推荐系统中的冷启动问题及解决方案

from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.metrics.pairwise import linear_kernel

# 示例：基于电影描述的内容推荐

# 假设我们有一个电影描述的DataFrame
movies = pd.DataFrame({
    'movie_id': [1, 2, 3, 4, 5],
    'title': ['Movie A', 'Movie B', 'Movie C', 'Movie D', 'Movie E'],
    'description': [
        'A science fiction movie about space exploration.',
        'A romantic comedy set in New York City.',
        'A thrilling action movie with car chases.',
        'A documentary about wildlife conservation.',
        'A horror movie with a haunted house.'
    ]
})

# 使用TF-IDF向量化电影描述
tfidf = TfidfVectorizer(stop_words='english')
tfidf_matrix = tfidf.fit_transform(movies['description'])

# 计算相似度矩阵
cosine_sim = linear_kernel(tfidf_matrix, tfidf_matrix)

# 基于电影ID推荐相似电影
def get_recommendations(title, cosine_sim=cosine_sim):
    idx = movies.index[movies['title'] == title].tolist()[0]
    sim_scores = list(enumerate(cosine_sim[idx]))
    sim_scores = sorted(sim_scores, key=lambda x: x[1], reverse=True)
    sim_scores = sim_scores[1:4]  # 推荐前3个相似电影
    movie_indices = [i[0] for i in sim_scores]
    return movies['title'].iloc[movie_indices]

# 推荐与Movie A相似的电影
recommended_movies = get_recommendations('Movie A')
print("推荐电影:", recommended_movies.tolist())

import networkx as nx

# 示例：基于社交网络的推荐

# 构建一个简单的社交网络图
G = nx.Graph()

# 添加用户节点和社交关系
G.add_edges_from([
    ('User1', 'User2'),
    ('User2', 'User3'),
    ('User3', 'User4'),
    ('User4', 'User5'),
    ('User1', 'User5')
])

# 为每个用户添加喜欢的物品
user_likes = {
    'User1': ['Item1', 'Item2'],
    'User2': ['Item2', 'Item3'],
    'User3': ['Item3', 'Item4'],
    'User4': ['Item4', 'Item5'],
    'User5': ['Item5', 'Item1']
}

# 新用户User6加入社交网络，且与User1和User2建立了连接
G.add_node('User6')
G.add_edges_from([('User6', 'User1'), ('User6', 'User2')])

# 根据社交网络推荐物品给User6
def recommend_items(user, G, user_likes):
    # 获取用户的朋友
    friends = list(G.neighbors(user))
    # 收集朋友喜欢的物品
    items = set()
    for friend in friends:
        items.update(user_likes.get(friend, []))
    return list(items)

# 为User6推荐物品
recommended_items = recommend_items('User6', G, user_likes)
print("推荐物品:", recommended_items)

# 示例：线性组合模型

# 假设我们有两种推荐算法的评分结果
user_id = 0
collaborative_scores = [3.0, 4.5, 2.0, 4.0, 5.0]  # 协同过滤评分
content_based_scores = [4.0, 3.5, 3.0, 4.5, 2.0]  # 基于内容的评分

# 线性组合
alpha = 0.6  # 协同过滤的权重
beta = 0.4   # 基于内容推荐的权重
final_scores = [alpha * c + beta * cb for c, cb in zip(collaborative_scores, content_based_scores)]

# 推荐评分最高的物品
recommended_item_index = final_scores.index(max(final_scores))
print("推荐物品索引:", recommended_item_index)

from sklearn.cluster import KMeans

# 示例：用户聚类推荐

# 假设我们有用户的基本信息数据
users = pd.DataFrame({
    'user_id': [1, 2, 3, 4, 5],
    'age': [25, 34, 22, 35, 28],
    'income': [50000, 60000, 45000, 70000, 52000],
    'gender': [1, 0, 1, 0, 1]  # 1: Male, 0: Female
})

# K-means聚类
kmeans = KMeans(n_clusters=2, random_state=0)
users['cluster'] = kmeans.fit_predict(users[['age', 'income', 'gender']])

# 新用户加入并分配到最近的聚类
new_user = pd.DataFrame({'age': [30], 'income': [55000], 'gender': [1]})
new_user_cluster = kmeans.predict(new_user[['age', 'income', 'gender']])

# 根据新用户的聚类推荐物品
cluster_group = users[users['cluster'] == new_user_cluster[0]]
print("推荐物品给新用户的聚类组:", cluster_group)

import pandas as pd

# 加载用户和物品数据
users = pd.DataFrame({
    'user_id': [1, 2, 3, 4, 5],
    'age': [25, 34, 22, 35, 28],
    'income': [50000, 60000, 45000, 70000, 52000],
    'gender': [1, 0, 1, 0, 1]
})

items = pd.DataFrame({
    'item_id': [1, 2, 3, 4, 5],
    'category': ['Electronics', 'Books', 'Clothing', 'Electronics', 'Books'],
    'price': [299.99, 19.99, 49.99, 199.99, 14.99]
})

from sklearn.preprocessing import StandardScaler
from sklearn.cluster import KMeans

# 标准化用户特征
scaler = StandardScaler()
users_scaled = scaler.fit_transform(users[['age', 'income', 'gender']])

# 用户聚类
kmeans = KMeans(n_clusters=2, random_state=0)
users['cluster'] = kmeans.fit_predict(users_scaled)

# 为新用户推荐物品
new_user = pd.DataFrame({'age': [30], 'income': [55000], 'gender': [1]})
new_user_scaled = scaler.transform(new_user[['age', 'income', 'gender']])
new_user_cluster = kmeans.predict(new_user_scaled)

# 根据聚类组推荐热门物品
cluster_group = users[users['cluster'] == new_user_cluster[0]]
recommended_items = items[items['category'].isin(['Electronics'])]  # 假设当前组用户对Electronics有偏好
print("推荐物品:", recommended_items)