Skip to content

rajesh-bhat/Capillary-ML-Hack

Repository files navigation

Capillary-ML-Hack

Problem statement

Capillary Hackathon - Recommender system for fashion retail. Develop an algorithm which will recommend best suited items from inventory to a user in order to improve his/her shopping experience.

Data description

Alt Text

Libraries used

  • implicit 0.3.8
  • pandas 0.23.4
  • scipy 1.2.0
  • tqdm 4.29.1

Final submission

User-Item matrix(R) was decomposed into lower dimensional user factors(U) and item factors(V). By randomly assigning the values in U & V, using alternating least squares iteratively U & V were computed such that we arrive closer to R = U x V.

Then, using transaction data, last purchased item was taken for each of the users and top 10 items were recommended accordingly based on item-item similarity scores.

Other approaches tried:

To be updated soon.

References

[1]https://medium.com/radon-dev/als-implicit-collaborative-filtering-5ed653ba39fe

Releases

No releases published

Packages

No packages published