支持向量机在柴油机尾气分析中的核模型选择

建立包含三个层次54个指标的长沙市“十二五”时期发展指标体系,通过运用长沙市2006~2009年的数据对各指标的得分值进行分析发现,“十二五”时期的发展不能过多注重经济增长方面的指标,而应更多考虑社会民生和区域创新方面的指标,在具体运用时应尽量避免主观赋权。

The Nuclear Model Selection of SVM in the Analysis of Diesel Engine Exhaust Emissions

The effect of ordinary kernels and their parameters of the Support Vector Machine (SVM) on classification are researched. Then, use Cross-Validation (CV) to obtain the optimum parameters of SVM with different ordinary kernel on the diesel engine exhaust emissions data, and 3 performance indexes of SVM, the CV accuracy on training data and the classification accuracy on testing data as well as the parameter optimizing time, under the corresponding optimum parameters respectively. Comparing the same type indexes, the results are, for the diesel engine exhaust emissions data, the RBF kernel model has higher accuracy than other models at the first index, but the lower one at the second; the Linear kernel model takes shorter time on parameter optimizing than other ones. Considering the learning and extrapolating ability as well as the parameter optimizing time, linear kernel is determined to be used in SVM in the analysis of diesel engine exhaust emissions.