I was curious about my watching habits, and since YouTube doesn't provice a service like this itself, I had to do it myself. YouTube Rewind exists, but that's not personal. What I did was like a 'YouTube 2020-2023 Recapped' Project. After gathering the data, I could ask questions to myself I was curious about and find ways to answer them. I wondered stuff like 'whose videos I watch the most?', or 'how much educational content do I watch by percentage?'.
The beginning data for this project was taken from Google Takeout
However, that only contained an .html file with only 5 colums. ('video_link', 'channel_link', 'video_title', 'channel_name', 'watch_date_time')
The column 'video_link' was essential though, as it made it possible to webscrape additional data about the videos.
('subscribed', 'video_length', 'view_count', 'like_count', 'comment_count', 'description', 'description_length', 'category', 'tags', 'video_quality', 'is_shorts', 'publish_date')
By doing all this, I got a .csv file with 17 columns and 52K rows with lots of data points I could potentially analyze.
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
import numpy as np
from matplotlib.ticker import MaxNLocator
import matplotlib.dates as mdates
import scipy.stats as stats
import re
import nltk
from nltk.corpus import stopwords# Load the CSV file into a DataFrame
csv_file_path = 'output2.csv' # Replace with the actual path to your CSV file
df = pd.read_csv(csv_file_path)
# Drop YouTube shorts
#df.drop(df[df['is_shorts'] == True].index, inplace=True) #Optional toggle that changes everything that comes after thisprint(df.head(3)) video_link \
0 https://www.youtube.com/watch?v=zlzzO1e6dws
1 https://www.youtube.com/watch?v=GUAQebYNzAQ
2 https://www.youtube.com/watch?v=GEDA4TmW44s
channel_link \
0 https://www.youtube.com/channel/UCRHZGz8g6b10r...
1 https://www.youtube.com/channel/UCausCHptqoa4o...
2 https://www.youtube.com/channel/UC4_SUDiYxC8xs...
video_title channel_name \
0 How to Download Your YouTube History! Downloa... Scoby Tech
1 I Got Vac Banned... Here's Why Gomer
2 Hydro Plant Surge Tank!!! #shorts Fearjames
watch_date_time subscribed video_length view_count like_count \
0 Dec 09, 2023, 19:03:02 False 02:08 21110.0 438000.0
1 Dec 09, 2023, 19:01:38 False 02:51 1429.0 28000.0
2 Dec 09, 2023, 17:57:36 True 00:38 1492.0 102000.0
comment_count description \
0 47.0 Hey guys my name is Scoby and in todays video ...
1 30.0 I swear to Gaben if I don't get my skins back ...
2 530.0 NaN
description_length category \
0 2770 Education
1 107 People & Blogs
2 0 People & Blogs
tags video_quality is_shorts \
0 ['scoby', 'youtube history', 'youtube history ... 1440p False
1 ['counter strike 2', 'vac ban cs2', 'van ban',... 1080p False
2 [] 1080p True
publish_date
0 Aug 22, 2018, 10:30:00
1 Dec 05, 2023, 11:51:45
2 Nov 05, 2023, 05:13:30
# Display DataFrame info
df.info()<class 'pandas.core.frame.DataFrame'>
RangeIndex: 52552 entries, 0 to 52551
Data columns (total 17 columns):
# Column Non-Null Count Dtype
--- ------ -------------- -----
0 video_link 52552 non-null object
1 channel_link 49947 non-null object
2 video_title 52552 non-null object
3 channel_name 49947 non-null object
4 watch_date_time 52547 non-null object
5 subscribed 52552 non-null bool
6 video_length 49840 non-null object
7 view_count 12871 non-null float64
8 like_count 49504 non-null float64
9 comment_count 6487 non-null float64
10 description 45412 non-null object
11 description_length 52552 non-null int64
12 category 49840 non-null object
13 tags 52552 non-null object
14 video_quality 48201 non-null object
15 is_shorts 52552 non-null bool
16 publish_date 49840 non-null object
dtypes: bool(2), float64(3), int64(1), object(11)
memory usage: 6.1+ MB
# Group the data by channel_name and count the number of videos watched for each channel
channel_watch_count = df.groupby('channel_name')['video_title'].count().reset_index()
# Sort the channels by the number of videos watched in descending order
sorted_channels = channel_watch_count.sort_values(by='video_title', ascending=False)
