Problem statement

Throughout this project we will use Collaborative filtering to predict the ratings that a user might give to a certain item based on the Amazon’s rating data set. This Project is built using the Surprise recommendation library

Dataset

About: The dataset is the actual ratings given to Amazon clothing, shoes, and jewelry category. It has a set of users and items and the ratings given by those users to some items. Our model will try to minimize the error between predicted ratings and actual ratings.

Download:

• Download the dataset from the following link and place it in the same directory of the project.
• Extract the file and make sure it has the following name:"Clothing_Shoes_and_Jewelry_5.json"
• Download the dataset from the direct link here

Project structure

Code files:

• UserDefinedAlgorithm.py: Implementation of the benchmark algorithm that is used in the project. Place it on the same project directory
• cf_recommendation_clothing_dataset.ipynb: Jupyter notebook that has the actual implementation of the project

Other files:

• proposal.pdf: The proposal that was previously accepted for this project
• report.pdf: The report of the project
• cf_recommendation_clothing_dataset.html: The HTML version of the exectuted notebook

Libraries Needed

Surprise:

• Surpise is an open source collaborative filtering library
• Install it as follows:
  • $ git clone https://github.com/NicolasHug/surprise.git $ python setup.py install
  • Important: DO NOT install using pip. The pip version is not the latest version and the project is using functionalities available in latest version only

Libraries:

• pandas
• Numpy
• matplotlib
• seaborn
• sklearn
• pickle
• Surprise (See below section of instruction on how to install it)