Anime-Recommendations-Project

Introduction

I was an avid anime fan and this lead me to pursue this project of creating a recommendation system for animes a user may enjoy based on their enjoyment of similar animes and similar people. This made me try to create both a content based filtering system and a collaborative filtering system to recommend animes to the user. What I would like to do is try combine both systems into one more effective system which eventually could be deployed as a website. The repo containing the deployment website files can be found here.

What I learnt:

• How to replace missing or sub-optimal values in a dataset
• Creating pairplot using seaborn to compare numerical data and distributions
• Use pandas and collections' counter to analyse frequency of values of multiple elements in strings
• Create content based recommendation system using the genres of a show and a user's ratings of watched shows
• Create a collaborative filtering recommendation system using scikit surprise and SVD

Included Files:

Notebooks

• Data Cleaning.ipynb (Notebook for cleaning the anime and ratings datasets that will be used)
• Exploratory Data Analysis.ipynb (Exploring the data and trying to extract insights and patterns before building the recommendation system)
• Content Based Filtering.ipynb (Notebook containing the code for the content based recommender)
• Collaborative Filtering.ipynb (Notebook containing the code for the collaborative filtering recommender)

Datasets

• anime.csv (Original anime dataset)
• rating.csv (Original user and rating data)
• cleaned_anime.csv (The cleaned and processed anime dataset after data cleaning)
• cleaned_rating.csv (The cleaned and processed rating dataset after data cleaning)

Resources and Datasets Used

Python Version: 3.8

Packages:

• pandas 1.13
• numpy 1.19.2
• matplotlib 3.3.2
• seaborn 0.11.0
• scikit surprise 1.1.1

Anime and Rating Datasets

For this project, I originally used an anime dataset by CooperUnion for data cleaning, data exploration, and model building. However, as I was developing the deployment website, I though maybe I should use a more up to data dataset. This led me to use Marlesson's data which will be use for deployment purposes.

• Original dataset: https://www.kaggle.com/CooperUnion/anime-recommendations-database.
• Deployment dataset: https://www.kaggle.com/marlesson/myanimelist-dataset-animes-profiles-reviews