hossam 0.4.15__py3-none-any.whl → 0.4.17__py3-none-any.whl

hossam/hs_plot.py

@@ -1,7 +1,8 @@
 # -*- coding: utf-8 -*-
 from __future__ import annotations
 from types import SimpleNamespace
-from typing import Callable
+from typing import Callable, Literal
+from itertools import combinations
 
 # ===================================================================
 import numpy as np
@@ -15,6 +16,7 @@ from pandas import DataFrame
 # ===================================================================
 from scipy.stats import t
 from scipy.spatial import ConvexHull
+from scipy.cluster.hierarchy import dendrogram as sch_dendrogram
 from statsmodels.graphics.gofplots import qqplot as sm_qqplot
 from statsmodels.nonparametric.smoothers_lowess import lowess as sm_lowess
 
@@ -22,7 +24,8 @@ from statsmodels.nonparametric.smoothers_lowess import lowess as sm_lowess
 from statannotations.Annotator import Annotator
 
 # ===================================================================
-from sklearn.cluster._kmeans import KMeans
+from sklearn.cluster import AgglomerativeClustering, KMeans
+from sklearn.decomposition import PCA
 
 from sklearn.metrics import (
     mean_squared_error,
@@ -34,6 +37,8 @@ from sklearn.metrics import (
     silhouette_samples,
 )
 
+from .hs_util import is_2d
+
 # ===================================================================
 config = SimpleNamespace(
     dpi=200,
@@ -749,7 +754,7 @@ def stackplot(
 # 산점도를 그린다
 # ===================================================================
 def scatterplot(
-    df: DataFrame,
+    df: DataFrame | None,
     xname: str,
     yname: str,
     hue=None,
@@ -769,7 +774,7 @@ def scatterplot(
     """산점도를 그린다.
 
     Args:
-        df (DataFrame): 시각화할 데이터.
+        df (DataFrame | None): 시각화할 데이터.
         xname (str): x축 컬럼.
         yname (str): y축 컬럼.
         hue (str|None): 범주 컬럼.
@@ -796,12 +801,12 @@ def scatterplot(
 
     if outline and hue is not None:
         # 군집별 값의 종류별로 반복 수행
-        for c in df[hue].unique():
+        for c in df[hue].unique():  # type: ignore
             if c == -1:
                 continue
 
             # 한 종류만 필터링한 결과에서 두 변수만 선택
-            df_c = df.loc[df[hue] == c, [xname, yname]]
+            df_c = df.loc[df[hue] == c, [xname, yname]] # type: ignore
 
             try:
                 # 외각선 좌표 계산
@@ -2559,7 +2564,7 @@ def distribution_plot(
 
 
 def silhouette_plot(
-    estimator: KMeans,
+    estimator: KMeans | AgglomerativeClustering,
     data: DataFrame,
     title: str | None = None,
     width: int = config.width,
@@ -2574,7 +2579,7 @@ def silhouette_plot(
     군집분석 결과의 실루엣 플롯을 시각화함.
 
     Args:
-        estimator (KMeans): 학습된 KMeans 군집 모델 객체.
+        estimator (KMeans | AgglomerativeClustering): 학습된 KMeans 또는 AgglomerativeClustering 군집 모델 객체.
         data (DataFrame): 군집분석에 사용된 입력 데이터 (n_samples, n_features).
         title (str, optional): 플롯 제목. None이면 자동 생성.
         width (int, optional): 플롯 가로 크기 (inch 단위).
@@ -2605,7 +2610,15 @@ def silhouette_plot(
     y_lower = 10
 
     # 클러스터링 갯수별로 fill_betweenx( )형태의 막대 그래프 표현.
-    for i in range(estimator.n_clusters):  # type: ignore
+    n_clusters: int = 0
+    if hasattr(estimator, "n_clusters") and estimator.n_clusters is not None:   # type: ignore
+        n_clusters = estimator.n_clusters  # type: ignore
+    elif hasattr(estimator, "n_clusters_") and estimator.n_clusters_ is not None:    # type: ignore
+        n_clusters = estimator.n_clusters_  # type: ignore
+    else:
+        n_clusters = len(np.unique(estimator.labels_))  # type: ignore
+
+    for i in range(n_clusters):  # type: ignore
         ith_cluster_sil_values = sil_values[estimator.labels_ == i]  # type: ignore
         ith_cluster_sil_values.sort()  # type: ignore
 
