Feat/target feat analysis

dataprep/eda/create_diff_report/__init__.py

@@ -22,6 +22,7 @@
 
 def create_diff_report(
     df_list: Union[List[pd.DataFrame], Dict[str, pd.DataFrame]],
+    target: Optional[str] = None,
     config: Optional[Dict[str, Any]] = None,
     display: Optional[List[str]] = None,
     title: Optional[str] = "DataPrep Report",
@@ -35,6 +36,8 @@ def create_diff_report(
     ----------
     df_list
         The DataFrames for which data are calculated.
+    target
+        Target feature to be compared against all other columns.
     config
         A dictionary for configuring the visualizations
         E.g. config={"hist.bins": 20}
@@ -63,7 +66,7 @@ def create_diff_report(
     _suppress_warnings()
     cfg = Config.from_dict(display, config)
 
-    components = format_diff_report(df_list, cfg, mode, progress)
+    components = format_diff_report(df_list, cfg, mode, progress, target)
 
     dict_stats = defaultdict(list)
 
@@ -82,16 +85,6 @@ def create_diff_report(
         "legend_labels": components["legend_lables"],
     }
 
-    # {% for div in value.plots[1] %}
-    #             <div class="vp-plot">
-    #                 {{ div }}
-    #                 {% if key in context.components.dfs[1].variables %}
-    #                 {{ context.components.dfs[1].variables[key].plots[1][loop.index0] }}
-    #                 {% endif %}
-    #             </div>
-
-    # return context
-
     template_base = ENV_LOADER.get_template("base.html")
     report = template_base.render(context=context, zip=zip)
     return Report(report)

dataprep/eda/create_diff_report/diff_formatter.py

@@ -68,6 +68,7 @@ def format_diff_report(
     cfg: Config,
     mode: Optional[str],
     progress: bool = True,
+    target: Optional[str] = None,
 ) -> Dict[str, Any]:
     """
     Format the data and figures needed by create_diff_report
@@ -83,6 +84,8 @@ def format_diff_report(
         Currently only the 'basic' is fully implemented.
     progress
         Whether to show the progress bar.
+    target
+        Target feature to be compared against all other columns.
 
     Returns
     -------
@@ -110,20 +113,44 @@ def format_diff_report(
         if mode == "basic":
             # note: we need the type ignore comment for mypy otherwise it complains because
             # it doesn't realize that we converted df_list to a list if it's a dictionary
-            report = format_basic(df_list, cfg)  # type: ignore
+            if target:
+                validate_target(target, df_list)
+            report = format_basic(df_list, target, cfg)  # type: ignore
         else:
             raise ValueError(f"Unknown mode: {mode}")
     return report
 
 
-def format_basic(df_list: List[pd.DataFrame], cfg: Config) -> Dict[str, Any]:
+def validate_target(target: str, df_list: List[pd.DataFrame]):
+    """
+    Helper function, verify that target column exists
+
+    Parameters
+    ----------
+    target
+        Target feature to be compared against all other columns.
+    df_list
+        The Dataframe for which data are calculated.
+    """
+    exists = False
+    for df in df_list:
+        if target in df.columns:
+            exists = True
+        break
+    if not exists:
+        raise ValueError(f"Sorry, {target} is not a valid column")
+
+
+def format_basic(df_list: List[pd.DataFrame], target: Optional[str], cfg: Config) -> Dict[str, Any]:
     """
     Format basic version.
 
     Parameters
     ----------
     df_list
         The DataFrames for which data are calculated.
+    target
+        Target feature to be compared against all other columns.
     cfg
         The config dict user passed in. E.g. config =  {"hist.bins": 20}
         Without user's specifications, the default is "auto"
@@ -158,7 +185,7 @@ def format_basic(df_list: List[pd.DataFrame], cfg: Config) -> Dict[str, Any]:
             # data = dask.compute(data)
             delayed_results.append(data)
 
-    res_plots = dask.delayed(_format_plots)(cfg=cfg, df_list=df_list)
+    res_plots = dask.delayed(_format_plots)(cfg=cfg, df_list=df_list, target=target)
 
     dask_results["df_computations"] = delayed_results
     dask_results["plots"] = res_plots
@@ -211,7 +238,7 @@ def basic_computations(df: EDAFrame, cfg: Config) -> Dict[str, Any]:
 
 
 def compute_plot_data(
-    df_list: List[dd.DataFrame], cfg: Config, dtype: Optional[DTypeDef]
+    pd_list: List[pd.DataFrame], cfg: Config, dtype: Optional[DTypeDef], target: Optional[str]
 ) -> Intermediate:
     """
     Compute function for create_diff_report's plots
@@ -226,9 +253,15 @@ def compute_plot_data(
         E.g.  dtype = {"a": Continuous, "b": "Nominal"} or
         dtype = {"a": Continuous(), "b": "nominal"}
         or dtype = Continuous() or dtype = "Continuous" or dtype = Continuous()
+    target
+        Target feature to be compared against all other columns.
     """
     # pylint: disable=too-many-branches, too-many-locals
 
+    df_list = list(map(to_dask, pd_list))
+    for i, _ in enumerate(df_list):
+        df_list[i].columns = df_list[i].columns.astype(str)
+
     dfs = Dfs(df_list)
     dfs_cols = dfs.columns.apply("to_list").data
 
@@ -249,6 +282,8 @@ def compute_plot_data(
             col_dtype = col_dtype[0]
 
         orig = [src for src, seq in labeled_cols.items() if col in seq]
+        if col == target and not is_dtype(col_dtype, Continuous_v1()):
+            raise ValueError("Sorry, target must be a numerical feature.")
 
         if is_dtype(col_dtype, Continuous_v1()):
             data.append((col, Continuous_v1(), diff_cont_calcs(srs.apply("dropna"), cfg), orig))
@@ -277,7 +312,9 @@ def compute_plot_data(
         elif is_dtype(dtp, DateTime_v1()):
             plot_data.append((col, dtp, dask.compute(*datum), orig))  # workaround
 
-    return Intermediate(data=plot_data, stats=stats, visual_type="comparison_grid")
+    return Intermediate(
+        data=plot_data, stats=stats, visual_type="comparison_grid", target=target, df_list=pd_list
+    )
 
 
 def _compute_variables(df: EDAFrame, cfg: Config) -> Dict[str, Any]:
@@ -407,14 +444,11 @@ def _format_variables(df: EDAFrame, cfg: Config, data: Dict[str, Any]) -> Dict[s
 
 
 def _format_plots(
-    df_list: Union[List[pd.DataFrame], Dict[str, pd.DataFrame]], cfg: Config
+    df_list: Union[List[pd.DataFrame], Dict[str, pd.DataFrame]], cfg: Config, target: Optional[str]
 ) -> Dict[str, Any]:
     """Formatting of plots section"""
-    df_list = list(map(to_dask, df_list))
-    for i, _ in enumerate(df_list):
-        df_list[i].columns = df_list[i].columns.astype(str)
 
-    itmdt = compute_plot_data(df_list=df_list, cfg=cfg, dtype=None)
+    itmdt = compute_plot_data(pd_list=df_list, cfg=cfg, dtype=None, target=target)
     return render_diff(itmdt, cfg=cfg)