workbench 0.8.193__py3-none-any.whl → 0.8.197__py3-none-any.whl

workbench/model_scripts/custom_models/proximity/proximity.py

@@ -2,22 +2,13 @@ import pandas as pd
 import numpy as np
 from sklearn.preprocessing import StandardScaler
 from sklearn.neighbors import NearestNeighbors
-from typing import List, Dict, Optional
+from typing import List, Dict, Optional, Union
 import logging
-import pickle
-import json
-from pathlib import Path
-from enum import Enum
 
 # Set up logging
 log = logging.getLogger("workbench")
 
 
-class ProximityType(Enum):
-    DISTANCE = "distance"
-    SIMILARITY = "similarity"
-
-
 class Proximity:
     def __init__(
         self,
@@ -26,7 +17,6 @@ class Proximity:
         features: List[str],
         target: Optional[str] = None,
         track_columns: Optional[List[str]] = None,
-        n_neighbors: int = 10,
     ):
         """
         Initialize the Proximity class.
@@ -37,64 +27,132 @@ class Proximity:
             features: List of feature column names to be used for neighbor computations.
             target: Name of the target column. Defaults to None.
             track_columns: Additional columns to track in results. Defaults to None.
-            n_neighbors: Number of neighbors to compute. Defaults to 10.
         """
         self.id_column = id_column
         self.target = target
         self.track_columns = track_columns or []
-        self.proximity_type = None
-        self.scaler = None
-        self.X = None
-        self.nn = None
 
         # Filter out non-numeric features
         self.features = self._validate_features(df, features)
 
         # Drop NaN rows and set up DataFrame
         self.df = df.dropna(subset=self.features).copy()
-        self.n_neighbors = min(n_neighbors, len(self.df) - 1)
+
+        # Compute target range if target is provided
+        self.target_range = None
+        if self.target and self.target in self.df.columns:
+            self.target_range = self.df[self.target].max() - self.df[self.target].min()
 
         # Build the proximity model
-        self.build_proximity_model()
+        self._build_model()
 
-    def _validate_features(self, df: pd.DataFrame, features: List[str]) -> List[str]:
-        """Remove non-numeric features and log warnings."""
-        non_numeric = df[features].select_dtypes(exclude=["number"]).columns.tolist()
-        if non_numeric:
-            log.warning(f"Non-numeric features {non_numeric} aren't currently supported...")
-            return [f for f in features if f not in non_numeric]
-        return features
+        # Precompute landscape metrics
+        self._precompute_metrics()
 
-    def build_proximity_model(self) -> None:
-        """Standardize features and fit Nearest Neighbors model."""
-        self.proximity_type = ProximityType.DISTANCE
-        self.scaler = StandardScaler()
-        self.X = self.scaler.fit_transform(self.df[self.features])
-        self.nn = NearestNeighbors(n_neighbors=self.n_neighbors + 1).fit(self.X)
+    def isolated(self, top_percent: float = 1.0) -> pd.DataFrame:
+        """
+        Find isolated data points based on distance to nearest neighbor.
+
+        Args:
+            top_percent: Percentage of most isolated data points to return (e.g., 1.0 returns top 1%)
+
+        Returns:
+            DataFrame of observations above the percentile threshold, sorted by distance (descending)
+        """
+        percentile = 100 - top_percent
+        threshold = np.percentile(self.df["nn_distance"], percentile)
+        isolated = self.df[self.df["nn_distance"] >= threshold].copy()
+        return isolated.sort_values("nn_distance", ascending=False).reset_index(drop=True)
 
