Scalable and Accurate Time Series Classification

This thesis focuses on time series classification, which aims to develop algorithms that learn to categorize temporally ordered data. It is an important area of machine learning research with a diverse range of applications, such as the classification of satellite images, medical and human activity data. This research addresses the lack of support for scalability and multivariate time series among state-of-the-art time series classifiers. It contributes two novel univariate algorithms that demonstrate state-of-the-art performance in accuracy while being several magnitudes faster than its competitors. It also contributes seven multivariate similarity measures and two ensembles for multivariate time series classification.