# Select the top 10 channels
top_10_channels = sorted_channels.head(10)
# Display the result without the index column and with a more descriptive column name
print("Top 10 YouTubers based on the number of videos + shorts watched:")
top_10_channels = top_10_channels.rename(columns={'video_title': 'watched_videos'})
print(top_10_channels[['channel_name', 'watched_videos']].to_string(index=False))Top 10 YouTubers based on the number of videos + shorts watched:
channel_name watched_videos
Mental Outlaw 445
The Gaming Merchant 383
penguinz0 313
SomeOrdinaryGamers 295
Also Gaming Merchant 268
Mok3ysnip3r 230
Linus Tech Tips 225
Daily Dose Of Internet 214
SeeOk - Clash Royale 211
Hero Hei 210
# Separate the data into two DataFrames based on is_shorts value
shorts_df = df[df['is_shorts'] == True]
not_shorts_df = df[df['is_shorts'] == False]
# Function to get top 10 YouTubers based on the number of videos watched
def get_top_youtubers(data):
channel_watch_count = data.groupby('channel_name')['video_title'].count().reset_index()
sorted_channels = channel_watch_count.sort_values(by='video_title', ascending=False)
return sorted_channels.head(10)
# Get top 10 YouTubers for shorts and not shorts
top_10_shorts = get_top_youtubers(shorts_df)
top_10_not_shorts = get_top_youtubers(not_shorts_df)
# Display the results for normal videos without the index column and with a more descriptive column name
print("\nTop 10 YouTubers for normal videos:")
top_10_not_shorts = top_10_not_shorts.rename(columns={'video_title': 'watched_videos'})
print(top_10_not_shorts[['channel_name', 'watched_videos']].to_string(index=False))
# Display the results for shorts without the index column and with a more descriptive column name
print("\nTop 10 YouTubers for shorts:")
top_10_shorts = top_10_shorts.rename(columns={'video_title': 'watched_shorts'})
print(top_10_shorts[['channel_name', 'watched_shorts']].to_string(index=False))Top 10 YouTubers for normal videos:
channel_name watched_videos
Mental Outlaw 444
penguinz0 313
SomeOrdinaryGamers 295
Also Gaming Merchant 267
The Gaming Merchant 254
Mok3ysnip3r 230
SeeOk - Clash Royale 211
Hero Hei 210
Pants are Dragon 209
Grrt 206
Top 10 YouTubers for shorts:
channel_name watched_shorts
camman18 166
Action Lab Shorts 155
filian 138
Doug Sharpe 138
The Gaming Merchant 129
mryeester 121
UFD Tech 89
JesseLHV 83
Vsauce 81
Ugo Lord: Modern Age Attorney 76
I knew Mental Outlaw would show up as my favorite Youtuber before I even started the project.
My hypothesis about them being my favorite channel turned out to be correct.
# Count the number of shorts and normal videos
shorts_count = df[df['is_shorts'] == True]['video_link'].nunique()
normal_count = df[df['is_shorts'] == False]['video_link'].nunique()
# Create a pie chart
labels = ['Shorts', 'Normal Videos']
counts = [shorts_count, normal_count]
plt.figure(figsize=(8, 8))
plt.pie(counts, labels=labels, autopct='%1.1f%%', startangle=140)
plt.title('Ratio of Shorts to Normal Videos Watched')
plt.show()
Graph looks like pacman lol
# Count the number of videos in each category
category_counts = df['category'].value_counts()
# Filter out small segments (e.g., less than 2%)
threshold = 0.02
small_segments = category_counts[category_counts / category_counts.sum() < threshold]
category_counts = category_counts[category_counts / category_counts.sum() >= threshold]
category_counts['Other'] = small_segments.sum()
# Display the main categories and their percentages
print("Main Categories:")
print(category_counts)
# Create a pie chart with improved label placement for the main categories
plt.figure(figsize=(10, 6))
plt.pie(category_counts, labels=None, autopct='%1.1f%%', startangle=140)
# Display category labels outside the pie chart with better positioning
plt.gca().legend(category_counts.index, loc='center left', bbox_to_anchor=(1, 0.5))
plt.title('Distribution of Video Categories')
plt.axis('equal') # Equal aspect ratio ensures that the pie is drawn as a circle.
