The Research On The Combined Forecast Model Of Total Power Of Agricultural Machinery In Xinjiang Corps

The total power of the corps'agricultural machinery reflects the overall level of agricultural mechanization development in the corps area,and is an important indicator of the agricultural mechanization development plan of the corps.However,the growth of the agricultural machinery total power influenced by various factors,the conventional single forecast model is adopted to predict and error precision is large,prediction effect is often not satisfactory.In recent years,with the continuous development of combination forecast theory,the combination forecast model is widely used in the prediction of total power of agricultural machinery for the high prediction accuracy,but there are still some,such as single forecasting model selection with no quantitative methods in the application of combination forecast model,and other issues need to be solved.With the related statistical data from 1989 to 2015 of agricultural machinery total power of Xinjiang construction corps as the basis,established Logistic regression model,index method,the cubic polynomial method,ARIMA model,quadratic polynomial method,exponential smoothing and gray model of seven kinds of single forecasting model,used inclusive testing principle to selected single forecasting model of seven kinds of models,for the four kinds of selected single prediction models,the entropy method,Shapley method and variation coefficient method are adopted to establish the non-optimal combination prediction model based on the model error information,based on the minimum absolute value of error and minimum error variance criterion to establish the optimal combination forecast model,selecting the MAPE(square absolute percentage error)evaluation index to evaluate forecasting results,used a combination model with high precision of 2016-2020 to predict the total power of agricultural machinery and used Markov model to predict the interval.The main conclusions are as follows:(1)Passed the inclusive inspection models of seven selected single forecasting models respectively are:Logistic regression model,cubic polynomial model,quadratic polynomial model and ARIMA model.(2)Different weight allocation methods were used to construct the optimal combination forecast model,the prediction accuracy from high to low respectively are:based on the Shapley value method of combination model,based on the variation coefficient method of combination model,based on entropy value method of combination model.(3)Built two kinds of optimal combination models with error sum of squares of the smallest and error absolute value the minimum criterions,through the contrast analysis of SSE,MSE and many other error indicators,get two optimal combination model prediction accuracy were higher than the four kinds of single forecasting model prediction accuracy in conclusion(1),and in the two kinds of optimal combination model,based on the combination model of error absolute value and minimum standards better prediction effect.(4)Selected the optimal combination model based on the Shapley value to predict the total power of agricultural machinery in 2016-2020,prediction results are:533.795,557.783,575.445,592.899,608.38510~4 kW.(5)Based the combination forecast model with Shapley value method as basis to analyze the forecast data from 1991 to 2015,and used the state of the prediction error as a reference index for classification,set up state transition matrix,to solve the prediction interval value in 2016-2020,using the Markov model range forecast can not only improves the reliability and stability of the medium and long term prediction,but also improves the practicability of forecasting result,and provides an important reference to formulate a long time,scientific and reasonable planning for the relevant department.