揭秘Python在推荐系统开发中的重要角色

import pandas as pd# 读取数据data = pd.read_csv("data.csv")# 打印数据前5行print(data.head())# 数据清洗# 删除空值data.dropna()# 数据处理# 数据转换data["price"] = data["price"].apply(lambda x: float(x.replace("$", "")))# 数据筛选filtered_data = data[data["price"] <ol start="2"><li>特征提取与表示<br>在推荐系统中，特征提取和表示是非常重要的任务。Python的机器学习库scikit-learn提供了丰富的特征提取和表示方法。例如，使用TF-IDF方法可以将文本数据转换为数字特征向量。示例如下：</li></ol><pre class="brush:python;toolbar:false;">from sklearn.feature_extraction.text import TfidfVectorizer# 文本数据text_data = [    "Python is a popular programming language",    "Machine learning is an important part of AI",    "Python and Machine learning are closely related"]# 使用TF-IDF方法提取特征vectorizer = TfidfVectorizer()features = vectorizer.fit_transform(text_data)# 打印特征向量print(features.toarray())

from sklearn.metrics.pairwise import cosine_similarityfrom sklearn.model_selection import train_test_split# 用户-物品评分矩阵rating_matrix = [[5, 3, 0, 1],                 [4, 0, 0, 1],                 [1, 1, 0, 5],                 [1, 0, 0, 4]]# 切分训练集和测试集train_matrix, test_matrix = train_test_split(rating_matrix, test_size=0.2)# 计算用户相似度user_similarity = cosine_similarity(train_matrix)# 预测用户对物品的评分def predict(user_id, item_id):    similarity_sum = 0    score_sum = 0    for u_id in range(len(train_matrix)):        if train_matrix[u_id][item_id] != 0:            similarity_sum += user_similarity[user_id][u_id]            score_sum += (user_similarity[user_id][u_id] * train_matrix[u_id][item_id])    return score_sum / similarity_sum if similarity_sum != 0 else 0# 对测试集进行评估total_error = 0for user_id in range(len(test_matrix)):    for item_id in range(len(test_matrix[user_id])):        if test_matrix[user_id][item_id] != 0:            predicted_score = predict(user_id, item_id)            error = abs(predicted_score - test_matrix[user_id][item_id])            total_error += error# 打印评估结果print("Mean Absolute Error:", total_error / len(test_data))