Update app.py

app.py

@@ -2,6 +2,8 @@ from pymongo import MongoClient
 from transformers import BertTokenizer, BertModel
 import torch
 from torch.nn import Embedding
+import numpy as np
+from sklearn.metrics.pairwise import cosine_similarity
 
 # MongoDB Atlas 연결 설정
 client = MongoClient("mongodb+srv://waseoke:[email protected]/test?retryWrites=true&w=majority&tls=true&tlsAllowInvalidCertificates=true")
@@ -107,4 +109,49 @@ for user_data in all_users:
         )
         print(f"Embedding saved to MongoDB Atlas for user_id {user_data['user_id']}.")
     except ValueError as e:
-        print(f"Skipping user_id {user_data['user_id']} due to error: {e}")
+        print(f"Skipping user_id {user_data['user_id']} due to error: {e}")
+
+# 사용자 맞춤 추천 함수
+def recommend_products_for_user(user_id, top_n=1):
+    try:
+        # MongoDB에서 사용자 임베딩 가져오기
+        user_embedding_data = user_embedding_collection.find_one({"user_id": user_id})
+        if not user_embedding_data:
+            print(f"User ID {user_id} embedding not found.")
+            return []
+
+        user_embedding = np.array(user_embedding_data["embedding"]).reshape(1, -1)
+
+        # 모든 상품 임베딩 가져오기
+        all_product_embeddings = list(product_embedding_collection.find())
+
+        # 상품 ID 및 임베딩 추출
+        product_ids = []
+        product_embeddings = []
+        for product_data in all_product_embeddings:
+            product_ids.append(product_data["product_id"])
+            product_embeddings.append(np.array(product_data["embedding"]))
+
+        product_embeddings = np.array(product_embeddings)
+
+        # Cosine Similarity 계산
+        similarities = cosine_similarity(user_embedding, product_embeddings).flatten()
+
+        # 유사도 정렬 및 상위 N개 선택
+        top_indices = similarities.argsort()[::-1][:top_n]
+        recommended_products = [(product_ids[i], similarities[i]) for i in top_indices]
+
+        print(f"Top {top_n} recommendations for User ID {user_id}:")
+        for product_id, similarity in recommended_products:
+            print(f"Product ID: {product_id}, Similarity: {similarity:.4f}")
+
+        return recommended_products
+
+    except Exception as e:
+        print(f"Error during recommendation for User ID {user_id}: {e}")
+        return []
+
+# 사용자 맞춤 추천 실행
+user_id_to_recommend = 1  # 추천할 사용자 ID
+top_n_recommendations = 1  # 추천 상품 개수
+recommended_products = recommend_products_for_user(user_id_to_recommend, top_n=top_n_recommendations)