-    def all_neighbors(self) -> pd.DataFrame:
+    def target_gradients(
+        self,
+        top_percent: float = 1.0,
+        min_delta: Optional[float] = None,
+        k_neighbors: int = 5,
+    ) -> pd.DataFrame:
         """
-        Compute nearest neighbors for all rows in the dataset.
+        Find compounds with steep target gradients (data quality issues and activity cliffs).
+
+        Uses a two-phase approach:
+        1. Quick filter using nearest neighbor gradient
+        2. Verify using k-neighbor median to handle cases where the nearest neighbor is the outlier
+
+        Args:
+            top_percent: Percentage of compounds with steepest gradients to return (e.g., 1.0 = top 1%)
+            min_delta: Minimum absolute target difference to consider. If None, defaults to target_range/100
+            k_neighbors: Number of neighbors to use for median calculation (default: 5)
 
         Returns:
-            DataFrame of neighbors and their distances.
+            DataFrame of compounds with steepest gradients, sorted by gradient (descending)
         """
-        distances, indices = self.nn.kneighbors(self.X)
+        if self.target is None:
+            raise ValueError("Target column must be specified")
+
+        epsilon = 1e-5
+
+        # Phase 1: Quick filter using precomputed nearest neighbor
+        candidates = self.df.copy()
+        candidates["gradient"] = candidates["nn_target_diff"] / (candidates["nn_distance"] + epsilon)
 
-        results = [
-            self._build_neighbor_result(
-                query_id=self.df.iloc[i][self.id_column], neighbor_idx=neighbor_idx, distance=dist
+        # Apply min_delta
+        if min_delta is None:
+            min_delta = self.target_range / 100.0 if self.target_range > 0 else 0.0
+        candidates = candidates[candidates["nn_target_diff"] >= min_delta]
+
+        # Get top X% by initial gradient
+        percentile = 100 - top_percent
+        threshold = np.percentile(candidates["gradient"], percentile)
+        candidates = candidates[candidates["gradient"] >= threshold].copy()
+
+        # Phase 2: Verify with k-neighbor median to filter out cases where nearest neighbor is the outlier
+        results = []
+        for _, row in candidates.iterrows():
+            cmpd_id = row[self.id_column]
+            cmpd_target = row[self.target]
+
+            # Get k nearest neighbors (excluding self)
+            nbrs = self.neighbors(cmpd_id, n_neighbors=k_neighbors, include_self=False)
+
+            # Calculate median target of k nearest neighbors
+            neighbor_median = nbrs.head(k_neighbors)[self.target].median()
+            median_diff = abs(cmpd_target - neighbor_median)
+
+            # Only keep if compound differs from neighborhood median
+            # This filters out cases where the nearest neighbor is the outlier
+            if median_diff >= min_delta:
+                mean_distance = nbrs.head(k_neighbors)["distance"].mean()
+
+                results.append(
+                    {
+                        self.id_column: cmpd_id,
+                        self.target: cmpd_target,
+                        "neighbor_median": neighbor_median,
+                        "neighbor_median_diff": median_diff,
+                        "mean_distance": mean_distance,
+                        "gradient": median_diff / (mean_distance + epsilon),
+                    }
+                )
+
+        # Handle empty results
+        if not results:
+            return pd.DataFrame(
+                columns=[
+                    self.id_column,
+                    self.target,
+                    "neighbor_median",
+                    "neighbor_median_diff",
+                    "mean_distance",
+                    "gradient",
+                ]
             )
-            for i, (dists, nbrs) in enumerate(zip(distances, indices))
-            for neighbor_idx, dist in zip(nbrs, dists)
-            if neighbor_idx != i  # Skip self
-        ]
 
-        return pd.DataFrame(results)
+        results_df = pd.DataFrame(results)
+        results_df = results_df.sort_values("gradient", ascending=False).reset_index(drop=True)
+        return results_df
 
     def neighbors(
         self,
-        id_or_ids,
+        id_or_ids: Union[str, int, List[Union[str, int]]],
         n_neighbors: Optional[int] = 5,
         radius: Optional[float] = None,
         include_self: bool = True,
@@ -104,9 +162,9 @@ class Proximity:
 
