Prediction of Electric Load Neural Network Prediction Model for Big Data

Abstract

Firstly, this article analyzes the distribution patterns and interrelationships of sales volume among different categories of vegetables and individual products. Based on the results of the stability test of the daily sales volume of six types of vegetables, the data of each type of vegetables is divided into 7 sub-data sets according to the week to explore the distribution patterns of vegetable sales volume and time. Subsequently, this article conducts a distribution test on 42 sub-data sets of the six types of vegetables, and finds that the Weibull distribution is the optimal fitting distribution. Spearman correlation coefficients are used to judge the correlation between the six types of vegetables based on scatter plots. The results show that there is weak correlation among different categories of vegetable products. An Apriori algorithm is used to analyze the correlation of 246 individual products, and 24 frequent itemsets are found. Then, a heatmap analysis is conducted on the individual products in each set, which shows strong correlation within the set. Finally, this article divides the daily sales volume of each vegetable category into two time scales, and uses the ARIMA(p,d,q) model and the ARIMA-LSTM model to predict the sales volume for the following week, which facilitates the construction of a sales volume-price model based on future predicted sales volume to predict the price of vegetables for the following week, and further analysis can be conducted.