Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies
Time-series forecasting is essential for predicting future trends based on historical data, with significant applications in meteorology, transportation, and finance. However, existing models often exhibit unsatisfactory performance in long-term forecasting scenarios. To address this limitation, we...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2025-06-01
|
| Series: | Applied Sciences |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2076-3417/15/11/6371 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1850130828310347776 |
|---|---|
| author | Jiahui Wang Wenqian Zhou Fangshu Chen Liming Wang Ruijun Pan Chengcheng Yu |
| author_facet | Jiahui Wang Wenqian Zhou Fangshu Chen Liming Wang Ruijun Pan Chengcheng Yu |
| author_sort | Jiahui Wang |
| collection | DOAJ |
| description | Time-series forecasting is essential for predicting future trends based on historical data, with significant applications in meteorology, transportation, and finance. However, existing models often exhibit unsatisfactory performance in long-term forecasting scenarios. To address this limitation, we propose the Time-Series Neural Networks via Ensemble of Short-Term Dependencies (TSNN-ESTD). This model leverages iTransformer as the base predictor to simultaneously train short-term and long-term forecasting models. The vanilla iTransformer’s linear decoding layer is optimized by replacing it with an LSTM layer, and an additional long-term model is introduced to enhance stability. The ensemble strategy employs short-term predictions to correct the bias in long-term forecasts. Our extensive experiments demonstrate that TSNN-ESTD reduces the MSE and MAE by 9.17% and 2.3% on five benchmark datasets. |
| format | Article |
| id | doaj-art-95fab44a227a4fd3a37d6b6cc84b80b5 |
| institution | OA Journals |
| issn | 2076-3417 |
| language | English |
| publishDate | 2025-06-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Applied Sciences |
| spelling | doaj-art-95fab44a227a4fd3a37d6b6cc84b80b52025-08-20T02:32:34ZengMDPI AGApplied Sciences2076-34172025-06-011511637110.3390/app15116371Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term DependenciesJiahui Wang0Wenqian Zhou1Fangshu Chen2Liming Wang3Ruijun Pan4Chengcheng Yu5School of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaSchool of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaSchool of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaSchool of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaSchool of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaSchool of Computer and Information Engineering, Shanghai Polytechnic University, Shanghai 201209, ChinaTime-series forecasting is essential for predicting future trends based on historical data, with significant applications in meteorology, transportation, and finance. However, existing models often exhibit unsatisfactory performance in long-term forecasting scenarios. To address this limitation, we propose the Time-Series Neural Networks via Ensemble of Short-Term Dependencies (TSNN-ESTD). This model leverages iTransformer as the base predictor to simultaneously train short-term and long-term forecasting models. The vanilla iTransformer’s linear decoding layer is optimized by replacing it with an LSTM layer, and an additional long-term model is introduced to enhance stability. The ensemble strategy employs short-term predictions to correct the bias in long-term forecasts. Our extensive experiments demonstrate that TSNN-ESTD reduces the MSE and MAE by 9.17% and 2.3% on five benchmark datasets.https://www.mdpi.com/2076-3417/15/11/6371time-series forecastingtime-series neural networkslong-term biasshort-term dependencyensemble learning |
| spellingShingle | Jiahui Wang Wenqian Zhou Fangshu Chen Liming Wang Ruijun Pan Chengcheng Yu Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies Applied Sciences time-series forecasting time-series neural networks long-term bias short-term dependency ensemble learning |
| title | Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies |
| title_full | Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies |
| title_fullStr | Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies |
| title_full_unstemmed | Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies |
| title_short | Mitigating Long-Term Forecasting Bias in Time-Series Neural Networks via Ensemble of Short-Term Dependencies |
| title_sort | mitigating long term forecasting bias in time series neural networks via ensemble of short term dependencies |
| topic | time-series forecasting time-series neural networks long-term bias short-term dependency ensemble learning |
| url | https://www.mdpi.com/2076-3417/15/11/6371 |
| work_keys_str_mv | AT jiahuiwang mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies AT wenqianzhou mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies AT fangshuchen mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies AT limingwang mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies AT ruijunpan mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies AT chengchengyu mitigatinglongtermforecastingbiasintimeseriesneuralnetworksviaensembleofshorttermdependencies |