SEQUENTIAL PATTERN ANALYSIS IN DYNAMIC BUSINESS ENVIRONMENTS

Abstract

OF THE DISSERTATION Sequential Pattern Analysis in Dynamic Business Environments By CHUANREN LIU Dissertation Advisor: Dr. Hui Xiong Sequential pattern analysis targets on finding statistically relevant temporal structures where the values are delivered in sequences. This is a fundamental problem in data mining with diversified applications in many science and business fields, such as multimedia analysis (motion gesture/video sequence recognition), marketing analytics (buying path identification), and financial modelling (trend of stock prices). Given the overwhelming scale and the dynamic nature of the sequential data, new techniques for sequential pattern analysis are required to derive competitive advantages and unlock the power of the big data. In this dissertation, we develop novel approaches for sequential pattern analysis with applications in dynamic business environments. Our major contribution is to identify the right granularity for sequential pattern analysis. We first show that the right pattern granularity for sequential pattern mining is often unclear due to the so-called “curse of cardinality”, which corresponds to a variety of difficulties in mining sequential patterns from massive data represented by a huge set of symbolic features. Therefore, pattern mining with the original features may provide few clues on interesting temporal dynamics. To address this challenge, our approach, temporal skeletonization, reduces the representation of the sequential data by uncovering significant, hidden temporal structures. Furthermore, the right granularity is also critical for sequential pattern modelling. Particularly, ii there are often multiple granularity levels accessible for estimating statistical models with the sequential data. However, on one hand, the patterns at the lowest level may be too complicated for the models to produce application-enabling results; and on the other hand, the patterns at the highest level may be as trivial as common sense, which are already known without analyzing the data. To dig out the most value from the data, we propose to construct the modelling granularity in a data-driven manner balancing between the above two extremes. By identifying the right pattern granularity for both sequential pattern mining and modelling, we have successful applications in B2B (Business-to-Business) marketing analytics, healthcare operation and management, and modeling of the product adoption in digit markets, as three case studies in dynamic business environments.