배포 v1.2.3

배포 v1.2.3

Author: skqorrla

.github/workflows/Deploy-ML-EC2.yml

@@ -36,7 +36,11 @@ jobs:
           cd MLOps
           
           # Docker 컨테이너 재시작 (docker-compose가 있다면)
-          docker compose down -v || true
-          docker compose up -d --build
+          if [ -f docker-compose.yml ]; then
+            docker compose down
+            docker compose rm -f
+            docker rmi $(docker images -q) -f
+            docker compose up -d --build
+          fi
           
           echo "Deployment completed successfully!"

MLOps/app/routers/recommendation.py

@@ -15,11 +15,17 @@
 async def recommend_places(userid: str, query: str):
     """Place ID 리스트 기반 추천 API"""
     try:
-        top_3_place_ids, other_places_data = await deepctr_service.rank_places_by_ctr(userid, query)
+        top_places_data, other_places_data = await deepctr_service.rank_places_by_ctr(userid, query)
 
         # 스키마에 맞게 데이터 형식 변환
-        recommend_places = [{"id": place_id} for place_id in top_3_place_ids]
-        normal_places = [{"id": place['place_id']} for place in other_places_data]
+        recommend_places = [
+            {"id": p.get("id"), "category": p.get("category"), "subcategory": p.get("subcategory")}
+            for p in top_places_data
+        ]
+        normal_places = [
+            {"id": p.get("id"), "category": p.get("category"), "subcategory": p.get("subcategory")}
+            for p in other_places_data
+        ]
 
         return ReccomendResponse(
             success=True,

MLOps/app/schema/recommendation_schema.py

@@ -14,7 +14,29 @@ class ReccomendResponse(BaseModel):
         ...,
         description="추천 장소 목록",
         example={
-            "recommend": [{"id": "place_id_1"}, {"id": "place_id_2"}, {"id": "place_id_3"}],
-            "normal": [{"id": "place_id_4"}, {"id": "place_id_5"}]
+            "recommend": [
+                {
+                    "id": "place_id_1",
+                    "category": "관광지",
+                    "subcategory": "자연관광지"
+                },
+                {
+                    "id": "place_id_2", 
+                    "category": "음식점&카페",
+                    "subcategory": "한식"
+                }
+            ],
+            "normal": [
+                {
+                    "id": "place_id_3",
+                    "category": "문화시설",
+                    "subcategory": "박물관"
+                },
+                {
+                    "id": "place_id_4",
+                    "category": "쇼핑",
+                    "subcategory": "전문매장/상가"
+                }
+            ]
         }
     )

MLOps/app/services/deepctr_service.py

@@ -26,7 +26,7 @@ async def rank_places_by_ctr(self, userid: str, query: str):
         places_df = pd.DataFrame(places_list)
 
         # DataFrame의 컬럼 이름 변경
-        places_df = places_df.rename(columns={'uuid': 'place_id', 'name': 'place_name'})
+        places_df = places_df.rename(columns={'uuid': 'id', 'name': 'place_name'})
 
         # 사용자 정보를 DataFrame의 모든 행에 추가
         if not user_info_df.empty:
@@ -39,19 +39,21 @@ async def rank_places_by_ctr(self, userid: str, query: str):
             places_df['gender'] = 1 # 남성을 기본값으로 가정
             places_df['like_list'] = "[1,2,3,4,5]"
 
-        input_data = places_df
-                
+        # 원본 데이터에서 응답에 필요한 컬럼 보존
+        output_columns = ['id', 'category', 'subcategory']
+        preserved_df = places_df[output_columns]
+
         # 모델 예측
-        predictions = self.model.predict(input_data)
+        predictions = self.model.predict(places_df)
 
-        # input data에 예측값 추가
-        input_data['ctr'] = predictions
+        # 예측 결과를 원본 데이터에 추가
+        preserved_df['ctr'] = predictions
 
         # 예측값 기준 정렬
-        input_data = input_data.sort_values(by='ctr', ascending=False)
+        sorted_df = preserved_df.sort_values(by='ctr', ascending=False)
 
         # 상위 3개 장소 반환, 나머지도 반환
-        top_places = input_data.head(3)['place_id'].tolist()
-        other_places = input_data.iloc[3:].to_dict('records')
+        top_places = sorted_df.head(3).to_dict('records')
+        other_places = sorted_df.iloc[3:].to_dict('records')
         return top_places, other_places