Skip to content

Latest commit

 

History

History
45 lines (27 loc) · 1.62 KB

README.md

File metadata and controls

45 lines (27 loc) · 1.62 KB
Raw

Content-Based Filtering

To recommend relevant items to the user, based on their previous actions (e.g., purchase) or explicit feedback (e.g., preference).

Disclaimer: this repository is based on my own research using various knowledge bases, including Google's machine learning course.

Concepts

Note: The model makes user-specific recommendation without using any information from other users.

Feature matrix

cols \ rows feature 1 feature 2 ... feature n
item 1 (0, 1) (0, 1) ... (0, 1)
item 2 (0, 1) (0, 1) ... (0, 1)
... ... ... ... ...
item m (0, 1) (0, 1) ... (0, 1)

and

cols \ rows feature 1 feature 2 ... feature n
user (only one) (0, 1) (0, 1) ... (0, 1)

Note: Here, the dot product between [the user embedding x] and [an item embedding y] gives the similarity measure.

Pros and Cons

  • Pros: (a) can make user-specific recommendations for large number of users quickly, even when other users do not share similar interests.

  • Cons: (a) requiring human expert's domain knowledge of the features, and (b) difficult to expand beyond user's existing interests (these limitations may be addressed in collaborative filtering)

Reference