Platform For AI:Prophet

Limits

You can use Prophet based only on the computing resources of MaxCompute.

Configure the component in the PAI console

• Input ports

  Input port (left-to-right)	Data type	Recommended upstream component	Required
  Input Data	N/A	Read Table Read CSV File Feature engineering Data preprocessing	Yes

• Component parameters

  Tab	Parameter	Description
  Field Setting	valueCol	The data type is STRING, and the data format is MTable. You can use the MATBEL aggregation component to construct the data. You can use the data column of the Datetime type as the time series. Example of MTable: {"data":{"ds":["2019-05-07 00:00:00.0","2019-05-08 00:00:00.0"],"val":[8588.0,8521.0]},"schema":"ds TIMESTAMP,val DOUBLE"}.
  	reservedCols	The columns that are reserved for the algorithm.
  Parameter Setting	predictionCol	The name of the prediction result column.
  	cap	The upper limit of the predicted value.
  	changepoint_prior_scale	Default value: 0.05.
  	change_point_range	The proportion of trend change points. Default value: 0.8.
  	changepoints	The list of change points. Separate multiple change points with commas (,). Example: 2021-05-02,2021-05-07.
  	daily_seasonality	Specifies whether to fit seasonality by day. Default value: auto.
  	floor	The lower limit of the predicted value.
  	growth	The type of the trend. Valid values: LINEAR (default value) Logistic Flat
  	holidays	Separate multiple holidays with spaces. Example: playoff:2021-05-03,2021-01-03 superbowl:2021-02-07,2021-11-02.
  	holidays_prior_scale	Holiday model parameters. Default value: 10.0.
  	include_history	Specifies whether to predict the value that corresponds to the date in the original data.
  	interval_width	The uncertainty interval. Default value: 0.8.
  	mcmc_samples	The number of samples used for Bayesian inference. The value of this parameter is an integer. If this parameter is set to 0, the maximum a posteriori probability (MAP) estimation is performed. The default estimate value is 100.
  	n_change_point	The maximum number of change points. Default value: 25.
  	predictNum	Valid values: (0, inf). Default value: 12.
  	predictionDetailCol	The name of the prediction details column.
  	seasonality_mode	The seasonality mode. Valid values: ADDITIVE (default value) MULTIPLICATIVE
  	seasonality_prior_scale	The parameter of the seasonality model. Default value: 10.0.
  	stanInit	The initial value. This parameter is empty by default.
  	uncertaintySamples	Default value: 1000. Samples are used to calculate statistical metrics. If you do not need to calculate statistical metrics and you want to accelerate the prediction, set this parameter to 0.
  	weekly_seasonality	Specifies whether to fit seasonality by week. Default value: auto.
  	yearly_seasonality	Specifies whether to fit seasonality by year. Default value: auto.
  	numThreads	The number of threads of the component.
  Execution Tuning	Number of Workers	The number of cores. This parameter must be used together with the Memory per worker, unit MB parameter. The value of this parameter must be a positive integer. Valid values: [1,9999].
  	Memory per worker, unit MB	The memory size of each worker node. Valid values: 1024 to 64 × 1024. Unit: MB.

Configure the component by using code

You can copy the following code to the PyAlink Script component to allow the PyAlink Script component to function in the same manner as the Prophet component.

import time, datetime
import numpy as np
import pandas as pd

downloader = AlinkGlobalConfiguration.getPluginDownloader()
downloader.downloadPlugin('tf115_python_env_linux')

data = pd.DataFrame([
			[1,  datetime.datetime.fromtimestamp(1), 10.0],
			[1,  datetime.datetime.fromtimestamp(2), 11.0],
			[1,  datetime.datetime.fromtimestamp(3), 12.0],
			[1,  datetime.datetime.fromtimestamp(4), 13.0],
			[1,  datetime.datetime.fromtimestamp(5), 14.0],
			[1,  datetime.datetime.fromtimestamp(6), 15.0],
			[1,  datetime.datetime.fromtimestamp(7), 16.0],
			[1,  datetime.datetime.fromtimestamp(8), 17.0],
			[1,  datetime.datetime.fromtimestamp(9), 18.0],
			[1,  datetime.datetime.fromtimestamp(10), 19.0]
])

source = dataframeToOperator(data, schemaStr='id int, ts timestamp, val double', op_type='batch')

source.link(GroupByBatchOp()
			.setGroupByPredicate("id")
			.setSelectClause("id, mtable_agg(ts, val) as data")
		).link(ProphetBatchOp()
			.setValueCol("data")
			.setPredictNum(4)
			.setPredictionCol("pred")
		).link(FlattenMTableBatchOp()
					.setSelectedCol("pred_detail")
          .setSchemaStr("ds timestamp, yhat double")
    ).print()

References

• You can use the MTable Expander component to expand a MTable to a table. For more information, see MTable Expander.

• For information about Machine Learning Designer components, see Overview of Machine Learning Designer.

• Machine Learning Designer provides various preset algorithm components. You can select a component for data processing based on your business scenarios. For more information, see Component reference: overview of all components.