最近在网上看到别人做的爬取微信聊天记录并分析聊天内容,GitHub上试着运行了一下,这好东西肯定要分享出来给各位,总结一下几年的微信聊天内容,废话不多说,下面一步步来。先展示一下,我和我对象的聊天内容分析:
源代码和出处:GitHub - LC044/WeChatMsg: 提取微信聊天记录,将其导出成HTML、Word、CSV文档永久保存,对聊天记录进行分析生成年度聊天报告
大家记得给作者点点star,督促作者开发更优的信息抓取功能。
下载微信聊天记录爬取程序:(软件安全正常,直接无视安全问题)
https://github.com/LC044/WeChatMsg/releases/download/v1.0.6/MemoTrace-1.0.6.exe
电脑需要登录微信,如果电脑微信聊天记录不齐全,可以通过手机进行微信聊天记录迁移。
打开软件,随后点击获取信息,获取手机号、微信昵称、wxid等内容,之后点击开始启动就行。
若出现wxid或微信路径无法获取问题,查看解决办法("留痕"使用教程 (lc044.love)),一般都是没问题的。
选择 “数据 --> 批量导出”,选择你想要导出的联系人信息。导出格式选择csv格式,方便我们后续利用python进行数据分析:
导出后的结果在程序同目录下的“data --> 聊天记录“文件中,我们需要csv文件,记住csv文件的地址,自此微信聊天记录爬取结束。
PS:上述软件也可以进行数据分析,作者也贴出年度报告,各位可以尝试一下,不过内容较少且存在乱码。
环境配置:python3.8(3.10matplotlib不兼容问题) numpy pandas seaborn jieba july wordcloud
接下来直接内容分析代码,代码中需要根据你的CSV文件地址修改以及聊天双方名字修改:
import matplotlib.pyplot as plt
import pandas as pd
import re
import july
import jieba
from july.utils import date_range
import seaborn as sns
from scipy.stats import norm
import numpy as np
from wordcloud import WordCloud
from collections import Counter
def set_chinese_font():
# 设置中文字体
plt.rcParams['font.sans-serif'] = ['SimHei'] # 设置中文字体为黑体
plt.rcParams['axes.unicode_minus'] = False # 用来正常显示负号
def read_chat_data(file_path):
# 读取CSV文件
df = pd.read_csv(file_path)
return df
def preprocess_data(df):
# 数据预处理
df = df[df['Type'] == 1] # 只保留文本聊天
selected_columns = ['IsSender', 'StrContent', 'StrTime']
df = df[selected_columns] # 只取'IsSender','StrContent','StrTime'列
df['StrTime'] = pd.to_datetime(df['StrTime'])
df['Date'] = df['StrTime'].dt.date
return df
def plot_chat_frequency_by_day(df):
# 每天聊天频率柱状图
chat_frequency = df['Date'].value_counts().sort_index()
chat_frequency.plot(kind='bar', color='#DF9F9B')
total_messages = len(df)
date_labels = [date.strftime('%m-%d') for date in chat_frequency.index]
plt.text(30, 1300, '消息总数:{0}条'.format(total_messages), ha='left', va='top', fontsize=10, color='black')
plt.text(30, 1250, '起止时间:{0} --- {1}'.format(date_labels[0], date_labels[-1]), ha='left', va='top', fontsize=10,
color='black')
plt.xlabel('Date')
plt.ylabel('Frequency')
plt.title('Chat Frequency by Day')
plt.xticks(range(1, len(date_labels), 7), date_labels[::7])
plt.xticks(fontsize=5)
plt.show()
def plot_calendar_heatmap(df):
# 制作日历热力图
df['Date'] = pd.to_datetime(df['Date'])
start_date = df['Date'].min()
end_date = df['Date'].max()
dates = date_range(start_date, end_date)
july.heatmap(dates=dates,
data=df['Date'].value_counts().sort_index(),
cmap='Pastel1',
month_grid=True,
horizontal=True,
value_label=False,
date_label=False,
weekday_label=True,
month_label=True,
year_label=True,
colorbar=False,
fontfamily="monospace",
fontsize=12,
title=None,
titlesize='large',
dpi=100)
plt.tight_layout()
plt.show()
def analyze_message_comparison(df):
# 双方信息数量对比
sent_by_me = df[df['IsSender'] == 1]['StrContent']
sent_by_others = df[df['IsSender'] == 0]['StrContent']
count_sent_by_me = len(sent_by_me)
count_sent_by_others = len(sent_by_others)
labels = ['你的名字', '聊天对象的名字']
sizes = [count_sent_by_me, count_sent_by_others]
colors = ['#FF6347', '#9ACD32']
explode = (0, 0.05)
plt.rc('font', family='YouYuan')
plt.pie(sizes, explode=explode, labels=labels, colors=colors, autopct='%1.1f%%', shadow=True, startangle=90)
plt.axis('equal')
plt.title('Comparison of the number of chats')
plt.legend()
plt.show()
def analyze_hourly_chat_frequency(df):
# 根据一天中的每一个小时进行统计聊天频率,并生成柱状图
df['DateTime'] = pd.to_datetime(df['StrTime'])
df['Hour'] = df['DateTime'].dt.hour
hourly_counts = df['Hour'].value_counts().sort_index().reset_index()
hourly_counts.columns = ['Hour', 'Frequency']
plt.figure(figsize=(10, 8))
plt.rc('font', family='YouYuan')
ax = sns.barplot(x='Hour', y='Frequency', data=hourly_counts, color="#E6AAAA")
