COMP9313 week8b Pipeline

Pipeline:  

  1. process and learn from data 

  2. is a sequence of stages, the stage could be either a Transformer or an Estimator

  3. input: DataFrame  output: DataFrame

  4. Transformer: transform() processing data

  5. Estimator: fit()  learn from data and predict

  6. example:  

    WordTokenizer, countVectors, label_stringIdx 作为 Transformer

    nb_model 作为 Estimator 

    建立pipeline

 

 

Stacking:

  1.  base model

  2. meta model: 由 BaseModel的结果为features训练而成

  3. 

  

    

原文地址:https://www.cnblogs.com/ChevisZhang/p/13360092.html