• Paper title: Extracting Discriminative Shapelets from Heterogeneous Sensor Data
• Authors/Affiliations: Om P. Patri, Abhishek B. Sharma, Haifeng Chen, Guofei Jiang, Anand V. Panangadan, Viktor K. Prasanna. University of Southern California and NEC Laboratories America
• Paper Source:https://ieeexplore.ieee.org/document/7004344
• Topic tags: [data mining][outlier detection][Multivariate Data][Time Series Shapelets][Shapelet Forests]
• Summary of the paper
◦ What is it?
Shapelets for multivariate time series classification. In this work, the authors explore the efficacy of shapelets for solving multivariate time series classification problem for complex physical systems.
While multivariate time series classification is a well-studied problem, the authors focused on that problem(i.e. multivariate time series classification) in the context of complex real world engineering applications such as manufacturing plants and automobiles.
The user is required to specify the minL and maxL parameters that denote the minimum and maximum length of shapelets.
◦ How is it great compared to the related works?
The authors claim that most of the work on shapelet mining is aimed at univariate time series data. They proposed a new approach called Shapelet Forests which is able to find shapelets from multivariate time series.
◦ What are the key technical differentiators?
The proposed algorithm ( i.e. Shapelets Forest (SF)) can work directly with multivariate data when others ( Naive Shapelets (NS) and Concatenated Shapelets (CS) ) convert multivariate time series data to an equivalent univariate representation.and this process require s high computational complexity. Hence the claim of superiority of the Shapelets Forest (SF) method.
◦ How did they validate the advantages?
The authors showed that their SF algorithm outperforms the baseline approaches of NS and CS using data from the two real-world engineering applications .
• Paper title: Extracting Discriminative Shapelets from Heterogeneous Sensor Data • Authors/Affiliations: Om P. Patri, Abhishek B. Sharma, Haifeng Chen, Guofei Jiang, Anand V. Panangadan, Viktor K. Prasanna. University of Southern California and NEC Laboratories America • Paper Source: https://ieeexplore.ieee.org/document/7004344 • Topic tags: [data mining][outlier detection][Multivariate Data][Time Series Shapelets][Shapelet Forests] • Summary of the paper ◦ What is it?
◦ How is it great compared to the related works? The authors claim that most of the work on shapelet mining is aimed at univariate time series data. They proposed a new approach called Shapelet Forests which is able to find shapelets from multivariate time series.
◦ What are the key technical differentiators? The proposed algorithm ( i.e. Shapelets Forest (SF)) can work directly with multivariate data when others ( Naive Shapelets (NS) and Concatenated Shapelets (CS) ) convert multivariate time series data to an equivalent univariate representation.and this process require s high computational complexity. Hence the claim of superiority of the Shapelets Forest (SF) method.
◦ How did they validate the advantages? The authors showed that their SF algorithm outperforms the baseline approaches of NS and CS using data from the two real-world engineering applications .