1017 lines
40 KiB
Python
1017 lines
40 KiB
Python
"""
|
||
推荐引擎模块 - 基于基座架构
|
||
结合机器学习和AI分析的混合推荐系统
|
||
"""
|
||
|
||
from typing import Dict, List, Optional, Any, Tuple
|
||
import json
|
||
import numpy as np
|
||
from sklearn.feature_extraction.text import TfidfVectorizer
|
||
from sklearn.metrics.pairwise import cosine_similarity
|
||
from sklearn.cluster import KMeans
|
||
from sklearn.preprocessing import StandardScaler
|
||
import joblib
|
||
from datetime import datetime, timedelta
|
||
from core.base import BaseModule, ModuleType, UserData, AnalysisResult, BaseConfig
|
||
from pathlib import Path
|
||
import logging
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
|
||
class RecommendationEngine(BaseModule):
|
||
"""推荐引擎模块"""
|
||
|
||
def __init__(self, config: BaseConfig):
|
||
super().__init__(config, ModuleType.RECOMMENDATION)
|
||
self.model_path = Path(config.model_path)
|
||
self.model_path.mkdir(parents=True, exist_ok=True)
|
||
|
||
# 推荐模型组件
|
||
self.tfidf_vectorizer = None
|
||
self.user_clustering_model = None
|
||
self.food_similarity_matrix = None
|
||
self.user_preference_model = None
|
||
|
||
# 食物数据库
|
||
self.food_database = self._load_food_database()
|
||
|
||
# 餐食搭配模板
|
||
self.meal_templates = self._load_meal_templates()
|
||
|
||
# 推荐配置
|
||
self.max_recommendations = config.max_recommendations
|
||
self.min_training_samples = config.min_training_samples
|
||
self.model_update_threshold = config.model_update_threshold
|
||
|
||
def initialize(self) -> bool:
|
||
"""初始化推荐引擎"""
|
||
try:
|
||
self.logger.info("推荐引擎初始化中...")
|
||
|
||
# 加载或训练模型
|
||
self._load_or_train_models()
|
||
|
||
self.is_initialized = True
|
||
self.logger.info("推荐引擎初始化完成")
|
||
return True
|
||
except Exception as e:
|
||
self.logger.error(f"推荐引擎初始化失败: {e}")
|
||
return False
|
||
|
||
def process(self, input_data: Any, user_data: UserData) -> AnalysisResult:
|
||
"""处理推荐请求"""
|
||
try:
|
||
recommendation_type = input_data.get('type', 'meal_recommendation')
|
||
|
||
if recommendation_type == 'meal_recommendation':
|
||
result = self._generate_meal_recommendations(input_data, user_data)
|
||
elif recommendation_type == 'food_similarity':
|
||
result = self._find_similar_foods(input_data, user_data)
|
||
elif recommendation_type == 'preference_update':
|
||
result = self._update_user_preferences(input_data, user_data)
|
||
elif recommendation_type == 'model_retrain':
|
||
result = self._retrain_models(input_data, user_data)
|
||
else:
|
||
result = self._create_error_result("未知的推荐类型")
|
||
|
||
return AnalysisResult(
|
||
module_type=self.module_type,
|
||
user_id=user_data.user_id,
|
||
input_data=input_data,
|
||
result=result,
|
||
confidence=result.get('confidence', 0.5)
|
||
)
|
||
except Exception as e:
|
||
self.logger.error(f"处理推荐请求失败: {e}")
|
||
return self._create_error_result(str(e))
|
||
|
||
def _generate_meal_recommendations(self, input_data: Dict, user_data: UserData) -> Dict[str, Any]:
|
||
"""生成餐食推荐"""
|
||
meal_type = input_data.get('meal_type', 'lunch')
|
||
preferences = input_data.get('preferences', {})
|
||
context = input_data.get('context', {})
|
||
|
||
try:
|
||
# 生成完整餐食搭配推荐
|
||
meal_combinations = self._generate_meal_combinations(user_data, meal_type, preferences, context)
|
||
|
||
return {
|
||
'success': True,
|
||
'recommendations': meal_combinations,
|
||
'reasoning': self._generate_meal_reasoning(meal_combinations, user_data, meal_type),
|
||
'confidence': self._calculate_recommendation_confidence(user_data),
|
||
'metadata': {
|
||
'meal_type': meal_type,
|
||
'combination_count': len(meal_combinations)
|
||
}
|
||
}
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"生成餐食推荐失败: {e}")
|
||
return self._create_error_result(f"推荐生成失败: {str(e)}")
|
||
|
||
def _generate_meal_combinations(self, user_data: UserData, meal_type: str,
|
||
preferences: Dict, context: Dict) -> List[Dict[str, Any]]:
|
||
"""生成基于用户数据的动态餐食搭配组合"""
|
||
combinations = []
|
||
|
||
try:
|
||