plt.show()
# Display the details of the 'Other' category
if 'Other' in category_counts.index:
other_details = df[df['category'].isin(small_segments.index)]
print("\nDetails of 'Other' category:")
print(other_details[['category', 'video_title']])
else:
print("\nNo 'Other' category present.")Main Categories:
Gaming 20808
Entertainment 8276
People & Blogs 5592
Science & Technology 4064
Education 2931
Music 2416
Film & Animation 2050
Comedy 1639
Other 2064
Name: category, dtype: int64
Details of 'Other' category:
category video_title
3 Travel & Events The Reason Warts Shouldn't Be Cut Off 🤔
15 Pets & Animals Cat Wall Training #outdoorsavannah
28 Travel & Events Why American Eggs Have To be Refrigerated 😬
47 Pets & Animals russian tortoise drinking water | sulcata tor...
58 Pets & Animals Demonstration of Spin 1/2
... ... ...
52471 Howto & Style Finding Out I'm Pregnant *AFTER 9 YRS*
52474 Pets & Animals watame factory
52478 News & Politics Süt fabrikasında banyo skandalı!
52479 News & Politics Bu eve nasıl 'az hasarlı' raporu verilir?
52489 News & Politics Scientists discover bizarre hell planet where...
[2064 rows x 2 columns]
# Extract and flatten the video tags
all_tags = [tag for tags in df['tags'].dropna() for tag in eval(tags)]
# Create a DataFrame to count tag occurrences
tag_counts = pd.Series(all_tags).value_counts()
# Display the top N tags
top_n = 14
top_tags = tag_counts.head(top_n)
# Plot a bar chart
plt.figure(figsize=(12, 6))
top_tags.plot(kind='bar', color='skyblue')
plt.title(f'Top {top_n} Video Tags')
plt.xlabel('Tag')
plt.ylabel('Occurrences')
plt.xticks(rotation=45, ha='right')
plt.show()
# Convert video length to total seconds
def convert_to_seconds(time_str):
if pd.isna(time_str):
return np.nan
minutes, seconds = map(int, time_str.split(':'))
return minutes * 60 + seconds
df['video_length_seconds'] = df['video_length'].apply(convert_to_seconds)
# Split the dataframe for all videos and for non-shorts videos
df_all_videos = df
df_shorts = df[df['is_shorts'] == True]
df_not_shorts = df[df['is_shorts'] == False]
# Function to create a histogram plot with toggle for mean or median
def create_histogram(dataframe, title, ax, bin_size, binmax, use_median=True):
# Calculate central tendency based on the toggle
if use_median:
central_value = dataframe['video_length_seconds'].median()
central_label = 'Median'
else:
central_value = dataframe['video_length_seconds'].mean()
central_label = 'Mean'
# Binning configuration
bins_range = (0, binmax)
# Plot histogram
counts, bins, _ = ax.hist(dataframe['video_length_seconds'].dropna(), bins=bin_size, range=bins_range, color='skyblue', edgecolor='black')
# Central tendency line
ax.axvline(central_value, color='red', linestyle='dashed', linewidth=2, label=f'{central_label}: {int(central_value)} seconds')
# Title and labels
ax.set_title(title)
ax.set_xlabel('Video Length (seconds)')
ax.set_ylabel('Frequency')
ax.legend()
# Ticks and grid
ax.xaxis.set_major_locator(MaxNLocator(integer=True))
ax.grid(True, linestyle='--', linewidth=0.5, alpha=0.7)
# Enable minor ticks on the x-axis
plt.minorticks_on()
# Annotation for the bin with the maximum count
max_count_index = np.argmax(counts)
max_bin_value = bins[max_count_index]
ax.annotate(f'{int(max_bin_value)} seconds', xy=(max_bin_value, counts[max_count_index]),
xytext=(8, 3), textcoords='offset points',
arrowprops=dict(arrowstyle='->', connectionstyle='arc3,rad=.5'), fontsize=8)
# Add a toggle for median and mean
use_median_toggle = True # Set to False to use mean instead of median | extreme outliers ruined the mean results, that's why I opted to use median.
