scala - DataFrames for Apache Spark MLLib online real-time single-item prediction, seriously? -
having trained mllib model, have convert every incoming item in online system dataframe call predict/transform-like method on it? looks overkill high risk of time/memory-related issues.
can build/transform/predict 1 row somehow, or that? (solution in scala preferable)
i've found text authors they've implemented of own this: https://www.oreilly.com/ideas/training-and-serving-nlp-models-using-spark-mllib
Comments
Post a Comment