Automated analysis of heterogeneous sequential data

Currently, heterogeneous collections of sequential data are generated in important application domains such as resource usage, sports, predictive maintenance, and health. This has spurred substantial interest in data science, which concerns itself with discovering useful patterns and models in data. However, the data science process is laborious, difficult, and repetitive. This has motivated substantial interest in automating aspects of data science. However, current work has a major limitation, namely it assumes standard tabular data whereas heterogeneous sequential data is highly non-standard, which introduces substantial complexities for analysis. This project will focus on trying to automate data science tasks such as feature construction for heterogeneous sequential data.