Run the API Server using the following commands.
First, activate the virtual environment using,
.\venv\Scripts\activate
Then, install necessary modules.
pip install -r .\requirements.txt
Then run the server using,
flask run
This server has one API route, /predict. This route
is for getting recommendations based on the given book
title. The names must be from the dataset goodbooks-10k.
Working on extending it to a custom dataset, and running a check to see if the name is within the database before running the recommender.
Please check requirements.txt for a full list of
dependencies. Of course, you need Python 3.11 to run
the server. I tested on Python 3.11.8
This model uses the goodbooks-10k dataset. It gives recommendations using the cosine similarity measure metric.
The parameters used in this model are
- Genre
- Title
- Description
Feature extraction is done using multi label binarization (for genre), and vectorization with a bag-of-words approach (title and description)
-
Create Content-Based Filtering Model
-
Create a Collaborative filtering model using data gathered from the user data in the goodbooks-10k dataset, our own generated data, and data from the Sri Lankan National Bibliography.
-
Combine the Collaborative Filtering model and the Content- Based filtering model, to create a hybrid recommendation model.
-
Improve the model using Matrix Factorization and Neural Net techniques.