Making Predictions

import chainladder as cl
import pandas as pd

This example demonstrates how you can create development patterns at a particular index grain and apply them to another.

In this example, we can apply both an industry derived LDF or a company derived LDF to a specific triangle.

clrd = cl.load_sample("clrd")["CumPaidLoss"]
clrd_wc = clrd[clrd["LOB"] == "wkcomp"]

industry_wc = clrd_wc.sum()

allstate_wc_industry_pattern = (
    cl
    .Chainladder()
    .fit(industry_wc)
    .predict(clrd_wc.loc["Allstate Ins Co Grp"])
    .ultimate_
)
allstate_wc_company_pattern = (
    cl.Chainladder().fit(clrd_wc.loc["Allstate Ins Co Grp"]).ultimate_
)

df = pd.DataFrame({
    "allstate_wc_industry_pattern": allstate_wc_industry_pattern.to_frame().iloc[:, 0],
    "allstate_wc_company_pattern": allstate_wc_company_pattern.to_frame().iloc[:, 0],
}).fillna(0)
df.index = pd.to_datetime(df.index).year
df.index.name = "Origin Year"
df.reset_index()

	Origin Year	allstate_wc_industry_pattern	allstate_wc_company_pattern
0	1988	325322.000000	325322.000000
1	1989	276660.740492	276863.571154
2	1990	264559.787229	268960.552632
3	1991	252620.869201	258402.289056
4	1992	173864.635638	180150.887317
5	1993	99493.416499	104286.313106
6	1994	112350.433124	119003.414168
7	1995	123829.928837	132157.175486
8	1996	83778.038787	90947.646834
9	1997	2837.012601	3110.281691

import matplotlib.pyplot as plt

plot_df = df[["allstate_wc_industry_pattern", "allstate_wc_company_pattern"]].rename(
    columns={
        "allstate_wc_industry_pattern": "Industry Pattern",
        "allstate_wc_company_pattern": "Company Pattern",
    }
)

# Create bar chart
plot_df.plot(kind="bar")

# Labels and title
plt.title("Allstate WC Ultimate Comparison")
plt.ylabel("Ultimate Loss")
plt.xticks(rotation=0)

plt.tight_layout()
plt.show()