Temperature prediction model based on BP-Holt-winters

• 1、College of Science, Liaoning Technical University, Fuxin 123000

Abstract：In modern times, the prediction of the monthly average temperature in each region is mainly to grid the surface of the earth, collect the values of the meteorological elements at each point, and perform interpolation calculations to obtain the value of the temperature in a future month. Therefore, this paper presents a new Holt-winters model based on BP neural network, referred to as BP-Holt-winters model (BP Neural Network and Holt-winters algorithm) for short-to-medium-term temperature prediction. This algorithm is completely Big data processing method based on historical data, the algorithm dimension is small, and the data is easy to collect. BP Neural Network has the function of realizing any complex non-linear mapping, and it can automatically extract "reasonable" solution rules by learning the instance set with correct answers To obtain data with relatively small errors, and the third-order exponential smoothing algorithm can prevent the problem of excessive boundary error of the BP algorithm, which greatly reduces the error of the entire algorithm. Therefore, the prediction method has high reliability and can be relatively accurate Forecast the future precipitation of various cities.

Keywords： Information and Computing Science Temperature forecast BP Neural Network Holt-winters Mega data