Python TMDB电影数据集多维度关联规则分析（python商务大数据分析）

# -*- coding: utf-8 -*- """ Created on Mon Nov 18 17:56:46 2024 @author: 62561 """ # 导入必要的库 from mlxtend.frequent_patterns import apriori, association_rules import pandas as pd import numpy as np import pandas as pd import matplotlib.pyplot as plt import seaborn as sns from mlxtend.frequent_patterns import apriori, association_rules from sklearn.preprocessing import LabelEncoder from sklearn.preprocessing import MultiLabelBinarizer #数据读取 movies=pd.read_csv('tmdb_5000_movies.csv') movies.info() movies.isnull().sum() movies.describe() #数据预处理 movies.query('budget==0').head(2) movies.drop(['homepage','id','keywords','tagline','overview','spoken_languages','status'],axis=1,inplace = True) movies.shape movies.dropna(inplace=True) movies.shape movies.drop_duplicates(keep='first',inplace=True) for col in movies.columns: zero=len(movies[movies[col]==0]) print('{},零售数目：{}'.format(col,zero)) movies=movies.drop(index=movies.query('budget==0 or revenue==0').index) movies.shape movies.describe() #属性构造：1-预算除以电影收入得到新属性：预算收益率 movies['ReturnRate']=1-movies['budget']/movies['revenue'] movies.describe() #数据信息统计 #相关系数矩阵，即给出了任意两个变量之间的相关系数 alld=movies.corr() #计算收入与其他字段相关系数 movies['release_year']= [i.year for i in pd.to_datetime(movies['release_date'])] a=movies[['budget','popularity','vote_average','vote_count','runtime','release_year','revenue']].corr() #查看收入前十的电影信息 top10=movies.sort_values('revenue',ascending = False).head(10) top10[['budget','genres','original_title','release_date','revenue','runtime','vote_average','vote_count','popularity']] #查看电影年份分布 movies['release_year'].value_counts() movies['release_year'].hist(bins=25) #查看电影时长分布 movies['runtime'].value_counts() movies['runtime'].hist(bins=30) #查看预算分布 movies['budget'].hist(bins=15) #查看评分分布 movies['vote_average'].hist(bins=20) #查看评论次数分布 movies['vote_count'].hist(bins=20) #影响收入的因素分析 #绘制电影收入与影响因子散点图 plt.rcParams['font.sans-serif'] = [u'SimHei'] list1= ['budget','popularity','vote_count','vote_average','runtime','release_year'] fig,axes = plt.subplots(2,3,figsize=(12,8),dpi = 70) for i in range(2): for j in range(3): axes[i,j].scatter(x= list1[3*i+j], y='revenue', data=movies,s = 3) axes[i,j].set_title(list1[3*i+j]) fig.suptitle('影响电影要房的因素') fig.show() #分组分析电影收入与影响因子的关系 movies=movies[(movies['release_year']>2000)&(movies['release_year']<=2015)] movies.shape edge=[2000,2005,2010,2015] moviesclass=['2001-2005年','2006-2010年','2011-2015年'] movies['fiveyearclass']=pd.cut(movies.loc[:,'release_year'],edge,labels=moviesclass,include_lowest=True) movies.head() edge = movies['revenue'].quantile([0,0.25,0.5,0.75,1]).values moviesclass=['Low','Medium','Moderately High','High'] movies['revenueclass']= pd.cut(movies.loc[:,'revenue'],edge, labels=moviesclass, include_lowest = True) movies.head () d_summary=movies.groupby(['fiveyearclass','revenueclass']).median() d_summary list1=['budget','popularity','vote_count','vote_average','runtime','release_year'] pos=list(range(len(d_summary[:'2001-2005年']))) width=0.2 fig,ax=plt.subplots(2,3,figsize=(10,5)) plt.subplots_adjust(wspace=0.4,hspace =0.5) for i in range(2): for j in range(3): ax[i,j].bar(pos,d_summary.loc['2001-2005年'][list1[3*i+j]], width, label='2001-2005年') ax[i,j].bar([p + width for p in pos],d_summary.loc['2006-2010年'][list1[3*i+j]],width, label='2006-2010年') ax[i,j].bar([p + width*2 for p in pos], d_summary.loc['2011-2015年'][list1[3*i+j]], width, label='2011-2015年') ax[i,j].set_ylabel(list1[3*i+j]) ax[i,j].set_xlabel('revenue') ax[i,j].set_title('revenue&'+list1[3*i+j]) ax[i,j].set_xticks([p + 1.5 * width for p in pos]) ax[i,j].set_xticklabels(['低','中下','中上','高']) ax[i,j].legend() ax[i,j].grid() #分析电影类型对电影收入的影响 def counttype(movies): moviesgeners={} for i in movies.genres: for j in eval(i): if j['name']not in moviesgeners: moviesgeners[j['name']]=1 else: moviesgeners[j['name']]+=1 return moviesgeners plt.figure(figsize=(10,5)) moviesgeners=counttype(movies) moviesgeners=list(moviesgeners.items()) moviesgeners.sort(key=lambda tup:tup[1],reverse=True) x=[i[0] for i in moviesgeners] y=[i[1] for i in moviesgeners] plt.bar(x,y) plt.xticks(rotation=45) plt.xlabel("电影类型") plt.ylabel("电影数量") plt.title("电影类型分布条形图") plt.show() top100=movies.sort_values('popularity',ascending=False).head(100) moviesgeners=counttype(top100) moviesgeners moviesclass=['Low','Medium','Moderately High','High'] groupmovies=movies.groupby('revenueclass') fig,axes = plt.subplots(2,2,figsize =(10,8)) plt.subplots_adjust(wspace=0.2,hspace=0.2) i= 0 for j in range(2): for k in range(2): data=groupmovies.get_group(moviesclass[i]) moviesCounttype = counttype(data) moviesCounttype=list(moviesCounttype.items()) moviesCounttype.sort(key=lambda tup: tup[1],reverse = True) moviesCounttype= moviesCounttype[:5] x=[i[0] for i in moviesCounttype] y=[i[1] for i in moviesCounttype] axes[j,k].bar(x,y,width =0.5) axes[j,k].title.set_text(moviesclass[i]) i+= 1 import pandas as pd import json import random from efficient_apriori import apriori # 确保将 genres 列转换为列表 movies['genres'] = movies['genres'].apply(json.loads) # 根据 vote_count 和 popularity 生成模拟交易数据 def generate_transactions(movies, num_users=200): transactions = [] for _ in range(num_users): user_movies = movies.sample(frac=1, weights='popularity').head(random.randint(1, 10)) watched_genres = set() for _, row in user_movies.iterrows(): watched_genres.update([genre['name'] for genre in row['genres']]) transactions.append(list(watched_genres)) return transactions random.seed(42) transactions = generate_transactions(movies) # 使用 efficient-apriori 进行关联规则分析 min_support = 0.5 # 降低支持度 min_confidence = 0.9 # 降低置信度 # 生成频繁项集和关联规则 itemsets, rules = apriori(transactions, min_support=min_support, min_confidence=min_confidence) # 打印有趣的关联规则 for rule in rules: lhs = ', '.join(rule.lhs) # 规则的前提（左侧项集） rhs = ', '.join(rule.rhs) # 规则的结果（右侧项集） print(f"If a user watches {lhs}, they are likely to also watch {rhs} (confidence: {rule.confidence:.2f})") import networkx as nx import matplotlib.pyplot as plt from efficient_apriori import apriori # 假设 transactions 是之前代码中准备好的交易数据列表 # itemsets, rules = apriori(transactions, min_support=0.05, min_confidence=0.7) # 函数来绘制关联规则网络图 def plot_rules(rules): G = nx.DiGraph() # 添加节点和边 for rule in rules: antecedents = rule.lhs # 使用 lhs (left-hand side) 获取前件 consequents = rule.rhs # 使用 rhs (right-hand side) 获取后件 for antecedent in antecedents: G.add_node(antecedent, label=antecedent) for consequent in consequents: G.add_node(consequent, label=consequent) for antecedent in antecedents: for consequent in consequents: G.add_edge(antecedent, consequent, weight=rule.confidence) #