Mutual information based feature subset selection in multivariate time series classification

Ircio, Josu; Lojo, Aizea; Mori, Usue; Lozano, José A.

doi:10.1016/j.patcog.2020.107525

Cited by 44 publications

(20 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To avoid this, some studies have used a k-nearest neighbour (k-NN) approach to calculate the MI which avoids the need to calculate the probability distribution function and therefore can be used on the original multidimensional feature subset [11,16]. In a very recent development [12], the authors propose a Filter method for feature subset selection where they assign a score function to assess the relevance of each feature subset. Mutual information based on a k-NN strategy is used to measure the information shared between the two time series.…”

Section: Mutual Information Based Methodsmentioning

confidence: 99%

“…It is worth comparing these results with the feature subset selection technique proposed by Ircio et al, which uses mutual information based on k-NN as it is directly comparable [12]. The authors show that on average for 8 out of the 16 datasets for DT W D and 4 out of the 16 datasets for DT W I the accuracy can be maintained or improved [12]. In our evaluations, MSTS improves or maintains the accuracy for 11 out of the 19 datasets for DTW and for 12 out of the 19 datasets for MiniRocket.…”

Section: Cfs For Time Seriesmentioning

confidence: 98%

“…In FSS of MTS data, a feature refers to an univariate time series [12,26,27,6]. Many current approaches use feature extraction where the time series is transformed or derived and these extracted values are termed as the features.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Correlation Based Feature Subset Selection for Multivariate Time-Series Data

Kathirgamanathan¹,

Cunningham²

2021

Preprint

View full text Add to dashboard Cite

Correlations in streams of multivariate time series data means that typically, only a small subset of the features are required for a given data mining task. In this paper, we propose a technique which we call Merit Score for Time-Series data (MSTS) that does feature subset selection based on the correlation patterns of single feature classifier outputs. We assign a Merit Score to the feature subsets which is used as the basis for selecting 'good' feature subsets. The proposed technique is evaluated on datasets from the UEA multivariate time series archive and is compared against a Wrapper approach for feature subset selection. MSTS is shown to be effective for feature subset selection and is in particular effective as a data reduction technique. MSTS is shown here to be computationally more efficient than the Wrapper strategy in selecting a suitable feature subset, being more than 100 times faster for some larger datasets while also maintaining a good classification accuracy.

show abstract

Section: Mutual Information Based Methodsmentioning

confidence: 99%

Section: Cfs For Time Seriesmentioning

confidence: 98%

See 1 more Smart Citation

Correlation Based Feature Subset Selection for Multivariate Time-Series Data

Kathirgamanathan¹,

Cunningham²

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…[24][25][26] The filter method 24 is independent of the used classifier. 27 It firstly selects the features of the data set, and then selects the optimal feature subset after sorting all the features. The wrapper method needs to use the determined learning algorithm repeatedly to search the feature subsets.…”

Section: Introductionmentioning

confidence: 99%

Ensemble mutation slime mould algorithm with restart mechanism for feature selection

Jia

Zhang

Zheng

et al. 2021

Int J of Intelligent Sys

View full text Add to dashboard Cite

Existing data acquisition technologies desire further improvement to meet the increasing need for big, accurate, and high-quality data collection. Most of the collected data have redundant information such as noise. To improve the classification accuracy, the dimensionality reduction technique, which is also known as the feature selection, is a necessity for data processing. In this paper, the slime mould algorithm (SMA) is optimized by introducing the composite mutation strategy (CMS) and restart strategy (RS). The improved SMA is named CMSRSSMA, which stands for the CMS, RS, and the improved SMA. The CMS is utilized to increase the population diversity, and the RS is used to avoid the local optimum. In this paper, the CEC2017 benchmark function is used to test the effectiveness of the proposed CMSRSSMA. Then, the CMSRSSMA-SVM model is proposed for feature selection and parameter optimization simultaneously.The performance of the model is tested by 14 data sets from UCI data repository. Experimental results show that the proposed method is superior to other algorithms in terms of classification accuracy, number of features and fitness value on most selected data sets.

show abstract

“…In addition, compared to the mutual information based methods, which can only work with the discrete or categorical features [15,16], the proposed method is applicable to both numerical (including discrete and continuous) and categorical features. The proposed method is also closely related to two recent advanced topics in the feature selection field, which are the multi-label feature selection [17,18,19] and multivariate time series feature selection [20,21], respectively. The fundamental issue that is addressed under the two research topics is to develop methods that can deal with the features and response that have to be represented in a matrix form.…”

Section: Introductionmentioning

confidence: 99%

Orthogonal Least Squares Based Fast Feature Selection for Linear Classification

Zhang,

Lang

2021

Preprint

View full text Add to dashboard Cite

An Orthogonal Least Squares (OLS) based feature selection method is proposed for both binomial and multinomial classification. The novel Squared Orthogonal Correlation Coefficient (SOCC) is defined based on Error Reduction Ratio (ERR) in OLS and used as the feature ranking criterion. The equivalence between the canonical correlation coefficient, Fisher's criterion, and the sum of the SOCCs is revealed, which unveils the statistical implication of ERR in OLS for the first time. It is also shown that the OLS based feature selection method has speed advantages when applied for greedy search. The proposed method is comprehensively compared with the mutual information based feature selection methods in 2 synthetic and 7 real world datasets. The results show that the proposed method is always in the top 5 among the 10 candidate methods. Besides, the proposed method can be directly applied to continuous features without discretisation, which is another significant advantage over mutual information based methods.

show abstract

Mutual information based feature subset selection in multivariate time series classification

Cited by 44 publications

References 39 publications

Correlation Based Feature Subset Selection for Multivariate Time-Series Data

Correlation Based Feature Subset Selection for Multivariate Time-Series Data

Ensemble mutation slime mould algorithm with restart mechanism for feature selection

Orthogonal Least Squares Based Fast Feature Selection for Linear Classification

Contact Info

Product

Resources

About