首页|基于客户行为数据的卷烟价值感FCD-KNN聚类评估方法

基于客户行为数据的卷烟价值感FCD-KNN聚类评估方法

扫码查看
卷烟产品特征众多,如果将所有特征都纳入评估范围,不仅会增加评估的复杂度和工作量,还会因信息冗余而影响评估结果的分离性与准确性,对此研究了基于客户行为数据的卷烟价值感FCD-KNN聚类评估方法.首先标准化处理客户行为数据并量化部分非数值型数据,利用FCD方法(频繁项集挖掘,Frequent Combining Diging)对客户行为数据进行挖掘,提取出多个与卷烟价值感相关的特征;然后在KNN聚类算法(K近邻算法,K Nearest Neighbor)的作用下,通过计算特征相似度,基于高相似度选择与价值感关联性强、能够准确反映产品价值的特征,构建卷烟价值感评估体系.实验结果表明,该方法的Davies-Bouldin指数数值较小,而Calinski-Harabasz指数数值较高,评估结果的分离性与准确性得到保证.
FCD-KNN Clustering Evaluation Method for Cigarette Value Perception Based on Customer Behavior Data
There are many features of cigarettes.If all features are included in the evaluation scope,it will not only increase the complexity and workload of the evaluation,but also affect the separability and accuracy of the evaluation results due to information redundancy.Therefore,a FCD-KNN clustering evaluation method for cig-arette value perception based on customer behavior data was studied.Firstly,after standardizing customer be-havior data and quantifing some non numerical data,FCD(Frequent Combining Digiting)method was used to mine customer behavior data and extract multiple features related to cigarette value perception.Then,under KNN(K Nearest Neighbor)clustering algorithm,a cigarette value perception evaluation system was construc-ted by calculating feature similarity and selecting features with strong correlation with value perception and ac-curate reflection of product value based on high similarity.The experimental results show that the Davies-Bouldin index of this method was relatively small,while the Calinski-Harabasz index was relatively high,en-suring the separability and accuracy of the evaluation results.

Customer behavior dataFCD-KNN clustering algorithmCigarette value perceptionSimi-larity features

杨蕾、张涛、詹建波、陶鹰、蒋梦菲、何雪峰

展开 >

云南中烟工业有限责任公司技术中心,云南 昆明 650231

客户行为数据 FCD-KNN聚类算法 卷烟价值感 相似度特征

云南中烟工业有限责任公司科技资助项目

2023CP04

2024

云南师范大学学报(自然科学版)
云南师范大学

云南师范大学学报(自然科学版)

CSTPCD
影响因子:0.54
ISSN:1007-9793
年,卷(期):2024.44(5)