fix: add correct naming

NSL/aufgaben/src/uebung1_kmeans_fleet.py

@@ -14,6 +14,7 @@ Umbau aus dem Lösungs-Notebook (lektion_1.ipynb):
 """
 
 import matplotlib
+
 matplotlib.use("Agg")  # headless: nur in Dateien rendern
 
 from pathlib import Path
@@ -29,7 +30,7 @@ from scipy.spatial.distance import cdist, pdist
 # EINSTELLUNGEN / PFADE
 # =========================================================
 
-BASE = Path(__file__).resolve().parent.parent        # .../aufgaben
+BASE = Path(__file__).resolve().parent.parent  # .../aufgaben
 DATA = BASE / "data" / "data_delivery_fleet.tsv"
 PLOTS = BASE / "plots"
 PLOTS.mkdir(exist_ok=True)
@@ -107,8 +108,10 @@ plt.close(fig)
 
 # ---- Within-Cluster Sum of Squares (Inertia) ----
 Ks = range(1, 10)
-inertia = [KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data).inertia_
+inertia = [
-           for k in Ks]
+    KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data).inertia_
+    for k in Ks
+]
 
 fig, ax = plt.subplots()
 ax.bar(Ks, inertia)
@@ -122,12 +125,14 @@ plt.close(fig)
 # ---- Percentage of Variance Explained ----
 # TSS = WCSS + BSS  ->  Variance Explained = BSS / TSS
 ks = range(1, 10)
-models = [KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data) for k in ks]
+models = [
+    KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data) for k in ks
+]
 centroids = [m.cluster_centers_ for m in models]
 
 # Abstand jedes Punkts zum nächsten Zentrum -> WCSS pro k
 dist = [np.min(cdist(data, cent, "euclidean"), axis=1) for cent in centroids]
-wcss = np.array([np.sum(d ** 2) for d in dist])
+wcss = np.array([np.sum(d**2) for d in dist])
 tss = np.sum(pdist(data) ** 2) / data.shape[0]
 bss = tss - wcss
 
@@ -143,11 +148,14 @@ plt.close(fig)
 
 # ---- Silhouette Score (erst ab k >= 2 definiert) ----
 ks_sil = range(2, 10)
-silhouette = [silhouette_score(
+silhouette = [
-                  data,
+    silhouette_score(
-                  KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data).labels_,
+        data,
-                  metric="euclidean")
+        KMeans(n_clusters=k, n_init=10, random_state=RANDOM_SEED).fit(data).labels_,
-              for k in ks_sil]
+        metric="euclidean",
+    )
+    for k in ks_sil
+]
 
 fig, ax = plt.subplots()
 ax.plot(list(ks_sil), silhouette, marker="o")
@@ -166,7 +174,9 @@ print(f"Bestes k nach Silhouette: {best_k}")
 # =========================================================
 
 k_means_final = KMeans(n_clusters=best_k, n_init=10, random_state=RANDOM_SEED).fit(data)
-save_kmeans(data, k_means_final, "07_final.png", title=f"Finales Clustering (k={best_k})")
+save_kmeans(
+    data, k_means_final, "07_final.png", title=f"Finales Clustering (k={best_k})"
+)
 
 print(f"Plots gespeichert in: {PLOTS}")
 
@@ -177,6 +187,7 @@ print(f"Plots gespeichert in: {PLOTS}")
 # Im Lösungs-Notebook enthalten, gilt aber als veraltet (vgl. Chiang & Mirkin).
 # Silhouette/Elbow werden bevorzugt -> standardmässig deaktiviert (RUN_HARTIGAN).
 
+
 def hartigan_k(data, threshold=12):
     """'Korrekte' Clusterzahl nach Hartigan: erstes k mit H <= threshold."""
     inertia_list = np.zeros(len(data) + 1)
@@ -186,7 +197,9 @@ def hartigan_k(data, threshold=12):
         kmn = KMeans(n_clusters=num + 1, n_init=10, random_state=RANDOM_SEED).fit(data)
         inertia_list[num + 1] = kmn.inertia_
         if num > 0:
-            h_rule = ((float(inertia_list[num]) / inertia_list[num + 1]) - 1) * (len(data) - num - 1)
+            h_rule = ((float(inertia_list[num]) / inertia_list[num + 1]) - 1) * (
+                len(data) - num - 1
+            )
             print(f"k={num}, H={h_rule:.2f}")
         num += 1
     if h_rule > threshold: