Early classification of temporal sequences has applications in, for example, health informatics, intrusion detection, anomaly detection, and scientific and engineering sequence data monitoring. Comparing to learning conventional sequence classifiers, learning early classifiers is a more challenging task and has not been systematically studied before. In this work, we identify the problem of early classification and develop a series of classifiers for temporal sequence early classification. The proposed classifiers are designed for different types of temporal sequences including symbolic sequences and time series. Furthermore, the proposed classifiers have several desirable characteristics which are useful in different application scenarios. We evaluate our approaches on a broad range of real data sets and demonstrate that the classifiers can achieve competitive classification accuracies with great earliness. Also, the classifiers can extract interpretable features from sequences for better understanding.
Copyright is held by the author.
The author granted permission for the file to be printed and for the text to be copied and pasted.
Supervisor or Senior Supervisor
Thesis advisor: Pei, Jian
Member of collection