Spam Email Classifier – Neural Network in Java
Bernette Chan
Built a spam email classifier from scratch in Java as part of my Machine Learning course. The project involved implementing a feedforward neural network to distinguish spam from non-spam messages, with a focus on understanding the core mechanics of neural networks.
I contributed to parsing the dataset into features and targets, coding key components of the backpropagation function, and implementing a hyperparameter tuning class
to optimize learning rate and hidden layer size. The network was built using an object-oriented approach, with Node classes representing individual neurons.
Our final model achieved 74% precision, 60% recall, and a 67% F1 score on the test set, demonstrating solid performance for a custom-built model without external libraries.