MANSI DUBEY

Detail-oriented Data Scientist and Analyst with hands-on experience in Python, SQL, Machine Learning, Power BI, and Tableau. Skilled in building end-to-end predictive models, data wrangling, exploratory data analysis, feature engineering, and developing actionable business insights from large datasets. Experienced across fintech, banking, and retail domains, with successful projects in customer churn prediction, credit risk analysis, and marketing optimization. Passionate about applying ML algorithms to real-world business problems with strong data storytelling and visualization capabilities.

Zone 1 Power Consumption Prediction – Wellington, NZ  Term Deposit Subscription Prediction – Banking  Credit Risk Prediction – Loan Approvals  Retail Customer Churn Prediction

· • Built ML models to predict electricity consumption using 52K+ environmental records.

· • Performed EDA on weather and air quality factors; selected Random Forest (R² = 0.64, MAE = 3.3K).

· • Communicated results via stakeholder-ready Jupyter notebook and presentation.

· • Tech Stack: Python, Pandas, Scikit-learn, Matplotlib, Seaborn, Jupyter Notebook

· • GitHub: https://github.com/Mansi-web-25/wellington-power-prediction

· • Modeled customer subscription behavior (45K+ records, 17 features).

· • Applied OneHotEncoding, handled class imbalance, and delivered 90.35% accuracy, 50.76% F1-score.

· • Derived insights for targeted marketing and budget optimization.

· • Tech Stack: Python, Scikit-learn, Pandas, Matplotlib, Seaborn, Jupyter Notebook

· • GitHub: https://github.com/Mansi-web-25/banking-term-deposit-prediction

· • Predicted loan defaults using classification models (89% accuracy, 80% recall).

· • Delivered Tableau dashboards enabling underwriters to segment risk profiles.

· • Helped reduce expected defaults by ~20%.

· • Tech Stack: Python, Pandas, Scikit-learn, Tableau, Excel

· • GitHub: https://github.com/Mansi-web-25/-Credit-Risk-Prediction-for-Loan-Approvals

· • Identified high-risk churn customers with 87% accuracy using behavioral features.

· • Created interactive Power BI dashboards leading to 10% retention increase.

· • Tech Stack: Python, Scikit-learn, Power BI, Pandas, Jupyter Notebook

· • GitHub: https://github.com/Mansi-web-25/Customer-Retention-Prediction-for-Retail-Business