Enter Applicant Information

Enter Age

Choose a gender

Select highest education

Enter Annual Income ($)

Enter Employment Experience in Years

Select Home Ownership

Enter Loan Amount ($)

Select Loan Intent

Enter Loan Interest Rate (%)

Enter Loan's Percentage of Income (Decimanl Equivalent Form)

Enter Credit Bureau's Record of Credit History Length in Years

Enter Credit Score

Previous Loan Defaults on File

(~30 seconds)

Results

Loan Application Data Pipeline Documentation

Pipeline Overview

This pipeline preprocesses the loan application dataset to prepare it for machine learning modeling. It handles categorical encoding, outlier detection and treatment, and feature scaling.

Pipeline Diagram

Step-by-Step Design Choices

1. Gender Mapping (`person_gender`)

Transformer: CustomMappingTransformer('person_gender', {'male': 0, 'female': 1})),
Design Choice: Binary encoding of gender with female as 1 and male as 0
Rationale: Simple categorical mapping that preserves the binary nature of the feature without increasing dimensionality.

2. Education Mapping (`person_education`)

Transformer: CustomMappingTransformer('person_education', {'High School': 0, 'Associate': 1, 'Bachelor': 2, 'Master': 3, 'Doctorate': 4}))
Design Choice: Ordinal encoding of education class from lowest (High School) to highest (Doctorate)
Rationale: Preserves the ordering of education levels. This representation aligns with the assumption that higher education levels may positively influence loan approval likelyhood.

3. Target Encoding for Home Ownership Column (`person_home_ownership`)

Transformer: CustomTargetTransformer(col='person_home_ownership')
Design Choice: Target encoding.
Rationale:
Replaces the categorical 'person_home_ownership' feature with its relationship to the target variable

4. Target Encoding for Loan Intent Column (`loan_intent`)

Transformer: CustomTargetTransformer(col='loan_intent')
Design Choice: Target encoding.
Rationale:
Replaces the categorical 'loan_intent' feature with its relationship to the target variable

5. Previous Loan Defaults on File Mapping (`previous_loan_defaults_on_file`)

Transformer: CustomMappingTransformer('previous_loan_defaults_on_file', {'Yes': 1, 'No': 0}),
Design Choice: Binary encoding of whether the applicant has a previous loan default on file with yes as 1 and no as 0
Rationale: Simple categorical mapping that preserves the binary nature of the feature without increasing dimensionality.

6. Outlier Treatment for Age (`person_age`)

Transformer: CustomTukeyTransformer(target_column='person_age', fence='outer'),
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

6. Outlier Treatment for Income (`person_income`)

Transformer: CustomTukeyTransformer(target_column='person_income', fence='outer'),
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

7. Outlier Treatment for Employment Experience (`person_emp_exp`)

Transformer: CustomTukeyTransformer(target_column='person_emp_exp', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

8. Outlier Treatment for Loan Interest Rate (`loan_int_rate`)

Transformer: CustomTukeyTransformer(target_column='loan_int_rate', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

9. Outlier Treatment for Loan Amount (`loan_amnt`)

Transformer: CustomTukeyTransformer(target_column='loan_amnt', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

10. Outlier Treatment for Loan Percentage of Applicant Income (`loan_percent_income`)

Transformer: CustomTukeyTransformer(target_column='loan_percent_income', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

11. Outlier Treatment for Credit History Length (`cb_person_cred_hist_length`)

Transformer: CustomTukeyTransformer(target_column='cb_person_cred_hist_length', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

12. Outlier Treatment for Credit Score (`credit_score`)

Transformer: CustomTukeyTransformer(target_column='credit_score', fence='outer')
Design Choice: Tukey method with outer fence for identifying extreme outliers
Rationale: Outer fence preserves most of the original distribution while handling extreme values

13. Age Scaling (`scale_age`)

Transformer: CustomRobustTransformer(target_column='person_age')
Design Choice: Robust scaling for Age feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for Age which may not follow normal distribution

14. Income Scaling (`person_income`)

Transformer: CustomRobustTransformer(target_column='person_income')
Design Choice: Robust scaling for income feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for income which typically follows a right-skewed distribution

15. Applicant Exployment Experience Scaling (`person_emp_exp`)

Transformer: CustomRobustTransformer(target_column='person_emp_exp')
Design Choice: Robust scaling for employment experience feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for employment experience which may not follow a normal distribution

16. Loan Interest Rate Scaling (`loan_int_rate`)

Transformer: CustomRobustTransformer(target_column='loan_int_rate')
Design Choice: Robust scaling for loan interest rate feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for loan interest rate which may not follow a normal distribution

17. Loan Amount Scaling (`loan_amount`)

Transformer: CustomRobustTransformer(target_column='loan_amount')
Design Choice: Robust scaling for loan amount feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for loan amount which may not follow a normal distribution

