C4.5决策树算法的阈值自适应色谱峰研究与实现

Study and Implementation of Threshold Adaptive Chromatographic Peak based on C4.5 Decision Tree Algorithm

变压器油色谱在线监测装置在长期运行过程中,存在峰位置偏移情况。针对这一现象,将C4.5决策树算法引入色谱峰定性领域。将峰高、峰宽、峰面积和峰中点位置作为色谱峰辨识的特征属性。利用决策树对于根结点选取时,采用二分法对连续属性进行离散化处理,得到特征属性的自适应阈值,达到数据正确分类的效果。利用色谱峰测试样本集Q 1至Q 6的6组数据对该算法进行测试,准确率达到95.23%,可有效避免因峰位的前后移动、峰形的扩展收缩导致的对气体色谱峰的误判和漏判现象。

In the long running process of on-line chromatographic monitoring device for transformer oil,the peak position was shifted.In view of this phenomenon,the C4.5 decision tree algorithm was introduced into chromatographic peak qualitative field.Taking peak height,peak width,peak area and midpoint position as the characteristic attributes of chromatographic peak identification,the dichotomy was used to discretize continuous attributes when selecting root nodes by decision tree,so as to obtain the adaptive threshold of characteristic attributes and achieve the result of correct data classification.The algorithm was tested with 6 groups of data from sample set Q1 to Q6 of chromatographic peak test and the accuracy rate reached to 95.23%.The results show that the algorithm is effective in peak qualitative accuracy,which effectively avoids misjudgment and missing judgment of gas chromatographic peak caused by forward and backward movement of peak position and expansion and contraction of peak shape.