         Args:
             id_or_ids: Single ID or list of IDs to look up
-            n_neighbors: Number of neighbors to return (default: 5)
+            n_neighbors: Number of neighbors to return (default: 5, ignored if radius is set)
             radius: If provided, find all neighbors within this radius
-            include_self: Whether to include self in results (if present)
+            include_self: Whether to include self in results (default: True)
 
         Returns:
             DataFrame containing neighbors and distances
@@ -123,38 +181,6 @@ class Proximity:
         query_df = self.df[self.df[self.id_column].isin(ids)]
         query_df = query_df.set_index(self.id_column).loc[ids].reset_index()
 
-        # Use the core implementation
-        return self.find_neighbors(query_df, n_neighbors=n_neighbors, radius=radius, include_self=include_self)
-
-    def find_neighbors(
-        self,
-        query_df: pd.DataFrame,
-        n_neighbors: Optional[int] = 5,
-        radius: Optional[float] = None,
-        include_self: bool = True,
-    ) -> pd.DataFrame:
-        """
-        Return neighbors for rows in a query DataFrame.
-
-        Args:
-            query_df: DataFrame containing query points
-            n_neighbors: Number of neighbors to return (default: 5)
-            radius: If provided, find all neighbors within this radius
-            include_self: Whether to include self in results (if present)
-
-        Returns:
-            DataFrame containing neighbors and distances
-        """
-        # Validate features
-        missing = set(self.features) - set(query_df.columns)
-        if missing:
-            raise ValueError(f"Query DataFrame is missing required feature columns: {missing}")
-
-        id_column_present = self.id_column in query_df.columns
-
-        # Handle NaN rows
-        query_df = self._handle_nan_rows(query_df, id_column_present)
-
         # Transform query features
         X_query = self.scaler.transform(query_df[self.features])
 
@@ -167,30 +193,71 @@ class Proximity:
         # Build results
         results = []
         for i, (dists, nbrs) in enumerate(zip(distances, indices)):
-            query_id = query_df.iloc[i][self.id_column] if id_column_present else f"query_{i}"
+            query_id = query_df.iloc[i][self.id_column]
 
             for neighbor_idx, dist in zip(nbrs, dists):
                 neighbor_id = self.df.iloc[neighbor_idx][self.id_column]
 
-                # Skip if neighbor is self and include_self is False
+                # Skip self if requested
                 if not include_self and neighbor_id == query_id:
                     continue
 
                 results.append(self._build_neighbor_result(query_id=query_id, neighbor_idx=neighbor_idx, distance=dist))
 
-        results_df = pd.DataFrame(results).sort_values([self.id_column, "distance"]).reset_index(drop=True)
-        return results_df
+        df_results = pd.DataFrame(results)
+        df_results["is_self"] = df_results["neighbor_id"] == df_results[self.id_column]
+        df_results = df_results.sort_values([self.id_column, "is_self", "distance"], ascending=[True, False, True])
+        return df_results.drop("is_self", axis=1).reset_index(drop=True)
 
-    def _handle_nan_rows(self, query_df: pd.DataFrame, id_column_present: bool) -> pd.DataFrame:
-        """Drop rows with NaN values in feature columns and log warnings."""
-        rows_with_nan = query_df[self.features].isna().any(axis=1)
+    def _validate_features(self, df: pd.DataFrame, features: List[str]) -> List[str]:
+        """Remove non-numeric features and log warnings."""
+        non_numeric = [f for f in features if f not in df.select_dtypes(include=["number"]).columns]
+        if non_numeric:
+            log.warning(f"Non-numeric features {non_numeric} aren't currently supported, excluding them")
+        return [f for f in features if f not in non_numeric]
 
