铸坯质量预测的混合式特征选择方法

Hybrid Feature Selection Method for Slab Quality Prediction

为了降低铸坯质量预测的错误率,本文以铸坯夹杂缺陷为例,提出了一种混合式特征选择方法。首先,采集某钢厂铸坯生产过程数据,根据冶金机理,得到铸坯夹杂缺陷的影响因素,构造原始特征集;其次,采用过滤式与包装式相结合的混合式特征选择方法对特征进行选择,获取特征子集;最后,通过预测结果的分类错误率来评价所提方法优劣。结果表明,混合式特征选择方法相比单一的特征选择方法有效地降低了预测的分类错误率。将经过混合式特征选择后的特征子集作为预测模型的输入,分类错误率为9.8%,与使用原始特征集相比降低了23.3%,与单独采用过滤式方法相比降低了13.5%,与单独采用包装式方法相比降低了8.2%。

In order to reduce the error rate of slab quality prediction,a hybrid feature selection method is proposed in this paper.Firstly,the data of slab production process in a steel plant is collected,and the influencing factors of slab inclusion defects are obtained according to the metallurgical mechanism,and the original feature set is constructed;secondly,the hybrid feature selection method combining filtering and packaging is used to select features and obtain feature subset;finally,the advantages and disadvantages of the proposed method are evaluated by the classification error rate of the prediction results.The results show that the hybrid feature selection method can effectively reduce the prediction error rate compared with the single feature selection method.When the feature subset after hybrid feature selection is used as the input of the prediction model,the classification error rate is 9.8%,which is 23.3%lower than the original feature set,13.5%lower than the filter method alone,and 8.2%lower than the packaging method alone.