Actuarial & Financial Modeling

@tool(name="ActuarialModelingTool", description="Perform actuarial analysis on historical claims & persistency data(product, metric, start_year, end_year, csv_path). Metric: LossRatio, PersistencyRate, ClaimsSeverity, ClaimsFrequency.", show_result=True)
def ActuarialModelingTool(
    product: str = None,
    metric: str = "LossRatio",
    start_year: int = 2015,
    end_year: int = 2020,
    csv_path: str = None,
) -> str:
    """
    Perform actuarial analysis on historical claims & persistency data.

    Args:
        product: Filter results for a specific insurance product(e.g. 'Whole Life', 'Annuity', 'Health'). Defaults to None (all products).
        metric: The actuarial metric to calculate. Options: LossRatio, PersistencyRate, ClaimsSeverity, ClaimsFrequency.
        start_year: Starting year of analysis. Defaults to 2015.
        end_year: Ending year of analysis. Defaults to 2020.
        csv_path: Path to CSV file with historical claims & persistency data.

    Returns:
        str: JSON string with computed actuarial results.
    """
    csv_path = csv_path or _path("agent3.csv")
    try:
        df = pd.read_csv(csv_path)
        df = df[(df["Year"] >= start_year) & (df["Year"] <= end_year)]
        if product:
            df = df[df["Product"].str.lower() == product.lower()]
        results = []
        for _, row in df.iterrows():
            record = {"Year": int(row["Year"]), "Product": row["Product"]}
            if metric == "LossRatio":
                record["LossRatio"] = round(row["Claims_Paid"] / row["Premiums_Collected"], 4)
            elif metric == "PersistencyRate":
                record["PersistencyRate"] = row["Persistency_Rate"]
            elif metric == "ClaimsSeverity":
                record["ClaimsSeverity"] = round(row["Claims_Paid"] / row["Claims_Count"], 2) if row["Claims_Count"] > 0 else None
            elif metric == "ClaimsFrequency":
                record["ClaimsFrequency"] = round(row["Claims_Count"] / row["Policy_Count"], 4) if row["Policy_Count"] > 0 else None
            else:
                return json.dumps({"error": f"Invalid metric '{metric}'"})
            results.append(record)
        return json.dumps(results, indent=2)
    except Exception as e:
        return json.dumps({"error": str(e)})