Recommender system

User 3200 | 2/11/2016, 11:54:26 AM

Hello Everyone , I'm trying to do a recommender system with gui, Once a customer(person) logs in he should get a recommendation. Can any one help me . :(


User 4 | 2/15/2016, 10:54:26 PM

Hi @Venki, can you elaborate on what you are trying to do and which part you are struggling with?

Generally speaking you'll want to have an application that consumes recommendations (makes queries to a hosted recommender model over the internet) and has application logic to present those recommendations to the user. Within your application code, you can know which user is currently logged in, and thus which user ID to query the recommender with.

To get started building recommenders with GraphLab Create, take a look at this video or the gallery items related to recommenders. Many of these examples could get you up and running with a Recommender model that can be queried for a user.

Once you have a recommender, to integrate it with an application you'll probably want to deploy it to a Predictive Service. See the Predictive Services chapter of the userguide for more info on this; but the basic idea is that you are taking the trained Recommender (a Python object) and putting on a reliable, fault-tolerant hosted service, so that it can be queried from a production application. The hosted model is queryable using a REST-like interface over HTTP, so any application that can make HTTP requests should be able to consume it.

As for how to build the GUI portion, it would be specific to the application you are integrating the recommender into. Many popular choices include web apps (dynamic HTML with JavaScript), mobile apps (often Objective-C for iOS, Java for Android, or potentially another language choice, paired with the mobile OS APIs), and desktop apps (often C++ with a UI toolkit like QT, but more often lately JavaScript and HTML wrapped in something like Electron ).

User 3200 | 2/18/2016, 6:27:12 AM

Actually i wanted to do a desktop app, for my college mini project , like i have a movie data set and i need to recommend a new movie to the user. with GUI . and dint know how to integrate graphlab with UI.

User 4 | 2/18/2016, 9:31:15 AM

@Venki No problem! Does the userguide chapter on Predictive Services help? I think that would be the most straightforward approach to integrate with a desktop app: the app can make HTTP requests to the Predictive Service, and then use the responses to determine what to show to the user.

User 3200 | 3/1/2016, 10:39:51 AM

Ya thanks Zach