Arya Bhatta Journal of Mathematics and Informatics
Year : 2023, Volume : 15, Issue : 1
First page : ( 11) Last page : ( 20)
Print ISSN : 0975-7139. Online ISSN : 2394-9309.
Article DOI : 10.5958/2394-9309.2023.00002.1

Stride Window Approach in Multiclass Classification for Temporal Sensor Data Augmentation

Jaiswal Rashi^*

ICT Research Lab, Department of Computer Science, University of Lucknow (UP) India

^*Email : rashijaiswal.rj95@gmail.com

Online Published on 08 June, 2023.

Abstract

In the real world, the sensors are used to fetch the data to perform different tasks such as analysis, sense to make a decision, and prediction purposes. Existing methods for sensor data are used to perform various tasks such as classification and regression. Various methods are available to compute the temporal sensor data with data augmentation as the sequential data. However, existing methods augmented a lot of messy and biased data that affects the outcome. In this paper, we propose a stride window approach for dealing with sensor data augmentation to perform multiclass classification. The stride window approach reduces the biased data and controls the overlapping in data augmentation and preprocessing. Illustrate the proposed method and justified by visualization and quantitative results on an open-source Human activity recognition dataset. The performance analysis has been done for multi-class classification by measuring the Accuracy and F1-Score, the results show that the proposed stride window approach outperforms with approx. 1%-10% more performance score than the resembling and time slice approaches for multiclass classification. The proposed approach can be used to generate augmented data concerning sensor data in defense applications for accurate monitoring and controlling.

Top

Keywords

Stride window Approach, Multiclass Classification, Data Augmentation, Temporal Data, Sensor Data, Human Activity Recognition.

Top