Random forest-based medical geography: geographical distribution of reference values for LP-PLA2 in healthy adults in China
摘要
Lipoprotein-associated phospholipase A2 (LP-PLA2), a vascular-specific inflammatory marker, is associated with coronary heart disease (CHD), a highly prevalent condition among middle-aged and older adults. It explores the relationship between environmental factors in environmentscience, geography, statistics and medicine and the reference value of LP-PLA2 in China. A geographic distribution prediction model for Lp-PLA2 medical reference values was developed using optimized machine learning approaches. The support vector machine (SVM) model employed a polynomial kernel function, with hyperparameters tuned via grid search and cross-validation to determine the optimal classifier configuration. Concurrently, the random forest algorithm was applied to mitigate overfitting and perform regression analysis on the dataset. Further considering the influence of geographical factors on the 100 LP-PLA2 medical reference values in city and county units from 2000 to 2024, using 24 geographical factors as independent variables, to fit 2317 LP-PLA2 medical reference values, and then explore the spatial variation characteristics of LP-PLA2 medical reference values. Further spatial interpolation was visualized on the national map. The medical reference value of LP-PLA2 was significantly correlated with longitude, altitude, annual average wind speed, topsoil clay percentage, topsoil reference capacity, topsoil organic matter content, and total convertible amount of topsoil, respectively. The geospatial distribution map shows that the whole value was high in the east and low in the west, with significant differences in longitude. The geographical characteristics of any given region, when analyzed through either the SVM prediction model or the derived geographical distribution map, enable precise determination of region-specific Lp-PLA2 reference values. This establishes an evidence-based foundation for clinical decision-making and research design in healthcare institutions and medical research organizations.