Analysis of Measure Fluctuation Based on Adtributor Algorithm
Zhao Yanjie, Duan Xiaochen, Kang Lin · 2020
In order to be able to quickly and easily discover the root-cause of data when fluctuation occurs, this paper introduces an Adtributor algorithm based on the explanatory power and surprise value, and it is selected from the forecast value compared with, the calculation of non-mutually exclusive dimension and the selection of algorithm threshold three aspects to optimize Adtributor algorithm. Through two comparative experiments, it's proved that the calculation of non-mutually exclusive dimensions uses the sum of the measure corresponding to each element and the algorithm threshold compute by the cardinality of dimensions. Finally, the practical application of Adtributor algorithm in data platform is introduced.