基于混合属性的产品优化聚类算法

针对分类研究中采用单一类型数据造成的结果失真, 提出了综合考虑产品属性和销售时间序列的两阶段优化聚类算法。分别采用基于属性的相似性排序及时间序列的分层优化聚类实现产品单独聚类, 然后基于初始聚类结果及参数化的动态相对权重提出考虑噪声数据处理的分层聚类方法实现产品综合优化分类。企业实例应用研究表明综合聚类模型及两阶段算法在聚类精度及时间复杂度上具有明显的优势, 相对权重的动态参数化设置有效解决了不同产品间个性化特征的差异表示。通用数据集的仿真进一步验证了算法在解决混合属性产品聚类问题时的优越性及广泛适用性。

Optimization clustering algorithm based on mixed attributes

This paper developed a two-phase optimization clustering algorithm considering both product attributes and time-series, aiming at revising the distortion results caused by using uniform attribute data in product classification. Firstly, it used attribute-based similarity sorting and time series hierarchical optimization to gain products separate clustering, and then based on the initial clustering results and parameterized dynamic relative weights, it proposed a hierarchical clustering algorithm with noise data processing to achieve optimized classification. Enterprise application testifies the superiority of the integrated clustering model and the two-phase algorithm in accuracy and time complexity, and the dynamic parameterized relative weights provide an effective solution to the discrepancy of personalized features between different products. Finally, common data sets simulation further validates the algorithm superiority and applicability in solving mixed-attribute product clustering problem.