18. Loan Percentage of Applicant Income Scaling (`loan_percent_income`)

Transformer: CustomRobustTransformer(target_column='loan_percent_income')
Design Choice: Robust scaling for loan percentage of income feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for loan percentage of income which may not follow a normal distribution

19. Credit History Length Scaling (`cb_person_cred_hist_length`)

Transformer: CustomRobustTransformer(target_column='cb_person_cred_hist_length')
Design Choice: Robust scaling for credit history length feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for credit history length which may not follow a normal distribution

20. Credit History Length Scaling (`credit_score`)

Transformer: CustomRobustTransformer(target_column='credit_score')
Design Choice: Robust scaling for credit score feature
Rationale:
Robust to outliers compared to standard scaling
Uses median and interquartile range instead of mean and standard deviation
Appropriate for credit score which may not follow a normal distribution

Ensemble:

LGB Threshold Table

threshold	precision	recall	f1	auc	accuracy
0.00	0.22	1.00	0.36	0.97	0.22
0.05	0.43	1.00	0.60	0.97	0.71
0.10	0.46	1.00	0.63	0.97	0.74
0.15	0.48	1.00	0.65	0.97	0.76
0.20	0.50	1.00	0.67	0.97	0.78
0.25	0.53	1.00	0.69	0.97	0.80
0.30	0.56	0.98	0.71	0.97	0.82
0.35	0.59	0.97	0.73	0.97	0.84
0.40	0.63	0.95	0.76	0.97	0.86
0.45	0.66	0.94	0.77	0.97	0.88
0.50	0.69	0.92	0.79	0.97	0.89
0.55	0.71	0.91	0.80	0.97	0.90
0.60	0.73	0.88	0.80	0.97	0.90
0.65	0.76	0.86	0.81	0.97	0.91
0.70	0.81	0.83	0.82	0.97	0.92
0.75	0.86	0.78	0.82	0.97	0.92
0.80	0.88	0.72	0.79	0.97	0.92
0.85	0.94	0.64	0.76	0.97	0.91
0.90	0.98	0.54	0.69	0.97	0.90
0.95	0.99	0.33	0.50	0.97	0.85
1.00	0.00	0.00	0.00	0.97	0.78

LGB eli5 Explanation

KNN Threshold Table

threshold	precision	recall	f1	auc	accuracy
0.00	0.22	1.00	0.36	0.94	0.22
0.05	0.36	1.00	0.53	0.94	0.60
0.10	0.44	0.98	0.61	0.94	0.72
0.15	0.50	0.96	0.66	0.94	0.78
0.20	0.58	0.92	0.71	0.94	0.83
0.25	0.64	0.88	0.74	0.94	0.86
0.30	0.69	0.83	0.75	0.94	0.88
0.35	0.73	0.78	0.76	0.94	0.89
0.40	0.77	0.71	0.74	0.94	0.89
0.45	0.81	0.64	0.72	0.94	0.89
0.50	0.85	0.57	0.68	0.94	0.88
0.55	0.86	0.48	0.62	0.94	0.87
0.60	0.90	0.40	0.55	0.94	0.86
0.65	0.94	0.34	0.50	0.94	0.85
0.70	0.95	0.26	0.40	0.94	0.83
0.75	0.93	0.19	0.32	0.94	0.82
0.80	0.96	0.11	0.20	0.94	0.80
0.85	1.00	0.05	0.10	0.94	0.79
0.90	1.00	0.02	0.04	0.94	0.78
0.95	0.00	0.00	0.00	0.94	0.78
1.00	0.00	0.00	0.00	0.94	0.78

KNN Shap Explanation

Logistic Regression Threshold Table

threshold	precision	recall	f1	auc	accuracy
0.00	0.22	1.00	0.36	0.96	0.22
0.05	0.48	1.00	0.65	0.96	0.76
0.10	0.53	0.99	0.69	0.96	0.80
0.15	0.56	0.97	0.71	0.96	0.83
0.20	0.60	0.94	0.73	0.96	0.84
0.25	0.64	0.90	0.75	0.96	0.87
0.30	0.67	0.86	0.75	0.96	0.88
0.35	0.70	0.86	0.77	0.96	0.89
0.40	0.73	0.82	0.77	0.96	0.89
0.45	0.77	0.80	0.78	0.96	0.90
0.50	0.80	0.75	0.77	0.96	0.90
0.55	0.82	0.72	0.76	0.96	0.90
0.60	0.84	0.64	0.73	0.96	0.89
0.65	0.87	0.59	0.70	0.96	0.89
0.70	0.89	0.48	0.63	0.96	0.87
0.75	0.92	0.41	0.57	0.96	0.86
0.80	0.94	0.32	0.48	0.96	0.84
0.85	0.93	0.26	0.40	0.96	0.83
0.90	0.94	0.15	0.26	0.96	0.81
0.95	0.90	0.04	0.08	0.96	0.79
1.00	0.00	0.00	0.00	0.96	0.78

