Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility
Abstract The problem of automatic and accurate forecasting of time‐series data has always been an interesting challenge for the machine learning and forecasting community. A majority of the real‐world time‐series problems have non‐stationary characteristics that make the understanding of trend and s...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2021-09-01
|
| Series: | CAAI Transactions on Intelligence Technology |
| Subjects: | |
| Online Access: | https://doi.org/10.1049/cit2.12002 |
| _version_ | 1828319243228151808 |
|---|---|
| author | Rohit Kaushik Shikhar Jain Siddhant Jain Tirtharaj Dash |
| author_facet | Rohit Kaushik Shikhar Jain Siddhant Jain Tirtharaj Dash |
| author_sort | Rohit Kaushik |
| collection | DOAJ |
| description | Abstract The problem of automatic and accurate forecasting of time‐series data has always been an interesting challenge for the machine learning and forecasting community. A majority of the real‐world time‐series problems have non‐stationary characteristics that make the understanding of trend and seasonality difficult. The applicability of the popular deep neural networks (DNNs) as function approximators for non‐stationary TSF is studied. The following DNN models are evaluated: Multi‐layer Perceptron (MLP), Convolutional Neural Network (CNN), and RNN with Long Short‐Term Memory (LSTM‐RNN) and RNN with Gated‐Recurrent Unit (GRU‐RNN). These DNN methods have been evaluated over 10 popular Indian financial stocks data. Further, the performance evaluation of these DNNs has been carried out in multiple independent runs for two settings of forecasting: (1) single‐step forecasting, and (2) multi‐step forecasting. These DNN methods show convincing performance for single‐step forecasting (one‐day ahead forecast). For the multi‐step forecasting (multiple days ahead forecast), the methods for different forecast periods are evaluated. The performance of these methods demonstrates that long forecast periods have an adverse effect on performance. |
| first_indexed | 2024-04-13T17:53:40Z |
| format | Article |
| id | doaj.art-cbe930f3c28a4559b24e7734efa704c9 |
| institution | Directory Open Access Journal |
| issn | 2468-2322 |
| language | English |
| last_indexed | 2024-04-13T17:53:40Z |
| publishDate | 2021-09-01 |
| publisher | Wiley |
| record_format | Article |
| series | CAAI Transactions on Intelligence Technology |
| spelling | doaj.art-cbe930f3c28a4559b24e7734efa704c92022-12-22T02:36:37ZengWileyCAAI Transactions on Intelligence Technology2468-23222021-09-016326528010.1049/cit2.12002Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatilityRohit Kaushik0Shikhar Jain1Siddhant Jain2Tirtharaj Dash3Department of Computer Science and Information Systems Birla Institute of Technology and Science Pilani K.K. Birla Goa Campus Zuarinagar Goa IndiaDepartment of Computer Science and Information Systems Birla Institute of Technology and Science Pilani K.K. Birla Goa Campus Zuarinagar Goa IndiaDepartment of Computer Science and Information Systems Birla Institute of Technology and Science Pilani K.K. Birla Goa Campus Zuarinagar Goa IndiaDepartment of Computer Science and Information Systems Birla Institute of Technology and Science Pilani K.K. Birla Goa Campus Zuarinagar Goa IndiaAbstract The problem of automatic and accurate forecasting of time‐series data has always been an interesting challenge for the machine learning and forecasting community. A majority of the real‐world time‐series problems have non‐stationary characteristics that make the understanding of trend and seasonality difficult. The applicability of the popular deep neural networks (DNNs) as function approximators for non‐stationary TSF is studied. The following DNN models are evaluated: Multi‐layer Perceptron (MLP), Convolutional Neural Network (CNN), and RNN with Long Short‐Term Memory (LSTM‐RNN) and RNN with Gated‐Recurrent Unit (GRU‐RNN). These DNN methods have been evaluated over 10 popular Indian financial stocks data. Further, the performance evaluation of these DNNs has been carried out in multiple independent runs for two settings of forecasting: (1) single‐step forecasting, and (2) multi‐step forecasting. These DNN methods show convincing performance for single‐step forecasting (one‐day ahead forecast). For the multi‐step forecasting (multiple days ahead forecast), the methods for different forecast periods are evaluated. The performance of these methods demonstrates that long forecast periods have an adverse effect on performance.https://doi.org/10.1049/cit2.12002time seriesforecasting theoryrecurrent neural netslearning (artificial intelligence)econophysics |
| spellingShingle | Rohit Kaushik Shikhar Jain Siddhant Jain Tirtharaj Dash Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility CAAI Transactions on Intelligence Technology time series forecasting theory recurrent neural nets learning (artificial intelligence) econophysics |
| title | Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility |
| title_full | Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility |
| title_fullStr | Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility |
| title_full_unstemmed | Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility |
| title_short | Performance evaluation of deep neural networks for forecasting time‐series with multiple structural breaks and high volatility |
| title_sort | performance evaluation of deep neural networks for forecasting time series with multiple structural breaks and high volatility |
| topic | time series forecasting theory recurrent neural nets learning (artificial intelligence) econophysics |
| url | https://doi.org/10.1049/cit2.12002 |
| work_keys_str_mv | AT rohitkaushik performanceevaluationofdeepneuralnetworksforforecastingtimeserieswithmultiplestructuralbreaksandhighvolatility AT shikharjain performanceevaluationofdeepneuralnetworksforforecastingtimeserieswithmultiplestructuralbreaksandhighvolatility AT siddhantjain performanceevaluationofdeepneuralnetworksforforecastingtimeserieswithmultiplestructuralbreaksandhighvolatility AT tirtharajdash performanceevaluationofdeepneuralnetworksforforecastingtimeserieswithmultiplestructuralbreaksandhighvolatility |