Prediction of Electric Load Neural Network Prediction Model for Big Data

Abstract

The characteristics of non-stationarity, non-linearity, and long-memory of the stock index series make it challenging to forecast. In order to improve the prediction accuracy of the existing models, this paper proposes a new ensemble prediction model of CSI 300 index returns by integrating variational mode technique and long short-term memory, which consists of particle swarm optimization (PSO), variational mode decomposition (VMD), sample entropy (SE) and long and short-term memory networks (LSTM). The stock index prices of the CSI 300 index during three consecutive months from October 2021 to December 2021 are selected as the research sample, and 48 sets of data are obtained in the 5-minute set every day, totaling more than 2,900 trading data as the modeling object, and the volume-weighted average price (VWAP) index is introduced to portray investor behavior. The experimental comparison shows that this method gives the smallest root mean square error and the best prediction fit, which significantly outperforms the existing portfolio model and has significant prediction advantages.