sns.kdeplot(df['Hour'], color='#C64F4F', linewidth=1, ax=ax.twinx())
plt.title('Chat Frequency by Hour')
plt.xlabel('Hour of the Day')
plt.ylabel('Frequency')
plt.show()
def is_chinese_word(word):
for char in word:
if not re.match(r'[\u4e00-\u9fff]', char):
return False
return True
def correct(a, stop_words):
b = []
for word in a:
if len(word) > 1 and is_chinese_word(word) and word not in stop_words:
b.append(word)
return b
def word_fre_draw(a, str):
a_counts = Counter(a)
top_30_a = a_counts.most_common(30)
words, frequencies = zip(*top_30_a)
# 绘制水平柱状图
plt.figure(figsize=(10, 15))
plt.barh(words, frequencies, color='skyblue')
plt.xlabel('Frequency')
plt.ylabel('Words')
plt.title('Top 30 Words in Chat Messages for {0}'.format(str))
plt.show()
def word_frequency_analysis(df):
sent_by_me_text = ' '.join(df[df['IsSender'] == 1]['StrContent'].astype(str))
sent_by_others_text = ' '.join(df[df['IsSender'] == 0]['StrContent'].astype(str))
all_text = ' '.join(df['StrContent'].astype(str))
words = list(jieba.cut(all_text, cut_all=False))
my_words = list(jieba.cut(sent_by_me_text, cut_all=False))
others_words = list(jieba.cut(sent_by_others_text, cut_all=False))
with open('stopwords_hit.txt', encoding='utf-8') as f: # 添加屏蔽词汇
con = f.readlines()
stop_words = set() # 集合可以去重
for i in con:
i = i.replace("\n", "") # 去掉读取每一行数据的\n
stop_words.add(i)
Words = correct(words, stop_words)
My_words = correct(my_words, stop_words)
others_words = correct(others_words, stop_words)
words_space_split = ' '.join(Words)
word_fre_draw(Words, 'All')
word_fre_draw(My_words, '你的名字')
word_fre_draw(others_words, '他/她的名字')
return words_space_split
def word_cloud(words_space_split):
wordcloud = WordCloud(font_path='C:\Windows\Fonts\STCAIYUN.TTF',
width=800, height=600,
background_color='white',
max_words=200,
max_font_size=100,
).generate(words_space_split)
plt.figure(figsize=(10, 8))
plt.imshow(wordcloud, interpolation='bilinear')
plt.axis('off')
plt.show()
def analyze_weekly_contribution(df):
df['Weekday'] = df['StrTime'].dt.day_name()
# 计算每天的消息数量
weekday_counts = df['Weekday'].value_counts().reindex([
"Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday", "Sunday"
])
# 找出频率最高的那天
max_day = weekday_counts.idxmax()
# 制作饼状图
plt.figure(figsize=(8, 8))
explode = [0.1 if day == max_day else 0 for day in weekday_counts.index] # 突出显示频率最高的那天
plt.pie(weekday_counts, labels=weekday_counts.index, explode=explode, autopct='%1.1f%%',
startangle=140, colors=plt.cm.Paired.colors)
plt.title('Distribution of Messages During the Week')
plt.show()
def analyze_most_active_day_and_month(df):
df['Date'] = pd.to_datetime(df['Date'])
df['YearMonth'] = df['Date'].dt.to_period('M')
df['Day'] = df['Date'].dt.date
daily_counts = df['Day'].value_counts()
max_day = daily_counts.idxmax()
max_day_count = daily_counts.max()
monthly_counts = df['YearMonth'].value_counts()
max_month = monthly_counts.idxmax()
max_month_count = monthly_counts.max()
print(f"Most active day: {max_day}, with {max_day_count} messages.")
print(f"Most active month: {max_month}, with {max_month_count} messages.")
if __name__ == "__main__":
set_chinese_font()
df = read_chat_data('CSV文件') # 加载数据集
df = preprocess_data(df) # 数据预处理
plot_chat_frequency_by_day(df) # 绘制每日聊天频率柱状图
plot_calendar_heatmap(df) # 绘制日历热力图
analyze_message_comparison(df) # 消息占比对比
analyze_hourly_chat_frequency(df) # 每小时聊天频率柱状图
words = word_frequency_analysis(df) # 词汇频率分析
word_cloud(words) # 词云制作
analyze_weekly_contribution(df) # 每周聊天频率
analyze_most_active_day_and_month(df) # 聊天最多的月和天
文件中引用有停词文件,可以从GitHub上下载你想使用的(差不多都一样,可以在文件中添加新的屏蔽词语)。停词文件和代码文件放在同一目录下:
GitHub - goto456/stopwords: 中文常用停用词表(哈工大停用词表、百度停用词表等)
然后直接运行代码就可以等着一张一张的图片展示啦
各位有任何问题评论区欢迎提问