Analysis and Prediction Model of Big Data-Driven Electricity Consumption Behavior in the Context of Power IoT Environment

With the easing of the contradiction between power supply and demand, the power industry has shifted from solving the problem of power shortage to a stage of structural adjustment, improving quality, and promoting comprehensive benefits for the entire industry. The Internet of Things uses information sensing devices to exchange and communicate information about any item, achieving intelligent identification, positioning, tracking, monitoring, and management. Electricity consumption behavior is subjective and variable, and the existing problems are that the data collected from electricity consumption behavior is not rich, the ability to process data is insufficient, and the predictive ability of user behavior is not accurate enough. Therefore, this article proposes to further improve the analysis and prediction of user behavior using big data-driven technology. With the help of the big environment of the power Internet of Things, a usable model is constructed to achieve accurate analysis of electricity consumption behavior, and the predicted results are generally consistent with the actual results. This article mainly uses survey methods to analyze user electricity consumption behavior, and verifies the accuracy of the prediction model through experimental simulation. By comparing the error between predicted electricity consumption and actual electricity consumption, the objective function and minimum leaf size are adjusted, and the proportion of instances in the test set, validation set, and training set is compared when the error is consistent. The experimental results show that the error with the highest number of user instances in the test set is 0.3703, and when the errors are 3.157 and 3.467, the number of user instances is 0.