食品溯源时序数据的函数型聚类分析

Functional Cluster Analysis of Time Series Data in Food Traceability

借助函数型数据分析的方法和思想,在一个一般性的框架下讨论对猪肉链时序数据的聚类问题,提出了函数型聚类分析在食品追溯时序数据分析上的应用方法:把时序数据看成一个完整的关于时间的函数对象,而非个体观测值的简单排列,将离散数据转化为函数数据;用基函数展开系数向量的距离代替原函数之间的距离,减少了大量数值积分,简化了运算。通过对复杂的、时序性强的溯源数据进行函数型聚类分析,把复杂离散的数据聚类为连续的分类信息,使得溯源数据的可用性极大增强,可以为决策者和进一步的分析提供科学依据。

A general scheme of time series data in the food supply chain is proposed. By using functional cluster analysis, the time series data are treated as a complete object of a time function, rather than a simple arrangement of individual observations. In this scheme, the discrete data are transformed into functional data while the distance of the origin function is replaced with the distance of the expansion coefficients＇ vector of the base function. In this way, the system can reduce the large number of numerical integration and simplify the calculation. Experiments indicate that the availability of the traceability data is enhanced significantly after discrete data are clustered into sequential information.