# Create figure and a single set of axes
fig, axs = plt.subplots(figsize=(10, 6))
# Plot histograms
create_histogram(df_all_videos, 'Histogram of Video Duration (All Videos)', axs, 40, 600, use_median=use_median_toggle)
# Set x ticks every 60 seconds
plt.xticks(np.arange(0, 601, 60))
# Display the plot
plt.show()
# Create figure and a single set of axes
fig, axs = plt.subplots(figsize=(10, 6))
# Call the function with a single set of axes
create_histogram(df_not_shorts, 'Histogram of Video Duration (Videos Only)', axs, 40, 780, use_median=use_median_toggle)
# Set x ticks every 60 seconds
plt.xticks(np.arange(0, 801, 60))
# Display the plot
plt.show()
seed_value = 8
shuffled_df_not_shorts_15_seconds = df_not_shorts[df_not_shorts['video_length_seconds'] == 15].sample(frac=1, random_state=seed_value)
# Display the first few rows of the shuffled DataFrame
print(shuffled_df_not_shorts_15_seconds.head()) video_link \
49838 https://www.youtube.com/watch?v=bPgi39vBeGM
49011 https://www.youtube.com/watch?v=1PQ2LC5O3dE
49069 https://www.youtube.com/watch?v=17UOd-EqZPU
44072 https://www.youtube.com/watch?v=ull5YaEHvw0
755 https://www.youtube.com/watch?v=BJUbLM8coho
channel_link \
49838 https://www.youtube.com/channel/UCF5g30i--vn34...
49011 https://www.youtube.com/channel/UC9wxb6Q1-QqTb...
49069 https://www.youtube.com/channel/UCXXU8EZGvlw5x...
44072 https://www.youtube.com/channel/UCgypNw4RJdgOA...
755 https://www.youtube.com/channel/UCOG9EOz-VHx6I...
video_title channel_name \
49838 Mr Krabs Overdoses on ketamine speedrun Death... ImJackTV
49011 [Touhou Parody Short] Okuu's Commercial Holman Animations
49069 jett cranos
44072 Are you from Russia? Raezra
755 Chainsaw Man Author Tries To Levitate El Alumno de Atrás
watch_date_time subscribed video_length view_count \
49838 Dec 21, 2020, 21:47:59 False 00:15 NaN
49011 Jan 05, 2021, 00:35:05 False 00:15 NaN
49069 Jan 04, 2021, 12:27:05 True 00:15 NaN
44072 Apr 06, 2021, 21:57:13 False 00:15 NaN
755 Oct 28, 2023, 01:12:47 False 00:15 5095.0
like_count comment_count \
49838 16000.0 NaN
49011 63000.0 NaN
49069 2000.0 NaN
44072 458000.0 NaN
755 251000.0 NaN
description description_length \
49838 Sub to me please 16
49011 「 DOUBLE SUN POWER ! !」\nOriginal Audio by 'Ol... 174
49069 GPU: GeForce RTX 2080\nCPU: Intel(R) Core(TM) ... 182
44072 This video is not mine it is from Anomaly's ch... 239
755 Tatsuki Fujimoto is the mangaka of Chainsaw Ma... 271
category tags \
49838 Entertainment ['Gaming', 'Game', 'Mr Krabs Overdoses on keta...
49011 Film & Animation []
49069 Gaming ['#GeForceGTX', '#ShotWithGeForce', '#valorant...
44072 Gaming ['Anomaly', 'CS:GO', 'Russia', 'Anomaly CS:GO']
755 Entertainment []
video_quality is_shorts publish_date video_length_seconds
49838 1080p False Dec 17, 2020, 12:35:34 15.0
49011 1080p False Aug 28, 2017, 08:51:00 15.0
49069 1440p False Jan 03, 2021, 08:27:25 15.0
44072 720p False Mar 17, 2018, 14:04:20 15.0
755 1080p False Dec 13, 2021, 14:46:59 15.0
The abundance of videos around the 15 second mark for non-short videos concerned me, so I analyzed it manually, and it seems legit.
