class %ML.AutoML.Provider
extends %ML.Provider
Implements the AutoML provider
• parameter PROVIDERNAME = "AutoML";
Provider name
• property initialized
as %Boolean [ InitialExpression = 0 ];
• method %BeginTraining(model As %ML.Model, data As %SQL.StatementResult, trainingrun As %ML.TrainingRun, ByRef name As %String = "", ByRef trainkey)
as %Status
Train an ML model
name is no longer used. trainingrun.name is already defined
• method %DataFrameToTempFile(tfn As %Integer, df As %SYS.Python, fieldnames As %List, positions As %List, types As %List, isPredict As %List)
as %Status
Update temp file #tfn using the data in DataFrame df
Inputs:
tfn: Temp file number
df: a Python DataFrame
fieldnames=$lb(field1, ...): A $List of strings that indicates names of fields in df that will be added to temp file #tfn
positions=$lb(pos1, ...): A list of integers that indicates the corresponding positions of each df field in temp file #tfn
types=$lb(type1, ...): A list of integers that indicates the corresponding ObjectScript type of each df field in temp file #tfn
isPredict=$lb(predict1, ...): A list of integers that indicates if each df field is predict or probablity. If predict=1, this is predict, otherwise, probability
• classmethod %GetDefaultSettings(ByRef settings As %DynamicObject)
Adds the default settings for AutoML to the settings dynamic object
• classmethod %ImportPackage(pkgname As %String, Output pkg)
as %Status
• method %OnInit()
as %Status
Initialize an ML provider
• method %PredictAll(trainedmodel As %ML.AutoML.TrainedModel, tfn As %Integer, argspos As %List, predpos As %List, probpos As %List, expr As %String = "", mtorder As %List, mtunary As %List)
as %Status
Bulk Predict
• classmethod %ResultSetMetaData(data As %SQL.StatementResult, ByRef info As %RegisteredObject, ByRef columns As %List, ByRef types As %List)
as %Status
Determine the metadata for a result set
• method %ResultSetToDataFrame(data As %SQL.StatementResult, ByRef info As %RegisteredObject, ByRef df As %RegisteredObject, ByRef count As %Integer, predictingColumn As %String)
as %Status
Convert an IRIS result set into a dataframe.
If the label column, predictingColumn, is defined,then rows with missing values in the label column will be excluded from the dataframe.
• method %StartProfiler(options As %String, ByRef profiler As %SYS.Python)
as %Status
Start the Python profiler
• method %StopProfiler(profiler As %SYS.Python, ByRef sortby As %String = "CUMULATIVE", ByRef results As %String)
as %Status
Stop the Python profiler
• method %TSDataFrameToTempFile(tfn As %Integer, df As %SYS.Python, tsheaders As %SYS.Python, datetimecolumn As %String, channelColumns As %List, channelTypes As %List, mtorder As %List, mtunary As %List)
as %Status
Update temp file #tfn using the data in DataFrame df acquired from TimeSeries predictions
Inputs:
tfn: Temp file number
df: a Python DataFrame
headers: IRIS table column names
pcTypes: datetime column name
• method %TempFileToDataFrame(columns As %List, types As %List, tfn As %Integer, argspos As %List, ByRef df As %SYS.Python, ByRef count As %Integer, mtorder As %List, mtunary As %List)
as %Status
Convert an IRIS temp file into Python Pandas DataFrame data
• method %WaitForTraining(ByRef trainkey, trainingrun As %ML.TrainingRun, ByRef trainedmodel As %ML.TrainedModel, timeoutMS As %Integer = -1)
as %Status
Check for training complete
• method pyval2str(pyval)
as %String
Convert a python value to an SQL string
