This project explores advanced machine learning techniques by building a recommendation system using factorization machines to predict user preferences. The goal was to create a robust system capable of learning intricate relationships between users, movies, and metadata.
- Utilized a dataset containing user ratings, movie metadata (e.g., genres, runtime, year), and categorical features (e.g., movie type: tvSeries, short, video) to support comprehensive learning.
- Captured explicit features (e.g., genres like Action, Drama) and implicit interactions (e.g., user preferences for specific time periods or genres).
- Implemented a factorization machine with TensorFlow, fine-tuning hyperparameters to ensure efficient feature interaction learning.
- Achieved high predictive accuracy by validating results with metrics like RMSE and precision at K.
This project highlights my ability to address real-world machine learning challenges, from data preprocessing to model development and evaluation. It demonstrates my proficiency in Python, TensorFlow, and Scikit-learn, as well as my understanding of recommendation systems and collaborative filtering techniques.
Through this project, I gained hands-on experience tackling complex ML problems, aligning with applications in cutting-edge machine learning solutions.