论文部分内容阅读
客户特征提取是整个客户行为分析过程中的重要环节。由于客户特征提取时获得的数据具有多共同特征及大噪声等特点,使得在客户行为分析中进行客户特征提取存在较大误差。采用UCI机器学习数据库中有多个共同特征的数据集分别对典型特征提取算法进行实验对比及分类规则提取结果分析,验证了FC-GMDH算法在特征提取精度和抗干扰方面具有明显的优势,在客户行为分析时取得满意的特征提取效果。
Customer feature extraction is an important part of the entire customer behavior analysis process. Due to the characteristics of multi-common features and big noises, the data obtained when customer features are extracted, there is a big error in extracting customer features in customer behavior analysis. Using the datasets with many common features in UCI machine learning database, the typical feature extraction algorithm is compared with the result of classification rule extraction. The results show that FC-GMDH algorithm has obvious advantages in feature extraction accuracy and anti-interference. Customer behavior analysis obtained satisfactory feature extraction.