Browse Prior Art Database

Online advertising effects estimation using sequential clicking data

IP.com Disclosure Number: IPCOM000236000D
Publication Date: 2014-Apr-02
Document File: 4 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

Method for constructing and integrating a user's sequence of advertising click events for purchasing propensity scoring and ad effects scoring in an online advertising environment. It integrate different types of ad clicks including display ad, sponsored search ad, classified ad, and other types to construct the ad click sequence; the click action and purchase action's reward are smoothly propogated to multiple clicked ads instead of the "last-click" advertisment: i) click-to-click contribution scoring: given an ad click event, its precedding ads that have been clicked are assigned a certain score based on their effect to the occurence of the current click action; ii) click-to-purchase contribution scoring:given an purchase event, its precedding advertisings that have been clicked are assigned a certain score based on their effect to the occurence of the current purchase action. In addition, a new click metrics for scroing the effects of the clicks based on the triggering process probabilistic point process model.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 61% of the total text.

Page 01 of 4

Online advertising effects estimation using sequential clicking data

In targeted online advertising, it is very important to estimate the effect of online advertisings to the aiming metrics, such as click-thru-rate (CTR), conversion rate (CVR). Existing work use user and page feature to estimate these metrics and further integrate it into the computational advertisement system. The limitation of the existing work is they calculate the conversion rate directly using the current advertising click statistics while ignoring the effect from earlier previous clicks from other online advertising publisher .

Method 1: sequence mining (ignore time span but order sensitive)

input: time stamp ordered click sequence associated with user profile (optional)

modeling: perform frequency mining to uncover the promising click sequences that lead to final purchase scoring: generate a score for each sequence for the online user for purchasing prediction

Method 2: stochastic process modeling (time sensitive) input: time stamp associated click sequence

modeling: stochastic process e.g. cox model

scoring: generate a score for purchasing prediction

Background

- different types of adervertisings, such as search advertisement (sponsored links on the search results pages) and display advertisement (images, grahics, banners in the web)

- different online channels

Motivation

-the effects of various types of advertisments towards client's purchase/conversion shall be considered and modeled j...