Application Research Of Prediction Method Based On LSTM

In recent years,with the continuous development of computer technology,neural network has been widely used in prediction,especially Long Short-Term Memory(LSTM)has been applied in various fields and disciplines.LSTM is an improved form on the basis of recurrent neural network,which is very suitable for dealing with the prediction of long-period time series.In addition,it has many advantages such as fast prediction speed and high accuracy,and benefits from these advantages.LSTM prediction method is widely used in weather forecast,stock forecast,behavior forecast and many other fields.Based on this,this paper adopts LSTM to establish the corresponding prediction model,and uses the corresponding professional methods to study the development law of natural things and predict their future trend,so as to prevent and take certain measures as early as possible,to slow down or avoid the adverse effects brought by the development of things.In this paper,fires in Australia and vegetation coverage in Gansu Province of China were selected as the research subjects.The main work is as follows:(1)Fire prediction in Australia based on LSTM.This article uses the Australian fire as a starting point to study Australian vegetation.Firstly,we distinguish the nodes in Australia based on climate and vegetation.After the nodes are determined,by collecting the historical daily maximum temperature,daily minimum precipitation,and fire information of each node,the data is processed into time series data.Then,we predict the fire possibility of each node based on time series data of multivariate LSTM prediction model.Compared with the BP neural network and Autoregressive Integrated Moving Average Model(ARIMA)algorithm,the multivariate LSTM has higher accuracy in predicting the Australian fire.Based on the effectiveness of this method.Using LSTM to design and implement a system which can predict fire.The system includes registration and login modules,data processing module,historical data query module,result display module,etc.The system can query historical data and predict the fire situation of different nodes.(2)Prediction of vegetation coverage in Gansu Province based on LSTM.Firstly,we collect the vegetation remote sensing satellite data of Gansu Province,and then the vegetation satellite images are processed as convenient time series data.By considering that the amount of image data is too large to be directly used for calculation,Principal Component Analysis(PCA)is used to reduce the dimension of vegetation data in Gansu Province.The data after dimensionality reduction is processed into a format that can be processed by LSTM.Next,LSTM is used to predict the data.Finally,the predicted results of LSTM output are transformed into visual images.Through the effective combination of these algorithms,we can learn the historical vegetation coverage in Gansu and study the change trend of vegetation coverage in the future,which is beneficial to the desertification prevention in Gansu.