Delete Max Silohuette and Show Cluster Density

app.py

@@ -617,36 +617,58 @@ def compute_global_regression(df_combined, embedding_cols, tsne_params, df_f1, r
 
         silhouette = np.max(silhouette_vals)
     
-    # inertias = []
-    # K = range(1, 20)
+    inertias = []
+    K = range(1, 20)
 
-    # for k in K:
-    #     kmeans = KMeans(n_clusters=k, random_state=42)
-    #     kmeans.fit(reduced_real)
-    #     inertias.append(kmeans.inertia_)
+    for k in K:
+        kmeans = KMeans(n_clusters=k, random_state=42)
+        kmeans.fit(reduced_real)
+        inertias.append(kmeans.inertia_)
 
 
-    # kl = KneeLocator(K, inertias, curve="convex", direction="decreasing")
-    # elbow_k = kl.elbow
+    kl = KneeLocator(K, inertias, curve="convex", direction="decreasing")
+    elbow_k = kl.elbow
 
 
-    silhouettes_test = []
-    K = range(2, 20)
+    # silhouettes_test = []
+    # K = range(2, 20)
 
-    for k in K:
-        kmeans = KMeans(n_clusters=k, random_state=42, n_init=10)
-        labels = kmeans.fit_predict(X)
-        sil = silhouette_score(X, labels)
-        silhouettes_test.append(sil)
+    # for k in K:
+    #     kmeans = KMeans(n_clusters=k, random_state=42, n_init=10)
+    #     labels = kmeans.fit_predict(X)
+    #     sil = silhouette_score(X, labels)
+    #     silhouettes_test.append(sil)
 
-    inertias = silhouettes_test
-    best_k = K[np.argmax(silhouettes_test)]
-    elbow_k = best_k
+    # inertias = silhouettes_test
+    # best_k = K[np.argmax(silhouettes_test)]
+    # elbow_k = best_k
 
     kmeans_opt = KMeans(n_clusters=elbow_k, random_state=42, n_init=10)
     labels_opt = kmeans_opt.fit_predict(X)
     silhouette_opt = silhouette_score(X, labels_opt)
 
+    centers = kmeans_opt.cluster_centers_
+    radii, densities, volumes = [], [], []
+
+    for i in range(elbow_k):
+        cluster_points = reduced_real[labels_opt == i]
+        n_points = len(cluster_points)
+        
+        # Distancias euclídeas al centroide
+        dists = np.linalg.norm(cluster_points - centers[i], axis=1)
+        
+        # Radio máximo (engloba todo el cluster)
+        r = dists.max()
+        radii.append(r)
+        
+        # Volumen de la esfera
+        V = (4/3) * np.pi * (r ** 3)
+        volumes.append(V)
+        
+        # Densidad = número de puntos / volumen
+        density = n_points / V if V > 0 else np.nan
+        densities.append(density)
+
 
     dfs_reduced, unique_subsets = split_versions(df_combined, reduced)
     
@@ -727,6 +749,9 @@ def compute_global_regression(df_combined, embedding_cols, tsne_params, df_f1, r
         "silhouette_opt": silhouette_opt,
         "elbow_k": elbow_k,
         "classes_k": labels_opt,
+        "centers_k": centers,
+        "radii_k": radii,
+        "densities_k": densities,
     }
     
     if reduction_method == "PCA":
@@ -888,6 +913,19 @@ def run_model(model_name):
     st.write(f"Silhouette Score: {result['silhouette_opt']:.2f}")
     st.write(f"Optimal number of clusters (k) from Elbow Method: {result['elbow_k']:.2f}")
 
+    if "radii_k" in result and "densities_k" in result:
+        st.subheader("Cluster Geometry (Radius & Density)")
+
+        df_clusters = pd.DataFrame({
+            "Cluster": np.arange(len(result["radii_k"])),
+            "Radius": np.round(result["radii_k"], 4),
+            "Density": np.round(result["densities_k"], 6)
+        })
+
+        # Mostrar tabla
+        st.dataframe(df_clusters, use_container_width=True)
+        # st.table(df_clusters)
+
     # # Mostrar los plots de loadings si se usó PCA (para el conjunto combinado)
     # if reduction_method == "PCA" and result.get("pca_model") is not None:
     #     # pca_model = result["pca_model"]