This report is made to ensure the continuity of business processes, preserve and share internal organization expertise within the Group, as well as to prove the feasibility and validity of the developed model.
To achieve the goals, the Report provides insights into some essential methodological approaches to modeling.
{{ model_aim }}
The report has the following structure:
Key model identifiers are given in Table 1.
Parameter | Value |
---|---|
Model name / ID | {{ model_name }} |
Report version | {{ report_version }} |
Model customer | {{ zakazchik }} |
Name of the Group member and high level department | {{ high_level_department }} |
Model developer (if an external developer is involved, it is necessary to indicate Company's name) | {{ ds_name }} |
Description and specification of datasets that are used to develop and test the model are presented in Table 2.
Parameter | Training set | Test set | |||
Target event description | {{ target_descr }} | ||||
Non-target event description | {{ non_target_descr }} | ||||
Observations count in sets | {{ count_train }} | {{ count_test }} | |||
Target events count | {{ train_target_cnt }} | {{ test_target_cnt }} | |||
Non-target events count | {{ train_nontarget_cnt }} | {{ test_nontarget_cnt }} | |||
Mean of target in sets | {{ train_target_perc }}% | {{ test_target_perc }}% |
Predictions of the model trained on the training set are shown in Table 3.
Parameter | Training set | Test set |
AUC | {{ train_auc_full }}% | {{ test_auc_full }}% |
Gini | {{ train_gini_full }}% | {{ test_gini_full }}% |
Coefficients of the regression model are shown in Table 4.
Feature name | Regression coefficient |
{{ pair[0] }} | {{ pair[1] }} |
Statistics related to the number of missing values between training and test sets are shown in Table 5.
Feature name | Number of valid values in the training set | Number of valid values in the test set | Number of missing values in the training set | Number of missing values in the test set | Percentage of valid values in the training set | Percentage of valid values in the test set | Difference in percentage of valid values in sets |
{{ val[0] }} | {{ val[1] }} | {{ val[2] }} | {{ val[3] }} | {{ val[4] }} | {{ val[5] }} | {{ val[6] }} | {{ val[7] }} |
Feature name | P-value |
{{ val[0] }} | {{ val[1] }} |
Feature name | P-value |
{{ val[0] }} | {{ val[1] }} |
Required to train model with a parameter regularized_refit=False
{% endif %}Gini for the training set
Gini for the test set
Feature name | VIF value |
{{ val[0] }} | {{ val[1] }} |
Two or more features required for VIF calculation
{% endif %}Total PSI
Feature name | PSI value |
{{ val[0] }} | {{ val[1] }} |
PSI for non-target events
Feature name | PSI value |
{{ val[0] }} | {{ val[1] }} |
PSI for target events
Feature name | PSI value |
{{ val[0] }} | {{ val[1] }} |
Grouping by predictions on the training set (total)
Grouping by predictions of target and non-target events on the training set
Grouping by predictions on the test set (total)
Grouping by predictions of target and non-target events on the test set
PSI by grouped predictions of model
Total PSI | {{ psi_binned_total }} |
PSI for non-target events | {{ psi_binned_zeros }} |
PSI for target events | {{ psi_binned_ones }} |
To calculate PSI, you must first call fit() and predict_proba()
{% endif %}ScoreBin | count | mean | std | min | 25% | 50% | 75% | max |
{{ val[0] }} | {{ val[1] }} | {{ val[2] }} | {{ val[3] }} | {{ val[4] }} | {{ val[5] }} | {{ val[6] }} | {{ val[7] }} | {{ val[8] }} |
ScoreBin | count | mean | std | min | 25% | 50% | 75% | max |
{{ val[0] }} | {{ val[1] }} | {{ val[2] }} | {{ val[3] }} | {{ val[4] }} | {{ val[5] }} | {{ val[6] }} | {{ val[7] }} | {{ val[8] }} |
Variable | Value | WOE | COEF | POINTS |
{{ val[0] }} | {{ val[1] }} | {{ val[2] }} | {{ val[3] }} | {{ val[4] }} |
Feature | |
{{ val[0] }} | {{ val[1] }} |
Feature | Contribution to ROC AUC |
{{ val[0] }} | {{ val[1] }} |
Evaluation requires at least 2 features in the final model
{% endif %} {% else %}Required to train the model with a parameter regularized_refit=False
{% endif %}