# 1. 基于用户历史数据生成搭配
|
||
historical_combinations = self._generate_historical_combinations(user_data, meal_type)
|
||
combinations.extend(historical_combinations)
|
||
|
||
# 2. 基于用户偏好生成个性化搭配
|
||
personalized_combinations = self._generate_personalized_combinations(user_data, meal_type, preferences)
|
||
combinations.extend(personalized_combinations)
|
||
|
||
# 3. 基于相似用户生成搭配
|
||
similar_user_combinations = self._generate_similar_user_combinations(user_data, meal_type)
|
||
combinations.extend(similar_user_combinations)
|
||
|
||
# 4. 如果没有足够数据，使用模板生成
|
||
if len(combinations) < 3:
|
||
template_combinations = self._generate_template_combinations(user_data, meal_type)
|
||
combinations.extend(template_combinations)
|
||
|
||
# 5. 去重和排序
|
||
combinations = self._deduplicate_and_rank_combinations(combinations, user_data)
|
||
|
||
# 6. 确保至少有一些推荐
|
||
if not combinations:
|
||
combinations = self._generate_fallback_combinations(meal_type)
|
||
|
||
return combinations[:5] # 返回前5个最佳搭配
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"生成餐食搭配失败: {e}")
|
||
# 返回基础推荐
|
||
return self._generate_fallback_combinations(meal_type)
|
||
|
||
def _generate_historical_combinations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于用户历史数据生成搭配"""
|
||
combinations = []
|
||
|
||
# 获取该餐次的所有历史记录
|
||
meal_records = [meal for meal in user_data.meals if meal.get('meal_type') == meal_type]
|
||
|
||
if not meal_records:
|
||
return combinations
|
||
|
||
# 分析高频搭配
|
||
food_combinations = {}
|
||
for meal in meal_records:
|
||
foods = meal.get('foods', [])
|
||
if len(foods) >= 2 and meal.get('satisfaction_score', 0) >= 4:
|
||
# 生成2-3食物组合
|
||
for i in range(len(foods)):
|
||
for j in range(i+1, min(i+3, len(foods))):
|
||
combo = tuple(sorted(foods[i:j+1]))
|
||
food_combinations[combo] = food_combinations.get(combo, 0) + 1
|
||
|
||
# 选择出现频率最高的搭配
|
||
sorted_combinations = sorted(food_combinations.items(), key=lambda x: x[1], reverse=True)
|
||
|
||
for combo, count in sorted_combinations[:2]: # 取前2个高频搭配
|
||
if count >= 2: # 至少出现2次
|
||
combination = {
|
||
"name": f"历史搭配{len(combinations)+1}",
|
||
"description": f"基于您{meal_type}的历史偏好",
|
||
"foods": [],
|
||
"total_calories": 0,
|
||
"nutrition_score": 0,
|
||
"personalization_score": 1.0,
|
||
"categories": ["历史数据"],
|
||
"source": "historical"
|
||
}
|
||
|
||
for food_name in combo:
|
||
food_info = self._get_food_info(food_name)
|
||
if food_info:
|
||
combination["foods"].append(food_info)
|
||
combination["total_calories"] += food_info.get("calories", 0)
|
||
|
||
if len(combination["foods"]) >= 2:
|
||
combination["nutrition_score"] = self._calculate_nutrition_score(combination["foods"])
|
||
combinations.append(combination)
|
||
|
||
return combinations
|
||
|
||
def _generate_personalized_combinations(self, user_data: UserData, meal_type: str, preferences: Dict) -> List[Dict[str, Any]]:
|
||
"""基于用户偏好生成个性化搭配"""
|
||
combinations = []
|
||
|
||
# 获取用户喜爱的食物
|
||
favorite_foods = self._get_user_favorite_foods(user_data, meal_type)
|
||
|
||
if len(favorite_foods) >= 2:
|
||
combination = {
|
||
"name": "个性化搭配",
|
||
"description": f"基于您的{meal_type}偏好定制",
|
||
"foods": [],
|
||
"total_calories": 0,
|
||
"nutrition_score": 0,
|
||
"personalization_score": 0.9,
|
||
"categories": ["个性化"],
|
||
"source": "personalized"
|
||
}
|
||
|
||
# 选择2-4种用户喜爱的食物
|
||
selected_foods = favorite_foods[:min(4, len(favorite_foods))]
|
||
for food in selected_foods:
|
||
food_info = self._get_food_info(food)
|
||
if food_info:
|
||
combination["foods"].append(food_info)
|
||
combination["total_calories"] += food_info.get("calories", 0)
|
||
|
||
if len(combination["foods"]) >= 2:
|
||
combination["nutrition_score"] = self._calculate_nutrition_score(combination["foods"])
|
||
combinations.append(combination)
|
||
|
||
return combinations
|
||
|
||
def _generate_similar_user_combinations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于相似用户生成搭配"""
|
||
return [] # 暂时返回空列表
|
||
|
