This project explores supervised machine learning techniques to predict whether an individual will be approved for a loan, based on key financial and behavioral features.
- Michael Olson – @michaelolson01
- Bee Xiong - @Bee-Xiong
- Nick Sigwanz - @NickSigwanz
- John Quinlan - No known github page.
Project completed as part of ICS 352 – Artificial Intelligence at Metropolitan State University.
The task is to estimate loan approval outcomes using:
- Annual income
- Number of credit inquiries
- Average monthly balance in a checking account
- Outstanding debt
We applied and compared the following ensemble classifiers:
- Random Forest
- AdaBoost
- Gradient Boosting
- Python
- NumPy
- Pandas
- Matplotlib
- Scikit-learn
- Google Colab
-
Data Preprocessing
- Cleaned and transformed financial data
- Normalized features and handled missing values
-
Exploratory Data Analysis
- Visualized distributions, feature correlations, and class balance
-
Model Training & Evaluation
- Trained multiple ensemble classifiers using
train_test_split - Performed hyperparameter tuning with
GridSearchCV - Evaluated using classification reports and cross-validation
- Trained multiple ensemble classifiers using
Each model was assessed using:
- Accuracy
- Precision, Recall, and F1-Score
- Confusion Matrix
The models achieved solid performance, with ensemble methods showing strong generalization across the test data.
This project was completed as the final assignment for the Artificial Intelligence course in the M.S. Software Engineering program at Metropolitan State University. It demonstrates practical machine learning skills applied to a realistic financial decision-making problem.
- Incorporate SHAP or feature importance visualizations
- Deploy as a Streamlit app or API
- Expand dataset and explore fairness metrics
ICS-352-AI-Final-Project.ipynb– Main analysis notebookREADME.md– Project documentation