data-analysis

## Instructions When the user asks for claims analysis, loss ratio calculations, premium comparisons, coverage utilization reports, or risk metrics, follow this workflow: ### 1. Data Ingestion - If the user has data from the **crm** skill (customer/policy records), use that as input - If the user provides CSV/JSON data inline, write it to `/tmp` first using `code_interpreter` - If no data is available, offer to generate realistic sample insurance data for demonstration ### 2. Insurance-Specific Analyses #### Claims Analysis ```python import pandas as pd df = pd.DataFrame(claims_data) print(f"Total Claims: {len(df)}") print(f"Open Claims: {len(df[df['status'] == 'Open'])}") print(f"Average Claim Amount: ${df['amount'].mean():,.2f}") print(f"Total Incurred: ${df['amount'].sum():,.2f}") print(f"\nClaims by Status:\n{df['status'].value_counts()}") print(f"\nClaims by Product Line:\n{df.groupby('product_type')['amount'].agg(['count', 'sum', 'mean'])}") ``` #### Loss Ratio Calculation ```python def loss_ratio(incurred_losses, earned_premiums): """Loss ratio = Incurred Losses / Earned Premiums""" ratio = incurred_losses / earned_premiums * 100 status = "Profitable" if ratio < 70 else "Borderline" if ratio < 100 else "Unprofitable" return round(ratio, 1), status ratio, status = loss_ratio(incurred_losses=850000, earned_premiums=1200000) print(f"Loss Ratio: {ratio}% — {status}") ``` #### Premium Comparison ```python # Compare coverage options side by side options = pd.DataFrame([ {"Plan": "Basic", "Premium": 120, "Deductible": 2500, "Limit": 100000}, {"Plan": "Standard", "Premium": 200, "Deductible": 1000, "Limit": 250000}, {"Plan": "Premium", "Premium": 350, "Deductible": 500, "Limit": 500000}, ]) print(options.to_string(index=False)) ``` #### Coverage Utilization - Claims frequency by policy type - Average severity (claim amount) by product line - Utilization rate (claims filed / policies in force) ### 3. Visualizations Generate charts with matplotlib and save to `/tmp`: ```python import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt # Claims by product line fig, ax = plt.subplots(figsize=(10, 6)) claims_by_type.plot(kind="bar", ax=ax) ax.set_title("Claims by Product Line") ax.set_ylabel("Number of Claims") plt.tight_layout() plt.savefig("/tmp/claims_by_product.png", dpi=150) plt.close() print("Chart saved: /tmp/claims_by_product.png") ``` Common insurance charts: - **Bar chart**: Claims count/amount by product line or status - **Line chart**: Monthly claims trend, loss ratio over time - **Pie chart**: Claims distribution by category (≤6 categories) - **Stacked bar**: Open vs closed claims by month - **Heatmap**: Claims by region and product type ### 4. Risk Metrics Provide insights based on the data: - **Loss ratio trend**: Is it improving or deteriorating? - **Claims frequency**: Number of claims per 1,000 policies - **Average severity**: Mean claim amount by product line - **Large loss identification**: Flag claims significantly above average - **Concentration risk**: Are claims concentrated in specific regions, products, or time periods? ### 5. Output - Print key findings as text - Save charts to `/tmp` with descriptive names - Reference file paths for download - If analysis produces a transformed dataset, save as CSV to `/tmp` ## Chaining - `code_interpreter` — runs the Python code - `crm` — provides source customer and policy data - `rag-search` — provides policy wording context for claims analysis - `file-sharing` — delivers charts and exports ## Constraints - Pre-installed libraries: pandas, numpy, matplotlib. Additional libraries can be installed at runtime via pip. - Files must be written to `/tmp` - Max execution time: 30 seconds per code block - Always include appropriate caveats on actuarial estimates