Prediction of Anime Rating with Hybrid Artificial Neural Networks and Convolutional Neural Networks

Al Kautsar, M. Nurudduja
Anggraini, Violita
Basirun, Arif Reza
Rifai, Achmad Pratama

SITEKIN: Jurnal Sains, Teknologi dan Industri
Vol 22, No 1 (2024): December 2024

This study proposes an innovative approach to predict anime scores by leveraging a combination of Artificial Neural Networks (ANN) and Convolutional Neural Networks (CNN). Tabular data such as source, number of episodes, type, and genre are incorporated alongside the image representation of anime into a holistic model. Evaluation results on the test set show satisfactory performance, with an average loss value of 0.673, Mean Absolute Error (MAE) of 0.654, and Mean Absolute Percentage Error (MAPE) of 9.44%. Training and validation graphs reflect the model's convergence without significant signs of overfitting or underfitting. The integration of information from both data sources yields a model capable of providing accurate predictions of anime scores, contributing to an understanding of trends and preferences in the anime industry, and opening opportunities for the development of similar models in the field of score prediction or other quality evaluations.