aidlab/AiDA_Python
6
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

TASK:冷启动热度推荐；

Browse Source
This commit is contained in:
shahaibo
2025-06-10 12:02:14 +08:00
parent fe1219a59e
commit 18c95a88b9
3 changed files with 42 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files

41
app/api/api_recommendation.py
View File

@@ -134,21 +134,36 @@ async def get_recommendations(user_id: int, category: str, brand_id: int, brand_
# 合并分数 # 合并分数
if brand_id is not None: if brand_id is not None:
if brand_id is not None: brand_idx_feature = matrix_data["brand_index_map"].get(brand_id)
brand_idx_feature = matrix_data["brand_index_map"].get(brand_id)
if brand_idx_feature is not None and valid_sketch_idxs_feature: brand_feat_valid = (
raw_brand_feat_scores = matrix_data["brand_feature_matrix"][ matrix_data["brand_feature_matrix"].size > 0 and # 矩阵非空
brand_idx_feature, valid_sketch_idxs_feature] brand_idx_feature is not None and
raw_brand_feat_scores = (raw_brand_feat_scores - np.min(raw_brand_feat_scores)) / ( valid_sketch_idxs_feature # 有可用索引
np.max(raw_brand_feat_scores) - np.min(raw_brand_feat_scores) + 1e-8) )
processed_brand_feat = raw_brand_feat_scores
final_scores = processed_inter + 0.3 * ((1 - brand_scale) * processed_feat + brand_scale * processed_brand_feat) if brand_feat_valid:
else: raw_brand_feat_scores = matrix_data["brand_feature_matrix"][
final_scores = processed_inter + 0.3 * processed_feat brand_idx_feature, valid_sketch_idxs_feature
]
raw_brand_feat_scores = (raw_brand_feat_scores - np.min(raw_brand_feat_scores)) / (
np.max(raw_brand_feat_scores) - np.min(raw_brand_feat_scores) + 1e-8
)
processed_brand_feat = raw_brand_feat_scores
# 如果 processed_feat 是空的，替换为全 0，避免 shape 不一致
if processed_feat.size == 0:
processed_feat = np.zeros_like(processed_brand_feat)
final_scores = processed_inter + 0.3 * (
(1 - brand_scale) * processed_feat + brand_scale * processed_brand_feat
)
else: else:
final_scores = processed_inter + 0.3 * processed_feat # brand 信息不可用
final_scores = processed_inter + 0.3 * processed_feat if processed_feat.size > 0 else processed_inter
else: else:
final_scores = processed_inter + 0.3 * processed_feat final_scores = processed_inter + 0.3 * processed_feat if processed_feat.size > 0 else processed_inter
valid_sketch_idxs = matrix_data["cached_valid_idxs"][cache_key] valid_sketch_idxs = matrix_data["cached_valid_idxs"][cache_key]
# 概率采样 # 概率采样

2
app/service/recommend/scheduled_task.py
View File

@@ -527,7 +527,7 @@ def update_heat_matrices():
if __name__ == "__main__": if __name__ == "__main__":
try: try:
# update_user_matrices() update_user_matrices()
update_heat_matrices() update_heat_matrices()
# scheduler = BlockingScheduler() # scheduler = BlockingScheduler()
# scheduler.add_job(update_user_matrices, 'cron', hour=12, timezone='Asia/Shanghai') # scheduler.add_job(update_user_matrices, 'cron', hour=12, timezone='Asia/Shanghai')

15
app/service/recommend/service.py
View File

@@ -55,9 +55,20 @@ def load_resources():
matrix_data["feature_matrix"] = np.load(f"{RECOMMEND_PATH_PREFIX}feature_matrix.npy", allow_pickle=True) matrix_data["feature_matrix"] = np.load(f"{RECOMMEND_PATH_PREFIX}feature_matrix.npy", allow_pickle=True)
matrix_data["brand_feature_matrix"] = np.load(f"{RECOMMEND_PATH_PREFIX}brand_feature_matrix.npy", allow_pickle=True) brand_feature_path = f"{RECOMMEND_PATH_PREFIX}brand_feature_matrix.npy"
if os.path.exists(brand_feature_path):
matrix_data["brand_feature_matrix"] = np.load(brand_feature_path, allow_pickle=True)
else:
logger.warning("brand_feature_matrix 文件不存在，使用空数组")
matrix_data["brand_feature_matrix"] = np.array([])
matrix_data["brand_index_map"] = np.load(f"{RECOMMEND_PATH_PREFIX}brand_index_map.npy",allow_pickle=True).item() # brand_index_map
brand_index_path = f"{RECOMMEND_PATH_PREFIX}brand_index_map.npy"
if os.path.exists(brand_index_path):
matrix_data["brand_index_map"] = np.load(brand_index_path, allow_pickle=True).item()
else:
logger.warning("brand_index_map 文件不存在，使用空字典")
matrix_data["brand_index_map"] = {}
matrix_data["user_index_feature"] = np.load(f"{RECOMMEND_PATH_PREFIX}user_index_feature_matrix.npy", allow_pickle=True).item() matrix_data["user_index_feature"] = np.load(f"{RECOMMEND_PATH_PREFIX}user_index_feature_matrix.npy", allow_pickle=True).item()