||
def _generate_template_combinations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于模板生成搭配（当历史数据不足时）"""
|
||
combinations = []
|
||
|
||
templates = self.meal_templates.get(meal_type, [])
|
||
|
||
for template in templates[:2]: # 只使用前2个模板
|
||
combination = {
|
||
"name": template["name"],
|
||
"description": f"营养均衡的{meal_type}搭配",
|
||
"foods": [],
|
||
"total_calories": 0,
|
||
"nutrition_score": 0,
|
||
"personalization_score": 0.3,
|
||
"categories": template["categories"],
|
||
"source": "template"
|
||
}
|
||
|
||
# 为每个类别选择常见食物
|
||
for category in template["categories"]:
|
||
food = self._select_common_food_for_category(category, meal_type)
|
||
if food:
|
||
combination["foods"].append(food)
|
||
combination["total_calories"] += food.get("calories", 0)
|
||
|
||
if len(combination["foods"]) >= 2:
|
||
combination["nutrition_score"] = self._calculate_nutrition_score(combination["foods"])
|
||
combinations.append(combination)
|
||
|
||
return combinations
|
||
|
||
def _select_common_food_for_category(self, category: str, meal_type: str) -> Dict[str, Any]:
|
||
"""为特定类别选择常见食物"""
|
||
common_foods = {
|
||
"主食": ["米饭", "面条", "面包", "粥"],
|
||
"蛋白质": ["鸡蛋", "鸡肉", "牛肉", "豆腐"],
|
||
"蔬菜": ["青菜", "白菜", "西红柿"],
|
||
"饮品": ["牛奶", "酸奶", "汤"],
|
||
"水果": ["苹果", "香蕉"],
|
||
"坚果": ["坚果"],
|
||
"小食": ["饼干"]
|
||
}
|
||
|
||
foods = common_foods.get(category, [])
|
||
if foods:
|
||
# 根据餐次选择合适的主食
|
||
if category == "主食" and meal_type == "breakfast":
|
||
food_name = "面包" if "面包" in foods else foods[0]
|
||
elif category == "主食" and meal_type == "lunch":
|
||
food_name = "米饭" if "米饭" in foods else foods[0]
|
||
elif category == "主食" and meal_type == "dinner":
|
||
food_name = "粥" if "粥" in foods else foods[0]
|
||
else:
|
||
food_name = foods[0]
|
||
|
||
return self._get_food_info(food_name)
|
||
|
||
return None
|
||
|
||
def _deduplicate_and_rank_combinations(self, combinations: List[Dict], user_data: UserData) -> List[Dict]:
|
||
"""去重和排序搭配组合"""
|
||
# 去重：基于食物组合去重
|
||
unique_combinations = []
|
||
seen_combinations = set()
|
||
|
||
for combo in combinations:
|
||
food_names = tuple(sorted([f["name"] for f in combo["foods"]]))
|
||
if food_names not in seen_combinations:
|
||
seen_combinations.add(food_names)
|
||
unique_combinations.append(combo)
|
||
|
||
# 排序：综合得分
|
||
def score_combination(combo):
|
||
personal_score = combo.get("personalization_score", 0) * 0.4
|
||
nutrition_score = combo.get("nutrition_score", 0) * 0.3
|
||
satisfaction_score = sum(f.get("satisfaction_score", 4) for f in combo["foods"]) / len(combo["foods"]) * 0.3
|
||
return personal_score + nutrition_score + satisfaction_score
|
||
|
||
return sorted(unique_combinations, key=score_combination, reverse=True)
|
||
|
||
def _generate_meal_reasoning(self, combinations: List[Dict], user_data: UserData, meal_type: str) -> str:
|
||
"""生成动态餐食推荐理由"""
|
||
if not combinations:
|
||
return "暂无推荐"
|
||
|
||
top_combination = combinations[0]
|
||
foods = top_combination["foods"]
|
||
|
||
reasoning_parts = []
|
||
|
||
# 基于数据来源生成理由
|
||
source = top_combination.get("source", "unknown")
|
||
if source == "historical":
|
||
reasoning_parts.append("基于您的历史用餐偏好")
|
||
elif source == "personalized":
|
||
reasoning_parts.append("基于您的个人喜好")
|
||
elif source == "template":
|
||
reasoning_parts.append("营养均衡搭配")
|
||
|
||
# 满意度理由
|
||
avg_satisfaction = sum(f.get("satisfaction_score", 4) for f in foods) / len(foods)
|
||
if avg_satisfaction > 4.0:
|
||
reasoning_parts.append(f"历史满意度: {avg_satisfaction:.1f}分")
|
||
|
||
# 食物搭配理由
|
||
food_names = [f["name"] for f in foods]
|
||
reasoning_parts.append(f"推荐搭配: {', '.join(food_names)}")
|
||
|
||
return "，".join(reasoning_parts) if reasoning_parts else "营养搭配推荐"
|
||
|
||
def _get_user_favorite_foods(self, user_data: UserData, meal_type: str) -> List[str]:
|
||
"""获取用户喜爱的食物"""
|
||
favorite_foods = []
|
||
|
||
# 从历史餐食记录中提取
|
||
for meal in user_data.meals:
|
||
