What distances measures perform well on content-based Recommendation Systems?

Question

I want to implement a content based recommendation system that provides a list of recommended books based on user input.

I`ll be using TF-IDF to determine how important a word is to a given book and will create a Book Characteristic Vector for every Book.

I need to create a similarity matrix to determine to determine possible pair of books. I came across Euclidean Distance for doing that. Any other methods better than Euclidean?

Hey John! Just saw this question posted, actually I am working on something like this. Were you able to complete this task? — rampuriyaaa, May 19 '15 at 07:12
Yes, I completed it, but the results were about 40-50% accurate. — John Constantine, May 20 '15 at 02:16
Ok! Is there any live site where you have implemented this.If it is can you please sent the link? — rampuriyaaa, May 25 '15 at 11:06

neuromouse · Answer 1 · 2016-03-12T17:55:57.347

1

These are some good distance measures you might try:

(generalized) Jaccard distance
Manhattan distance
Hellinger distance
cosine similarity

edited Mar 12 '16 at 17:55

answered Mar 11 '16 at 20:10

neuromouse

921
1
12
32

What distances measures perform well on content-based Recommendation Systems?

1 Answers1