Browse Prior Art Database

A System and Method for Recommending Multiple Offerings based on Detection of Inter-related Events Disclosure Number: IPCOM000238429D
Publication Date: 2014-Aug-26
Document File: 5 page(s) / 420K

Publishing Venue

The Prior Art Database


Methods are introduced to detect a set of inter-related events and take a set of coordinated actions. This is an important problem in event management with wide applications. Prior work here have focused on detecting single events and on taking actions immediately based on detected events. Our work takes a holistic, long-term, approach and can help even when events may get missed or one can take actions optimizing across multiple observations.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 43% of the total text.

Page 01 of 5

A System and Method for Recommending Multiple Offerings based on Detection of Inter-related Events

In many situations, recommending actions based on detected events is required. A few common examples are online commerce and public surveillance. A simple approach to address it is that as soon as one detects an event, the system recommends an immediate action using rules or Markov Decision Processes (MPDs), which is a well-studied formalism for state-action pair learning

In a large range of situations, one wants to detect a set of inter-related events and take a set of coordinated actions. For example, the primary event of interest is hard to detect but detecting the successor may be easier; and then based on successor detection, one can infer the former and still take actions. As illustration, pregnancy leads to childbirth but one may miss it. Once we know about childbirth, we can still make recommendation for mother health. Another illustration is job offer leading to house purchase. We can detect house purchase and make recommendation for personal loans.

More generally, the detection of an event suggests follow-up events for which one wants to take more profitable actions. We describe methods that deal with detection of inter-related events and taking multiple actions based on them.

Prior Art


Series of Dynamic Targeted Recommendations, ECWeb 2002

     Summary: talks about a series of recommendation to users as they visit a website in a session
Gap: event trail is limited to web-site navigation and one session; in our case, event trails can be anytime - e.g., across different sessions, episodes, epochs Personalized tag recommendation using graph-based ranking on multi-type interrelated objects, SIGIR 2009 ( )

Summary: does tag recommendation for web sites, the latter are inter-related. Gap: event trail is not covered

A method and a system for generating dynamic recommendations in a distributed networking systemWO 2013136308 A1

     Summary: Builds a dynamic profile of users based on social actiivity and recommends using it
Relevance: irrelevant to subject matter
A system and method for dynamic member segmentation and targetingCA 2779971 A1

Summary: Describes targeting users based of segmentation rules

Relevance: irrelevant

The problems with these are that they

Ignore relationships between events and thus can detect fewer situations to take actions Are not able to optimize decisions
Do not provide enhanced flexibility to a user to control system actions

Solution Details



Page 02 of 5

First we define 3 important concepts and then give our solution.

Important Concepts

a) Event trail - Pre-known event sequences of interest, ET = {T1, T2, …, Tm} where order suggests time


Length of sequences

All sequences of same length, k

Sequences of variable lengths
Degree of overlap of events among trails
Non-overlapping events
Overlapping events
In illustrations, we will consider trail...