if meal.get('meal_type') == meal_type and meal.get('satisfaction_score', 0) >= 4:
|
||
favorite_foods.extend(meal.get('foods', []))
|
||
|
||
# 统计频率
|
||
food_counts = {}
|
||
for food in favorite_foods:
|
||
food_counts[food] = food_counts.get(food, 0) + 1
|
||
|
||
# 按频率排序
|
||
sorted_foods = sorted(food_counts.items(), key=lambda x: x[1], reverse=True)
|
||
return [food for food, count in sorted_foods[:10]]
|
||
|
||
def _get_food_info(self, food_name: str) -> Dict[str, Any]:
|
||
"""获取食物信息"""
|
||
# 从食物数据库获取信息
|
||
food_info = self.food_database.get(food_name, {})
|
||
return {
|
||
"name": food_name,
|
||
"calories": food_info.get("calories", 100),
|
||
"category": food_info.get("category", "其他"),
|
||
"nutrition": food_info.get("nutrition", {}),
|
||
"satisfaction_score": food_info.get("avg_satisfaction", 4.0)
|
||
}
|
||
|
||
def _calculate_nutrition_score(self, foods: List[Dict]) -> float:
|
||
"""计算营养得分"""
|
||
if not foods:
|
||
return 0.0
|
||
|
||
# 简单的营养评分逻辑
|
||
score = 0.0
|
||
categories = set()
|
||
|
||
for food in foods:
|
||
category = food.get("category", "其他")
|
||
categories.add(category)
|
||
score += food.get("satisfaction_score", 4.0)
|
||
|
||
# 多样性加分
|
||
diversity_bonus = min(len(categories) * 0.1, 0.3)
|
||
|
||
return (score / len(foods)) + diversity_bonus
|
||
|
||
def _generate_fallback_combinations(self, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""生成备选推荐（当其他方法都失败时）"""
|
||
combinations = []
|
||
|
||
# 基础推荐搭配
|
||
fallback_combinations = {
|
||
"breakfast": [
|
||
{"name": "营养早餐", "foods": ["面包", "鸡蛋", "牛奶"]},
|
||
{"name": "健康早餐", "foods": ["燕麦粥", "香蕉", "酸奶"]}
|
||
],
|
||
"lunch": [
|
||
{"name": "均衡午餐", "foods": ["米饭", "鸡肉", "青菜"]},
|
||
{"name": "轻食午餐", "foods": ["面条", "牛肉", "西红柿"]}
|
||
],
|
||
"dinner": [
|
||
{"name": "清淡晚餐", "foods": ["粥", "豆腐", "白菜"]},
|
||
{"name": "温馨晚餐", "foods": ["饺子", "汤"]}
|
||
],
|
||
"snack": [
|
||
{"name": "健康小食", "foods": ["苹果", "坚果"]},
|
||
{"name": "休闲小食", "foods": ["酸奶", "饼干"]}
|
||
]
|
||
}
|
||
|
||
templates = fallback_combinations.get(meal_type, [])
|
||
|
||
for template in templates:
|
||
combination = {
|
||
"name": template["name"],
|
||
"description": f"营养均衡的{meal_type}搭配",
|
||
"foods": [],
|
||
"total_calories": 0,
|
||
"nutrition_score": 4.0,
|
||
"personalization_score": 0.2,
|
||
"categories": ["基础推荐"],
|
||
"source": "fallback"
|
||
}
|
||
|
||
for food_name in template["foods"]:
|
||
food_info = self._get_food_info(food_name)
|
||
if food_info:
|
||
combination["foods"].append(food_info)
|
||
combination["total_calories"] += food_info.get("calories", 0)
|
||
|
||
if len(combination["foods"]) >= 2:
|
||
combinations.append(combination)
|
||
|
||
return combinations
|
||
|
||
def _get_historical_recommendations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于历史数据的推荐"""
|
||
recommendations = []
|
||
|
||
# 分析用户历史餐食
|
||
historical_meals = [meal for meal in user_data.meals if meal.get('meal_type') == meal_type]
|
||
|
||
if not historical_meals:
|
||
return recommendations
|
||
|
||
# 统计用户喜欢的食物
|
||
food_scores = {}
|
||
for meal in historical_meals:
|
||
satisfaction = meal.get('satisfaction_score', 3) # 默认3分
|
||
for food in meal.get('foods', []):
|
||
if food not in food_scores:
|
||
food_scores[food] = []
|
||
food_scores[food].append(satisfaction)
|
||
|
||
# 计算平均满意度
|
||
for food, scores in food_scores.items():
|
||
avg_score = np.mean(scores)
|
||
if avg_score >= 3: # 只推荐满意度>=3的食物
|
||
recommendations.append({
|
||
'food': food,
|
||
'score': avg_score,
|
||
'type': 'historical',
|
||
'reason': f'历史满意度: {avg_score:.1f}分'
|
||
})
|
||
|
||
return sorted(recommendations, key=lambda x: x['score'], reverse=True)
|
||
|
||
def _get_similar_user_recommendations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于相似用户的推荐"""
|
||
recommendations = []
|
||
|
||
if not self.user_clustering_model:
|
||
return recommendations
|
||
|
||
try:
|
||
# 获取用户特征向量
|
||
user_features = self._extract_user_features(user_data)
|
||
|
||
# 找到相似用户
|
||
