Hello,
Can you please help me to understand this LARS?
This is a fraudulent claim dataset, and I need to examine the results.
Thank you very much.
Results - Node: LARS Diagram: Data Preparation and Analysis X File Edit View Window Output 20 X| Parameter Estimate (Absolute Values) Q X| Coefficient Paths 419 SO 75.95 0. 07826 1. 75956 0. 4777 10. 7405 149 0. 03652 Standardized Estimate v 0.2- 420 59.95 0. 00000 1. 59951 0. 0000 9. 7635 150 0. 0345 421 57.956 1. 36025 1. 57956 8. 3030 9. 6417 150 0. 02865 422 52.982 0.93331 1. 52982 5. 6970 9. 3381 150 0. 02677 0.100 0.050 - 423 70 42. 049 0. 0000 1. 42049 . 090 8. 670 150 0. 02516 0.000 - 424 75 32.633 0. 00867 1. 32633 0. 0529 8. 0960 150 0. 02325 425 80 24.969 0. 10055 1.24969 0. 6138 7. 6282 0. 01687 0.1 - 426 85 17. 615 D. 00000 1. 17615 0. 0000 7. 1793 150 0. 01574 GENDER_F INTERCEPT_ 427 90 11. 078 0. 00000 . 11078 0. 0000 6. 7803 150 0. 01146 CLAIM CAUSE_HAIL VEHICLE_CLASS_SUV 428 5.230 0. 00000 1. 05230 0. 0000 6. 4233 150 0. 00544 CLAIM CAUSE_COLLISION Standardized Estimate 100 0.000 0. 00000 1. 00000 0. 0000 6. 1041 149 0. 00376 VEHICLE_CLASS_SPORTS CAR EMPLOYMENT_STATUS_EMPLOYED VEHICLE_CLASS_FOUR-DOOR CAR 431 432 Data Role-VALIDATE Target Variable-Fraudulent_Claim Target Label-Fraudulent_Claim 433 434 Mean 0.1 435 Cumulative Cumulative Number of Posterior 436 epth Gain Lift Lift Response Response Observations Probability 437 438 5 51. 146 1. 51146 1. 51146 9. 3617 9. 3617 101 0. 17327 Effect 10 20 30 40 88. 005 2. 25233 1. 88005 13.9505 1. 6447 100 0. 16222 Step Fit Statistics Selected Variables DO X Target Target Label Fit Statistics Statistics Label Train Validation Test Effect Variable Class Level Standardized Estimate Estimate Fraudulent CI... Fraudulent Cl.. ASE Average Square. 0.053783 0.055497 GENDER F GENDER -0.11506 -0.055101 Fraudulent CL... Fraudulent CL.. DIV Divisor for ASE 5996 4004 VEHICLE CLASS SUV VEHICLE CLASS SUV -0.09646 -0.058270 it CL. Fraudulent CL... MAX Maximum Abs. .98225 0.977687 CLAIM CAUSE HAIL CLAIM CAUSE HAIL 0.056004 0.028942 Fraudulent CL... Fraudulent CI... NOBS Sum of Freque 2998 2002 VEHICLE CLASS SPORTS CAR VEHICLE CLASS SPORTS CAR -0.0372 -0.041112 Fraudulent CI... Fraudulent CI. RASE Root Average .. 0.231912 0.235577 CLAIM CAUSE COLLISION CLAIM CAUSE COLLISION 0.033136 0.016130 t CI.. Fraudulent Cl. Sum of Square. 322.483 222.2082 EMPLOYMENT STATUS EMPLOYED EMPLOYMENT STATUS EMPLOYED -0.00787 -0.003895 Fraudulent CI... Fraudulent CI... DISF Frequency of 2998 2002 VEHICLE CLASS FOUR-DOOR CAR VEHICLE CLASS FOUR-DOOR CAR 0.002462 0.001180 Fraudulent CL... Fraudulent CI... MISC Misclassification 0.061041 0.061938 INTERCEPT INTERCEPT 0.087715 Fraudulent CI... Fraudulent Cl.. WRONG Number of Wr.. 183 124 Score Rankings Overlay: Fraudulent_Claim Q X| Iteration Plot Cumulative Lift SBC v 2.5 -8500 - 2.0- Cumulative Lift 15 - -8600 1,0- 20 40 100 