Prediction of User's Web-Browsing Behavior: Application of Markov Model

Area of Science:

• Computer Science
• Artificial Intelligence
• Data Mining

Background:

• Web prediction, a classification task, aims to forecast user navigation patterns based on historical data.
• Existing methods face a trade-off between model complexity and prediction accuracy.
• Scalability issues arise with traditional Markov models due to the increasing number of possible navigation paths.

Purpose of the Study:

• To analyze and compare Markov models and all-Kth Markov models for web prediction.
• To propose a modified Markov model addressing scalability challenges.
• To introduce a two-tier prediction framework for enhanced efficiency.

Main Methods:

• Analysis and comparison of Markov models and all-Kth Markov models.
• Development of a modified Markov model to reduce path complexity.
• Implementation of a two-tier prediction framework utilizing example classifiers.
• Evaluation using standard benchmark datasets and association rule mining.

Main Results:

• The modified Markov model effectively reduces the number of paths without compromising prediction accuracy.
• The two-tier framework improves prediction time significantly.
• Higher orders of the all-Kth Markov model correlate with improved prediction accuracy.
• Experimental results validate the proposed techniques.

Conclusions:

• The modified Markov model offers a scalable solution for web prediction.
• The two-tier framework provides a practical approach to accelerate web prediction.
• The study highlights the benefits of higher-order models for prediction accuracy.