@@ -2626,18 +2639,21 @@ def silhouette_plot(
     ax.set_xlabel("The silhouette coefficient values")  # type: ignore
     ax.set_ylabel("Cluster label")  # type: ignore
     ax.set_xlim([-0.1, 1])  # type: ignore
-    ax.set_ylim([0, len(data) + (estimator.n_clusters + 1) * 10])  # type: ignore
+    ax.set_ylim([0, len(data) + (n_clusters + 1) * 10])  # type: ignore
     ax.set_yticks([])  # type: ignore
     ax.set_xticks([0, 0.2, 0.4, 0.6, 0.8, 1])  # type: ignore
 
     if title is None:
-        title = "Number of Cluster : " + str(estimator.n_clusters) + ", Silhouette Score :" + str(round(sil_avg, 3))  # type: ignore
+        title = "Number of Cluster : " + str(n_clusters) + ", Silhouette Score :" + str(round(sil_avg, 3))  # type: ignore
 
     finalize_plot(ax, callback, outparams, save_path, True, title)  # type: ignore
 
 
+# ===================================================================
+# 군집분석 결과 시각화
+# ===================================================================
 def cluster_plot(
-    estimator: KMeans | None = None,
+    estimator: KMeans | AgglomerativeClustering | None = None,
     data: DataFrame | None = None,
     xname: str | None = None,
     yname: str | None = None,
@@ -2747,8 +2763,11 @@ def cluster_plot(
     )
 
 
+# ===================================================================
+# 군집분석 결과의 실루엣 플롯과 군집 산점도를 한 화면에 함께 시각화
+# ===================================================================
 def visualize_silhouette(
-    estimator: KMeans,
+    estimator: KMeans | AgglomerativeClustering,
     data: DataFrame,
     xname: str | None = None,
     yname: str | None = None,
@@ -2765,7 +2784,7 @@ def visualize_silhouette(
     군집분석 결과의 실루엣 플롯과 군집 산점도를 한 화면에 함께 시각화함.
 
     Args:
-        estimator (KMeans): 학습된 KMeans 군집 모델 객체.
+        estimator (KMeans | AgglomerativeClustering): 학습된 KMeans 또는 AgglomerativeClustering 군집 모델 객체.
         data (DataFrame): 군집분석에 사용된 입력 데이터 (n_samples, n_features).
         xname (str, optional): 산점도 x축에 사용할 컬럼명. None이면 첫 번째 컬럼 사용.
         yname (str, optional): 산점도 y축에 사용할 컬럼명. None이면 두 번째 컬럼 사용.
@@ -2811,3 +2830,211 @@ def visualize_silhouette(
     )
 
