基于类别区分度和关联性分析的综合特征选择

提出一种基于类别区分度和关联性分析的综合特征选择算法。利用类别区分度提取具有较强类别区分能力的特征词,降低特征空间的稀疏性,通过特征的关联性分析衡量特征与类别的相关性及特征之间的冗余度,选出具有类别代表性且相互之间不存在冗余的特征词。实验结果表明,该算法能有效提高分类器性能。

Research on Measurable Nonlinear Relationship Between Phytoplankton Biomass and Environmental Factors in Bohai Bay

B ased on the data of phytoplankton and environmental factors in the Bohai Bay, the dependence between the concentration of phytoplankton and environmental factors is analysed by linear correlation coefficient, rank correlation coefficient and Hoeffding test of independence .The result shows that wind-speed, air-pressure, surface temperature, field pH, salinity, DO, silicate and NO3- have a great impact on the concentration of phytoplankton.