Since the project is quite large, it may not load the first time. If there is an error, refresh the page

This project focuses on the study of an anime dataset and building a recommendation system based on collaborative filtering (Collaborative Filtering). The main goal is to analyze data, identify patterns, and provide users with personalized recommendations for anime content.

The dataset contains over 109 million rows, providing extensive insights into user preferences and behaviors.

During this stage, we conducted the following steps:

• Initial overview of the structure and types of data.
• Detection and handling of missing values.
• Identification of anomalies and outliers.
• Examination of distributions of key variables (such as ratings, genres, number of episodes).
• Creation of visualizations to understand relationships between parameters.

To build recommendations, we applied the technique of Collaborative Filtering. This method takes into account interactions between users and content. It considers the preferences of other users whose tastes are similar to those of the current user and suggests new anime that aligns with these preferences.

The project represents a comprehensive research effort aimed at enhancing user interaction with anime platforms by offering personalized recommendations.

The notebook with all stages of the project can be found here

• Датасет был взят с соревнования на Kaggle
• Python 11.x
• library versions
• download model

Anime_Recommedation_System/
│
├── data                  
│   └── *.csv                      
├── DS_fn.ipynb           
├── .gitignore              
├── README.md               
├── requirements.txt         
└── image.png   

anime_model.h5 - pre-trained model.
In order to load correctly, you need to execute the following code:

import keras
import tensorflow as tf
import tensorflow_text
import tensorflow_hub as hub
import pickle

# Путь к файлу с моделью
model_path = 'YOUR_PATH/anime_model.h5'

# Загрузка модели из файла
model = keras.models.load_model(
    model_path,
    custom_objects=dict(KerasLayer=hub.KerasLayer)
)

• Solution
• Libraries
• Kaggle
• download model

The anime recommendation system project has been successfully implemented and is ready for use by users. The system helps users find new and interesting anime series based on their preferences and interests. We value each piece of feedback and take user input into account when improving the quality of recommendations.