无法使用 Matplotlib 和 Python 在 x 轴上绘制带有时间的直方图

Question

I am trying to plot the number of times a user tweeted at specific times of the day. I plan to plot these on a histogram/bar chart with 24 "bins" - one for each hour.

I have the data in a Pandas dataframe in two columns - the tweet and the time of the tweet (as a datetime object).

I have converted the Time column into a Pandas time, however I am having a hard time plotting correctly. If I set the value of bins to be 24, then I get the following chart ( here ) which doesn't look correct. Firstly the chart looks wrong, but secondly the x-axis has horrible formatting.

I would like to try to resolve these two issues. Firstly the data isn't being plotted correctly and secondly the horizontal axis formatting is incorrect.

I have plotted the data using Google Sheets and the correct chart should look like this . I don't mind if the values are % of total or absolute volume.

Code to generate plots can be found here. generate_data.py and plot_data.py

Any help is hugely appreciated.

plot_data.py

import datetime
import matplotlib.dates as mdates
import matplotlib.pyplot as plt
import pandas as pd
import random

import generate_data


screen_name = "@joebiden"
data = generate_data.get_data(screen_name, save_output=True)

df = pd.DataFrame(data)
df["Time"]= pd.to_datetime(data["Time"], format="%H:%M") 

fig, ax = plt.subplots(1,1)
bin_list = [datetime.time(x) for x in range(24)]

ax.hist(df["Time"], bins=24, color='lightblue')
plt.show()

generate_data.py

import json
import re
from datetime import datetime

import tweepy

import common_words
import twitter_auth 



def create_connection():
    auth = tweepy.OAuthHandler(twitter_auth.CONSUMER_KEY, twitter_auth.CONSUMER_SECRET)
    auth.set_access_token(twitter_auth.ACCESS_KEY, twitter_auth.ACCESS_SECRET)
    return tweepy.API(auth, wait_on_rate_limit=True, wait_on_rate_limit_notify=True)


def retrieve_next_set_of_tweets(screen_name, api, count,max_id):
    '''Return next 200 user tweets'''
    return api.user_timeline(screen_name=screen_name,count=count, tweet_mode='extended', max_id=max_id)


def get_tweet_times(screen_name, api):
    user_tweet_count = api.get_user(screen_name).statuses_count

    all_tweets = {'Tweet':[], 'Time':[]}
    block_of_tweets = api.user_timeline(screen_name=screen_name,count=200, tweet_mode='extended')
    all_tweets["Tweet"].extend([tweet.full_text for tweet in block_of_tweets])
    all_tweets["Time"].extend([tweet.created_at for tweet in block_of_tweets])
    oldest = block_of_tweets[-1].id - 1

    
    while block_of_tweets:    
        try:
            block_of_tweets = retrieve_next_set_of_tweets(screen_name, api, 200, oldest)
            oldest = block_of_tweets[-1].id - 1
        except IndexError: #Reached limit of 3245
            pass
        # all_tweets.update({tweet.full_text: tweet.created_at.time() for tweet in block_of_tweets}) 
        all_tweets["Tweet"].extend([tweet.full_text for tweet in block_of_tweets])
        all_tweets["Time"].extend([tweet.created_at for tweet in block_of_tweets])
 
    return all_tweets


def get_all_tweets(screen_name, api):
    user_tweet_count = api.get_user(screen_name).statuses_count

    all_tweets = []
    block_of_tweets = api.user_timeline(screen_name=screen_name,count=200, tweet_mode='extended')
    all_tweets.extend([tweet.full_text for tweet in block_of_tweets])
    oldest = block_of_tweets[-1].id - 1
    

    while block_of_tweets:    
        try:
            block_of_tweets = retrieve_next_set_of_tweets(screen_name, api, 200, oldest)
            oldest = block_of_tweets[-1].id - 1
        except IndexError: #Reached limit of 3245
            pass

        all_tweets.extend([tweet.full_text for tweet in block_of_tweets]) 
    return all_tweets


def parse_all_tweets(tweet_list, max_words_to_show=50):
    tweet_dict = {}
    regex = re.compile('[^a-zA-Z ]')
    for tweet in tweet_list: 
        text = regex.sub("", tweet).lower().strip().split()
        
        for word in text:
            if word in common_words.MOST_COMMON_ENGLISH_WORDS: continue
            if word in tweet_dict.keys():
                tweet_dict[word] += 1
            else:
                if len(tweet_dict.items()) == max_words_to_show:
                    return tweet_dict
                tweet_dict[word] = 1
    return tweet_dict
    
    
def get_data(screen_name, words_or_times="t", save_output=False):
    api = create_connection()
    print(f"...Getting max of 3245 tweets for {screen_name}...")

