I've added a detaile error message, maybe it makes sense for someone:
DataSource.Error: An error happened while reading data from the provider: '{"error":{"code":"UnknownError","pbi.error":{"code":"UnknownError","parameters":{},"details":[]}}}'
section Section1;
shared #"Online Visitor" = let
Source = Csv.Document(Web.Contents("https://raw.githubusercontent.com/santoshc1/PowerBI-AI-samples/master/Tutorial_AutomatedML/online_shoppers_intention.csv"), [Delimiter = ",", Columns = 18, QuoteStyle = QuoteStyle.None]),
#"Promoted headers" = Table.PromoteHeaders(Source, [PromoteAllScalars = true]),
#"Changed column type" = Table.TransformColumnTypes(#"Promoted headers", {{"Administrative_Duration", type number}, {"Informational_Duration", type number}, {"ProductRelated_Duration", type number}, {"BounceRates", type number}, {"ExitRates", type number}, {"PageValues", type number}, {"SpecialDay", type number}, {"Weekend", type logical}, {"Revenue", type logical}})
in
#"Changed column type";
shared #"Purchase Intent Prediction Training Data" = let
Source = #"Online Visitor",
#"Selected columns" = Table.SelectColumns(Source, {"Administrative", "Administrative_Duration", "Informational", "Informational_Duration", "ProductRelated", "ProductRelated_Duration", "BounceRates", "ExitRates", "PageValues", "SpecialDay", "Month", "OperatingSystems", "Browser", "Region", "TrafficType", "VisitorType", "Weekend", "Revenue"}),
#"Removed nulls" = Table.SelectRows(#"Selected columns", each [Revenue] <> null),
#"Sampled input" = AI.SampleStratifiedWithHoldout("Revenue", Table.RowCount(#"Removed nulls"), #"Removed nulls")
in
#"Sampled input";
shared #"Purchase Intent Prediction" = let
Source = #"Purchase Intent Prediction Training Data",
#"Invoked TrainPrediction" = AIInsights.Contents(){[Key = "AI.Execute"]}[Data]("AI.TrainPrediction", "Regular", [labelColumnName = "Revenue", data = Source]),
#"Selected training schema columns" = Table.SelectColumns(#"Invoked TrainPrediction", {"TrainingId", "Model", "Stats", "GlobalExplanation", "TrainingSchema", "TrainingAUC", "LabelColumn"})
in
#"Selected training schema columns";
shared #"Purchase Intent Prediction.Score" = let
ApplyScoringFunction = (inputQuery as table, newColumn as text, decisionThreshold as number) => let
MlModel = #"Purchase Intent Prediction",
MlModelJson = try Text.FromBinary(Json.FromValue(MlModel{0})) otherwise "InvalidModel",
Source = inputQuery,
SelectedBaseEntityColumns = {"Administrative", "Administrative_Duration", "Informational", "Informational_Duration", "ProductRelated", "ProductRelated_Duration", "BounceRates", "ExitRates", "PageValues", "SpecialDay", "Month", "OperatingSystems", "Browser", "Region", "TrafficType", "VisitorType", "Weekend", "Revenue"},
InputRowCount = Table.RowCount(Source),
InputTableType = Value.Type(Source),
SelectedColumnsTypes = List.Transform(SelectedBaseEntityColumns, each Type.TableColumn(InputTableType, _)),
ScoringFunction =
let
ScoringFunctionScalarType = type function (row as record) as any,
VectorizedScoringFunction = (input as table) =>
let
ExpandedColumns = Table.ExpandRecordColumn(input, "row", SelectedBaseEntityColumns),
ExpandedColumnsWithTypes = Table.TransformColumnTypes(ExpandedColumns, List.Zip({SelectedBaseEntityColumns, SelectedColumnsTypes})),
ErrorList = List.Repeat({[Output = null]}, InputRowCount),
Result = if MlModelJson <> "InvalidModel" then (try Table.ToRecords(AIInsights.Contents(){[Key = "AI.Execute"]}[Data]("AI.ScorePrediction", "Vectorized", [data = ExpandedColumns, scoreParameters = MlModelJson])) otherwise ErrorList) else ErrorList
in
Result,
ScalarVectorScoringFunction = Function.ScalarVector(ScoringFunctionScalarType, VectorizedScoringFunction)
in
ScalarVectorScoringFunction,
AddScoringColumn = Table.AddColumn(Source, newColumn, each ScoringFunction(_)),
ExpandResultColumns = Table.ExpandRecordColumn(AddScoringColumn, newColumn, {"PredictionScore", "PredictionExplanation"}, {Text.Combine({newColumn, "PredictionScore"}, "."), Text.Combine({newColumn, "PredictionExplanation"}, ".")}),
LabeledOutput = Table.AddColumn(ExpandResultColumns, Text.Combine({newColumn, "Outcome"}, "."), each Record.Field(_, Text.Combine({newColumn, "PredictionScore"}, ".")) >= decisionThreshold * 100),
ReplacedErrors = Table.ReplaceErrorValues(LabeledOutput, {{Text.Combine({newColumn, "Outcome"}, "."), null}, {Text.Combine({newColumn, "PredictionScore"}, "."), null}, {Text.Combine({newColumn, "PredictionExplanation"}, "."), null}}),
TransformTypes = Table.TransformColumnTypes(ReplacedErrors, {{Text.Combine({newColumn, "Outcome"}, "."), type logical}, {Text.Combine({newColumn, "PredictionScore"}, "."), type text}, {Text.Combine({newColumn, "PredictionExplanation"}, "."), type text}})
in
TransformTypes
in
ApplyScoringFunction;
shared #"Purchase Intent Prediction_PollingQuery" = let
Source = "536eb714-1488-49b5-9ac6-598e40d1780b"
in
Source;
shared #"Purchase Intent Prediction Testing Data" = let
Source = #"Purchase Intent Prediction Training Data",
#"Filtered rows" = Table.SelectRows(Source, each ([__IsTraining__] = false)),
#"Invoked Scoring" = #"Purchase Intent Prediction.Score"(#"Filtered rows", "Purchase Intent PredictionOutput", 0.5)
in
#"Invoked Scoring";
shared #"Purchase Intent Prediction Preview" = let
Source = PowerBI.Dataflows(),
Workspace = Source{[workspaceId = "94fe567d-5d7c-496c-85c7-30650d1271ce"]}[Data],
Dataflow = Workspace{[dataflowId = "90099937-edac-4f34-8f73-3fca50551482"]}[Data],
Preview = Dataflow{[entity = "Purchase Intent Prediction"]}[Data]
in
Preview;
shared #"Purchase Intent Prediction Testing Data Preview" = let
Source = PowerBI.Dataflows(),
Workspace = Source{[workspaceId="94fe567d-5d7c-496c-85c7-30650d1271ce"]}[Data],
Dataflow = Workspace{[dataflowId="90099937-edac-4f34-8f73-3fca50551482"]}[Data],
Preview = Dataflow{[entity="Purchase Intent Prediction Testing Data"]}[Data]
in
Preview;
