Name
Abstract | ||
|---|---|---|
The ability to accurately forecast browsing patterns in an e-business website can enhance E-commerce sales in three ways: converting browsers into buyers, increasing cross-sell, and building loyalty. Markov model is suitable for predicting future user requests. We propose a scalable Hybrid Order Markov Tree (HOMT) recommendation algorithm. It uses a pattern tree to compactly store all Web access sequences. Then an order-by-order incremental updating approach is used to create an all-order Markov model tree. Finally a hybrid order Markov model online prediction method is adopted for online recommendation. Experiments confirm that HOMT has high coverage, short prediction time and high prediction accuracy The offline pattern discovery phase of HOMT is scalable, better than the traditional Markov models. And as a whole, the online phase of HOMT performs much better than the previously proposed methods, including traditional Markov models. It can be widely used in pre-fetching, recommendation and Web prediction, etc. |
Year | Venue | Keywords |
|---|---|---|
2004 | IC'04: PROCEEDINGS OF THE INTERNATIONAL CONFERENCE ON INTERNET COMPUTING, VOLS 1 AND 2 | Markov,recommendation,Web usage mining |
Field | DocType | Citations |
Forward algorithm,Markov model,Computer science,Markov chain,Algorithm,Markov blanket,Hidden Markov model,Incremental decision tree | Conference | 0 |
PageRank | References | Authors |
0.34 | 1 | 3 |
Name | Order | Citations | PageRank |
|---|---|---|---|
| Ying-ji Li | 1 | 0 | 0.34 |
| Hong Peng | 2 | 14 | 10.33 |
| Qi Lun Zheng | 3 | 2 | 2.10 |