Depth 10 20 30 40 TRAIN VALIDATEOutput 10 Variable Summary Measurement Frequency Role Level Count INTERVAL INPUT INTERVAL INPUT NOMINAL 12 REJECTED NOMINAL TARGET BINARY 25 26 Model Events 27 Number Measurement of 30 Target Event Level Levels Order Label 31 32 Fraudulent_Claim BINARY 2 Descending Fraudulent_Claim 33 34 35 Predicted and decision variables Type Variable Label TARGET Fraudulent_Claim Fraudulent_Claim PREDICTED P_Fraudulent_ClaimY Predicted: Fraudulent_Claim-Y RESIDUAL R_Fraudulent_ClaimY Residual: Fraudulent_Claim=Y PREDICTED P_Fraudulent_Claim Predicted: Fraudulent_Claim=N RESIDUAL R_Fraudulent_ClaimN Residual: Fraudulent_Claim-N FROM F_Fraudulent_Claim From: Fraudulent_Claim INTO I_Fraudulent_Claim Into: Fraudulent_Claim The STDIZE Procedure Location and Scale Heasures Location = mean Scale = standard deviation Name Location Scale N Label Annual_Premium 1129. 889260 312. 414269 2998 Annual_Premium Clain_Amount 797. 421747 688. 767255 2998 Claim_Amount IMP_Outstanding_Balance 23800 13817 998 Imputed: Outstanding_Balance Income 38327 34559 2998 Income Monthly_Premium 94. 157438 26. 034522 2998 Monthly_Premium Months_Since_Last_Claim 15. 065710 11. 099995 2998 Months_Since_Last_Claim 57 Months_Since_Policy_Inception 48. 270847 28. 172902 2998 Months_Since_Policy_Inception 68Output 73 The GLMSELECT Procedure 74 75 Data Set WORK. THE TRAINZ 76 validation Data Set WORK. _TMP_VALIDATE2 77 Dependent Variable targetcode 78 Selection Method LAR 79 Stop at Specified Number of Effects 200 80 Choose Criterion SBC 81 Effect Hierarchy Enforced None 82 83 84 Observation Profile for Analysis Data 85 86 Number of Observations Read 2998 87 Number of Observations Used 2998 88 Number of Observations Used for Training 2998 89 90 91 Observation Profile for Validation Data 92 93 Number of Observations Read 2002 94 Number of Observations Used 2002 95 96 97 Class Level Information 98 99 Class Levels Values 100 101 Clain_Cause COLLISION FIRE HAIL OTHER SCRATCH/DENT 102 Claim Date 01/15/2019 12/01/2018 12/15/2018 103 Clain_Report_Type AGENT BRANCH CALL CENTER WEB 104 Employment_Status DISABLED EMPLOYED MEDICAL LEAVE RETIRED UNEMPLOYED 105 Gender FM 106 IMP_Education 5 BACHELOR COLLEGE DOCTOR HIGH SCHOOL OR BELOW 107 MASTER 108 IMP_Location 3 RURAL SUBURBAN URBAN 109 Marital_Status DIVORCED MARRIED SINGLE 110 State_Code IA KS HO NE OK 111 Vehicle_Class FOUR-DOOR CAR LUXURY CAR LUXURY SUV SPORTS CAR SUV 112 TWO-DOOR CAR 113 Vehicle_Model CHEVROLET FORD HONDA TOYOTA 114 Vehicle_Size COMPACT LUXURY MIDSIZE 115 116 117 Dimensions 118 119 Number of Effects 20 120 Number of Effects after Splits 44 121 Number of Parameters 44Output 127 The GLMSELECT Procedure 128 129 LAR Selection Summary 130 131 Effect Number Validation 132 Step Entered Effects In SBC ASE ASE 133 134 Intercept -8563. 8689 0.0573 0. 0581 135 136 Gender_F -8574. 3971 0. 0570 0. 