-        if rows_with_nan.any():
-            log.warning(f"Found {rows_with_nan.sum()} rows with NaNs in feature columns:")
-            if id_column_present:
-                log.warning(query_df.loc[rows_with_nan, self.id_column])
+    def _build_model(self) -> None:
+        """Standardize features and fit Nearest Neighbors model."""
+        self.scaler = StandardScaler()
+        X = self.scaler.fit_transform(self.df[self.features])
+        self.nn = NearestNeighbors().fit(X)
 
-        return query_df.dropna(subset=self.features)
+    def _precompute_metrics(self, n_neighbors: int = 10) -> None:
+        """
+        Precompute landscape metrics for all compounds.
+
+        Adds columns to self.df:
+        - nn_distance: Distance to nearest neighbor
+        - nn_id: ID of nearest neighbor
+
+        If target is specified, also adds:
+        - nn_target: Target value of nearest neighbor
+        - nn_target_diff: Absolute difference from nearest neighbor target
+        """
+        log.info("Precomputing proximity metrics...")
+
+        # Make sure n_neighbors isn't greater than dataset size
+        n_neighbors = min(n_neighbors, len(self.df) - 1)
+
+        # Get nearest neighbors for all points (including self)
+        X = self.scaler.transform(self.df[self.features])
+        distances, indices = self.nn.kneighbors(X, n_neighbors=2)  # Just need nearest neighbor
+
+        # Extract nearest neighbor (index 1, since index 0 is self)
+        self.df["nn_distance"] = distances[:, 1]
+        self.df["nn_id"] = self.df.iloc[indices[:, 1]][self.id_column].values
+
+        # If target exists, compute target-based metrics
+        if self.target and self.target in self.df.columns:
+            # Get target values for nearest neighbor
+            nn_target_values = self.df.iloc[indices[:, 1]][self.target].values
+            self.df["nn_target"] = nn_target_values
+            self.df["nn_target_diff"] = np.abs(self.df[self.target].values - nn_target_values)
+
+            # Precompute target range for min_delta default
+            self.target_range = self.df[self.target].max() - self.df[self.target].min()
+
+        log.info("Proximity metrics precomputed successfully")
 
     def _build_neighbor_result(self, query_id, neighbor_idx: int, distance: float) -> Dict:
         """
@@ -204,111 +271,31 @@ class Proximity:
         Returns:
             Dictionary containing neighbor information
         """
-        neighbor_id = self.df.iloc[neighbor_idx][self.id_column]
         neighbor_row = self.df.iloc[neighbor_idx]
+        neighbor_id = neighbor_row[self.id_column]
 
         # Start with basic info
         result = {
             self.id_column: query_id,
             "neighbor_id": neighbor_id,
-            "distance": distance,
+            "distance": 0.0 if distance < 1e-5 else distance,
         }
 
-        # Columns to automatically include if they exist
-        auto_include = (
-            ([self.target, "prediction"] if self.target else [])
-            + self.track_columns
-            + [col for col in self.df.columns if "_proba" in col or "residual" in col or col == "outlier"]
-        )
+        # Add target if present
+        if self.target and self.target in self.df.columns:
+            result[self.target] = neighbor_row[self.target]
 
-        # Add values for existing columns
-        for col in auto_include:
+        # Add tracked columns
+        for col in self.track_columns:
             if col in self.df.columns:
                 result[col] = neighbor_row[col]
 
