall and AUC-ROC.
Key Results:
- LightGBM with SMOTEENN achieved 89% accuracy and 0.957 AUC-ROC.
- Random Forest with SMOTEENN achieved 93% default recall.
Tools & Technologies: Python, Machine Learning, Statistical Analysis, EDA, Data Visualization, Scikit-learn, LightGBM.
This project helped me apply statistical concepts and machine learning techniques to a real-world financial risk prediction problem.
#LoanDefaultPrediction #MachineLearning #Python #DataScience #StatisticalAnalysis #EDA #PredictiveAnalytics #MScStatistics #DataAnalytics
The objective was to analyse borrower-related factors and develop predictive models to identify the likelihood of loan default.
Key Highlights:
- Performed Exploratory Data Analysis (EDA).
- Conducted statistical hypothesis te… Recently, I completed an academic project on Loan Default Prediction using Machine Learning as part of my M.Sc. Statistics studies.
The objective was to analyse borrower-related factors and develop predictive models to identify the likelihood of loan default.
Key Highlights:
- Performed Exploratory Data Analysis (EDA).
- Conducted statistical hypothesis testing using Chi-Square tests.
- Analysed correlations and multicollinearity using VIF.
- Identified important predictors using Random Forest.
- Applied SMOTE and SMOTEENN to handle class imbalance.
- Built and compared Logistic Regression, Random Forest and LightGBM models.
- Evaluated model performance using Accuracy, Recall and AUC-ROC.
Key Results:
- LightGBM with SMOTEENN achieved 89% accuracy and 0.957 AUC-ROC.
- Random Forest with SMOTEENN achieved 93% default recall.
Tools & Technologies: Python, Machine Learning, Statistical Analysis, EDA, Data Visualization, Scikit-learn, LightGBM.
This project helped me apply statistical concepts and machine learning techniques to a real-world financial risk prediction problem.
#LoanDefaultPrediction #MachineLearning #Python #DataScience #StatisticalAnalysis #EDA #PredictiveAnalytics #MScStatistics #DataAnalytics
Loan_Default_Prediction.pdfPDF · 0.5 MB
1 appreciations