Long short-term memory autoencoder based network of financial indices
Abstract We present a novel approach for analyzing financial time series data using a Long Short-Term Memory Autoencoder (LSTMAE), a deep learning method. Our primary objective is to uncover intricate relationships among different stock indices, leading to the extraction of stock networks. We examin...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer Nature
2025-01-01
|
| Series: | Humanities & Social Sciences Communications |
| Online Access: | https://doi.org/10.1057/s41599-025-04412-y |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832571959644258304 |
|---|---|
| author | Kamrul Hasan Tuhin Ashadun Nobi Mahmudul Hasan Rakib Jae Woo Lee |
| author_facet | Kamrul Hasan Tuhin Ashadun Nobi Mahmudul Hasan Rakib Jae Woo Lee |
| author_sort | Kamrul Hasan Tuhin |
| collection | DOAJ |
| description | Abstract We present a novel approach for analyzing financial time series data using a Long Short-Term Memory Autoencoder (LSTMAE), a deep learning method. Our primary objective is to uncover intricate relationships among different stock indices, leading to the extraction of stock networks. We examine time series data spanning from 2000 to 2022, encompassing multiple financial crises within the S&P 500 stock indices. By training a modified LSTMAE with normalized stock index returns, we extract the inherent correlations embedded in the model weights. We create directional threshold networks by applying a fixed threshold, calculated as the sum of the mean and standard deviation of matrices from various years. Our investigation explores the topological characteristics of these threshold networks across different years. Notably, the observed network properties exhibit unique responses to the various financial crises that occurred between 2000 and 2022. Furthermore, our sector analysis reveals substantial sectoral influences during times of crisis. For example, during global financial crises, the financial sector assumes a prominent role, exerting significant influence on other sectors, particularly during the European Sovereign Debt (ESD) crisis. During the COVID-19 pandemic, the health care and consumer discretionary sectors are predominantly impacted by other sectors. Our proposed method effectively captures the underlying network structure of financial markets and is validated by a comprehensive analysis of network metrics, demonstrating its ability to identify significant financial crises over time. |
| format | Article |
| id | doaj-art-d2e14b80ff7d402d83ba4334fe36a96e |
| institution | Kabale University |
| issn | 2662-9992 |
| language | English |
| publishDate | 2025-01-01 |
| publisher | Springer Nature |
| record_format | Article |
| series | Humanities & Social Sciences Communications |
| spelling | doaj-art-d2e14b80ff7d402d83ba4334fe36a96e2025-02-02T12:13:04ZengSpringer NatureHumanities & Social Sciences Communications2662-99922025-01-0112111510.1057/s41599-025-04412-yLong short-term memory autoencoder based network of financial indicesKamrul Hasan Tuhin0Ashadun Nobi1Mahmudul Hasan Rakib2Jae Woo Lee3Department of Computer Science and Telecommunication Engineering, Noakhali Science and Technology UniversityDepartment of Computer Science and Telecommunication Engineering, Noakhali Science and Technology UniversityDepartment of Computer Science and Telecommunication Engineering, Noakhali Science and Technology UniversityDepartment of Physics, Inha UniversityAbstract We present a novel approach for analyzing financial time series data using a Long Short-Term Memory Autoencoder (LSTMAE), a deep learning method. Our primary objective is to uncover intricate relationships among different stock indices, leading to the extraction of stock networks. We examine time series data spanning from 2000 to 2022, encompassing multiple financial crises within the S&P 500 stock indices. By training a modified LSTMAE with normalized stock index returns, we extract the inherent correlations embedded in the model weights. We create directional threshold networks by applying a fixed threshold, calculated as the sum of the mean and standard deviation of matrices from various years. Our investigation explores the topological characteristics of these threshold networks across different years. Notably, the observed network properties exhibit unique responses to the various financial crises that occurred between 2000 and 2022. Furthermore, our sector analysis reveals substantial sectoral influences during times of crisis. For example, during global financial crises, the financial sector assumes a prominent role, exerting significant influence on other sectors, particularly during the European Sovereign Debt (ESD) crisis. During the COVID-19 pandemic, the health care and consumer discretionary sectors are predominantly impacted by other sectors. Our proposed method effectively captures the underlying network structure of financial markets and is validated by a comprehensive analysis of network metrics, demonstrating its ability to identify significant financial crises over time.https://doi.org/10.1057/s41599-025-04412-y |
| spellingShingle | Kamrul Hasan Tuhin Ashadun Nobi Mahmudul Hasan Rakib Jae Woo Lee Long short-term memory autoencoder based network of financial indices Humanities & Social Sciences Communications |
| title | Long short-term memory autoencoder based network of financial indices |
| title_full | Long short-term memory autoencoder based network of financial indices |
| title_fullStr | Long short-term memory autoencoder based network of financial indices |
| title_full_unstemmed | Long short-term memory autoencoder based network of financial indices |
| title_short | Long short-term memory autoencoder based network of financial indices |
| title_sort | long short term memory autoencoder based network of financial indices |
| url | https://doi.org/10.1057/s41599-025-04412-y |
| work_keys_str_mv | AT kamrulhasantuhin longshorttermmemoryautoencoderbasednetworkoffinancialindices AT ashadunnobi longshorttermmemoryautoencoderbasednetworkoffinancialindices AT mahmudulhasanrakib longshorttermmemoryautoencoderbasednetworkoffinancialindices AT jaewoolee longshorttermmemoryautoencoderbasednetworkoffinancialindices |