similar_users = self._find_similar_users(user_features)
|
||
|
||
# 基于相似用户的偏好推荐
|
||
for similar_user_id in similar_users:
|
||
similar_user_data = self._get_user_data_by_id(similar_user_id)
|
||
if similar_user_data:
|
||
user_recommendations = self._get_historical_recommendations(similar_user_data, meal_type)
|
||
for rec in user_recommendations:
|
||
rec['type'] = 'similar_user'
|
||
rec['reason'] = f'相似用户推荐'
|
||
recommendations.append(rec)
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"相似用户推荐失败: {e}")
|
||
|
||
return recommendations
|
||
|
||
def _get_content_based_recommendations(self, user_data: UserData, meal_type: str) -> List[Dict[str, Any]]:
|
||
"""基于内容相似性的推荐"""
|
||
recommendations = []
|
||
|
||
if not self.food_similarity_matrix:
|
||
return recommendations
|
||
|
||
try:
|
||
# 获取用户喜欢的食物
|
||
liked_foods = []
|
||
for meal in user_data.meals:
|
||
if meal.get('satisfaction_score', 0) >= 4: # 满意度>=4的食物
|
||
liked_foods.extend(meal.get('foods', []))
|
||
|
||
if not liked_foods:
|
||
return recommendations
|
||
|
||
# 基于食物相似性推荐
|
||
for liked_food in liked_foods:
|
||
if liked_food in self.food_similarity_matrix:
|
||
similar_foods = self.food_similarity_matrix[liked_food]
|
||
for food, similarity in similar_foods.items():
|
||
if food not in liked_foods: # 避免重复推荐
|
||
recommendations.append({
|
||
'food': food,
|
||
'score': similarity,
|
||
'type': 'content_based',
|
||
'reason': f'与{liked_food}相似'
|
||
})
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"内容推荐失败: {e}")
|
||
|
||
return recommendations
|
||
|
||
def _get_physiological_recommendations(self, user_data: UserData, context: Dict) -> List[Dict[str, Any]]:
|
||
"""基于生理状态的推荐"""
|
||
recommendations = []
|
||
|
||
# 获取生理状态信息
|
||
physiological_state = context.get('physiological_state', {})
|
||
needs = physiological_state.get('needs', [])
|
||
|
||
if not needs:
|
||
return recommendations
|
||
|
||
# 根据营养需求推荐食物
|
||
nutrition_food_mapping = {
|
||
'铁质': ['菠菜', '瘦肉', '红枣', '黑芝麻', '猪肝'],
|
||
'蛋白质': ['鸡蛋', '豆腐', '鱼肉', '鸡肉', '牛奶'],
|
||
'维生素C': ['橙子', '柠檬', '西红柿', '西兰花', '草莓'],
|
||
'叶酸': ['绿叶蔬菜', '豆类', '坚果', '菠菜', '芦笋'],
|
||
'维生素B': ['全谷物', '瘦肉', '蛋类', '香蕉', '土豆'],
|
||
'锌': ['牡蛎', '瘦肉', '坚果', '豆类', '南瓜子'],
|
||
'维生素E': ['坚果', '植物油', '鳄梨', '葵花籽', '杏仁'],
|
||
'镁': ['坚果', '绿叶蔬菜', '全谷物', '黑巧克力', '香蕉'],
|
||
'维生素B6': ['香蕉', '土豆', '鸡肉', '三文鱼', '鹰嘴豆'],
|
||
'钙质': ['牛奶', '豆腐', '绿叶蔬菜', '奶酪', '酸奶']
|
||
}
|
||
|
||
for need in needs:
|
||
foods = nutrition_food_mapping.get(need, [])
|
||
for food in foods:
|
||
recommendations.append({
|
||
'food': food,
|
||
'score': 0.8, # 生理需求推荐分数较高
|
||
'type': 'physiological',
|
||
'reason': f'补充{need}'
|
||
})
|
||
|
||
return recommendations
|
||
|
||
def _fuse_recommendations(self, recommendation_lists: List[List[Dict]], user_data: UserData) -> List[Dict[str, Any]]:
|
||
"""融合多种推荐结果"""
|
||
food_scores = {}
|
||
|
||
# 权重配置
|
||
weights = {
|
||
'historical': 0.4,
|
||
'similar_user': 0.2,
|
||
'content_based': 0.2,
|
||
'physiological': 0.2
|
||
}
|
||
|
||
for rec_list in recommendation_lists:
|
||
for rec in rec_list:
|
||
food = rec['food']
|
||
score = rec['score']
|
||
rec_type = rec['type']
|
||
|
||
if food not in food_scores:
|
||
food_scores[food] = {
|
||
'total_score': 0,
|
||
'count': 0,
|
||
'reasons': [],
|
||
'types': []
|
||
}
|
||
|
||
weight = weights.get(rec_type, 0.1)
|
||
food_scores[food]['total_score'] += score * weight
|
||
food_scores[food]['count'] += 1
|
||
food_scores[food]['reasons'].append(rec['reason'])
|
||
food_scores[food]['types'].append(rec_type)
|
||
|
||
# 转换为推荐列表
|
||
recommendations = []
|
||
for food, data in food_scores.items():
|
||
recommendations.append({
|
||
'food': food,
|
||
'score': data['total_score'],
|
||
'count': data['count'],
|
||
'reasons': data['reasons'],
|
||
'types': data['types']
|
||
})
|
||
|
||
return sorted(recommendations, key=lambda x: x['score'], reverse=True)
|
||
|
||
def _filter_and_rank_recommendations(self, recommendations: List[Dict],
