Portfolio Risk Profiling
Project Context
A wealth management firm wants to segment investors based on age, allocation mix, horizon, return
expectations, and risk tolerance. The cluster profiles will be used for advisory and product
personalization.
Dataset Reference
Specialization: Finance
Dataset file: datasets/finance_portfolio_risk_profiling.csv
Available columns:
investor_id,age,income_lpa,investment_horizon_years,equity_pct,debt_pct,gold_pct,cash_pct,monthly
_investment_inr,risk_tolerance_score,expected_return_pct,past_portfolio_return_pct,drawdown_toler
ance_pct,advisor_interactions,goal_type
Target And Predictor Variables
Recommended model family: Clustering
Target or modeling output: No single dependent target. Use investor and allocation features to create
risk clusters.
Suggested predictor variables: age, income_lpa, investment_horizon_years, equity_pct, debt_pct,
gold_pct, cash_pct, monthly_investment_inr, risk_tolerance_score, expected_return_pct,
past_portfolio_return_pct, drawdown_tolerance_pct, advisor_interactions
End To End Notebook Workflow
1. Create a Python notebook and begin with a title cell, project objective, business context, and team
member names.
2. Import pandas, numpy, matplotlib, seaborn, and relevant scikit-learn modules. Display package
versions if possible.
3. Load the CSV file, inspect shape, preview rows, review data types, and report missing values and
duplicate counts.
4. Clean the data by correcting data types, handling null values, standardizing category labels, and
documenting any assumptions.
5. Run preprocessing steps such as label encoding or one-hot encoding, scaling numeric variables where
required, and feature selection or feature engineering if justified.
6. Perform exploratory data analysis. Use describe, value_counts, correlations, grouped summaries, and
domain interpretation before any model building.
7. Use pandas operations that the course expects: filtering, sorting, groupby, agg, pivot_table, crosstab,
ranking, and derived columns.
8. Build at least 5 matplotlib visualizations. Include a mix of distribution charts, comparison charts, trend
or pattern charts, and at least one plot that directly supports a business decision.
9. Define the target and predictor set clearly. Split the data into training and test sets where supervised
learning is used.
10. Train the recommended model. Also train one comparison model and explain why the final choice is
stronger or weaker.
11. Evaluate results using suitable metrics. For regression use measures such as RMSE, MAE, and R-
squared. For classification use accuracy, precision, recall, F1-score, ROC-AUC, and confusion matrix. For
clustering use silhouette score and cluster profile interpretation.
12. Interpret the model in business language. Discuss feature importance, coefficients, cluster profiles,
or decision drivers.
13. End the notebook with managerial recommendations, limitations of the synthetic dataset, and ideas
for future improvement.
Required Analysis Tasks
1. Report shape, column list, data types, null counts, and duplicate count.
2. Show at least 3 cleaning or preprocessing actions and explain why they were needed.
3. Create at least 3 groupby summaries relevant to the specialization.
4. Create at least 2 pivot tables and sort them to show top and bottom performers.
5. Use filtering and sorting to isolate useful business cases such as high risk, high response, low
performance, or high opportunity observations.
6. Build at least 5 matplotlib charts with titles, readable axes, and short written interpretation below
each chart.
7. Compare the recommended model with one alternate model.
8. Present final findings as business recommendations rather than only technical output.
Project Questions To Answer
1. Which investor groups are conservative, balanced, and aggressive?
2. How do age, horizon, and asset allocation vary across clusters?
3. Which cluster is suitable for targeted advisory products?
Power BI Dashboard Requirements
Students must also create a Power BI dashboard using the same dataset.
1. Import the CSV into Power BI and clean obvious issues in Power Query if needed.
2. Create a data model that keeps the dataset usable for slicing by category, region, segment, period, or
other important dimensions.
3. Build KPI cards for the main business outcomes relevant to the project.
4. Add at least one bar or column chart, one line or trend view where applicable, one matrix or table,
one slicer panel, and one comparison visual.
5. Include filters such as region, channel, customer segment, product, policy type, department, route, or
equivalent based on the dataset.
6. Add a final page or section called Key Insights that summarizes the major takeaways from the
dashboard.
7. The dashboard should support the same story told in the Python notebook and not contradict the
notebook findings.
Submission Checklist
1. Python notebook with code, outputs, and explanations.
2. Power BI dashboard file and screenshots if required by faculty.
3. Short presentation summarizing problem, analysis, model, dashboard insights, and recommendations.
4. Clear statement of target variable and predictor variables used.
5. Final conclusion describing what managers should do next.