Research on time seriesprediction based on RBMO-BiLSTM-KAN

Series are widely present in various fields of the real world, usually accompanied by a large amount of high-dimensional data, so time series prediction has become an important research direction in this field. Traditional time series prediction methods are mainly based on the time dimension for analysis, and often ignore the potential impact of external factors on time series. This paper highlights one based on Red-billed Blue Magpie Optimizer Bidirectional Long Short-Term Memory Kolmogorov-Arnold Networks (RBMO-BiLSTM-KAN). The model introduces a bidirectional LSTM (BiLSTM) to capture the bidirectional dependencies in the time series, and extracts the correlation between features and learning input features through the KAN network. In the experimental part, we first determined the hyperparameters of the model, then verified the effectiveness of each component of the model, and finally proved the excellent prediction performance of the RBKAN model on high-dimensional feature datasets through comparative experiments.