Forecast Of The Increasing Trend Of Highway Freight Transport Volume Based On Cluster Analysis

In recent years,the state needs to fully consider the benefits brought by highway construction when formulating highway infrastructure construction plans for various provinces and cities and determining the scale of related investment.The benefits of highway construction come from road transportation,and the development of road transportation depends on road goods.Therefore,the research goal of this paper is to use a small number of models to predict the growth trend of highway freight transportation in 31 provinces,and to reduce the number of prediction models and model parameters by adopting the method of clustering first and then predicting,and simplifying the prediction work.The specific work content is as follows:According to the actual situation of the time series data of road cargo transportation,a time series clustering algorithm is proposed to divide 31 provinces.Firstly,an improved K-means algorithm that combines variance and distance is proposed.By combining variance with the Max-Min distance algorithm,the selection of the centroid is avoided to be random,and in each iteration,the median is selected instead of the mean as the new value.The cluster center can eliminate the influence of outliers on the clustering results.Then,in view of the high computational complexity of the commonly used time series similarity measurement method dynamic time warping,a fast dynamic time warp is proposed by constraining the search space.At the same time,the pruning strategy was introduced to further reduce the computational complexity on the basis of ensuring the calculation accuracy of the clustering distance.Finally,the monthly data of road freight volume of 31 provinces in the past ten years were selected for empirical analysis,and compared with the K-means algorithm and the time series clustering algorithm before improvement,it is confirmed that the time series clustering algorithm proposed in this paper can obtain better clustering results.According to the clustering results,the road freight transportation volume of each type of cluster center province is fitted and predicted,and then the growth trend of the remaining provinces in the class to which each cluster center belongs is grasped.Since a time series contains both linear and nonlinear characteristics,it is obviously not sufficient to use a single forecasting model to model the time series.In order to extract the information in the time series to the greatest extent,this paper adopts ARIMA-PSO-SVR combined prediction model for prediction,and compared with ARIMA model,SVR model and Stacking integration method,and evaluated the effectiveness of the model through the average absolute percentage error and root mean square error of the prediction effect evaluation indicators.The experimental results show that the combined forecasting model of ARIMA-PSO-SVR can more fully reflect the changing law of road freight transportation in the prediction of highway freight transportation,and has good validity and reliability.