0577 137 Vehicle_Class_SUV -8629. 8898 0. 0558 0. 0566 138 Claim_Cause HAIL -8629. 5606 0. 0556 0. 0565 139 Vehicle_Class_SPORTS CAR -8648. 2866 0. 0551 0. 0561 140 Claim_Cause_COLLISION -8652. 1660 0. 0549 0. 0559 141 Vehicle_Class_FOUR-DOOR CAR -8644. 1782 0. 0549 0. 0559 142 Employment_Status_EMPLOYED -8652. 3789- 0. 0546 0. 0557 143 Marital_Status_HARRIED -8646. 3425 0. 0546 0. 0557 144 Vehicle_Size_COMPACT 10 -8641. 2040 0. 0545 0. 0556 145 10 Income 11 -8638. 5998 0. 0544 0. 0556 146 11 Vehicle_Class_LUXURY SUV 12 -8633. 5977 0. 0544 D. 0555 147 12 Claim_Date_12/01/2018 13 -8632. 7113 0. 0542 0. 0555 148 13 Claim_Cause_OTHER 14 -8627. 4778 0. 0542 0. 0554 149 14 Vehicle_Class_LUXURY CAR 15 -8620.9126 0. 0542 0. 0554 150 15 IMP_Outstanding_Balance 16 -8613. 7134 0. 0542 0. 0554 151 16 Marital_Status_DIVORCED 17 -8607. 0612 0. 0541 0. 0554 152 17 Employment_Status_RETIRED 18 -8599. 5511 0. 0541 0. 0553 153 18 Claim_Amount 19 -8594. 5800 0. 0541 0. 0553 154 19 Claim_Date_01/15/2019 20 -8586. 7761 0. 0541 0. 0553 155 20 Vehicle_Model_HONDA 21 -8585.6688 0. 0539 0. 0552 156 21 State_Code_TA 22 -8579.9113 0. 0539 0. 0552 157 22 Claim_Report_Type_AGENT 23 -8577. 5503 0. 0538 0. 0551 156 23 IMP_Education_BACHELOR 24 -8569. 6012 0. 0538 0. 0551 159 24 IMP_Location_SUBURBAN 25 -8562. 0745 0.0538 0. 0551 160 25 Claim_Report_Type_BRANCH 26 -8556. 7595 0. 0537 0. 0551 161 26 IMP_Education_COLLEGE 27 -8548. 8087 0.0537 0. 0551 162 27 Months_Since_Last_Claim 28 -8544. 0164 0.0537 0. 0550 163 28 Claim_Report_Type_CALL CENTER 29 -8536. 8609 0. 0537 0. 0550 164 29 IMP_Education_DOCTOR 30 -8530. 7027 0. 0536 0. 0550 165 30 Annual_Premium 31 -6524. 0552 0. 0536 0. 0550 166 31 Months_Since_Policy_Inception 32 -8517. 9178 0. 0536 0. 0550 167 32 State_Code_KS 33 -8511. 3713 0.0536 0. 0550 168 33 State_Code_NE 34 -8504. 9908 0. 0535 0. 0550 169 34 Claim_Cause_FIRE 35 -8497. 0095 0. 0535 0. 0550 170 35 Vehicle Model_FORD 36 -6489. 1753 0. 0535 0. 0550 171 36 Employment_Status_MEDICAL LEAVE 37 -8481. 5021 0. 0535 0. 0550 172 37 Vehicle_Size_LUXURY 38 -8473. 5556 0. 0535 0. 0550 173 38 IMP_Location RURAL 39 -8466. 2943 0. 0535 0. 0550 174 39 IMP_Education_HIGH SCHOOL OR BELOW 40 -8459. 3887 0. 0535 0. 0550* 175 40 state_Code_MO 41 -8455. 7916 0. 0534 0. 0550 176 41 Vehicle Model_CHEVROLET 42 -8447. 8594 0. 0534 0. 0550 177 42 Employment_Status_DISABLED 43 -8440. 1515 0. 0534 0. 0550 178 179 * Optimal Value of Criterion 180 181 102 Selection stopped because all candidate effects for entry are linearly dependent on effects in the model. 183Output 178 179 * Optimal Value of Criterion 180 181 182 Selection stopped because all candidate effects for entry are linearly dependent on effects in the model. 183 184 185 186 187 188 The GLMSELECT Procedure 189 Selected Hodel 190 191 The selected model, based on SBC, is the model at Step 7. 