Since the project would like to know which type of video got the most views and likes, the histogram of those two columns were plotted to see if they have normal distribution. Since the number of views and likes are huge, when plotting the histograms, the log of those values were taken to have a better visual. Below are the histograms of views and likes.
Finding NA's is always important when we do data cleaning, so a for loop was used to iterate through all the columns of the dataset to show if there are any NA's inside that columns. And luckily, this dataset doesn't have any NA value.
For the category_id column, the original dataset are all numeric values that represent the id of each category. It's not easy to tell which category it is. Therefore, by going to the API website with the category chart, all the category id got changed into the category names.
I also want to know what percent each category have on these trending videos, so I do a pie chart of them.
I don't need the video id, title, tags, publish time, link and description columns. So I am going to remove those columns. Moreover,the dataset has three columns that shows if the video have disable comments, ratings, and deleted videos, so I will remove the rows with those values if the video has disabled comments, ratings, and videos. And then we will remove those three columns. Last but not least, since we have too many observations, I decide to narrow my data to only analyze the 2018 data.
For now, I have finished the initial data cleaning for my Youtube dataset. And I did the pie chart, the histograms again after my cleaning process. I began with 40949 rows, 16 columns and end with 30381 rows, 5 columns.