-        # Truncate very small distances to zero
-        result["distance"] = 0.0 if distance < 1e-7 else distance
-        return result
-
-    def serialize(self, directory: str) -> None:
-        """
-        Serialize the Proximity model to a directory.
-
-        Args:
-            directory: Directory path to save the model components
-        """
-        dir_path = Path(directory)
-        dir_path.mkdir(parents=True, exist_ok=True)
-
-        # Save metadata
-        metadata = {
-            "id_column": self.id_column,
-            "features": self.features,
-            "target": self.target,
-            "track_columns": self.track_columns,
-            "n_neighbors": self.n_neighbors,
-        }
-
-        (dir_path / "metadata.json").write_text(json.dumps(metadata))
-
-        # Save DataFrame
-        self.df.to_pickle(dir_path / "df.pkl")
-
-        # Save models
-        with open(dir_path / "scaler.pkl", "wb") as f:
-            pickle.dump(self.scaler, f)
-
-        with open(dir_path / "nn_model.pkl", "wb") as f:
-            pickle.dump(self.nn, f)
-
-        log.info(f"Proximity model serialized to {directory}")
-
-    @classmethod
-    def deserialize(cls, directory: str) -> "Proximity":
-        """
-        Deserialize a Proximity model from a directory.
-
-        Args:
-            directory: Directory path containing the serialized model components
-
-        Returns:
-            A new Proximity instance
-        """
-        dir_path = Path(directory)
-        if not dir_path.is_dir():
-            raise ValueError(f"Directory {directory} does not exist or is not a directory")
-
-        # Load metadata
-        metadata = json.loads((dir_path / "metadata.json").read_text())
-
-        # Load DataFrame
-        df_path = dir_path / "df.pkl"
-        if not df_path.exists():
-            raise FileNotFoundError(f"DataFrame file not found at {df_path}")
-        df = pd.read_pickle(df_path)
-
-        # Create instance without calling __init__
-        instance = cls.__new__(cls)
-        instance.df = df
-        instance.id_column = metadata["id_column"]
-        instance.features = metadata["features"]
-        instance.target = metadata["target"]
-        instance.track_columns = metadata["track_columns"]
-        instance.n_neighbors = metadata["n_neighbors"]
-
-        # Load models
-        with open(dir_path / "scaler.pkl", "rb") as f:
-            instance.scaler = pickle.load(f)
-
-        with open(dir_path / "nn_model.pkl", "rb") as f:
-            instance.nn = pickle.load(f)
-
-        # Restore X
-        instance.X = instance.scaler.transform(instance.df[instance.features])
-        instance.proximity_type = ProximityType.DISTANCE
+        # Add prediction/probability columns if they exist
+        for col in self.df.columns:
+            if col == "prediction" or "_proba" in col or "residual" in col or col == "in_model":
+                result[col] = neighbor_row[col]
 
-        log.info(f"Proximity model deserialized from {directory}")
-        return instance
+        return result
 
 
 # Testing the Proximity class
@@ -328,28 +315,15 @@ if __name__ == "__main__":
 
     # Test the Proximity class
     features = ["Feature1", "Feature2", "Feature3"]
-    prox = Proximity(df, id_column="ID", features=features, n_neighbors=3)
-    print(prox.all_neighbors())
-
-    # Test the neighbors method
-    print(prox.neighbors(1))
+    prox = Proximity(df, id_column="ID", features=features)
+    print(prox.neighbors(1, n_neighbors=2))
 
     # Test the neighbors method with radius
     print(prox.neighbors(1, radius=2.0))
 
-    # Test with data that isn't in the 'train' dataframe
-    query_data = {
-        "ID": [6],
-        "Feature1": [0.31],
-        "Feature2": [0.31],
-        "Feature3": [2.31],
-    }
-    query_df = pd.DataFrame(query_data)
-    print(prox.find_neighbors(query_df=query_df))  # For new data we use find_neighbors()
-
     # Test with Features list
-    prox = Proximity(df, id_column="ID", features=["Feature1"], n_neighbors=2)
-    print(prox.all_neighbors())
+    prox = Proximity(df, id_column="ID", features=["Feature1"])
+    print(prox.neighbors(1))
 
     # Create a sample DataFrame
     data = {
@@ -367,40 +341,9 @@ if __name__ == "__main__":
         features=["Feature1", "Feature2"],
         target="target",
         track_columns=["Feature1", "Feature2"],
-        n_neighbors=3,
     )
-    print(prox.all_neighbors())
-
-    # Test the neighbors method
     print(prox.neighbors(["a", "b"]))
 