192 193 194 Effects: Intercept Claim_Cause_COLLISION Claim_Cause_HAIL Employment_Status_EMPLOYED Gender_F Vehicle_Class_FOUR-DOOR CAR Vehicle_Class_SPORTS CAR Vehicle_Class_SUV 195 196 197 Analysis of Variance 198 199 Sum of Mean 200 Source Squares Square F Value 201 202 Model 8. 08828 1. 15547 21 . 10 203 Error 2990 163. 74127 0. 05476 204 Corrected Total 2997 171.82955 205 206 207 Root MSE 0. 23401 208 Dependent Mean 0. 06104 209 R-Square 0. 0471 210 Adj R-Sq 0. 0448 211 -5700. 42446 212 AIC -5700. 36422 213 SBC -8652. 37885 214 ASE (Train) 0. 05462 215 ASE (Validate) 0. 05570 216 217 218 Parameter Estimates 219 220 Parameter DF Estimate 221 222 Intercept 1 0. 087715 223 Claim_Cause_COLLISION 0. 016130 224 Clain_Cause HAIL 0. 02894 225 Employment_Status_EMPLOYED -0. 003895 226 Gender_F -0. 055101 227 Vehicle_Class_FOUR-DOOR CAR 0. 001180 228 Vehicle_Class_SPORTS CAR -0. 041112 229 Vehicle_Class_SUV -0. 058270 230 231 232 Score Information 233 234 Input Data Set WORK. _TMP_VALIDATE2 235 Output Data Set WORK. PRED_DATA 236 237 Number of Observations Read 2002 238 Number of Observations Scored 2002 239Model Information Data Set WORK. PRED DATA Score Results for DATA=WORK. _TMP_VALIDATEZ Response Variable targetcode Number of Response Levels 2 Model generalized logit Optimization Technique Newton-Raphson Number of Observations Read 2002 Number of Observations Used 2002 Response Profile Ordered Total Value targetcode Frequency 1878 124 Logits modeled use targetcode-l as the reference category. Model Convergence Status Convergence criterion (GCONV=1E-8] satisfied. Model Fit Statistics Intercept Intercept and Criterion Only Covariates AIC 931.998 831. 743 SC 937. 600 842.947 -2 Log L 929.998 827. 743 Testing Global Null Hypothesis: BETA-0 Test Chi-Square DF Pr > Chisq Likelihood Ratio 102. 2544 <. score c. wald type analysis of effects effect df chi-square pr> Chisq p_targetcode 1 73. 8677 <. analysis of maximum likelihood estimates standard wald parameter targetcode df estimate error chi-square pr> Chisq Intercept 1 5. 0137 0. 3274 234. 5737 <.0001 p_targetcode ratio estimates point wald effect targetcode estimate confidence limits association of predicted probabilities and observed responses percent concordant somers d discordant ganma tied tau-a pairs score output report fit statistics target-fraudulent_clain target label="Fraudulent_Claim" train validation ase_ average squared error div_ divisor for ase max maximum absolute nobs_ sum frequencies rase_ root sse sun errors disf_ frequency classified cases misc hisclassification rate wrong_ number wrong classifications table data role="TRAIN" variable="Fraudulent_Claim" outcome total percentage count event classification false true negative positive rankings mean cumulative posterior depth gain life lift response observations probability d. variable-fraudulent_claim label-fraudulent_claim b. assessment distribution range events nonevents>
