Recommendation Engine

This repo demonstrate a ecommerce recommendation engine which recommend items for the user based on their preference.

Here are the different notebooks:

• Data Processing: Loading and processing the data to prepare for input into model.
• Model Creation: This file is responsible for the creation of deep leaning model. This file use implicit collaborative filtering which is a Deep Learning approach to predict the items.
• Making Predictions: Making predictions of items user may like based on user_id and also item_id user purchased.

Data Source

This is a transnational data set which contains all the transactions occurring between 01/12/2010 and 09/12/2011 for a UK-based and registered non-store online retail.

https://www.kaggle.com/jihyeseo/online-retail-data-set-from-uci-ml-repo

Rest Api

Backend for web based recommendation system is also available here:

• Rest Api: I also created a rest api for the recommender using django and django-restframework which takes user_id as input and predict items user may also like.

The main contents of this project are following:

• Implementation of implicit collaborative filtering using LightFM framework.
• Different ways of recommendating items to uers(user-user or item-item)
• Implementation of demo program of the recommender system

Installing

I created a virtual environment in my machine, and run the code. To run the python code first create a virtual environment and install all dependencies by run the command:

$: pip install -r requirements.txt

and run jupyter nootebook

$:  jupyter nootebook

To make predictions run lightFM_retail_recom_sys_PREDICTIONS.ipynb

Requirements

• Python 3.6
• Jupyter Notebook

Dependencies

Choose the latest versions of any of the dependencies below:

• lightfm
• pandas
• numpy
• scipy
• sklearn