Enhancing LSTM-based Time Series Forecasting with RandomForest algorithm and Attention Mechanism

Accurate forecasting of time series data is a significant challenge thatdemands the capabilities of advanced machine learning models. In thisstudy, we introduce an innovative hybrid model combining Long Short-Term Memory (LSTM) with an attention mechanism, further enhancedby the Random Forest algorithm. This hybrid approach enhances themodel's ability to learn from complex features and long-termdependencies in time series data, providing more precise forecasts. Themain purpose of this study is to examine the effect of this modelcombination on improving the accuracy of forecasting time series data.We benchmarked the proposed model against several other approaches,including GRU, LSTM, and attention-based LSTM models. Theperformance of these models was evaluated using RMSE and MAEmetrics on two distinct datasets, revealing that the RF-ATLSTM modelachieved the highest accuracy in predicting electricity consumption andfinance forecasting. The findings suggest that the integration ofRandom Forest with LSTM and the attention mechanism not onlyboosts the neural network's learning capacity but also significantlyenhances the model's ability to minimize prediction errors.

کلیدواژه ها:

LSTM networks ، Time Series ، Hybrid Models