|
||
user_data: UserData, preferences: Dict) -> List[Dict[str, Any]]:
|
||
"""过滤和排序推荐"""
|
||
filtered = []
|
||
|
||
# 获取用户不喜欢的食物
|
||
dislikes = user_data.profile.get('dislikes', [])
|
||
allergies = user_data.profile.get('allergies', [])
|
||
|
||
for rec in recommendations:
|
||
food = rec['food']
|
||
|
||
# 过滤不喜欢的食物
|
||
if any(dislike in food for dislike in dislikes):
|
||
continue
|
||
|
||
# 过滤过敏食物
|
||
if any(allergy in food for allergy in allergies):
|
||
continue
|
||
|
||
# 应用用户偏好
|
||
if preferences:
|
||
if 'taste' in preferences:
|
||
taste = preferences['taste']
|
||
if taste == 'sweet' and not self._is_sweet_food(food):
|
||
rec['score'] *= 0.8
|
||
elif taste == 'spicy' and not self._is_spicy_food(food):
|
||
rec['score'] *= 0.8
|
||
|
||
filtered.append(rec)
|
||
|
||
return sorted(filtered, key=lambda x: x['score'], reverse=True)
|
||
|
||
def _generate_recommendation_reasoning(self, recommendations: List[Dict], user_data: UserData) -> str:
|
||
"""生成推荐理由"""
|
||
if not recommendations:
|
||
return "暂无推荐"
|
||
|
||
top_rec = recommendations[0]
|
||
reasons = top_rec.get('reasons', [])
|
||
|
||
if reasons:
|
||
return f"推荐{top_rec['food']}，理由：{'; '.join(reasons[:2])}"
|
||
else:
|
||
return f"推荐{top_rec['food']}，基于您的个人偏好"
|
||
|
||
def _calculate_recommendation_confidence(self, user_data: UserData) -> float:
|
||
"""计算推荐置信度"""
|
||
meal_count = len(user_data.meals)
|
||
feedback_count = len(user_data.feedback)
|
||
|
||
# 基于数据量计算置信度
|
||
if meal_count >= 15 and feedback_count >= 5:
|
||
return 0.9
|
||
elif meal_count >= 10 and feedback_count >= 3:
|
||
return 0.7
|
||
elif meal_count >= 5:
|
||
return 0.5
|
||
else:
|
||
return 0.3
|
||
|
||
def _extract_user_features(self, user_data: UserData) -> np.ndarray:
|
||
"""提取用户特征向量"""
|
||
features = []
|
||
|
||
# 基础特征
|
||
profile = user_data.profile
|
||
features.extend([
|
||
profile.get('age', 25),
|
||
profile.get('height', 165),
|
||
profile.get('weight', 60),
|
||
len(profile.get('allergies', [])),
|
||
len(profile.get('dislikes', []))
|
||
])
|
||
|
||
# 口味偏好特征
|
||
taste_prefs = profile.get('taste_preferences', {})
|
||
features.extend([
|
||
taste_prefs.get('sweet', 3),
|
||
taste_prefs.get('salty', 3),
|
||
taste_prefs.get('spicy', 3),
|
||
taste_prefs.get('sour', 3),
|
||
taste_prefs.get('bitter', 3),
|
||
taste_prefs.get('umami', 3)
|
||
])
|
||
|
||
# 餐食特征
|
||
features.extend([
|
||
len(user_data.meals),
|
||
np.mean([meal.get('satisfaction_score', 3) for meal in user_data.meals]) if user_data.meals else 3,
|
||
len(user_data.feedback)
|
||
])
|
||
|
||
return np.array(features)
|
||
|
||
def _find_similar_users(self, user_features: np.ndarray) -> List[str]:
|
||
"""找到相似用户"""
|
||
# 这里简化实现，实际应该基于用户聚类模型
|
||
return []
|
||
|
||
def _get_user_data_by_id(self, user_id: str) -> Optional[UserData]:
|
||
"""根据ID获取用户数据"""
|
||
# 这里需要从数据管理器获取，简化实现
|
||
return None
|
||
|
||
def _is_sweet_food(self, food: str) -> bool:
|
||
"""判断是否为甜食"""
|
||
sweet_keywords = ['甜', '糖', '蜂蜜', '果', '蛋糕', '巧克力', '冰淇淋']
|
||
return any(keyword in food for keyword in sweet_keywords)
|
||
|
||
def _is_spicy_food(self, food: str) -> bool:
|
||
"""判断是否为辣食"""
|
||
spicy_keywords = ['辣', '椒', '麻', '辛', '咖喱', '辣椒']
|
||
return any(keyword in food for keyword in spicy_keywords)
|
||
|
||
def _load_meal_templates(self) -> Dict[str, List[Dict]]:
|
||
"""加载基础餐食搭配模板（作为备选方案）"""
|
||
return {
|
||
"breakfast": [
|
||
{"name": "经典早餐", "categories": ["主食", "蛋白质", "饮品"]},
|
||
{"name": "健康早餐", "categories": ["谷物", "水果", "蛋白质"]},
|
||
{"name": "中式早餐", "categories": ["主食", "蛋白质", "蔬菜"]}
|
||
],
|
||
"lunch": [
|
||
{"name": "均衡午餐", "categories": ["主食", "蛋白质", "蔬菜"]},
|
||
{"name": "轻食午餐", "categories": ["主食", "蛋白质", "蔬菜"]},
|
||
{"name": "丰盛午餐", "categories": ["主食", "蛋白质", "蔬菜", "汤品"]}
|
||
],
|
||
"dinner": [
|
||
{"name": "清淡晚餐", "categories": ["主食", "蛋白质", "蔬菜"]},
|
||
{"name": "温馨晚餐", "categories": ["主食", "蛋白质", "蔬菜"]}
|
||
],
|
||
"snack": [
|