# Create figure and a single set of axes
fig, axs = plt.subplots(figsize=(10, 6))
# Call the function with a single set of axes
create_histogram(df_shorts, 'Histogram of Video Duration (Shorts Only)', axs, 12, 60, use_median=True)
# Set x ticks every 5 seconds (adjust as needed)
axs.set_xticks(np.arange(0, 61, 5))
# Enable minor ticks on the x-axis every 1 second (adjust as needed)
axs.xaxis.set_minor_locator(plt.MultipleLocator(1))
# Display the plot
plt.show()
# Calculate mode
video_lengths = df_shorts['video_length_seconds'].dropna()
mode_result = stats.mode(video_lengths, axis=None, keepdims=True) # Explicitly set keepdims
mode_value = mode_result.mode[0]
mode_count = mode_result.count[0]
# Display mode information
print(f"Mode: {mode_value} seconds (Count: {mode_count})")
Mode: 59.0 seconds (Count: 1413)
As I expected, YouTube shorts tend to aim around the 60 second mark, but the existance of even shorter shorts brings down the median to 39 seconds.
df['video_length_seconds'] = df['video_length'].apply(convert_to_seconds)
# Find the index of the video with the longest duration
max_length_index = df['video_length_seconds'].idxmax()
min_length_index = df['video_length_seconds'].idxmin()
# Get the row with the longest duration
longest_video = df.loc[max_length_index]
shortest_video = df.loc[min_length_index]
# Convert the duration of the longest video to days, hours, minutes, and seconds
duration_seconds = longest_video['video_length_seconds']
days, remainder = divmod(duration_seconds, 86400)
hours, remainder = divmod(remainder, 3600)
minutes, seconds = divmod(remainder, 60)
# Print information about the shortest video
print("Shortest Video:")
print(f"Video Title: {shortest_video['video_title']}")
print(f"Video Link: {shortest_video['video_link']}")
print(f"Video Length: {shortest_video['video_length']}")
print(f"Video Length (seconds): {shortest_video['video_length_seconds']} seconds")
# Print information about the longest video
print("\nLongest Video:")
print(f"Video Title: {longest_video['video_title']}")
print(f"Video Link: {longest_video['video_link']}")
print(f"Video Length: {longest_video['video_length']}")
print(f"Video Duration: {days} days, {hours} hours, {minutes} minutes, {seconds} seconds")Shortest Video:
Video Title: CHILIBEAN.mp4
Video Link: https://www.youtube.com/watch?v=lLFHnkgfnes
Video Length: 00:00
Video Length (seconds): 0.0 seconds
Longest Video:
Video Title: lofi hip hop radio - beats to relax/study to
Video Link: https://www.youtube.com/watch?v=5qap5aO4i9A
Video Length: 1250623:50
Video Duration: 868.0 days, 11.0 hours, 43.0 minutes, 50.0 seconds
Lo-fi hip hop radio livestreamed music 24/7 on this exact link for 2.37 years! That's so long that YouTube won't even let you watch it anymore...
# Convert 'publish_date' and 'watch_date_time' columns to datetime
df['publish_date'] = pd.to_datetime(df['publish_date'])
df['watch_date_time'] = pd.to_datetime(df['watch_date_time'])
# Sort DataFrame by 'publish_date' and 'watch_date_time' in ascending order
df_sorted_publish = df.sort_values(by='publish_date', ascending=True).reset_index(drop=True)
df_sorted_watch = df.sort_values(by='watch_date_time', ascending=True).reset_index(drop=True)
# Get the oldest video by publish date
oldest_by_publish = df_sorted_publish.iloc[0]
# Get the oldest video by watch time
oldest_by_watch = df_sorted_watch.iloc[0]
print("Oldest video by publish date:")
print(oldest_by_publish[['video_title', 'video_link', 'publish_date']])
print("\nOldest video by watch time:")
print(oldest_by_watch[['video_title', 'video_link', 'watch_date_time']])Oldest video by publish date:
video_title Me at the zoo
video_link https://www.youtube.com/watch?v=jNQXAC9IVRw
publish_date 2005-04-23 20:31:52
Name: 0, dtype: object
Oldest video by watch time:
video_title ina 1000iq braincell moment
video_link https://www.youtube.com/watch?v=Xw56mDCf8hw
watch_date_time 2020-11-05 04:11:51
Name: 0, dtype: object
Of course the oldest video by publish date is "Me at the zoo", it's the oldest video on YouTube after all and I have seen it.