-    # Time neighbors with all IDs versus calling all_neighbors
-    import time
-
-    start_time = time.time()
-    prox_df = prox.find_neighbors(query_df=df, include_self=False)
-    end_time = time.time()
-    print(f"Time taken for neighbors: {end_time - start_time:.4f} seconds")
-    start_time = time.time()
-    prox_df_all = prox.all_neighbors()
-    end_time = time.time()
-    print(f"Time taken for all_neighbors: {end_time - start_time:.4f} seconds")
-
-    # Now compare the two dataframes
-    print("Neighbors DataFrame:")
-    print(prox_df)
-    print("\nAll Neighbors DataFrame:")
-    print(prox_df_all)
-    # Check for any discrepancies
-    if prox_df.equals(prox_df_all):
-        print("The two DataFrames are equal :)")
-    else:
-        print("ERROR: The two DataFrames are not equal!")
-
-    # Test querying without the id_column
-    df_no_id = df.drop(columns=["foo_id"])
-    print(prox.find_neighbors(query_df=df_no_id, include_self=False))
-
     # Test duplicate IDs
     data = {
         "foo_id": ["a", "b", "c", "d", "d"],  # Duplicate ID (d)
@@ -409,17 +352,52 @@ if __name__ == "__main__":
         "target": [1, 0, 1, 0, 5],
     }
     df = pd.DataFrame(data)
-    prox = Proximity(df, id_column="foo_id", features=["Feature1", "Feature2"], target="target", n_neighbors=3)
+    prox = Proximity(df, id_column="foo_id", features=["Feature1", "Feature2"], target="target")
     print(df.equals(prox.df))
 
     # Test with a categorical feature
     from workbench.api import FeatureSet, Model
 
-    fs = FeatureSet("abalone_features")
-    model = Model("abalone-regression")
+    fs = FeatureSet("aqsol_features")
+    model = Model("aqsol-regression")
     features = model.features()
     df = fs.pull_dataframe()
     prox = Proximity(
         df, id_column=fs.id_column, features=model.features(), target=model.target(), track_columns=features
     )
-    print(prox.find_neighbors(query_df=df[0:2]))
+    print(prox.neighbors(df[fs.id_column].tolist()[:3]))
+
+    print("\n" + "=" * 80)
+    print("Testing isolated_compounds...")
+    print("=" * 80)
+
+    # Test isolated data in the top 1%
+    isolated_1pct = prox.isolated(top_percent=1.0)
+    print(f"\nTop 1% most isolated compounds (n={len(isolated_1pct)}):")
+    print(isolated_1pct[[fs.id_column, "nn_distance", "nn_id"]].head(10))
+
+    # Test isolated data in the top 5%
+    isolated_5pct = prox.isolated(top_percent=5.0)
+    print(f"\nTop 5% most isolated compounds (n={len(isolated_5pct)}):")
+    print(isolated_5pct[[fs.id_column, "nn_distance", "nn_id"]].head(10))
+
+    print("\n" + "=" * 80)
+    print("Testing target_gradients...")
+    print("=" * 80)
+
+    # Test with different parameters
+    gradients_1pct = prox.target_gradients(top_percent=1.0, min_delta=1.0)
+    print(f"\nTop 1% target gradients (min_delta=5.0) (n={len(gradients_1pct)}):")
+    print(
+        gradients_1pct[
+            [fs.id_column, model.target(), "neighbor_median", "neighbor_median_diff", "mean_distance", "gradient"]
+        ].head(10)
+    )
+
+    gradients_5pct = prox.target_gradients(top_percent=5.0, min_delta=5.0)
+    print(f"\nTop 5% target gradients (min_delta=5.0) (n={len(gradients_5pct)}):")
+    print(
+        gradients_5pct[
+            [fs.id_column, model.target(), "neighbor_median", "neighbor_median_diff", "mean_distance", "gradient"]
+        ].head(10)
+    )