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

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

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

A demand forecasting template based on prices and sales history would be really nice for retailers. Once the model is trained, they could adapt their prices and discounts in order to maximise their profit.

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.

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

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.

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

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.

ElasticSearch is gaining momentum in use-cases that aren't strictly search. Also includes geospatial support as well.

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

SaaS/hosted version of PredictionIO

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

It could be great to enable the CORS request ; configure the domains and the methods allowed

pagination at API/SDK calls for itemrec topn.

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..).

Add the ability to identify anomalies within a set of time series data.