..and the oldest video by watch time is on the day I decided to turn on my YouTube watch history again, I don't know why I had it frozen and cleared, I regret that decision now, I could've had way more data, although 3-4 years of data is already huge and even that took me a while to crunch through, 8 hours of processing time in total, ran it on two processes so it took only 4 in reality, if I could implement multiprocessing it would have been way faster though.
# Convert 'watch_date_time' to a datetime object
df['watch_date_time'] = pd.to_datetime(df['watch_date_time'])
# Create new columns for day of the week, year, and month
df['day_of_week'] = df['watch_date_time'].dt.day_of_week
df['year'] = df['watch_date_time'].dt.year
df['month'] = df['watch_date_time'].dt.month
# Group by day of the week, year, and month and count the number of videos watched
videos_by_day_of_week = df.groupby('day_of_week').size()
videos_by_year = df.groupby('year').size()
videos_by_month = df.groupby('month').size()
# Plotting
fig, axs = plt.subplots(3, 1, figsize=(12, 15))
# Plot for videos watched by day of the week
axs[0].bar(videos_by_day_of_week.index, videos_by_day_of_week.values, color='skyblue')
axs[0].set_title('Videos Watched by Day of the Week')
axs[0].set_xlabel('Day of the Week')
axs[0].set_ylabel('Number of Videos')
axs[0].set_xticks(range(7))
axs[0].set_xticklabels(['Monday', 'Tuesday', 'Wednesday', 'Thursday', 'Friday', 'Saturday', 'Sunday'])
# Plot for videos watched by year
axs[1].plot(videos_by_year.index, videos_by_year.values, marker='o', linestyle='-', color='green')
axs[1].set_title('Videos Watched by Year')
axs[1].set_xlabel('Year')
axs[1].set_ylabel('Number of Videos')
axs[1].xaxis.set_major_locator(plt.MaxNLocator(integer=True)) # Show every year as an integer
axs[1].tick_params(axis='x', rotation=45) # Rotate the year labels for better visibility
# Plot for videos watched by month
axs[2].plot(videos_by_month.index, videos_by_month.values, marker='o', linestyle='-', color='orange')
axs[2].set_title('Videos Watched by Month')
axs[2].set_xlabel('Month')
axs[2].set_ylabel('Number of Videos')
# Use month names as x-axis tick labels
month_names = ['Jan', 'Feb', 'Mar', 'Apr', 'May', 'Jun', 'Jul', 'Aug', 'Sep', 'Oct', 'Nov', 'Dec']
axs[2].set_xticks(range(1, 13))
axs[2].set_xticklabels(month_names)
plt.tight_layout()
plt.show()
Videos watched being low in 2020 is due to me turning on my watch history amidst 2020 instead of at Jan 1. Other than that, it's weird seeing how days of the week and years don't really matter, but for months it matters a lot. My hypothesis for watch count being low in summer is that I play games/hang out with friends when I have lots of free time rather than trying to help my boredom by myself.
video_watch_counts = df['video_link'].value_counts()
top_11_watched_videos = video_watch_counts.head(11)
top_10_watched_videos = top_11_watched_videos.iloc[:10].drop(top_11_watched_videos.index[4])
print("Top 10 Most-Watched Videos:")
for rank, (video_link, watch_count) in enumerate(top_10_watched_videos.items(), start=1):
video_name = df[df['video_link'] == video_link]['video_title'].iloc[0]
print(f"{rank}. Video Link: {video_link}\t Video Title: {video_name}\t Watch Count: {watch_count}")Top 10 Most-Watched Videos:
1. Video Link: https://www.youtube.com/watch?v=FtutLA63Cp8 Video Title: 【東方】Bad Apple!! ＰＶ【影絵】 Watch Count: 17
2. Video Link: https://www.youtube.com/watch?v=JH23_kuLNMg Video Title: SVRGE - Nocturnal Watch Count: 16
3. Video Link: https://www.youtube.com/watch?v=UnIhRpIT7nc Video Title: 稲葉曇『ラグトレイン』Vo. 歌愛ユキ Watch Count: 16
4. Video Link: https://www.youtube.com/watch?v=hsXeFqj5p7Q Video Title: Defqwop - Heart Afire (ft. Strix) Watch Count: 16
5. Video Link: https://www.youtube.com/watch?v=PvT8Kx1WC64 Video Title: GTA IV - Soviet Connection (New mixed Intro) Watch Count: 12
6. Video Link: https://www.youtube.com/watch?v=lX44CAz-JhU Video Title: SIAMÉS "The Wolf" [Official Animated Music Video] Watch Count: 11
7. Video Link: https://www.youtube.com/watch?v=NMRhx71bGo4 Video Title: Let It Happen Watch Count: 10
8. Video Link: https://www.youtube.com/watch?v=qi3o0c6-Q0o Video Title: what.mp4 Watch Count: 10
9. Video Link: https://www.youtube.com/watch?v=32RJPfqxwFo Video Title: Hololive Rhythm Heaven Remix 10 Watch Count: 10
*It makes sense that the videos I watched on repeat are music videos.
