【摘要】 集成学习是一种可以有效改善分类系统性能的数据挖掘方法。采用动态分类器集成选择算法对卷烟感官质量进行智能评估。产生包含多个基分类器的分类器池;根据基分类器在被测样本邻域内的表现选择满足要求的分类器;采用被选择的分类器产生最终的预测结果。为了验证该方法的有效性,采用国内某烟草公司提供的卷烟感官评估历史数据集进行了实验比较分析。实验结果表明,与其他方法相比,该方法获得的效果明显改善。更多还原

【Abstract】 Ensemble learning is an effective data mining technique to improve classification performance. Therefore, this paper adopts dynamic classifier ensemble selection algorithm to evaluate the sensory quality of cigarette. We produced a classifier pool with several single classifiers, and selected the classifiers that met the requirement according to the performance on the neighborhood of the test sample. The final prediction result was generated by the selected classifier. In order to verify the effectiveness of our method, we carry out the experimental comparison on a historical dataset of cigarette sensory evaluation from a tobacco company in China. The results indicate that our method acquires significant improvement compared with other approaches.更多还原