TMDB电影数据分析.

import pandas as pd import numpy as np import json #可视化包 import matplotlib.pyplot as plt import matplotlib #加载词云图 from wordcloud import WordCloud #更改字体 plt.rcParams['font.sans-serif'] = ['SimHei'] #读取演员列表信息和电影列表信息数据 credits = pd.read_csv('tmdb_5000_credits.csv',encoding='utf-8') movies = pd.read_csv('tmdb_5000_movies.csv',encoding='utf-8') json_cols = ['cast','crew'] #json #编码：把python对象编码转换json字符串 json.dumps #解码：把json字符串转换成python对象 json.loads for i in json_cols: credits[i] = credits[i].apply(json.loads) #用json方法汇总所有类型 def get_name(x): return ','.join([i['name'] for i in x]) # cast 提取演员 credits['cast'] = credits['cast'].apply(get_name) # crew 提取导演 def director(x): for i in x: if i['job'] =='Director': return i['name'] credits['crew'] = credits['crew'].apply(director) # print(credits.head()) #将‘crew'转换为’directo‘ credits.rename(columns = {'crew':'director'},inplace=True) # print(credits.head()) #movies 解析 json_colss = ['genres','keywords','spoken_languages','production_countries','production_companies'] for i in json_colss: movies[i]=movies[i].apply(json.loads) def get_names(x): return ','.join([i['name'] for i in x]) movies['genres']=movies['genres'].apply(get_names) movies['keywords']=movies['keywords'].apply(get_names) movies['spoken_languages']=movies['spoken_languages'].apply(get_names) movies['production_countries']=movies['production_countries'].apply(get_names) movies['production_companies']=movies['production_companies'].apply(get_names) # print((credits['title']==movies['title']).describe()) del movies['title'] #合并两张表 df = credits.merge(right=movies,how='inner',left_on='movie_id',right_on='id') # print(df.head()) del df['id'] del df['homepage'] del df['original_title'] del df['tagline'] df['release_date']=df['release_date'].fillna('2014-06-01')#填充缺失值 df['runtime']=df['runtime'].fillna(df.runtime.mean())#填充平均值 df[df[['release_date','runtime']].isnull().values==True] #获取电影类型 df['genres'][1].split(',') genre = set()#建立一个集合，会去重 for i in df['genres'].str.split(','): genre = set().union(i,genre) genre = list(genre) genre.remove('') # print (genre) #转化日期格式 df['release_year'] = pd.to_datetime(df.release_date, format= '%Y-%m-%d').dt.year df['release_month'] = pd.to_datetime(df.release_date, format='%Y-%m-%d').dt.month #判断每部电影属于什么类型，返回1 for genr in genre: df[genr] = df['genres'].str.contains(genr).apply(lambda x:1 if x else 0) #str.contains 字符串包含 df_gy = df.loc[:,genre] # print(df_gy) #建立包含电影类型和年份的数据框 # print(df_gy.index) df_gy.index = df['release_year'] # print(df_gy.head()) # df_gy.groupby('release_year').count().Adventure.plot(figsize=(12,6)) # plt.xticks(range(1915,2018,5)) # plt.title('电影数量的年度趋势图', fontsize = 20) # plt.xlabel('年份',fontsize = 20) # plt.ylabel('数量',fontsize = 20) # plt.show() df_gy1 = df_gy.sort_index(ascending= False)#sort_index 索引排序，默认升序,False 为降序。 # print(df_gy1.head()) df_gys = df_gy1.groupby('release_year').sum() # plt.figure(figsize=(12,6)) # plt.plot(df_gys,label = df_gys.columns)#label设置上方小图标签。 # plt.legend(df_gys) # plt.xticks(range(1915,2018,5)) # plt.title('电影类型随时间的变化',fontsize=20) # plt.xlabel('年份',fontsize = 20) # plt.ylabel('数量',fontsize = 20) # plt.show() #计算各类型的电影总数量 df_gysum = df_gys.sum().sort_values(ascending= True) # print(df_gysum) #电影类型数量条形图 # df_gysum.plot.barh(label = 'df_gys', figsize=(15,7)) # plt.title('电影类型的数量条形图',fontsize = 20) # plt.xlabel('数量',fontsize=20) # plt.xlabel('电影类型',fontsize=20) # plt.show() #各种类型所占比例 # b1 = df_gysum/df_gysum.sum() # #(每一块)之间距离，所占比例大于0.06的，往外高出一点 # explode = (b1>=0.06)/20+0.02 # # print(explode) # df_gysum.plot.pie(autopct = '%1.1f%%', figsize = (10,10),label = '',explode = explode) # plt.title('电影类型比例分布图',fontsize = 20) # plt.show() #电影类型与利润的关系 df['profit'] = df['revenue']-df['budget'] df_gen_pro = pd.DataFrame(index = genre) #每种类型的平均利润 list =[] for genr in genre: list.append(df.groupby(genr)['profit'].mean()) list2 = [] for i in range(len(genre)): list2.append(list[i][1]) df_gen_pro['mean_profit'] = list2 # print(df_gen_pro.head()) # df_gen_pro.sort_values(by='mean_profit',ascending=True).plot.barh(label='genre',figsize=(15,7)) # plt.title('电影类型的利润条形图',fontsize = 20) # plt.xlabel('利润',fontsize=20) # plt.ylabel('电影类型',fontsize=20) # plt.grid(True) # plt.show() #电影类型受欢迎条形图 df_gen_popu = pd.DataFrame(index=genre) list = [] for genr in genre: list.append(df.groupby(genr)['popularity'].mean()) list2=[] for i in range(len(genre)): list2.append(list[i][1]) df_gen_popu['mean_popularity']=list2 # df_gen_popu.sort_values(by='mean_popularity',ascending=True).plot.barh(label='genre',figsize=(15,7)) # plt.title('电影类型受欢迎条形图',fontsize=20) # plt.xlabel('受欢迎',fontsize=20) # plt.ylabel('电影类型',fontsize=20) # plt.grid(True) # plt.show() #电影预算与票房的关系 # plt.scatter(x=df.budget,y=df.revenue) # plt.xlabel('预算',fontsize=20) # plt.ylabel('票房',fontsize=20) # plt.title('电影预算与票房散点图',fontsize=20) # plt.show() #Universal Pictures和Paramount Pictures两个公司产出电影的情况对比 #提取两个公司 df['Universal Pictures']=df['production_companies'].str.contains('Universal Pictures').apply(lambda x:1 if x else 0) df['Paramount Pictures']=df['production_companies'].str.contains('Paramount Pictures').apply(lambda x:1 if x else 0) #计算公司利润 df['Universal_profit']=df['Universal Pictures']*df['profit'] df['Paramount_profit']=df['Paramount Pictures']*df['profit'] #计算两个公司的票房 df['Universal_revenue']=df['Universal Pictures']*df['revenue'] df['Paramount_revenue']=df['Paramount Pictures']*df['revenue'] companiesup = [df['Universal Pictures'].sum(),df['Paramount Pictures'].sum()] # print(companiesup) #比例图 # pd.Series(companiesup,index=['Universal Pictures','Paramount Pictures']).plot.pie(autopct='%1.1f%%',figsize=(5,5),label='') # plt.show() #两家公司的电影随时间变化的趋势对比 companydf = df[['release_year','Universal Pictures','Paramount Pictures','Universal_profit','Paramount_profit','Universal_revenue','Paramount_revenue']] # print(companydf.head()) companydf1 = companydf.sort_index() # print(companydf1.head()) #按年份计算总额 companydf2=companydf1.groupby('release_year').sum() # print(companydf2) # plt.figure(figsize=(12,6)) # plt.plot(companydf2.index,companydf2['Universal Pictures'],label=companydf2['Universal Pictures']) # plt.plot(companydf2.index,companydf2['Paramount Pictures'],label=companydf2['Paramount Pictures']) # plt.xticks(range(1915,2018,5)) # plt.title('Universal Pictures和Patamount Pictures电影年发行量趋势图对比',fontsize=22) # plt.xlabel('时间',fontsize=15) # plt.ylabel('发行量',fontsize=15) # plt.grid(True) # plt.legend(['Universal Pictures','Patamount Picture']) # plt.show() #关键词分析 keywords_list = [] for i in df['keywords']: keywords_list.append(i) lst = ''.join(keywords_list) #制作词云图 wc = WordCloud(background_color='black',max_words=3000, scale=1.5).generate(lst) plt.figure(figsize=(14,7)) plt.imshow(wc)#显示 plt.axis('off') p