Matrix and Tensor Approximation for Internet Latency Prediction and Online Purchase Prediction

• Author / Creator

  Liu, Bang

• In this thesis, I study the application of matrix and tensor approximation techniques to Internet latency prediction and purchase prediction in e-commerce. Traditional approaches for latency prediction suffer from the limitation of the Euclidean assumptions such as triangle inequality and symmetric latencies. I propose a new scheme that can decompose an incomplete latency matrix into a distance component and a network feature component, and apply low-rank matrix completion to complete the network feature matrix based on the fact that network conditions are correlated. In the second problem, I analyze the user behavior records collected from Alibaba group's mobile B2C platform and propose a novel model, named Multifaceted Factorizing Personalized Markov Chains, to jointly factorize the tensor of purchase records into hidden factors for various context information. Extensive evaluations based on real-world datasets show that our proposed approaches outperform traditional approaches in both problems.

• Subjects / Keywords

  • Matrix Factorization
  • Network Latency Prediction
  • Online Purchase Prediction

• Graduation date

  Fall 2015

• Type of Item

  Thesis

• Degree

  Master of Science

• DOI

  https://doi.org/10.7939/R3CC0V463

• License

  This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.