A simple recommender system based on Collaborative Filtering for implicit feedback dataset that accepts a user id as an input and return top 10 recommended items for that user
- First, clone this repo
git clone https://github.com/cruisybd/implicit-recommender-system/ - Install all the requirments,
pip install -r requirements.txt(you can do this in a virtual environment, just ensure you install the requirements after you have activated the virtual environment)
Please make sure you set up your environment correctly and install the requirements accordingly. Then, you can start using the app:
# on your terminal
cd ~/implicit-recommender-system
python item_recommender.py
# the program is building the recommender engine and once finished, will ask you to specify the user id you would like recommendations for
# if the user id you specify does not exist in the data, it will throw a warning and ask for another one
# once you are finished using it, exit the program via ctrl + c
The file data.csv is a comma-separated file containing a random sample of user-item interactions. Both users and items are recorded with a unique id.
A series of exploratory analysis was conducted prior to building the model. Please see this notebook
An explanation of the model, assumptions and design decisions are outlined in this notebook
The testing method is described in this notebook and the code used is test.py