• The data consists of different variables that may impact the credit score of different applicants.

• The machine learning model is built up by 2 scenarios (duration <= 12 months or > 12 months)

• The preliminary analysis is first conducted to examine outlier and overall data distribution via boxplot. Heatmap is also used to roughly examine the correlation between each - variable and the credit score result.

• The scorecardpy package especially woe (weight of evidence) is used to find out the impact level of different variables to result.

• Different model training algorithms such as linear regression and logistic regression are used.

• The performance is evaluated by KS (Kolmogorov–Smirnov) and AUC (Area under the ROC Curve) values.