# Filter videos with the "music" tag
music_df = df[df['tags'].apply(lambda tags: 'music' in tags)].copy()
# Extract information about each tag as a potential genre
music_df['tags'] = music_df['tags'].apply(lambda tags: [re.sub(r'\W+', '', tag.lower()) for tag in tags.split()])
# List of common genre names
common_genres = ['rock', 'pop', 'hiphop', 'rap', 'jazz', 'classical', 'country', 'metal', 'electronic', 'indie', 'breakcore', 'dubstep','melodic death metal', 'hyperpop', 'phonk', 'touhou', 'lofi']
# DO NOT INCLUDE ANIME| IT TOWERS ANYTHING ELSE
# Filter tags based on common genre names
music_df['tags'] = music_df['tags'].apply(lambda tags: [tag for tag in tags if tag in common_genres])
# Flatten the list of tags for counting
all_tags = [tag for tags in music_df['tags'] for tag in tags]
# Convert the list to a Pandas Series
tag_series = pd.Series(all_tags)
# Count occurrences of each tag
tag_counts = tag_series.value_counts()
# Visualize the distribution of music tags (top 10)
top_tags = tag_counts.head(15)
plt.figure(figsize=(10, 6))
top_tags.plot(kind='bar', color='skyblue')
plt.title('Top 15 Music Genres Watched')
plt.xlabel('Music Genre')
plt.ylabel('Number of Videos')
plt.show()
If you asked me in person I'd have just said 'electronic'...
Doing a 'pseudo-Spotify' analysis using YouTube was pretty cool.
# Convert 'watch_date_time' to a datetime object
df['watch_date_time'] = pd.to_datetime(df['watch_date_time'])
# Filter out shorts
non_shorts_df = df[df['is_shorts'] == False]
# Calculate the number of non-short videos watched per day
non_short_videos_per_day = non_shorts_df.groupby(non_shorts_df['watch_date_time'].dt.date)['video_link'].count()
# Find the day with the most non-short videos watched
most_watched_day = non_short_videos_per_day.idxmax()
most_watched_non_short_videos = non_shorts_df[non_shorts_df['watch_date_time'].dt.date == most_watched_day]
# Get the first and last watch date and time for the most-watched day
first_watch_time = most_watched_non_short_videos['watch_date_time'].min()
last_watch_time = most_watched_non_short_videos['watch_date_time'].max()
print("Day with the Most Non-Short Videos Watched:")
print(f"Date: {most_watched_day}")
print(f"Total Non-Short Videos Watched on that day: {non_short_videos_per_day[most_watched_day]}")
print(f"First Watch Time: {first_watch_time}")
print(f"Last Watch Time: {last_watch_time}")
# Calculate session length
session_length = last_watch_time - first_watch_time
print(f"Session Length: {session_length}")Day with the Most Non-Short Videos Watched:
Date: 2021-01-08
Total Non-Short Videos Watched on that day: 175
First Watch Time: 2021-01-08 00:03:33
Last Watch Time: 2021-01-08 23:58:02
Session Length: 0 days 23:54:29
I analyzed it manually as I couldn't believe the large number, and it turns out that it's a mix of having watched lots of videos that day on top of leaving music on to play in the background. Also it seems to have counted videos I watched the day before after midnight too.
