Add multiseries support to graph_prediction_vs_actual_over_time

evalml/model_understanding/visualizations.py

@@ -446,6 +446,7 @@ def graph_prediction_vs_actual_over_time(
  dates,
  )
 
+ fig = None
  if is_multiseries(pipeline.problem_type):
  id_list = (
  [single_series] if single_series is not None else data["series_id"].unique()
@@ -481,8 +482,6 @@ def graph_prediction_vs_actual_over_time(
  fig.update_yaxes(title_text=y.name)
  if single_series is not None:
  fig.update_layout(
- height=600,
- width=1000,
  title_text=f"Graph for Series {single_series}",
  )
  else:
@@ -491,32 +490,32 @@ def graph_prediction_vs_actual_over_time(
  width=1500,
  title_text="Graph for Multiseries",
  )
- return fig
-
- data = [
- _go.Scatter(
- x=data["dates"],
- y=data["target"],
- mode="lines+markers",
- name="Target",
- line=dict(color="#1f77b4"),
- ),
- _go.Scatter(
- x=data["dates"],
- y=data["prediction"],
- mode="lines+markers",
- name="Prediction",
- line=dict(color="#d62728"),
- ),
- ]
- # Let plotly pick the best date format.
- layout = _go.Layout(
- title={"text": "Prediction vs Target over time"},
- xaxis={"title": "Time"},
- yaxis={"title": "Target Values and Predictions"},
- )
+ else:
+ data = [
+ _go.Scatter(
+ x=data["dates"],
+ y=data["target"],
+ mode="lines+markers",
+ name="Target",
+ line=dict(color="#1f77b4"),
+ ),
+ _go.Scatter(
+ x=data["dates"],
+ y=data["prediction"],
+ mode="lines+markers",
+ name="Prediction",
+ line=dict(color="#d62728"),
+ ),
+ ]
+ # Let plotly pick the best date format.
+ layout = _go.Layout(
+ title={"text": "Prediction vs Target over time"},
+ xaxis={"title": "Time"},
+ yaxis={"title": "Target Values and Predictions"},
+ )
 
- return _go.Figure(data=data, layout=layout)
+ fig = _go.Figure(data=data, layout=layout)
+ return fig
 
 
 def get_linear_coefficients(estimator, features=None):

evalml/tests/model_understanding_tests/test_visualizations.py

@@ -472,57 +472,13 @@ class NotTSPipeline:
  )
 
 
-def test_graph_prediction_vs_actual_over_time_multiseries_single(
- multiseries_ts_data_stacked,
- go,
- component_graph_multiseries,
- pipeline_parameters_multiseries,
-):
- X, y = multiseries_ts_data_stacked
- X_train, _, y_train, _ = split_multiseries_data(
- X,
- y,
- "series_id",
- "date",
- )
- pipeline = MultiseriesRegressionPipeline(
- component_graph_multiseries,
- pipeline_parameters_multiseries,
- )
- pipeline.fit(X_train, y_train)
- fig = graph_prediction_vs_actual_over_time(
- pipeline,
- X,
- y,
- X_train,
- y_train,
- X["date"],
- "1",
- )
- assert isinstance(fig, go.Figure)
- fig_dict = fig.to_dict()
-
- assert fig_dict["layout"]["title"]["text"] == "Graph for Series 1"
- assert fig_dict["layout"]["xaxis"]["title"]["text"] == "Time"
- assert fig_dict["layout"]["yaxis"]["title"]["text"] == "target"
- assert len(fig_dict["data"]) == 2
-
- assert len(fig_dict["data"][0]["x"]) == len(X["date"].unique())
- assert len(fig_dict["data"][0]["y"]) == len(X["date"].unique())
- assert not np.isnan(fig_dict["data"][0]["y"]).all()
- assert fig_dict["data"][0]["name"] == "Series 1: Target"
-
- assert len(fig_dict["data"][1]["x"]) == len(X["date"].unique())
- assert len(fig_dict["data"][1]["y"]) == len(X["date"].unique())
- assert not np.isnan(fig_dict["data"][1]["y"]).all()
- assert fig_dict["data"][1]["name"] == "Series 1: Prediction"
-
-
+@pytest.mark.parametrize("single_series", ["0", None])
 def test_graph_prediction_vs_actual_over_time_multiseries(
  multiseries_ts_data_stacked,
  go,
  component_graph_multiseries,
  pipeline_parameters_multiseries,
+ single_series,
 ):
  X, y = multiseries_ts_data_stacked
  X_train, _, y_train, _ = split_multiseries_data(
@@ -543,17 +499,23 @@ def test_graph_prediction_vs_actual_over_time_multiseries(
  X_train,
  y_train,
  X["date"],
+ single_series=single_series,
  )
  assert isinstance(fig, go.Figure)
 
  fig_dict = fig.to_dict()
- assert fig_dict["layout"]["title"]["text"] == "Graph for Multiseries"
+
+ if single_series is not None:
+ assert fig_dict["layout"]["title"]["text"] == "Graph for Series 0"
+ assert len(fig_dict["data"]) == 2
+ else:
+ assert fig_dict["layout"]["title"]["text"] == "Graph for Multiseries"
+ # there's 5 series, and each series has two lines (one each for target/prediction)
+ assert len(fig_dict["data"]) == 10
+
  assert fig_dict["layout"]["xaxis"]["title"]["text"] == "Time"
  assert fig_dict["layout"]["yaxis"]["title"]["text"] == "target"
 
- # there's 5 series, and each series has two lines (one each for target/prediction)
- assert len(fig_dict["data"]) == 10
-
  curr_series = 0
  for i in range(len(fig_dict["data"])):
  assert len(fig_dict["data"][i]["x"]) == len(X["date"].unique())

evalml/utils/__init__.py

@@ -41,3 +41,4 @@
  _determine_downcast_type,
  _get_incompatible_nullable_types,
 )
+from evalml.utils.lazy_loader import LazyLoader