Business Context
An auto insurance provider wants to estimate the expected number of claims per policyholder based on driver age, vehicle type, and region. This helps in adjusting premium rates for different risk profiles using a standard Poisson regression.
Data Preparation
Generation of 1,000 policyholder records with age, car type, and claim counts derived from a Poisson distribution.
Copied!
| 1 | |
| 2 | DATA mycas.insurance_claims; |
| 3 | call streaminit(12345); |
| 4 | DO policy_id = 1 to 1000; |
| 5 | Age = 18 + floor(rand('UNIFORM') * 60); |
| 6 | CarType = ifn(rand('UNIFORM') > 0.3, 'Sedan', 'Sport'); |
| 7 | Region = ifn(rand('UNIFORM') > 0.5, 'Urban', 'Rural'); |
| 8 | RiskBase = -1.5 + 0.02 * Age + 0.5 * (CarType='Sport') + 0.3 * (Region='Urban'); |
| 9 | Lambda = exp(RiskBase); |
| 10 | NumClaims = rand('POISSON', Lambda); |
| 11 | OUTPUT; |
| 12 | END; |
| 13 | |
| 14 | RUN; |
| 15 |
Étapes de réalisation
1
Fit a standard Poisson regression model using Class variables for categorical inputs.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | countreg.countregFitModel / TABLE={name='insurance_claims'} class={'CarType', 'Region'} model={depVars={{name='NumClaims'}}, effects={{vars={'Age', 'CarType', 'Region'}}}, modeloptions={modeltype='POISSON'}}; |
| 4 | |
| 5 | RUN; |
| 6 |
2
Verify model convergence and parameter estimates for risk factors.
Copied!
| 1 | |
| 2 | PROC CAS; |
| 3 | countreg.countregFitModel / TABLE={name='insurance_claims'} class={'CarType'} model={depVars={{name='NumClaims'}}, effects={{vars={'CarType'}}}} outputTables={names={ParameterEstimates='params'}}; |
| 4 | |
| 5 | RUN; |
| 6 |
Expected Result
The model successfully converges. The 'ParameterEstimates' table shows positive coefficients for 'Sport' cars and 'Urban' regions, indicating higher expected claim counts, validating the business logic.