Nearest Neighbor Implementation

[Waqas Danish]

Hi,

I am planning to use Nearest Neighbor algorithm for predicting the user behavior. Basically, I have a large number of reviews along with their star ratings(1-5). For a new user that registers with my system, I want to show him the best matched results when that new user searches for an item. For example, UserA and UserB have rated three items in my system and UserC has only rated one. How can I be able to pair the users matching their interests based on the ratings they have given? I think Nearest Neighbor algorithm can do the job.

Any ideas/suggestions?
Regards,
Waqas Danish

[Rabbit]

In the absence of demographics and preference data, you can't use nearest neighbor. And it would seem you're missing that data since you didn't make any mention of it. In this case, you will have to use collaborative filtering.

[Waqas Danish]

Rabbit,

Thanks for your response. Can you please explain what do you exactly mean by demographies and preference data?

What CF algorithm should I use? SVD??

Regards,
Waqas Danish

[Rabbit]

Demographics is gender, age, job, income, educational attainment, etc. Preference data is whether or not they like fantasy, sci fi, romance, etc. SVD also requires this type of data to be collected. Item to item CF is your best bet but it's biggest weakness is it's inability to predict results with new data.