||
{"name": "健康小食", "categories": ["水果", "坚果"]},
|
||
{"name": "休闲小食", "categories": ["饮品", "小食"]}
|
||
]
|
||
}
|
||
|
||
def _load_food_database(self) -> Dict[str, Dict]:
|
||
"""加载食物数据库"""
|
||
return {
|
||
'米饭': {'calories': 130, 'protein': 2.7, 'carbs': 28, 'fat': 0.3, 'category': '主食'},
|
||
'面条': {'calories': 131, 'protein': 5, 'carbs': 25, 'fat': 1.1, 'category': '主食'},
|
||
'鸡蛋': {'calories': 155, 'protein': 13, 'carbs': 1.1, 'fat': 11, 'category': '蛋白质'},
|
||
'鸡肉': {'calories': 165, 'protein': 31, 'carbs': 0, 'fat': 3.6, 'category': '蛋白质'},
|
||
'鱼肉': {'calories': 206, 'protein': 22, 'carbs': 0, 'fat': 12, 'category': '蛋白质'},
|
||
'豆腐': {'calories': 76, 'protein': 8, 'carbs': 2, 'fat': 4.8, 'category': '蛋白质'},
|
||
'菠菜': {'calories': 23, 'protein': 2.9, 'carbs': 3.6, 'fat': 0.4, 'category': '蔬菜'},
|
||
'西兰花': {'calories': 34, 'protein': 2.8, 'carbs': 7, 'fat': 0.4, 'category': '蔬菜'},
|
||
'苹果': {'calories': 52, 'protein': 0.3, 'carbs': 14, 'fat': 0.2, 'category': '水果'},
|
||
'香蕉': {'calories': 89, 'protein': 1.1, 'carbs': 23, 'fat': 0.3, 'category': '水果'},
|
||
'牛奶': {'calories': 42, 'protein': 3.4, 'carbs': 5, 'fat': 1, 'category': '乳制品'},
|
||
'燕麦': {'calories': 389, 'protein': 17, 'carbs': 66, 'fat': 7, 'category': '主食'}
|
||
}
|
||
|
||
def _load_or_train_models(self):
|
||
"""加载或训练模型"""
|
||
try:
|
||
# 尝试加载现有模型
|
||
tfidf_path = self.model_path / 'tfidf_vectorizer.pkl'
|
||
if tfidf_path.exists():
|
||
self.tfidf_vectorizer = joblib.load(tfidf_path)
|
||
self.logger.info("TF-IDF向量化器加载成功")
|
||
|
||
# 构建食物相似性矩阵
|
||
self._build_food_similarity_matrix()
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"模型加载失败: {e}")
|
||
# 如果加载失败，使用默认配置
|
||
self._initialize_default_models()
|
||
|
||
def _build_food_similarity_matrix(self):
|
||
"""构建食物相似性矩阵"""
|
||
try:
|
||
if not self.tfidf_vectorizer:
|
||
self._initialize_default_models()
|
||
|
||
# 获取所有食物名称
|
||
food_names = list(self.food_database.keys())
|
||
|
||
# 使用TF-IDF计算相似性
|
||
food_features = self.tfidf_vectorizer.transform(food_names)
|
||
similarity_matrix = cosine_similarity(food_features)
|
||
|
||
# 构建相似性字典
|
||
self.food_similarity_matrix = {}
|
||
for i, food1 in enumerate(food_names):
|
||
similarities = {}
|
||
for j, food2 in enumerate(food_names):
|
||
if i != j:
|
||
similarities[food2] = similarity_matrix[i][j]
|
||
|
||
# 按相似性排序
|
||
sorted_similarities = sorted(similarities.items(), key=lambda x: x[1], reverse=True)
|
||
self.food_similarity_matrix[food1] = dict(sorted_similarities[:5]) # 只保留前5个相似食物
|
||
|
||
self.logger.info("食物相似性矩阵构建完成")
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"构建食物相似性矩阵失败: {e}")
|
||
self.food_similarity_matrix = {}
|
||
|
||
def _initialize_default_models(self):
|
||
"""初始化默认模型"""
|
||
try:
|
||
# 初始化TF-IDF向量化器
|
||
self.tfidf_vectorizer = TfidfVectorizer(
|
||
max_features=1000,
|
||
stop_words=None, # 中文不需要英文停用词
|
||
ngram_range=(1, 2)
|
||
)
|
||
|
||
# 使用食物名称训练
|
||
food_names = list(self.food_database.keys())
|
||
self.tfidf_vectorizer.fit(food_names)
|
||
|
||
# 保存模型
|
||
tfidf_path = self.model_path / 'tfidf_vectorizer.pkl'
|
||
joblib.dump(self.tfidf_vectorizer, tfidf_path)
|
||
|
||
self.logger.info("默认模型初始化完成")
|
||
|
||
except Exception as e:
|
||
self.logger.error(f"默认模型初始化失败: {e}")
|
||
|
||
def _find_similar_foods(self, input_data: Dict, user_data: UserData) -> Dict[str, Any]:
|
||
"""查找相似食物"""
|
||
target_food = input_data.get('food', '')
|
||
|
||
if not target_food or not self.food_similarity_matrix:
|
||
return self._create_error_result("无法找到相似食物")
|
||
|
||
similar_foods = self.food_similarity_matrix.get(target_food, {})
|
||
|
||
return {
|
||
'success': True,
|
||
'target_food': target_food,
|
||
'similar_foods': list(similar_foods.items()),
|
||
'confidence': 0.8
|
||
}
|
||
|
||
def _update_user_preferences(self, input_data: Dict, user_data: UserData) -> Dict[str, Any]:
|
||
"""更新用户偏好"""
|
||
feedback_data = input_data.get('feedback', {})
|
||
|
||
# 更新用户偏好模型
|
