performance_evaluation.py

import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import AgglomerativeClustering
from sklearn.metrics import silhouette_score
from sklearn.preprocessing import normalize
import umap

synonyms = pd.read_csv('Synonyms.csv')
synonyms.head()

# Load precomputed embeddings
news_embeddings = np.load("news_embeddings.npy")

# Normalize embeddings
normalized_embeddings = normalize(news_embeddings)

# Apply UMAP for dimensionality reduction
reducer = umap.UMAP(n_components=10, random_state=42)
umap_embeddings = reducer.fit_transform(normalized_embeddings)

# List to hold silhouette scores
silhouette_scores = []
cluster_range = range(2, 20)  # You can adjust the range based on your needs

# Iterate through different numbers of clusters
for num_clusters in cluster_range:
    # Perform Agglomerative Clustering
    clustering_model = AgglomerativeClustering(n_clusters=num_clusters, linkage='ward')
    cluster_labels = clustering_model.fit_predict(umap_embeddings)

    # Calculate silhouette score
    silhouette_avg = silhouette_score(umap_embeddings, cluster_labels)
    silhouette_scores.append(silhouette_avg)
    print(f"Number of Clusters: {num_clusters}, Silhouette Score: {silhouette_avg:.4f}")

# Plot silhouette scores for each number of clusters
plt.figure(figsize=(10, 6))
plt.plot(cluster_range, silhouette_scores, marker='o')
plt.title('Silhouette Scores for Different Numbers of Clusters (With UMAP)')
plt.xlabel('Number of Clusters')
plt.ylabel('Silhouette Score')
plt.xticks(cluster_range)  # Show all tick marks
plt.grid()
plt.show()

# Determine the optimal number of clusters
optimal_clusters = cluster_range[np.argmax(silhouette_scores)]
print(f"The optimal number of clusters is: {optimal_clusters}")


# Load precomputed embeddings
news_embeddings = np.load("news_embeddings.npy")

# Normalize embeddings
normalized_embeddings = normalize(news_embeddings)

# Define the number of clusters
num_clusters = 7

# Perform Agglomerative Clustering without UMAP
clustering_model_no_umap = AgglomerativeClustering(n_clusters=num_clusters, linkage='ward')
cluster_labels_no_umap = clustering_model_no_umap.fit_predict(normalized_embeddings)
silhouette_no_umap = silhouette_score(normalized_embeddings, cluster_labels_no_umap)
print(f"Silhouette Score without UMAP: {silhouette_no_umap}")

# Visualization of clustering results (using first two dimensions for a simple scatter plot)
plt.figure(figsize=(10, 7))
plt.scatter(normalized_embeddings[:, 0], normalized_embeddings[:, 1], c=cluster_labels_no_umap, cmap='viridis', alpha=0.6)
plt.title(f'Agglomerative Clustering Visualization (Silhouette Score: {silhouette_no_umap:.4f})')
plt.xlabel('Dimension 1')
plt.ylabel('Dimension 2')
plt.colorbar(label='Cluster Label')
plt.show()

# Perform UMAP and visualize
import umap

# Reduce dimensionality to 10 with UMAP (this step is optional for this part)
reducer = umap.UMAP(n_components=10)
reduced_embeddings = reducer.fit_transform(normalized_embeddings)

# Perform clustering on reduced embeddings
clustering_model_with_umap = AgglomerativeClustering(n_clusters=num_clusters, linkage='ward')
cluster_labels_with_umap = clustering_model_with_umap.fit_predict(reduced_embeddings)
silhouette_with_umap = silhouette_score(reduced_embeddings, cluster_labels_with_umap)
print(f"Silhouette Score after UMAP: {silhouette_with_umap}")

# Scatter plot of UMAP reduced embeddings
plt.figure(figsize=(10, 7))
plt.scatter(reduced_embeddings[:, 0], reduced_embeddings[:, 1], c=cluster_labels_with_umap, cmap='viridis', alpha=0.6)
plt.title(f'Agglomerative Clustering Visualization with UMAP (Silhouette Score: {silhouette_with_umap:.4f})')
plt.xlabel('UMAP Component 1')
plt.ylabel('UMAP Component 2')
plt.colorbar(label='Cluster Label')
plt.show()

# Compare scores
if silhouette_with_umap > silhouette_no_umap:
    print(f"UMAP improved the silhouette score by {silhouette_with_umap - silhouette_no_umap}")
else:
    print(f"UMAP did not improve the silhouette score. Difference: {silhouette_no_umap - silhouette_with_umap}")