A time series forest for classification and feature extraction

By Houtao Deng, George C. Runger, Eugene Tuv, Vladimir Martyanov. Information Sciences, 2013

622 citations recorded by the source corpus.

A tree-ensemble method, referred to as time series forest (TSF), is proposed for time series classification. TSF employs a combination of entropy gain and a distance measure, referred to as the Entrance (entropy and distance) gain, for evaluating the splits. Experimental studies show that the Entrance gain improves the accuracy of TSF. TSF randomly samples features at each tree node and has computational complexity linear in the length of time series, and can be built using parallel computing techniques. The temporal importance curve is proposed to capture the temporal characteristics useful for classification. Experimental studies show that TSF using simple features such as mean, standard deviation and slope is computationally efficient and outperforms strong competitors such as one-nearest-neighbor classifiers with dynamic time warping.

Subfields: Economics and Econometrics, Signal Processing

View Paper: https://www.semanticscholar.org/paper/9ad37f6e390f4eb4da0f82e312640c2b40919690

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Economics and Econometrics

In Signal Processing

Citations in the corpus

A shapelet transform for time series classification (2012, background): “Categories and Subject Descriptors H.2.8 [Database Management]: Database Applications Data Mining Keywords time series, shapelet, .lter, transformation 1.”
Time-Series Classification with COTE: The Collective of Transformation-Based Ensembles (2015, methodology): “The establishment of the UCR repository for TSC problems [1] has engendered growth in the number of algorithms proposed for TSC (for example, see [2], [3], [4], [5], [6], [7], [8], [9]).”
Ensemble learning: A survey (2018, background): “Diversity : The superior performance of ensemble models is achieved mainly due to the use of various “ inductive biases ” (Deng, Runger, Tuv, & Vladimir, 2013).”
Classification of time series by shapelet transformation (2013, background): “…alternative approaches, such as weighted dynamic time warping Jeong et al. (2010), support vector machines built on variable intervals Rodriguez and Alonso (2005), tree-based ensembles constructed on summary statistics Deng et al. (2011), and a fusion of alternative distance measures Buza (2011).”
Visualising forecasting algorithm performance using time series instance spaces (2017, methodology): “…on the research goals and domains, a variety of time series features have been developed to characterising time series in previous studies (e.g., Deng et al., 2013; Fulcher and Kang, Hyndman & Smith-Miles: 6 June 2016 4 Jones, 2014; Kang, Belušić, and Smith-Miles, 2014, 2015; Mörchen, 2003;…”
Highly Comparative Feature-Based Time-Series Classification (2014, methodology): “In this work, we automate the selection of features for time-series classiﬁcation by computing thousands of features from across the scientiﬁc time-series analysis literature and then selecting those with the best performance.”
The great time series classification bake off: a review and experimental evaluation of recent algorithmic advances (2016, methodology): “Deng et al. (2013) overcome the problem of the huge interval feature space by employing a random forest approach, using summary statistics of each interval as features.”