Quick Question about Collab

Part 1 (2019)
1

Hi All,

I am trying to implement a collaborative filter predicting the outcome of a fight between Fighter 1 and Fighter 2, given previous fight outcomes.
However it seems that the algorithm is built for users and items, and not users and users.

Does fast.ai support this? Any feedback would be great!

-------------- Additional Info ----------------

I am using the following dataset : https://www.kaggle.com/mdabbert/ultimate-ufc-dataset?select=ufc-master.csv

My teammate and I have already built a tabular NN including all the columns to predict the Winner for the validation set.
While going through the lecture series I saw similarities between what the collab filter was achieving and our desired outcome, with the only difference being my dataset is user user prediction.
I have trimmed the dataset down to R_Fighter, B_Fighter and Winner for simplicity.

Mark

2

I suggest to share your data set here, so some people can look at your data set or train with it. Technically, it should work because it is a tabular data set. However, it is better to say this after you trained with it.

1 Like
3

Thanks for your reply!
I’ve added the dataset and more info to my original post.