Recommendation-System

An online clothing store aims to increase the average bill per customer, we proposed the solution as a recommender system which suggests a set of related products for customers, we tested two models and prepared a demo using Flask.

Data:

We used the data of an online clothing store https://www.kaggle.com/ruichenyang/ecommerce-clothing-store

Models:

We tested two models :

• Baseline: Popularity based recommendation (recommend top purchased items)
• Collaborative Filtering: Singular Value Decomposition (SVD) based Recommendation

Code Structure:

• data_exploration.ipynb : a jupyter notebook to explore data and make some statistics
• generate_popularity_model.py : where we create popularity model, popularity_recommender.py is used inside it
• collaborative folder : which contains all notebooks and files to generate collaborative model
• models folder : which contains the two saved generated models (i.e collaborative + popularity)
• static folder: which contains the css of the demo pages
• templates folder : it contains our html pages
• data folder : it contains the link of our data, in addition to two csv files (orders_items + products' specification)
• demo_screenshots folder: contains screenshots from demo

Environment :

We used google colab for data exploration,as for model building python 3.7.5, the backend was written with Flask (Python) and the frontend using Bootstrap.

To test the program just run this:

python main.py

Demo Sample

After hitting submit, it shows the recommendations according to the selected model There is also an error page, in case of undefined product