Font Size: a A A

Study On Combination Model Of ARIMA And GM In Forecasting Population Mortality Rate Of China

Posted on:2010-05-15Degree:MasterType:Thesis
Country:ChinaCandidate:X D LiuFull Text:PDF
GTID:2167330332972511Subject:Epidemiology and Health Statistics
Abstract/Summary:PDF Full Text Request
Objective: To describe and analyze the trend of population mortality rate of China from 1975 to 2007 and explore the applicability of several mathematical models,so that to set up the best model to forecast the trend of population mortality rate in China and provide the theoretical basis for relative departments.Methods: The data this study used was population mortality rate of China from 1975 to 2007 from the Ministry of Health, "China's Health Statistics Year book 2008" . First, described the data of population mortality rate of China from 1975 to 2007 with average speed of development and average speed of growth and sequence chart. Based on the distribution character of population mortality rate of China from 1975 to 2007, ARIMA (Auto Regressive Integrated Moving Average model) model, Grey system forecasting (Grey Dynamics Model,GM) were used to fit the data .Then choosed the two models to make up four different Combination forecasting models . SSE , MAE , MSE , MAPE , MSPE were used to compare the prediction results to choose the best model.Results: (1) In general, the population mortality rate from 1975 to 2007 fluctuated obviously. In the years 1975-1979 there was a clear decline and there was a significant increase in1979-1983, and then began a slow decline. From 2004, the descending trend was obvious. The average speed of development was 99.8% and the average speed of growth was -0.2%. (2) The sum squares of Error of ARIMA model and GM(1,1) model were 0.5688 and 0.6572. (3) According to the comparison results of all the models involved, the Combined forecastting model based on Countdown method of variance was much better than other models. The sum squares of Error of this model was 0.3924. The model was X? = 0.5361X?1 +0.4639X?2. The predictive values of population mortality rate in China from 2008 to 2010 were 6.47‰,6.50‰,6.55‰.Couclusions: (1) The population mortality rate of China from 1975 to 2007 was a random, non-stationary time series. (2) The selection of mathematical model should be based on changes in the characteristics of data and application conditions of models. (3) ARIMA model was better than GM(1,1) model. The prediction results of population mortality rate were comparatively more precise by fitting the Combination model based on Countdown method of variance. So the Combination model based on Countdown method of variance was the best model in fitting and forecasting the population mortality rate of China. According to the results of its forecast,the population mortality rate of China in the next few years will continue the descending trend of previous years.Although it is impossible to predict accurately future changes and the exact value in population mortality rate, the results obtained are based on its historical trend, so the general direction will not have a great deviation. In this sense, the results of this study provide valuable reference for decision-making departments.
Keywords/Search Tags:Combination model, ARIMA model, GM(1,1) model, Population mortality rate
PDF Full Text Request
Related items