[recom] Add algorithm

recommander/Algo.py

@@ -0,0 +1,145 @@
+import numpy as np
+from sklearn.metrics.pairwise import cosine_similarity
+from scipy.cluster.hierarchy import linkage, fcluster, dendrogram
+import matplotlib.pyplot as plt
+import pandas as pd
+from sklearn.cluster import SpectralClustering
+from sklearn.cluster import KMeans
+
+def create_matches(data_matrix, party_size: int):
+    """
+
+    :param df: ranking for given user (first column of user x user similarity matrix)
+    :param party_size:
+    :return:
+    """
+    # max value of similarity can be 1, we weight it down to 0.7
+    weight_similarity = 0.2
+    weight_mood = 0.4
+    weight_sbl = 0.8
+    weight_stress = 0.6
+
+    pool_size = data_matrix.shape[0]
+
+    data_matrix["ranking"] = data_matrix["ranking"] * weight_similarity
+    data_matrix["mood"] *= weight_mood
+    data_matrix["battery"] *= weight_sbl
+    data_matrix["stress"] *= weight_stress
+
+    numeric_columns = data_matrix.select_dtypes(include=[int, float]).columns
+    data_matrix['row_sum'] = data_matrix[numeric_columns].sum(axis=1)
+    # mean = df['row_sum'].mean()
+    # df['row_sum'] -= mean
+
+    # users_below_zero = df[df['row_sum'] < 0]['user'].tolist()
+    # users_above_or_equal_zero = df[df['row_sum'] >= 0]['user'].tolist()
+    # matches = []
+    all = data_matrix['id'].tolist()
+
+    matches = []
+    while len(all) >= party_size + 2:
+        match = []
+        for _ in range(party_size // 2):
+            user = all.pop(0)
+            match.append(user)
+
+            user = all.pop(len(all)-1)
+            match.append(user)
+        matches.append(match)
+
+
+    if all:
+        matches.append(all)
+    print(matches)
+    for match in matches:
+        # match_id: int = get_match_id()
+        print("match:", match)
+        for user in match:
+            # upload (match_id, user)
+            pass
+
+
+data2 = pd.read_csv('output.csv')
+data = data2.iloc[:, 8:17]
+
+
+
+
+# Step 2: Calculate cosine similarity for each pair of test subjects
+
+similarity_matrix = cosine_similarity(data)
+ranking = data2
+ranking["ranking"] = similarity_matrix[0, :]
+ranking = ranking[["id", "ranking", "mood", "battery", "stress"]]
+create_matches(ranking, 4)
+
+"""
+# Step 3: Define a similarity threshold and filter out similar subjects
+similarity_threshold = 0.6  # Adjust this threshold as needed
+similar_subjects = []
+
+for i in range(data.shape[0]):
+    for j in range(i + 1, data.shape[0]):
+        if similarity_matrix[i, j] >= similarity_threshold:
+            similar_subjects.append((i, j))
+            
+
+df = pd.DataFrame(similarity_matrix)  
+df.to_csv('/Users/lisa/Desktop/simMatrix.csv', index=False)
+num_clusters = 10  # Create 10 groups
+
+# Step 3: Create a K-Means clustering model
+kmeans_model = KMeans(n_clusters=num_clusters, random_state=42)
+
+# Step 4: Perform clustering using the similarity matrix
+labels = kmeans_model.fit_predict(similarity_matrix)
+
+# Create a DataFrame to store cluster assignments
+cluster_data = pd.DataFrame({'Cluster': labels, 'Person': range(len(labels))})
+
+# Group individuals by cluster
+grouped = cluster_data.groupby('Cluster')['Person'].apply(list).reset_index(name='Members')
+
+# Print the table showing clusters and their members
+print(grouped)
+
+
+distance_threshold = 0.9  # Adjust this threshold as needed
+
+# Convert the similarity matrix into a distance matrix
+distance_matrix = 1 - similarity_matrix
+
+# Perform hierarchical clustering
+linkage_matrix = linkage(distance_matrix, method='single')  # You can choose a different linkage method
+
+# Assign individuals to clusters based on the distance threshold
+labels = fcluster(linkage_matrix, t=distance_threshold, criterion='distance')
+
+# Create a DataFrame to store the groups and their members
+group_data = pd.DataFrame({'Group': labels, 'Person': range(len(labels))})
+
+# Print the number of clusters
+num_clusters = group_data['Group'].nunique()
+print(f"Number of Clusters: {num_clusters}")
+
+# Print the clusters and their members
+grouped = group_data.groupby('Group')['Person'].apply(list)
+
+# Print the table showing group numbers and members
+table = pd.DataFrame({'Group Number': grouped.index, 'Members': grouped.values})
+print(table)
+
+# Plot the dendrogram for visualization
+dendrogram(linkage_matrix)
+plt.show()
+
+"""
+
+
+
+
+
+
+
+
+