Enhanced mixup for improved time series analysis
Time series data analysis is crucial for real-world applications. While deep learning has advanced in this field, it still faces challenges, such as limited or poor-quality data. In areas like computer vision, data augmentation has been widely used and highly effective in addressing similar issues....
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Universitas Ahmad Dahlan
2025-05-01
|
| Series: | IJAIN (International Journal of Advances in Intelligent Informatics) |
| Subjects: | |
| Online Access: | https://ijain.org/index.php/IJAIN/article/view/1592 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1839636411453538304 |
|---|---|
| author | Khoa Tho Anh Nguyen Khoa Nguyen Taehong Kim Ngoc Hong Tran Vinh Dinh |
| author_facet | Khoa Tho Anh Nguyen Khoa Nguyen Taehong Kim Ngoc Hong Tran Vinh Dinh |
| author_sort | Khoa Tho Anh Nguyen |
| collection | DOAJ |
| description | Time series data analysis is crucial for real-world applications. While deep learning has advanced in this field, it still faces challenges, such as limited or poor-quality data. In areas like computer vision, data augmentation has been widely used and highly effective in addressing similar issues. However, these techniques are not as commonly explored or applied in the time series domain. This paper addresses the gap by evaluating basic data augmentation techniques using MLP, CNN, and Transformer architectures, prioritized for their alignment with state-of-the-art trends in time series analysis rather than traditional RNN-based methods. The goal is to expand the use of data augmentation in time series analysis. The paper proposed EMixup, which adapts the Mixup method from image processing to time series data. This adaptation involves mixing samples while aiming to maintain the data's temporal structure and integrating target contributions into the loss function. Empirical studies show that EMixup improves the performance of time series models across various architectures (improving 23/24 forecasting cases and 12/24 classification cases). It demonstrates broad applicability and strong results in tasks like forecasting and classification, highlighting its potential utility across diverse time series applications. |
| format | Article |
| id | doaj-art-9aeb91cf7c3d49da877c191567d2fbb8 |
| institution | Matheson Library |
| issn | 2442-6571 2548-3161 |
| language | English |
| publishDate | 2025-05-01 |
| publisher | Universitas Ahmad Dahlan |
| record_format | Article |
| series | IJAIN (International Journal of Advances in Intelligent Informatics) |
| spelling | doaj-art-9aeb91cf7c3d49da877c191567d2fbb82025-07-07T23:40:27ZengUniversitas Ahmad DahlanIJAIN (International Journal of Advances in Intelligent Informatics)2442-65712548-31612025-05-0111222724010.26555/ijain.v11i2.1592333Enhanced mixup for improved time series analysisKhoa Tho Anh Nguyen0Khoa Nguyen1Taehong Kim2Ngoc Hong Tran3Vinh Dinh4School of Electrical Engineering and Computer Science, Vietnamese–German UniversityInformation and Communication Technology, Chungbuk National UniversityInformation and Communication Technology, Chungbuk National UniversitySchool of Electrical Engineering and Computer Science, Vietnamese–German UniversityVietnamese German University, AI VIETNAMTime series data analysis is crucial for real-world applications. While deep learning has advanced in this field, it still faces challenges, such as limited or poor-quality data. In areas like computer vision, data augmentation has been widely used and highly effective in addressing similar issues. However, these techniques are not as commonly explored or applied in the time series domain. This paper addresses the gap by evaluating basic data augmentation techniques using MLP, CNN, and Transformer architectures, prioritized for their alignment with state-of-the-art trends in time series analysis rather than traditional RNN-based methods. The goal is to expand the use of data augmentation in time series analysis. The paper proposed EMixup, which adapts the Mixup method from image processing to time series data. This adaptation involves mixing samples while aiming to maintain the data's temporal structure and integrating target contributions into the loss function. Empirical studies show that EMixup improves the performance of time series models across various architectures (improving 23/24 forecasting cases and 12/24 classification cases). It demonstrates broad applicability and strong results in tasks like forecasting and classification, highlighting its potential utility across diverse time series applications.https://ijain.org/index.php/IJAIN/article/view/1592time series augmentationmixed-sample augmentationtimes series forecasting |
| spellingShingle | Khoa Tho Anh Nguyen Khoa Nguyen Taehong Kim Ngoc Hong Tran Vinh Dinh Enhanced mixup for improved time series analysis IJAIN (International Journal of Advances in Intelligent Informatics) time series augmentation mixed-sample augmentation times series forecasting |
| title | Enhanced mixup for improved time series analysis |
| title_full | Enhanced mixup for improved time series analysis |
| title_fullStr | Enhanced mixup for improved time series analysis |
| title_full_unstemmed | Enhanced mixup for improved time series analysis |
| title_short | Enhanced mixup for improved time series analysis |
| title_sort | enhanced mixup for improved time series analysis |
| topic | time series augmentation mixed-sample augmentation times series forecasting |
| url | https://ijain.org/index.php/IJAIN/article/view/1592 |
| work_keys_str_mv | AT khoathoanhnguyen enhancedmixupforimprovedtimeseriesanalysis AT khoanguyen enhancedmixupforimprovedtimeseriesanalysis AT taehongkim enhancedmixupforimprovedtimeseriesanalysis AT ngochongtran enhancedmixupforimprovedtimeseriesanalysis AT vinhdinh enhancedmixupforimprovedtimeseriesanalysis |