This project explores the power of neural networks in predicting outcomes based on data. By applying simple feed-forward neural models, we strive to understand the nuances of the Loan dataset and predict decisions effectively.
The challenge was to apply neural network models to accurately predict outcomes from a given dataset. We aimed to leverage neural networks with regularization and Adam optimizer to classify data based on accuracy.
- Implemented feed-forward neural network models with varying hidden layers.
- Trained using binary cross-entropy and mean squared error loss with Adam optimizer.
- Evaluated models' performance using accuracy, loss, ROC curve, and confusion matrix.
Data analysis was conducted to understand feature correlations and distribution within the Loan dataset, influencing the neural network's decision-making process.
Performance was assessed through various metrics, with models trained on the Loan dataset achieving accuracies of 79.06% and 80.23%. Visualizations provided insights into model behavior and performance. Our visualizations, such as the ROC curve, confusion matrix, and decision boundary plots, provided valuable insights into the models' behavior and performance
- Loan Dataset: Models achieved 79.06% and 80.23% accuracies, showcasing effective data classification. Visualizations provided insights into model behavior. While Loan dataset accuracy was lower, our project highlights neural network potential and the importance of architecture and optimization.
- pandas
- google-colab
- keras
- tensorflow
- seaborn
- numpy
- scikit-learn
- matplotlib
Run the Tanuj_simple_feed_loan.ipynb file to execute this project. Refer the PPT for project flow