Logistic Regression eli5 Explanation

ANN Threshold Table

threshold	precision	recall	f1	auc	accuracy
0.00	0.22	1.00	0.36	0.96	0.22
0.05	0.32	1.00	0.49	0.96	0.54
0.10	0.50	1.00	0.66	0.96	0.78
0.15	0.55	0.98	0.71	0.96	0.82
0.20	0.58	0.96	0.73	0.96	0.84
0.25	0.62	0.92	0.74	0.96	0.86
0.30	0.68	0.90	0.77	0.96	0.88
0.35	0.71	0.86	0.78	0.96	0.89
0.40	0.74	0.83	0.79	0.96	0.90
0.45	0.78	0.80	0.79	0.96	0.91
0.50	0.81	0.77	0.79	0.96	0.91
0.55	0.82	0.72	0.77	0.96	0.90
0.60	0.84	0.64	0.73	0.96	0.89
0.65	0.88	0.59	0.70	0.96	0.89
0.70	0.90	0.51	0.65	0.96	0.88
0.75	0.94	0.39	0.55	0.96	0.86
0.80	0.95	0.32	0.48	0.96	0.85
0.85	0.96	0.22	0.35	0.96	0.82
0.90	0.93	0.06	0.11	0.96	0.79
0.95	0.00	0.00	0.00	0.96	0.78
1.00	0.00	0.00	0.00	0.96	0.78

Bank Loan Prediction

Enter Applicant Information

Results

LGB Threshold Table

LGB eli5 Explanation

KNN Threshold Table

KNN Shap Explanation

Logistic Regression Threshold Table

Logistic Regression eli5 Explanation

ANN Threshold Table

ANN Shap Explanation

Bank Loan Prediction

Enter Applicant Information

Results

Loan Application Data Pipeline Documentation

Pipeline Overview

Step-by-Step Design Choices

1. Gender Mapping (person_gender)

2. Education Mapping (person_education)

3. Target Encoding for Home Ownership Column (person_home_ownership)

4. Target Encoding for Loan Intent Column (loan_intent)

5. Previous Loan Defaults on File Mapping (previous_loan_defaults_on_file)

6. Outlier Treatment for Age (person_age)

6. Outlier Treatment for Income (person_income)

7. Outlier Treatment for Employment Experience (person_emp_exp)

8. Outlier Treatment for Loan Interest Rate (loan_int_rate)

9. Outlier Treatment for Loan Amount (loan_amnt)

10. Outlier Treatment for Loan Percentage of Applicant Income (loan_percent_income)

11. Outlier Treatment for Credit History Length (cb_person_cred_hist_length)

12. Outlier Treatment for Credit Score (credit_score)

13. Age Scaling (scale_age)

14. Income Scaling (person_income)

15. Applicant Exployment Experience Scaling (person_emp_exp)

16. Loan Interest Rate Scaling (loan_int_rate)

17. Loan Amount Scaling (loan_amount)

18. Loan Percentage of Applicant Income Scaling (loan_percent_income)

19. Credit History Length Scaling (cb_person_cred_hist_length)

20. Credit History Length Scaling (credit_score)

LGB Threshold Table

LGB eli5 Explanation

KNN Threshold Table

KNN Shap Explanation

Logistic Regression Threshold Table

Logistic Regression eli5 Explanation

ANN Threshold Table

ANN Shap Explanation

1. Gender Mapping (`person_gender`)

2. Education Mapping (`person_education`)

3. Target Encoding for Home Ownership Column (`person_home_ownership`)

4. Target Encoding for Loan Intent Column (`loan_intent`)

5. Previous Loan Defaults on File Mapping (`previous_loan_defaults_on_file`)

6. Outlier Treatment for Age (`person_age`)

6. Outlier Treatment for Income (`person_income`)

7. Outlier Treatment for Employment Experience (`person_emp_exp`)

8. Outlier Treatment for Loan Interest Rate (`loan_int_rate`)

9. Outlier Treatment for Loan Amount (`loan_amnt`)

10. Outlier Treatment for Loan Percentage of Applicant Income (`loan_percent_income`)

11. Outlier Treatment for Credit History Length (`cb_person_cred_hist_length`)

12. Outlier Treatment for Credit Score (`credit_score`)

13. Age Scaling (`scale_age`)

14. Income Scaling (`person_income`)

15. Applicant Exployment Experience Scaling (`person_emp_exp`)

16. Loan Interest Rate Scaling (`loan_int_rate`)

17. Loan Amount Scaling (`loan_amount`)

18. Loan Percentage of Applicant Income Scaling (`loan_percent_income`)

19. Credit History Length Scaling (`cb_person_cred_hist_length`)

20. Credit History Length Scaling (`credit_score`)