Research On Time Series Combination Forecast Model Based On Personnel Infrared Data

Prediction is an important data analysis method,which helps to mine the laws between data and estimate future trends.With the improvement of intelligent monitoring technology,the need for monitoring the use time of school rooms is becoming stronger and stronger.Compared with the traditional method of investigator statistics,the use of infrared sensor monitoring can not only make the statistical results more accurate,but also effectively reduce the labor force.Personnel infrared time series data are often complex and have strong uncertainty.The existing single prediction model is difficult to comprehensively predict it.After consulting the relevant literature on the combined prediction model and the infrared sensor data prediction model,it was decided to use the combined forecasting model to predict the infrared time series data of personnel.By predicting the infrared data of personnel,it is helpful to grasp the usage rules of school rooms,so as to make efficient time planning and reduce unnecessary waste of resources.There are many ways to combine models,and how to combine which single models is a question that needs to be studied.In response to the monitoring needs of school room usage,this paper proposes three combined forecasting models suitable for personnel infrared time series data.The content of the research and the contributions of this paper are:1)Aiming at the problem of poor temporal reasoning ability of Prophet model and lack of residual test,a forecast model composed of Seasonal Autoregressive Integrated Moving Average model and Prophet model is proposed.This model overcomes the shortcomings of the traditional equal-weight combination.First,use Seasonal Autoregressive Integrated Moving Average and Prophet model to predict,and then assign initial weights to the two models respectively,and continuously modify the weights within the maximum number of iterations to obtain the optimal combination of weights.Finally,the predicted values of the two models are multiplied by their corresponding optimal weights and summed.2)Aiming at the difficulty of obtaining Support Vector Machine model parameters.This paper proposes a Support Vector Machine combined forecasting model based on Ant Colony and Particle Swarm Optimization.The model is upgraded to a hybrid heuristic algorithm to optimize model parameters based on a single heuristic algorithm to optimize model parameters,which makes the prediction accuracy higher.First,the mean square error of the Support Vector Machine model is used as the fitness function,and then the better fitness and corresponding parameters in the Particle Swarm and Ant Colony algorithm are obtained.Finally,this set of parameters is substituted into the Support Vector Machine model for prediction.3)Aiming at the problem that it is difficult to capture the nonlinear relationship between the data of the traditional time series prediction model.The Autoregressive Integrated Moving Average model,parallel Long and Short-Term Memory networks,and Support Vector Machines were integrated into a prediction model for comprehensive time series analysis.The model first takes the predicted value of the Autoregressive Integrated Moving Average model as the linear component,and calculates its residual.Secondly,use the parallel structure of the Long and Short-Term Memory network and the Support Vector Machine model to predict the residual part,calculate the average value of the parallel model prediction value,and use this value as a non-linear component.Finally,the linear part and the non-linear part are summed as the final prediction result.The results of the three combined models are better than the corresponding single model in the prediction of the infrared time series data of personnel.The experimental results show that the three combined prediction models proposed in this paper have better accuracy in the field of room usage time prediction,which expands the application range of the combined prediction model.