||
# 这里可以实现更复杂的偏好学习算法
|
||
|
||
return {
|
||
'success': True,
|
||
'message': '用户偏好更新成功',
|
||
'confidence': 0.7
|
||
}
|
||
|
||
def _retrain_models(self, input_data: Dict, user_data: UserData) -> Dict[str, Any]:
|
||
"""重新训练模型"""
|
||
try:
|
||
# 检查是否有足够的数据进行重训练
|
||
total_samples = len(user_data.meals) + len(user_data.feedback)
|
||
|
||
if total_samples < self.model_update_threshold:
|
||
return {
|
||
'success': False,
|
||
'message': f'数据不足，需要至少{self.model_update_threshold}个样本',
|
||
'current_samples': total_samples
|
||
}
|
||
|
||
# 重新训练模型
|
||
self._load_or_train_models()
|
||
|
||
return {
|
||
'success': True,
|
||
'message': '模型重训练完成',
|
||
'confidence': 0.9
|
||
}
|
||
|
||
except Exception as e:
|
||
return self._create_error_result(f"模型重训练失败: {str(e)}")
|
||
|
||
def _create_error_result(self, error_message: str) -> Dict[str, Any]:
|
||
"""创建错误结果"""
|
||
return {
|
||
'success': False,
|
||
'error': error_message,
|
||
'message': f'推荐失败: {error_message}',
|
||
'confidence': 0.0
|
||
}
|
||
|
||
def cleanup(self) -> bool:
|
||
"""清理资源"""
|
||
try:
|
||
self.logger.info("推荐引擎清理完成")
|
||
return True
|
||
except Exception as e:
|
||
self.logger.error(f"推荐引擎清理失败: {e}")
|
||
return False
|
||
|
||
|
||
# 便捷函数
|
||
def generate_meal_recommendations(user_id: str, meal_type: str, preferences: Dict = None, context: Dict = None) -> Optional[Dict]:
|
||
"""生成餐食推荐"""
|
||
from core.base import get_app_core
|
||
|
||
if preferences is None:
|
||
preferences = {}
|
||
if context is None:
|
||
context = {}
|
||
|
||
app = get_app_core()
|
||
input_data = {
|
||
'type': 'meal_recommendation',
|
||
'meal_type': meal_type,
|
||
'preferences': preferences,
|
||
'context': context
|
||
}
|
||
|
||
result = app.process_user_request(ModuleType.RECOMMENDATION, input_data, user_id)
|
||
return result.result if result else None
|
||
|
||
|
||
def find_similar_foods(user_id: str, food: str) -> Optional[Dict]:
|
||
"""查找相似食物"""
|
||
from core.base import get_app_core
|
||
|
||
app = get_app_core()
|
||
input_data = {
|
||
'type': 'food_similarity',
|
||
'food': food
|
||
}
|
||
|
||
result = app.process_user_request(ModuleType.RECOMMENDATION, input_data, user_id)
|
||
return result.result if result else None
|
||
|
||
|
||
def update_user_preferences(user_id: str, feedback: Dict) -> Optional[Dict]:
|
||
"""更新用户偏好"""
|
||
from core.base import get_app_core
|
||
|
||
app = get_app_core()
|
||
input_data = {
|
||
'type': 'preference_update',
|
||
'feedback': feedback
|
||
}
|
||
|
||
result = app.process_user_request(ModuleType.RECOMMENDATION, input_data, user_id)
|
||
return result.result if result else None
|
||
|
||
|
||
def retrain_recommendation_model(user_id: str) -> Optional[Dict]:
|
||
"""重新训练推荐模型"""
|
||
from core.base import get_app_core
|
||
|
||
app = get_app_core()
|
||
input_data = {
|
||
'type': 'model_retrain'
|
||
}
|
||
|
||
result = app.process_user_request(ModuleType.RECOMMENDATION, input_data, user_id)
|
||
return result.result if result else None
|
||
|
||
|
||
if __name__ == "__main__":
|
||
# 测试推荐引擎
|
||
from core.base import BaseConfig, initialize_app, cleanup_app
|
||
|
||
print("测试推荐引擎...")
|
||
|
||
# 初始化应用
|
||
config = BaseConfig()
|
||
if initialize_app(config):
|
||
print("✅ 应用初始化成功")
|
||
|
||
# 测试餐食推荐
|
||
test_user_id = "test_user_001"
|
||
|
||
result = generate_meal_recommendations(test_user_id, "lunch", {"taste": "sweet"})
|
||
if result and result.get('success'):
|
||
recommendations = result.get('recommendations', [])
|
||
print(f"✅ 餐食推荐成功，推荐了{len(recommendations)}种食物")
|
||
for rec in recommendations[:3]:
|
||
print(f" - {rec['food']}: {rec.get('score', 0):.2f}")
|
||
|
||
# 测试相似食物查找
|
||
result = find_similar_foods(test_user_id, "米饭")
|
||
if result and result.get('success'):
|
||
similar_foods = result.get('similar_foods', [])
|
||
print(f"✅ 相似食物查找成功，找到{len(similar_foods)}种相似食物")
|
||
|
||
# 清理应用
|
||
cleanup_app()
|
||
print("✅ 应用清理完成")
|
||
else:
|
||
print("❌ 应用初始化失败")
|