TimeSequencePipeline


TimeSequencePipeline

TimeSequencePipeline integrates feature engineering and time sequence model into a data analysis pipeline. You can get an TimeSequencePipeline object after calling TimeSequencePredictor.fit() or load_ts_pipeline.

Methods

load_ts_pipeline

ts_pipeline = load_ts_pipeline(file)

Arguments

describe

Since you may get your ts_pipeline from a saved file which is a result pipeline of TimeSeqencePredictor. You can use describe method to get the initialization info for the TimeSeqencePredictor, including future_seq_len, dt_col, target_col, extra_features_col, drop_missing.

ts_pipeline.describe()

fit

Used for incremental fitting. Note that fit in TimeSequencePipeline doesn`t run in distributed mode.

ts_pipeline.fit(input_df, validation_df=None, mc=False, epoch_num=20)

Arguments

evaluate

ts_pipeline.evaluate(input_df,
                     metrics=["mse"],
                     multioutput=`raw_values`)

Arguments

predict

ts_pipeline.predict(input_df)

Arguments

save

ts_pipeline.save(ppl_file="my.ppl")

Arguments

config_save

ts_pipeline.config_save(config_file=="my.json")

Arguments