Introduction

The objective is to evaluate the approval likelihood of a loan application based on the applicant's characteristics. To accomplish this, I will utilize multiple models, such as Logistic Regression, Random Forest, and XGBoost. By employing these models, I can assess their respective accuracy metrics and determine which one performs the best. Subsequently, I will select the model with the highest accuracy and utilize it to make predictions on loan applications. 

The data for this analysis was sourced from Kaggle.

Data Preprocessing

Categorical outcomes visualization

Create new attributes and drop some columns

Numerical outcomes visualization and removal of outliers

Turn objects into numbers

Visualize the relationship between Income and Loan Amount for those whose Loans were approved

Correlation Matrix

Divide dataset into Features and Targets

Creating the models

Hyperparameter tuning of Logistic Regression using the GridSearchCV

Predictions with the model

Confusion Matrix

Classification report

ROC Curve

Feature Importance

Conclusion

Gender, dependents, self-employment, loan amount, and residing in rural areas are factors that decrease the likelihood of loan approval. Conversely, being married, having a graduate degree, possessing a good credit history, and having a higher income are factors that increase the odds of loan approval.