关键词: 判别分析 多元数据 定向判别 建模 最优分割

Abstract:

This paper introduces the principle, modeling flowchart, application flowchart and a practical example of a new algorithm called Directional Discrimin ant Analysis （DDA）, which may be used in multivariate sequence data. In DDA, class modeling and data discriminating are separated. The model for multi ivariate sequence data is built by multiple or stepwise discriminant analysis. In applying the model, the initial estimation of the samples＇ classifica  ation should be given according to the knowledge in the application field. Then the program selects the appropriate part of the model to discriminate th e classes of the data. In this way, we can solve the upside down problem of sample classification caused by the periodicity of multivariate time series  data.

Key words: （discriminant analysis, multivariate data, directional discriminant, modeling, best cutting）