A Study Of The Optimal-combined Forecasting Model On Chongqing Port’s Cargo Throughput

The development level of port cargo throughput plays a significant role inmaking the port’s development strategy and conducting the port’s scientific and rationallayout. Accurately predicting the cargo throughput of Chongqing port can provide theessential data to support the Chongqing port’s planning and development.The purposeof this study is to improve the accuracy and rationality of the port throughputprediction by constantly improving the predicting methods. With the premise ofguaranteeing complete statistical information, the research result of the paper willprovide a strong support for the construction and planning of the Upper Yangtze Rivershipping center.This paper summaries and reviews the port cargo throughput prediction models,including the causal analysis method, the time series method and the combinedforecasting model. Through investigating the modeling mechanism, the applicationscope as well as the advantages and disadvantages of these forecasting approaches, thecombined forecasting model founded on the single prediction model is finallyestablished in the paper.As for the single prediction model, we have chosen the GM (1,1) predictionmodel and the regression forecasting model. Due to the fact that the prediction samplesof the port cargo throughput are relatively few in number, which belongs to the issue of“the less data uncertainty”, it is more suitable to apply the GM (1,1) prediction modelto solve the problem. However, the GM (1,1) model, classified as the category of thetime series forecasting method, can only carry out the time extrapolation based on thethroughput’s own history trends. As a result, once the factors that affect cargothroughput undergo great changes, there will be a relatively large prediction error.While the regression model belongs to the category of causal analysis method, it fullytakes the hinterland’s economic and social indicators which are related to the cargothroughput into consideration. Therefore, the regression model has a greater advantagewhen comparing with the GM (1,1) model in this aspect. In addition, it is crucial todetermine the impact factors scientifically and rationally, because there are manyfactors influencing the cargo throughput. The paper has also brought in the principalcomponent factor analysis approach in order to settle the problem of the influencingfactor extraction in the regression prediction model, By means of the IBM spss19.0 software, we eventually extracted three indicators, namely Chongqing’s GDP,Chongqing’s investment in fixed assets and Chongqing’s cargo throughput. Aftermultifaceted and comprehensive comparisons and testing, we finally confirm the cargothroughput forecast model of the Chongqing port with Chongqing’s GDP asindependent variable.In the optimal-combined Chongqing port cargo throughput prediction model,selecting the weighting factors is the key focus. In this combined forecasting model,the weight coefficients at each point remain unchanged, but it is indeed difficult tohave every single prediction model at different points maintain exactly the sameprecision in practice. Therefore, the combined forecasting model should choosevariable weights when selecting the weighting coefficients. To meet this end, we haveadopted the induced ordered weighted average operator (i.e. IOWA operator),assigning a value to the parameter according to the order of fit precision in the sampleinterval at each point in the single prediction model and establishing theoptimal-combined forecasting model based on the principle of the minimum squarederror. Eventually, the practice proves that the accuracy of the newly created combinedforecasting model is significantly higher than the GM (1,1) prediction model and theregression forecasting model.