An Improved CNN–LSTM Model for Daily Gold Price Prediction

DOI

10.2991/978-94-6239-648-7_53How to use a DOI?

Keywords

Gold Price; Deep Learning; LSTM; CNN

Abstract

Gold prices play a pivotal role in the global financial system, but they often experience short-term volatility, long-term trends, and frequent regime switching. To address this, this paper proposes a new architecture, building on the CNN architecture and combining ordinary convolution with dilated convolution to create a parallel multi-scale convolution. This architecture then utilizes a learnable gating coefficient, which is then dot-producted with two branches before entering the LSTM. Ablation and comparative experiments were conducted with two training sets: test set ratios of 6:4 and 8:2. Experimental results show that, in the 6:4 ratio, the parallel dilated convolution without gates (DilatedCNN-LSTMs) performs best, achieving an RMSE of 13.11 and a MAE of 8.89. In the 8:2 ratio, the parallel model with gates (DilatedCNN-LSTMs) achieves the best RMSE of 15.31 and a MAE of 10.22. When the training set ratio is large, gating can effectively reduce bias and improve generalization. However, when the training set ratio is small, the increased degrees of freedom can lead to variance and training instability. This method requires no exogenous variables, is low-cost, and highly reproducible, making it valuable for daily gold price forecasting and risk management.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yi Lin
PY  - 2026
DA  - 2026/04/24
TI  - An Improved CNN–LSTM Model for Daily Gold Price Prediction
BT  - Proceedings of the International Workshop on Advances in Deep Learning for Image Analysis and Computer Vision (IWADIC 2025)
PB  - Atlantis Press
SP  - 480
EP  - 490
SN  - 2352-538X
UR  - https://doi.org/10.2991/978-94-6239-648-7_53
DO  - 10.2991/978-94-6239-648-7_53
ID  - Lin2026
ER  -