    if words_or_times == "t":
        all_tweets = get_tweet_times(screen_name, api)
        suffix = "tweet_times"
        
    elif words_or_times == "w":
        suffix = "ranked_words"
        parsed_tweets = parse_all_tweets(get_all_tweets(screen_name, api))
        parsed_tweets = {k:v for k,v in sorted(parsed_tweets.items(), key=lambda item: item[1], reverse=True)}

    else:
        return "...Error. Please enter 't' or 'w' to signify 'times' or 'words'."

    if save_output:
        f_name = f"{screen_name}_{suffix}.json"
        with open(f_name, "w") as f:
            json.dump(all_tweets if words_or_times == "t" else parsed_tweets, f, indent=4, default=str)
        
        print(f"...Complete! File saved as '{f_name}'")

    return all_tweets if words_or_times == "t" else parsed_tweets


if __name__ == "__main__":
    get_data(screen_name="@joebiden", save_output=True) 

Answer 1

OK. So you want to have time only from your date time. Try replacing

df["Time"]= pd.to_datetime(data["Time"], format="%H:%M")

With

df['Time'] = pd.to_datetime(df['Time'],format= '%H:%M' ).dt.time

Answer 2

I have tried plotting the data you have shared in the comment to the answer by Serge de Gosson de Varennes. The only thing I needed to change in your plot_data.py script was the date format where I added seconds. The rest worked as expected, the times are processed correctly for the x-axis.

Here is an example where the histogram is created with pd.Series.hist for convenience. The weights argument is included to produce a graph with percentages. Most of the code is for formatting:

import numpy as np                 # v 1.19.2
import pandas as pd                # v 1.1.3
import matplotlib.pyplot as plt    # v 3.3.2
import matplotlib.dates as mdates

# Import data from html table into pandas dataframe
url = 'https://docs.google.com/spreadsheets/d/e/2PACX-1vTc97VEzlfDP_jEkjC7dTbJzcLBLDQeFwPMg6E36BaiH5qkhnedSz8wsVGUMyW6kt85rD20BcTMbvqp/pubhtml'
table, = pd.read_html(url, header=[1], index_col=1)
df = table.iloc[:, 1:]

# Save time variable as a pandas series of datetime dtype
time_var = pd.to_datetime(df['Time'], format='%H:%M:%S')

# Plot variable with pandas histogram function
ax = time_var.hist(bins=24, figsize=(10,5), grid=False, edgecolor='white', zorder=2,
                   weights=np.ones(time_var.size)/time_var.size)

# Format x and y-axes tick labels
ax.xaxis.set_major_formatter(mdates.DateFormatter('%H:%M'))
ax.yaxis.set_major_formatter('{x:.1%}')

# Additional formatting
alpha = 0.3
ax.grid(axis='y', zorder=1, color='black', alpha=alpha)
for spine in ['top', 'right', 'left']:
    ax.spines[spine].set_visible(False)
ax.spines['bottom'].set_alpha(alpha)
ax.tick_params(axis='x', which='major', color=[0, 0, 0, alpha])
ax.tick_params(axis='y', which='major', length=0)
ax.set_title('Tweets sent per hour of day in UTC', fontsize=14, pad=20)
ax.set_ylabel('Relative frequency (% of total)', size=12, labelpad=10)

plt.show()

Because the counts are spread over 24 hours in this histogram, you may notice that the height of the bars are slightly different from those in the histogram in the image you have shared as a reference where the counts seem to be grouped into 23 bins instead of 24.

---

Reference: this answer by ImportanceOfBeingErnest