how to combine two similarity instances in Apache mahout

Question

I am new to Apache mahout and i have question in mind, please correct me if i am wrong.
Lets assume we have dataset and calculated recommendations from dataset and than displayed the results to users and when users come to visit our site again we will have new dataset ,so we have to again calculate recommendation according to new dataset.
we can do this by either combining the two datasets i.e the old dataset and new dataset but since we have done computations on old dataset,if we combine them than we have do same computation again on old dataset and this will become problem when data increases so wanted to ask is there any other way to do this and the other way I thought was to combine the similarities of two dataset because recommendation is based on similarities but i am not able to find anything on this.

Please Help me how to approach this problem,

Thanks

Answer 1

You want to calculate recommendations on a moving window of data. Decide your window and update frequencies. If you update daily and your window is 3 months, then you discard the oldest days worth of data every day and use the 3 months of most recent data every time you update.