     finalize_plot(ax)
+
+
+
+# ===================================================================
+# 덴드로그램 시각화
+# ===================================================================
+def dandrogram(
+    estimator: AgglomerativeClustering,
+    p: int = 30,
+    count_sort: Literal["ascending", "descending", False] = "ascending",
+    title: str | None = None,
+    width: int = config.width,
+    height: int = config.height,
+    dpi: int = config.dpi,
+    save_path: str | None = None,
+    callback: Callable | None = None,
+    ax: Axes | None = None
+) -> None:
+    """덴드로그램 시각화
+
+    Args:
+        estimator (AgglomerativeClustering): 학습된 AgglomerativeClustering 군집 모델 객체.
+        p (int): 덴드로그램에서 표시할 마지막 병합된 군집 수. 기본값 30.
+        count_sort (str): 'ascending' 또는 'descending'으로 병합 순서 정렬.
+        title (str|None): 그래프 제목.
+        palette (str|None): 팔레트 이름.
+        width (int): 캔버스 가로 픽셀.
+        height (int): 캔버스 세로 픽셀.
+        dpi (int): 그림 크기 및 해상도.
+        save_path (str|None): 저장 경로.
+        callback (Callable|None): Axes 후처리 콜백.
+        ax (Axes|None): 외부에서 전달한 Axes. None이면 새로 생성.
+
+    Returns:
+        None
+    """
+    # 덴드로그램을 그리기 위해 linkage 행렬 생성
+    counts = np.zeros(estimator.children_.shape[0]) # type: ignore
+    n_samples = len(estimator.labels_)
+
+    for i, merge in enumerate(estimator.children_): # type: ignore
+        current_count = 0
+        for child_idx in merge:
+            if child_idx < n_samples:
+                current_count += 1  # leaf node
+            else:
+                current_count += counts[child_idx - n_samples]
+        counts[i] = current_count
+
+        linkage_matrix = np.column_stack(
+            [estimator.children_, estimator.distances_, counts]
+        ).astype(float)
+
+    outparams = False
+
+    if ax is None:
+        fig, ax = get_default_ax(width, height, 1, 1, dpi)  # type: ignore
+        outparams = True
+
+
+    sch_dendrogram(
+        linkage_matrix,
+        ax=ax,
+        p=p,
+        truncate_mode="lastp" if p > 0 else None,
+        leaf_rotation=0,
+        leaf_font_size=8,
+        count_sort=count_sort,
+        color_threshold=None,
+        above_threshold_color="grey",
+    )
+
+    finalize_plot(ax, callback, outparams, save_path, True, title)  # type: ignore
+
+
+# ===================================================================
+# PCA 분석 결과에 대한 biplot 시각화
+# ===================================================================
+def pca_plot(
+    estimator: PCA,
+    data: DataFrame,
+    yname: str | None = None,
+    fields: list | tuple | list[list] | tuple[list] | list[tuple] | tuple[tuple] | None = None,
+    hue: str | None = None,
+    palette: str | None = None,
+    width: int = config.width,
+    height: int = config.height,
+    linewidth: float = config.line_width,
+    dpi: int = config.dpi,
+    save_path: str | None = None,
+    callback: Callable | None = None,
+) -> None:
+    """
+    PCA 분석 결과에 대한 biplot 시각화
+
+    Args:
+        estimator (PCA): 학습된 PCA 객체.
+        data (DataFrame): PCA에 사용된 원본 데이터.
+        yname (str | None): 종속변수 컬럼명.
+        fields (list | tuple | list[list] | tuple[list] | list[tuple] | tuple[tuple] | None): 시각화할 독립변수 목록. None이면 자동 탐지.
+        hue (str|None): 집단 구분 컬럼명.
+        palette (str|None): 팔레트 이름.
+        width (int): 캔버스 가로 픽셀.
+        height (int): 캔버스 세로 픽셀.
+        linewidth (float): 선 굵기.
+        dpi (int): 그림 크기 및 해상도.
+        save_path (str|None): 저장 경로.
+        callback (Callable|None): Axes 후처리 콜백.
+
+    Returns:
+        None
+    """
+    df = data.copy()
+    df_columns = df.columns.tolist()
+
+    # 종속변수가 지정되었다면 해당 컬럼 추출
+    yfield = None
+    if yname is not None and yname in data.columns:
+        yfield = df[[yname]].copy()
+        df = df.drop(columns=[yname])
+
+    # PCA 변환 수행
+    #display(df)
+    score = estimator.transform(df)
+    #print(score)
+
+    # 추정기로부터 PCA 결과 데이터 프레임 생성
+    pca_df = DataFrame(
+        data=score,
+        columns=[f"PC{i+1}" for i in range(estimator.n_components_)],
+    )
+    #display(pca_df)
+
+    # 종속변수 컬럼 추가
+    if yfield is not None:
+        pca_df[yname] = yfield
+
+    # 모든 컬럼명에 대한 조합 생성
+    if fields is None:
+        feature_cols = pca_df.columns.tolist()
+        if yname is not None and yname in feature_cols:
+            feature_cols.remove(yname)
+        fields = list(combinations(feature_cols, 2))
+
+    if not is_2d(fields):
+        fields = [fields]   # type: ignore
+
+    components = estimator.components_
+
+    x_index: int = 0
+    y_index: int = 0
+
+    def __callable(ax) -> None:
+        for i in range(n):
+            ax.arrow(
+                0,
+                0,
+                components[x_index, i],
+                components[y_index, i],
+                color="r",
+                head_width=0.007,
+                head_length=0.007,
+                linewidth=linewidth * 0.75,
+                alpha=0.75,
+            )
+            ax.text(
+                components[x_index, i] * 1.15,
+                components[y_index, i] * 1.15,
+                f"{df_columns[i]} ({components[x_index, i]:.2f})",
+                color="b",
+                ha="center",
+                va="center",
+            )
+
+        if callback is not None:
+            callback(ax)
+
+    for field_group in fields:  # type: ignore
+        x_index = int(pca_df.columns.get_loc(field_group[0]))   # type: ignore
+        y_index = int(pca_df.columns.get_loc(field_group[1]))   # type: ignore
+
+        xs = score[:, x_index]
+        ys = score[:, y_index]
+        n = score.shape[1]
+
+        scalex = 1.0 / (xs.max() - xs.min())
+        scaley = 1.0 / (ys.max() - ys.min())
+
+        title = "PCA Biplot"
+        if field_group is not None:
+            title += " - " + ", ".join(field_group)
+
+        scatterplot(
+            df=None,
+            xname=xs * scalex,
+            yname=ys * scaley,
+            hue=pca_df[hue] if hue is not None else None,
+            outline=False,
+            palette=palette,
+            width=width,
+            height=height,
+            linewidth=linewidth,
+            dpi=dpi,
+            save_path=save_path,
+            title=title,
+            callback=__callable,
+        )
+