论文部分内容阅读
神经网络和统计分析所构建的分类器均为复杂算式,难以体现专业知识;而分类规则直接以属性值为条件,确定个体类别,易于专业分析。对于连续属性的样本数据,本文应用基于信息熵的Chi-merge方法将其离散化,并将提取最优规则转换为组合优化问题,进而采用遗传算法求解。为此,本文将规则提取演绎为种群进化,并设计了个体适应度函数。由此提取出最优的分类规则,经过修剪处理后,与判别准则一起构成模式分类器。本文将其应用于橄榄油产地判别,所建立的分类器简单明了,规则数少,性能良好,适用于化学模式分类。
The classifiers constructed by neural network and statistical analysis are complex algorithms and difficult to reflect the professional knowledge. The classification rules directly determine the individual category according to the attribute value and are easy to be professionally analyzed. For the sample data with continuous attributes, this paper uses Chi-merge method based on information entropy to discretize it, and transforms the optimal rule into combinatorial optimization problem, and then uses genetic algorithm to solve it. For this reason, this article deduces rule extraction as population evolution, and designs individual fitness function. Therefore, the optimal classification rules are extracted, and after the trimming process, the classification rules are formed together with the criteria. In this paper, it is applied to the identification of olive oil producing areas. The established classifier is simple and clear, with few rules and good performance, suitable for chemical pattern classification.