Predictive modelling to score customer leads using data analytics using an end-to-end automated, sampled approach with iterative local and global optimization
Embodiments of the present invention disclose system to determine the best model to perform lead scoring for a given data set. The system can perform a multi-step iterative procedure including variable selection, feature set selection, training data selection, model development, model validation and process optimization. The system also performs local and global optimizations iteratively to determine the best possible model for a given scenario.
1. A computer-implemented method for lead scoring, comprising:
preparing a data set by using a median value for one or more missing values of variables in case a number of missing values of the variables is within a threshold level;
performing random sampling of the data set to generate training and test data;
building a model based on the training and test data; refining the model by using a true positive rate (TPR) and a true negative rate (TNR); and
validating the model by simultaneously optimizing a difference between the TPR and a validation TPR and a difference between the TNR and a validation TNR.
2. The computer-implemented method of claim 1 , wherein the
TPR
=
Number
of
Enquiries
Predicted
as
Potential
Orders
Number
of
Enquiries
Actually
Converted
to
Orders
.
3. The computer-implemented method of claim 1 , wherein the
TNR
=
Number
of
Enquiries
Predicted
as
Potential
Drops
Number
of
Enquiries
Actually
Confirm
as
Drops
.
4. The computer-implemented method of claim 1 , wherein the data set is further filtered to remove variables based on their usefulness.
5. The computer-implemented method of claim 1 , wherein the preparing of the data set further comprises identifying independent variables and dependent variables.
6. The computer-implemented method of claim 1 , further includes selecting variables, succeeding the step of preparing the data set, which includes:
ignoring variables with missing values up to a threshold percentage level;
creating new variables and dummy variables; and
grouping the variables based on their conversion levels.
7. The computer-implemented method of claim 1 , wherein the validation of the model further comprises:
determining a model from a plurality of models using a lift chart; and
performing a concordance test to validate the model.
8. The computer-implemented method of claim 7 , further comprises, succeeding the validation of the built model:
prediction of lead scores for the data set; and
tracking the predicted lead scores against an actual data.
9. The computer-implemented method of claim 8 , further comprises building a new model or updating the built model based on a determination if the predicted lead scores are below a threshold.
10. A system to score lead comprising:
preparation of a data set by using a median value for one or more missing values of variables in case a number of missing values of the variables is within a threshold level;
perform random sampling of the data set to generate training and test data;
build a model based on the training and test data; and
refine the model by using a true positive rate (TPR) and a true negative rate (TNR); and
validate the model by simultaneously optimizing a difference between the TPR and a validation TPR and a difference between the TNR and a validation TNR.
11. The system of claim 10 , wherein the
TNR
=
Number
of
Enquiries
Predicted
as
Potential
Drops
Number
of
Enquiries
Actually
Confirmed
as
Drops
.
12. The system of claim 10 , wherein the data set is further filtered to remove variables based on their usefulness.
13. The system of claim 10 , wherein the preparation of the data set further comprises identifying independent variables and dependent variables.
14. The system of claim 13 , further includes selection of variables, succeeding the preparation of the data set, the selection of variables comprises:
ignore variables with missing values up to a threshold percentage level;
creation new variables and dummy variables; and
group the variables based on their conversion levels.
15. The system of claim 10 , wherein the validation of the built model further comprises:
determination of a model from a plurality of models using a lift chart; and
perform a concordance test to validate the model.
16. The system of claim 15 , further comprises, succeeding the validation of the built model:
prediction of lead scores for the data set; and
tracking the predicted lead scores against an actual data.
17. The system of claim 16 , further comprises building a new model or updating the built model based on a determination if the predicted lead scores are below a threshold.
18. The computer-implemented method of claim 1 , wherein the model is validated by decreasing the difference between the TPR and the validated TPR and decreasing the difference between the TNR and the validated TNR iteratively.
19. The system of claim 10 , wherein the
TPR
=
Number
of
Enquiries
Predicted
as
Potential
Orders
Number
of
Enquiries
Actually
Converted
to
Orders
.