Based based on a description of the item and a profile of the user’s preference

PredictionIO algorithms currently integrated do not make use of custom attributes. When content based algorithms are integrated, they will impact similarities and recommendations.

https://groups.google.com/forum/#!topic/predictionio-user/Nj2lNRAkjxs
https://groups.google.com/forum/#!topic/predictionio-user/eRPo5uaBf5A
https://groups.google.com/forum/#!topic/predictionio-user/jEbDgWWoaTc
https://groups.google.com/forum/#!topic/predictionio-user/Gip5mcSr6Ug
https://groups.google.com/forum/#!topic/predictionio-user/CiVukwb1ipo

provide algorithms to recommend items to new users and solve cold start problem

https://groups.google.com/forum/#!topic/predictionio-user/0RitqAIvx3E

Classification engine will be useful for applications like spam detection

https://groups.google.com/forum/#!topic/predictionio-user/MighHxY4HFA
https://groups.google.com/forum/#!topic/predictionio-user/yMkhdPppjaI

Most users browses products before they sign in on my site. It would be great if anonymized user logging is supported.

Useful for friend recommendations or (PYMK) "people you may know" based on user-to-user actions like "friend" and "follow".

https://groups.google.com/forum/#!topic/predictionio-user/mYZIShQ1yJs
https://groups.google.com/forum/#!topic/predictionio-user/iNumdrdqiX8

Sentiment analysis engine will be useful for to identifying and extracting subjective information from textual data.

PredictionIO comes with a few built-in actions, such as like, rate and view. It would be nice to allow developers to define custom action types.

Is there a way to return the associated rating of an item together with the item, while asking for recommendations?

Has scoring been implemented yet in results? I'm looking for results that would be something like:

{ "pio_iids":[["item2",.09876], ["item3",72635], ["item8",.0765]] }

I'm not interested in this specific format, just something that would have scoring.

https://groups.google.com/d/msg/predictionio-user/u3dy3wBGyM0/HWuQa5R_LyMJ

Trend detection engine to detect trends in time series data.

Given the number of components needed to make PredictionIO work, consider supporting a Openshift deployment either via a Cartridge or DIY approach that is explained in a Blog post

It makes it easier to monitor training and errors from scheduler log files

https://groups.google.com/forum/#!topic/predictionio-user/4Ol9fqS9MLE
https://groups.google.com/forum/#!topic/predictionio-user/3r3WIWQqtaA
https://groups.google.com/forum/#!topic/predictionio-user/_RqAPUGFb5c

SaaS/hosted version of PredictionIO

All of your algorithm are providing recommendation base on previous behaviors of user. What about an algorithm to recommend items base on user's demographic information. It's good to solve problems like recommend items between different location (city, country, etc..).

I would like to add create app, engine, set algo params, etc... to the SDK as well as PredictionIO.

https://groups.google.com/forum/#!topic/predictionio-user/Xn7jl_kuS_s

pagination at API/SDK calls for itemrec topn.

can this be used for detecting unusual behaviour - like to detect fraud

Our users are typically very very consistent in their usage patterns
- this one always uses firefox on weekday mornings from this ip
- this one always uses ie on weekday afternoons from these 2 ips
- etc

We would like to be able to detect if the info for the current login is unusual based on the previous logins, and require additional authentication if so.

https://groups.google.com/forum/#!topic/predictionio-user/Oa6a7d_dcms

Does prediction.io support sales predictions? Example: I've timeline with sales per day and I'd like to predict future sales - how many sales will be tomorrow and so. Thank's for answer.

is there a way to ad a new user to the db, without deleting everything about this user in case he already exists

https://groups.google.com/forum/#!msg/predictionio-user/cXKya1j6Co8/6VYy3tqiwN0J