lenskit.batch.predict

[Márcia Barros]

Hi, 

Would it be possible to provide an example of how to use lenskit.batch.predict? 

I'm using it like this, following the example you have for lenskit.batch.recommend:

def eval_prediction(aname, algo, train, test):


    model = algo.fit(train)


    pred = batch.predict(model, test)


    # add the algorithm
    pred['Algorithm'] = aname




    return pred

This is returning a pandas dataframe with user  item  rating  prediction Algorithm, and then I use the prediction and the rating columns to calculate measures such as RMSE and MAE. 

I also use this results to get the topk for each user and to calculate the Precision and Recall. 

However, the results for different algorithms are returning almost the same. 

Example:

itemitem algorithm

precision:  0.6782234173273409

recall:  0.7296116298361521

f-measure:  0.632498476558497

nDCG:  0.7585345378190396

RMSE  0.9011650664582945

MAE  0.7024033793043883

useruser algorithm

precision:  0.6731332794694407

recall:  0.726284079214471

f-measure:  0.6283638681052782

nDCG:  0.7548728363422198

RMSE  0.9279957720764701

MAE  0.723097540584582

Maybe I'm not using the batch predict correctly. 

Best Regards, 

Márcia Barros 

[Michael Ekstrand]

Márcia,

You are using it correctly for estimating prediction accuracy (e.g. RMSE, MAE). Depending on your data set size, those differences are entirely within what I would expect.

Recommendation, however, needs to use the batch.recommend method. The reason is that we need to score far more than the test items: we need to score and rank a larger set of candidates.

In our documentation updates for 0.6, though, we will add examples to the documentation.

--
You received this message because you are subscribed to the Google Groups "LensKit Recommender Toolkit Development and Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lenskit-recsy...@googlegroups.com.
To post to this group, send email to lenskit...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lenskit-recsys/6ef1d191-6ec6-4fa5-b7ae-8a033fc483e9%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

--

Michael D. Ekstrand — michael...@boisestate.edu — https://md.ekstrandom.net
Assistant Professor, Dept. of Computer Science, Boise State University

People and Information Research Team (PIReT) — http://coen.boisestate.edu/piret/

[Márcia Barros]

Thanks :)

I'll use batch.recommend.

Best Regards, 

Márcia