# 示例：使用混合推荐策略缓解冷启动问题
from surprise import Dataset, KNNBasic
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import cosine_similarity

def hybrid_recommendation(user_id, photo_features, ratings_data):
    # 协同过滤部分
    trainset = ratings_data.build_full_trainset()
    sim_options = {'name': 'cosine', 'user_based': False}
    algo = KNNBasic(sim_options=sim_options)
    algo.fit(trainset)
    cf_pred = algo.predict(user_id, photo_features['id']).est
    
    # 内容过滤部分
    tfidf = TfidfVectorizer()
    features_matrix = tfidf.fit_transform(photo_features['description'])
    user_profile = get_user_profile(user_id)  # 获取用户偏好向量
    content_sim = cosine_similarity(user_profile, features_matrix)[0]
    
    # 混合推荐
    hybrid_score = 0.7 * cf_pred + 0.3 * content_sim
    return hybrid_score

# 示例：加入时间衰减因子
import numpy as np
from datetime import datetime

def time_aware_recommendation(photo, current_date):
    # 照片拍摄日期
    photo_date = datetime.strptime(photo['date'], '%Y-%m-%d')
    
    # 当前日期与照片日期的天数差
    days_diff = (current_date - photo_date.date()).days
    
    # 时间衰减因子 (半衰期设为365天)
    time_factor = np.exp(-np.log(2) * days_diff / 365)
    
    # 结合内容相关性
    content_score = calculate_content_score(photo)
    
    return content_score * time_factor

# 示例：使用MMR(Maximal Marginal Relevance)算法平衡相关性与多样性
def mmr_recommendation(photos, query_embedding, lambda_param=0.5):
    selected = []
    remaining = photos.copy()
    
    while remaining:
        # 计算相关性分数
        rel_scores = [cosine_similarity(query_embedding, p['embedding']) for p in remaining]
        
        # 计算多样性分数
        if selected:
            div_scores = [max(cosine_similarity(p['embedding'], s['embedding']) for s in selected) for p in remaining]
        else:
            div_scores = [0] * len(remaining)
            
        # MMR分数
        mmr_scores = [lambda_param * rel - (1 - lambda_param) * div for rel, div in zip(rel_scores, div_scores)]
        
        # 选择最高分
        best_idx = np.argmax(mmr_scores)
        selected.append(remaining.pop(best_idx))
    
    return selected