OtakuNet is an Content-Based Recommendation System App that recommends animes to users based on their personal ratings of genres.
Using only NumPy for faster linear algebra computations and Pandas for data organization, OtakuNet is trained on a neural network framework built from scratch. The model is trained using a dual neural network architecture to train both user and anime data.
- Some feature engineering techniques, such as Bayesian Rating, weighted average, and z-score normalization applied to the data for training.
- Implements only NumPy to create a neural network framework from scratch, specifically tuned towards training a dual neural network architecture to take in anime and user data to make predictions. It first independently trains the user and anime as two independent neural networks, and then combines those new learned features together to make a final prediction.
- Based on data of the features/genres of the animes and a new user input, the model uses this to compute a prediction of a recommended anime based on the user's genre ratings.
- Final trained parameters stored in pickle files to be accessed by the GUI app for predictions.
- Uses
customtkinterframework to create the GUI elements - Includes three screens: a Home screen, Prompt screen, and Recommend screen
- Prompt screen asks user for input values for their genre ratings.
- Recommend screen uses these inputs to compute a prediction, sort the highest predictions, and displays the recommended animes with their names, genres, and images.
Clone this repository:
https://github.com/kseto06/OtakuNet
For model functionalities:
pip install numpy
pip install pandas
pip install pickle
Then, navigate to the gui folder from the parent (OtakuNet) directory:
cd gui
To run the app:
- Install GUI dependencies
pip install -r requirements.txt
- Activate the GUI environment:
source env/bin/activate
- Run the app:
python app.py