Titanic Survival Neural Network

Using Kaggle’s classic Titanic dataset I built a lightweight neural-network classifier in Python/Keras to predict passenger survival. After a few rounds of feature engineering (fare binning, one-hot encoding for cabins and titles) and dropout tuning, the model steadily improved to ~78% accuracy on the public leaderboard.

The notebook (linked above) walks through the full workflow—data cleaning, exploratory analysis, model architecture, and evaluation. I keep iterating on this project to compare modern architectures (TabNet, gradient boosting) against the baseline dense network.