Applied Data Science & Analytics – Internship Portfolio
This portfolio by Ashna Imtiaz showcases the projects completed during my Data Science & Analytics internship at DevelopersHub Corporation. Each project reflects hands-on applications of machine learning, explainable AI, time-series forecasting, and business intelligence—translating theory into practical, impactful solutions.
Project 1: Bank Marketing Prediction
Objective
Predict customer subscription to a bank term deposit based on demographic, financial, and interaction features.
Approach
  • Data preprocessing (encoding, scaling, EDA)
  • ML model training (Logistic Regression, Random Forest)
  • Evaluation (confusion matrix, ROC-AUC)
  • Explainable AI with SHAP for transparent predictions
Results
  • Achieved high accuracy with the best-performing model.
  • SHAP explained individual customer predictions, demonstrating clarity in decision-making.
Key Skills: Data Preprocessing, ML Modeling, Explainable AI (SHAP), Classification.
Click here
Project 2: Customer Segmentation using K-Means
Objective
Segment customers into distinct groups to enable targeted and effective marketing strategies.
Approach
  • Extensive EDA with visualizations and statistical insights.
  • K-Means clustering applied, with optimal K identified via Elbow Method.
  • PCA used for 2D/3D visualization of clusters.
  • Proposed tailored marketing strategies for each segment.
Results
  • Customers clearly segmented by Age, Income, and Spending Score.
  • Visualizations confirmed clear separation of groups.
  • Generated actionable business insights for marketing campaigns.
Key Skills: Unsupervised Learning, Clustering, PCA, Business Insights.
Project 3: Energy Consumption Time Series Forecasting
1
Objective
Forecast short-term household energy usage based on historical consumption patterns.
2
Feature Engineering
Resampled data and engineered critical time-based features (hour of day, weekday/weekend, holidays).
3
Model Application
Applied ARIMA, Prophet, and XGBoost models for comparative analysis and optimal prediction.
4
Results
XGBoost provided superior predictive performance. Prophet effectively captured seasonal patterns, crucial for energy planning.
Key Skills: Time Series Forecasting, Feature Engineering, ARIMA, Prophet, XGBoost.
Project 4: Loan Default Risk Prediction with Business Cost Optimization
Objective
Predict loan defaults and optimize the decision threshold to minimize financial losses for the institution.
Approach
  • Robust data preprocessing (imputation, encoding, scaling, class balancing).
  • Trained Logistic Regression and CatBoost models.
  • Defined custom business costs for False Positives (risky loan approved) and False Negatives (safe loan rejected).
  • Optimized thresholds based on a comprehensive cost analysis.
Results
  • CatBoost exhibited the best predictive performance.
  • Business cost optimization significantly reduced total potential losses.
  • Visualizations (confusion matrix, cost vs. threshold curve) provided strong support for decision-making.
Key Skills: Classification, Imbalanced Data Handling, Cost-sensitive Learning, Threshold Optimization.
Project 5: Global Superstore Interactive Dashboard
Objective
Develop an interactive Business Intelligence (BI) dashboard for analyzing sales, profit, and customer segment performance.
Approach
Cleaned and prepared the Global Superstore dataset. Built a Streamlit + Plotly dashboard with dynamic filters (Region, Category, Sub-Category).
Key Visualizations
Displayed KPIs: Total Sales, Profit, Quantity Sold, Avg. Discount. Visualized sales trends, regional distribution, and customer segmentation.
Results: The interactive dashboard enabled dynamic exploration of sales and profit, providing clear insights into region- and segment-wise performance, serving as a business-ready tool for decision-makers.
Key Skills: Data Visualization, BI Dashboarding, Streamlit, Plotly.
Internship Takeaways
Comprehensive ML
Gained hands-on experience in Supervised (classification, forecasting) and Unsupervised (clustering) Machine Learning.
Actionable Insights
Mastered Explainable AI (SHAP) for transparent predictions and Business Cost Optimization for real-world decision-making.
Data Storytelling
Proficient in advanced Data Visualization and BI dashboards (Streamlit + Plotly) to communicate complex findings.
End-to-End Workflow
Executed the complete ML workflow from EDA and preprocessing to modeling, evaluation, and deriving actionable insights.
Conclusion: Bridging Theory and Practice
This internship provided a robust platform to apply theoretical knowledge to real-world data challenges across diverse domains. From predictive modeling and customer segmentation to time series forecasting and interactive dashboard development, each project refined my technical acumen and problem-solving capabilities.
My work demonstrates a strong foundation in end-to-end ML workflows, a commitment to generating transparent and cost-effective solutions, and the ability to translate complex data into actionable business intelligence. I am confident that these skills make me a valuable asset to any data-driven team.
Let's Connect!
I am always eager to discuss new challenges and opportunities in data science and analytics. Feel free to reach out!
Made with