Comparative Analysis of ARIMA and Deep Learning Models for Time Series Prediction

Chenyu Jiang

2024

Abstract

Time series analysis is crucial for forecasting future trends across various complex real-world domains, including finance, healthcare, and energy management. This study evaluates the performance of traditional and deep learning approaches for time series prediction, comparing the autoregressive Integrated Moving Average (ARIMA) model with Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) architectures. ARIMA, which is designed for linear and stationary data, was tested on the Corona Virus Disease (COVID)-19 dataset to predict infection rates. While ARIMA achieved reasonable success, its limitations became apparent in handling non-linear data. Conversely, RNN and LSTM models excelled in capturing complex non-linear patterns and temporal dependencies, demonstrating superior performance in forecasting a large-cap stock dataset. The experimental results revealed that LSTM significantly outperformed ARIMA in prediction accuracy. This underscores the growing need to integrate statistical models like ARIMA with deep learning techniques to enhance time series forecasting. The findings are particularly relevant for forecasting applications across industries, suggesting that hybrid models, which balance interpretability with predictive performance, may offer the most effective solutions.

Paper Citation

in Harvard Style

Jiang C. (2024). Comparative Analysis of ARIMA and Deep Learning Models for Time Series Prediction. In Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML; ISBN 978-989-758-754-2, SciTePress, pages 306-310. DOI: 10.5220/0013516200004619

in Bibtex Style

@conference{daml24,
author={Chenyu Jiang},
title={Comparative Analysis of ARIMA and Deep Learning Models for Time Series Prediction},
booktitle={Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML},
year={2024},
pages={306-310},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013516200004619},
isbn={978-989-758-754-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 2nd International Conference on Data Analysis and Machine Learning - Volume 1: DAML
TI - Comparative Analysis of ARIMA and Deep Learning Models for Time Series Prediction
SN - 978-989-758-754-2
AU - Jiang C.
PY - 2024
SP - 306
EP - 310
DO - 10.5220/0013516200004619
PB - SciTePress