PROGRESS IN GEOGRAPHY ›› 2014, Vol. 33 ›› Issue (5): 636-646.doi: 10.11820/dlkxjz.2014.05.005

• Population and Health Geography • Previous Articles     Next Articles

Application of combination forecasting model in geographic distribution of reference value of women’s peak expiratory flow rate

XUE Ranyin, GE Miao, HE Jinwei, HU Yanyu, GU Linlin, YANG Shaofang   

  1. Health Geography Institute of Tourism and Environment College, Shaanxi Normal University, Xi'an 710119, China
  • Received:2013-10-01 Revised:2013-12-01 Online:2014-05-25 Published:2014-05-28

Abstract: With the development of geography and people's overall health concerns, medical geography as an emerging discipline has also experienced rapid developments. Given that most of the existing medical reference values take into little consideration the influence of geographical factors, a more comprehensive and scientific method should be developed to take these into account. This article takes healthy adult women's lung peak expiratory flow rate reference value as an example, using 3809 cases of healthy adult women's peak expiratory flow reference value collected throughout China to analyze the impact of geographic factors, calculate the differences of different regions' reference values, and explore the mechanism of geographical factor's influence on medical reference value in an effort to improve the methods of medical reference value analysis through analyzing the relationship between geographical factors and medical reference values. As a first step, correlation analysis was used to analyze the relationship between peak expiratory flow value of adult women and the selected 25 indicators of geographical factors. Based on the result, seven geographic indicators (latitude, altitude, average temperature, annual average relative humidity, annual rainfall, topsoil gravel percentage, and topsoil reference capacity) that have significant correlation with peak expiratory flow reference value were extracted for further analysis. Second, Moran's I (spatial autocorrelation module), one of the ArcGIS software's analytical tools, was used to determine if this group of data is impacted by spatial and geographical factors. Third, using the data for the seven selected indicators, ridge regression analysis and SVR (support vector regression) were used to create two regression models and interpolate values. Then the results of these two prediction models were given different weights to establish the optimal combination forecasting model of spatial differences. Student's T test was used to compare the accuracy of ridge regression analysis, SVR and the combination forecasting model. Meanwhile, differences between the true values and results of the above three models were also considered for evaluating the performance of the models. Finally, a spatial difference prediction map was made. Based on this map and the results of correlation analyses, this article discusses why and how these geographical factors influence human tissues/organs and medical reference values. The output of this study indicates that the relationship between geographical factors and healthy adult Chinese women's lung peak expiratory flow rate should not be overlooked. The selected geographical factors (classify into terrain, climatic and soil factors) affect the lung tissue, especially the structure and function of the bronchi, because different living environments impact human tissues and organs differently, and humans living in different regions develop some differences in tissues and organs. The result of this research also shows that the combination forecasting model, which combined ridge regression and SVR, performed better than the individual prediction methods. Combination forecasting model not only can be used in traditional prediction exercises using temporal data: it is also possible to use this method to predict differences in geographic distribution or spatial data. It has the potential to be further expanded and utilized.

Key words: China, combination forecasting model, geographical distribution, peak expiratory flow rate, ridge regression, support vector regression

CLC Number: 

  • R188