Robust credit risk model that go beyond traditional credit scoring methods in banks
BSD-2-CLAUSE License
👨💻This project implements a bank credit risk prediction model using a Decision Tree classifier built with Python's scikit-learn library. It also demonstrates the deployment of this model as a web application using Flask for user interaction.
🔭 Built robust credit risk models that go beyond traditional credit scoring methods in banks. This model can analyze vast datasets of historical loan data, customer profiles, and relevant factors, leading to more accurate predictions.
⚡This project demonstrates proficiency in machine learning, data analysis, data visualization, and model building for real-world financial applications.
Performed data cleaning techniques like finding Null values, Outliers, Imputations and then performed EDA over the data in google colab using python.
In Feature engineering performed Chi square test, Anova test, VIF, Multicollinearity, Label Model seencoding, One hot encoding and Data wrangling by which some of the columns were removed which were not useful for further process of model building and finally we had 37 columns.
After performing all necessary actions created a new csv file named 'Clean_Bank_Data' which was used further in deployment of model.
To deploy this model we used pycharm platform and flask framework.
Loads the data from the CSV file "Clean_Bank_Data".
Encode categorical features using LabelEncoder from scikit-learn.
Split the data into training and testing sets (8:2).
Random Forest Algorithm, Desicion tree algorithm and logistic regression algorithms were used on training data as it was a task of classification.
Desicion Tree Algorithm gave the best results of classification report .
Evaluate the model's performance on the testing set.
Save the trained model using pickle as "model.pkl".