# Filter out shorts
non_shorts_df = df[df['is_shorts'] == False]
# Define the allowed resolutions
allowed_resolutions = ['144p', '360p', '720p', '1080p', '1440p', '2160p']
# Filter out video qualities not in the allowed resolutions
filtered_df = non_shorts_df[non_shorts_df['video_quality'].isin(allowed_resolutions)]
# Count the occurrences of each video quality
video_quality_counts = filtered_df['video_quality'].value_counts()
# Create a list to explode specific slices
explode = [0.1 if resolution in ['144p'] else 0 for resolution in video_quality_counts.index]
# Create a pie chart
plt.figure(figsize=(8, 8))
plt.pie(video_quality_counts, labels=video_quality_counts.index, autopct='%1.1f%%', startangle=140, explode=explode)
plt.title('Distribution of Allowed Video Qualities Watched')
plt.show()
There being more 4K videos than 2K videos in my watch history is interesting. Also I feel like my dataset is so large that I can come to general conclusions just from my own data.
# Remove newline characters from the description
df['clean_description'] = df['description'].replace('\n', ' ', regex=True)
# Compute the length of the cleaned description
df['clean_description_length'] = df['clean_description'].str.len()
# Find the index of the video with the longest cleaned description
max_clean_desc_index = df['clean_description_length'].idxmax()
# Get the row with the longest cleaned description
video_with_longest_clean_desc = df.loc[max_clean_desc_index]
# Print information about the video with the longest cleaned description
print("Video with the Longest Cleaned Description:")
print(f"Video Link: {video_with_longest_clean_desc['video_link']}")
print(f"Cleaned Description Length: {video_with_longest_clean_desc['clean_description_length']} characters")
print(f"Cleaned Description:\n{video_with_longest_clean_desc['clean_description']}")Video with the Longest Cleaned Description:
Video Link: https://www.youtube.com/watch?v=QQZ_k15sPag
Cleaned Description Length: 9634.0 characters
Cleaned Description:
• I Can't Post Selfies Anymore... he said he isnt there, heres proof he is\nobviously this is a scene from smiling friends, and its not muta ofc, just some character.\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n\n • You found moonzeri!!
That is actually the LONGEST description, you can click on the video and see how long it is.
Turns out, I don't watch as much educational content as I thought I did. I honestly expected to see more of it as that's what I've been interested in lately, but 3 years of entertainment content squashed those results.
I've obtained great insights about my viewing habits, maybe even learned a life-changing lesson or two about how now I feel bad about the type of content I watch and how I should transition towards more useful and educative content for the better. Also the amount is insane too..
Anothing finding for me was that... All of this was very fun! Since I was analyzing my own data, my curiosity about my own self gave me lots of inner motivation to dig down deeper and explore more. Every part of the process was fun (except waiting 4 + 4 = 4 hours for my data to be processed and pulling an all nighter to finish the project in a day), since I'm working with YouTube videos I've watched, some parts were nostalgic, and I dug out hidden gems and funny videos time and time again.
The way I scraped like, view and comment counts seemed to return valid results so I didn't notice them being kind of broken, the data for them in some columns are correct while some are straight up wrong, that's why I decided to avoid doing any analysis regarding those. I'd fix them and do all of this again but it takes 8 hours to process 52000 videos by sending requests and webscraping the response and I only have 4 hours to submit this.
Webscraping is really limited and hard to do on a complex site like YouTube, some of my functions occasionally return false results or are unable to return results as YouTube's page source is really complicated, variable and big. Maybe if I could have used YouTube's API as they let you do 10k requests for free, but I didn't have enough time to spread that over 6+ days. My scripts will stop working when YouTube gets updated too, I should've used BeautifulSoup a little bit better and MAYBE used Selenium, although it would be much slower, at least it would be much more consistent and correct.
Data was not a limitation in my project, I have 52 THOUSAND videos spread across 3-4 years, my data still has lots of untouched potential and I could analyze it further in the future.
But one problem was that the watch history YouTube provides isn't really a watch history, it's more like a 'history of videos you clicked on and stayed at least 3 seconds looked at it list'. If YouTube takeout provided 'where I left off' on a video, I'd be able to do a lot with it, like calculate how much actual time I spent watching stuff.