|
Activity Number:
|
436
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Wednesday, August 5, 2009 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Statistical Learning and Data Mining
|
| Abstract - #304688 |
|
Title:
|
Session-Centric Page Sequence Clustering for Improving Web Experience
|
|
Author(s):
|
Lakshminarayan K. Choudur*+
|
|
Companies:
|
Hewlett-Packard Laboratories
|
|
Address:
|
14231 Tandem Boulevard, Austin, TX, 78728,
|
|
Keywords:
|
Markov chains ; web mining ; customer satisfaction ; classification ; clustering ; data mining
|
|
Abstract:
|
The Website is an important medium for e-commerce sales and support. Companies spend millions on their design and operation and gather data about every aspect of customer visits. Unfortunately the data has not translated into sales effectiveness or quality of customer experience. Managers rely on ambiguous statistics from web logs or qualitative usability tests using small samples. We propose quantitative techniques to track website performance relative to sales and customer difficulties. Unlike web logs, we analyze complex behavior rather than statistics like page views, purchases, etc. Unlike qualitative techniques we leverage data from on-line customers. Our technique using Markov chains incorporates survey-satisfaction to train a classifier that predicts customer satisfaction based on sequences of customer page views. We deliver ~80% accuracy.
|
