Using LARS Regression for Objective FunctionEfficient Frontier Analytic TeamJuly , 2010
GoalFind the best set of weights across the whole loan channelthat best capture the loan intent and conversion behaviors to help optimize SEM optimization.
Correlation Analysis Auto TD and UW_TD have positive correlation with Total.Apps
Total.Loans has strong correlation with Total Amount, so we will only put Total.Loans in the objective function. Variable Selection: LARS AlgorithmModel 2 captures the best relationship between Total Loan and other variable best. By using this model, we can give best prediction of Total Loan with lowest predictive error based on other variables.
Recommended Objective FunctionTotal.Apps captures the volume of loan applications

Weight presentation

  • 1.
    Using LARS Regressionfor Objective FunctionEfficient Frontier Analytic TeamJuly , 2010
  • 2.
    GoalFind the bestset of weights across the whole loan channelthat best capture the loan intent and conversion behaviors to help optimize SEM optimization.
  • 3.
    Correlation Analysis AutoTD and UW_TD have positive correlation with Total.Apps
  • 4.
    Total.Loans has strongcorrelation with Total Amount, so we will only put Total.Loans in the objective function. Variable Selection: LARS AlgorithmModel 2 captures the best relationship between Total Loan and other variable best. By using this model, we can give best prediction of Total Loan with lowest predictive error based on other variables.
  • 5.
    Recommended Objective FunctionTotal.Appscaptures the volume of loan applications

Editor's Notes

  • #5 Step 2 gives lowest Prediction error so we only include two variables, Total.Apps and Approved.