Personalized recommendation for e-commerce

Hey foxes

For e-commerce usually they get click stream data which describe the customer clicks
“X person clicked on the y brand”
customer_id, event_type, brand_id
" X person clicked on Y product"
customer_id, event_type, product_id
mostly these data represented in firebase

in case we want to build a recommendation system in such scenario when the customer click on specific brand we send the customer id and brand id to the recommender ai personalized system , then this system send back the list of the brand items ranked

Any idea what could be the flow?

“”
Am sharing mine and i want to hear from the professionals and please suggest a project or article talk about the same
1- will embed the clicks data to vector database
2- when the customer click will embed the inputs too
3- using the vector search to query the vector database
4- then query LLM along with pre defined prompt
5- send the response to the frontend which should be list of items ranked
“”"

Hi Fahd,

I studied a similar case in my master’s program. I recommend that you start with a basic model, for example:

  • Collaborative Filtering:
    • User-Based Collaborative Filtering : Recommendations are made based on the similarity between users. If two users have similar tastes, recommendations for one will be based on the preferences of the other.
    • Item-Based Collaborative Filtering : Recommendations are made based on the similarity between items. If a user has shown interest in certain items, similar items will be recommended.

If you need more ideas about “Recommender Systems,” you can search using ChatGPT. I think an LLM can help, but there are other tools specialized in your use case.

Regarding your question, I think you can use the following flow:

  • Data Collection : Collect user interaction data (e.g., clicks, purchases, ratings) and store it in a database or data warehouse.
  • Data Preprocessing : Clean and preprocess the data using ETL (Extract, Transform, Load) processes.
  • Model Training : Periodically train or update the recommendation model using historical data.
  • Model Serving : Deploy the trained model to serve real-time recommendations.
  • Recommendation Delivery : Provide recommendations to users through an API or other interfaces.

This was supportive and helpful thanks