Precision and Recall for User Based Recommender System using Binary Data

[Shady Hanna <sh...@gmail.com>]

Hi everyone ,

I am trying to develop a user based recommender system using binary data.
The data has the user ID and product id which the user has bought and the
preference is always 1 since I don't have ratings in my dataset. If the
user did not buy an item, it is not included in the dataset.

I am using in Mahout:

   1. Boolean Data Model
   2. LogLikelihood Similarity
   3. ThresholdUserNeighborhood
   4. GenericBooleanPrefUserBasedRecommender
   5. IRS evaluator

I created a loop to try all the values of the threshold between 0.0 and
1.0, and got all the results of the precision and recall and I got really
strange results. In this link there is the csv file with the threshold as
first column and the precision and recall:

http://www.filedropper.com/loglikelihood-threshold

Can someone please tell me what these results mean, am I doing something
wrong ?

Thank you so much for your support,
Best Regards,
Shady