کنفرانس مهندسی برق ایران

صفحه اصلی / سی و چهارمین کنفرانس بین المللی مهندسی برق

Multivariate Gold Price Forecasting Using Attention LSTM-CBAM Network

نویسندگان :

AKHTAR PEZHMAN¹ ELHAM SHABANINIA² JAVAD MAHMOODI³

1- دانشگاه تحصیلات تکمیلی صنعتی و فناوری پیشرفته کرمان 2- دانشگاه تحصیلات تکمیلی صنعتی و فناوری پیشرفته کرمان 3- دانشگاه آزاد اسلامی واحد کرمان

کلمات کلیدی :

gold price forecasting،multivariate time series،attention mechanism،long short-term memory (LSTM)،convolutional block attention module (CBAM)

چکیده :

Gold prices have always been one of the influential factors in global financial markets, and it is natural that accurately predicting them is not an easy task. In recent years, researchers have turned to machine learning methods because these methods are generally better at analyzing complex behaviors and extreme market fluctuations. However, most traditional models that rely on just one prediction method still struggle with complex market patterns and cannot accurately identify all the intricate details of the data. For this reason, this study proposes a hybrid LSTM–CBAM model to address some of these limitations and achieve more accurate forecasting. The proposed framework for predicting gold prices is based on three input features by combining long short-term memory networks (LSTM) and convolutional block attention module (CBAM), allowing the model to focus on important temporal dependencies and prominent feature patterns. The data used were collected daily from the ForexSB website over the period from January 1, 2020, to October 16, 2024. The results show that the proposed hybrid model performs better than standard LSTM in detecting complex temporal patterns, relationships between features, and predicting gold prices.