Branching Time Series Prediction Method based on CNN and LSTM

Time series data, an increasingly common form of data in real-world applications, is typically time-dependent. Time series forecasting seeks to extract potential information from time series data to forecast future trends. The complexity and nonlinearity of current time series data make it difficult for traditional forecasting algorithms to handle them, hence this paper suggests a branching time series forecasting method based on convolutional neural network (CNN) and long short-term memory network (LSTM). The method initially uses CNN to extract features from time series data, then uses CNN and LSTM to map these features to future prediction values, and lastly inputs to the fully connected layer to obtain the prediction results. According to the experimental findings, the branching prediction method based on CNN and LSTM has superior time series prediction accuracy and stability.