This project was developed as part of the Mathematics III course. It involves the creation of neural networks to analyze and predict specific outcomes using two different datasets. The aim was to apply theoretical concepts in a practical setting, combining data preprocessing, neural network implementation, and performance evaluation.
- Prudente Tomás
- Bompensieri Josefina
- Nicolás Cernadas
- Catalina Correa
A neural network was built to predict the potability of water using a dataset provided in a CSV file titled 'Water Potability'. The model classifies whether a given sample of water is potable based on the features provided.
Another neural network was developed to predict fraudulent transactions in a dataset titled 'Card Transactions'. The goal was to classify transactions as fraudulent or legitimate based on their characteristics.
-
Data Cleaning and Preparation
- Missing values (NaN) and outliers were handled by replacing them with the median of the respective feature to ensure stability during training.
-
Neural Network Implementation
- Models were created using both manual implementations and Python libraries such as Scikit-Learn and TensorFlow. This dual approach was used to validate the results and compare performance.
- Separate models were designed for each dataset using both libraries.
-
Documentation and Analysis
- Detailed analyses of both datasets and their neural network models are provided in the Research folder. These documents include extended explanations of the studies conducted but exclude code.
- Jupyter notebooks include all functions, implementation details, and graphical visualizations.
Thank you for reviewing this project! Feel free to explore the folders for further details about the work conducted.