带特征染色体的遗传算法与支持向量机结合进行生物水质预警的研究

近年来水污染事件频发, 给人类饮用水安全带来巨大隐患, 而生物监测法能够从生物学角度对水质状况做出综合分析. 选取斑马鱼作为受试生物, 当水质发生变化时, 鱼群在毒性物质作用下产生应激反应, 此时运用图像处理技术量化鱼类群体行为参数; 并用带特征染色体的遗传算法与支持向量机相融合的方法进行特征选择和SVM参数优化, 分析影响水质预警的关键量化参数. 结果显示, 新方法与网格搜索法、不带特征染色体的遗传算法与SVM结合的方法相比较, 有更高的准确率.

Implementation of GA-based Feature Subset Selection Combined with SVM for Biological Water Quality Monitoring

In recent years, the water pollution incidents have been reported more frequently than ever before, imposing hazards to the drinking water. Bio-monitoring method can be used to perform a comprehensive check and analysis of water quality. In this article, Danio rerio is selected as a testing organism, the fish responding to stress under the action of toxic substances is observed. Image processing technique is used to quantify the fish population parameters. The support vector machine (SVM) and genetic algorithm with characteristic chromosome are applied for feature selection and SVM parameter optimization simultaneously, analyzing the key quantitative parameters issuing the water quality warning. The method reported in this work achieves higher accuracy than both Grid algorithm and the genetic algorithm without characteristic chromosome