> For the complete documentation index, see [llms.txt](https://technocollabs.gitbook.io/reddit-comments-upvotes-project/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://technocollabs.gitbook.io/reddit-comments-upvotes-project/master.md).
# Regarding Reddit Data Set and features.
## Background
Reddit is a popular social media site. On this site, users post *threads* in various *subreddits* like to one below\.A thread in the “AskReddit” subreddit
Users can comment on threads or other comments. They can also give *upvotes* or *downvotes* to other threads and comments.
Our goal is to predict the number of upvotes that comments will receive.
## Data
The data, a pickle file containing 1,205,039 rows (comments) that occurred in May of 2015, is hosted on google drive and can be downloaded using this link.
## Data set link :
The target variable and relevant features that will be used for modeling are listed below. They can be divided into several categories.
{% file src="/files/-MQ6o0OHrXYLtSSdGXXY" %}
### Target variable
* **score**: number of upvotes on the comment
### Comment level features
* **gilded**: the number of gilded tags (premium likes) on the comment
* **distinguished**: the type of user on the page. Either ‘moderator’, ‘admin’, or ‘user’
* **controversiality**: a Boolean indicating whether (1) or not (0) comment is controversial (popular comments that are getting close to the same amount of upvotes as downvotes)
* **over\_18**: Whether or not the thread has been marked as NSFW
* **time\_lapse**: the time in seconds between comment and the first comment on the thread
* **hour\_of\_comment**: the hour of day comment was posted
* **weekday**: the day of week comment was posted
* **is\_flair**: whether or not there is flair text for the comment ()
* **is\_flair\_css**: whether or not there is a CSS class for the comment flair
* **depth**: depth of comment in thread (number of parent comments that comment has)
* **no\_of\_linked\_sr:** number of subreddits mentioned in the comment
* **no\_of\_linked\_urls:** number of urls linked in the comment
* **subjectivity**: number of instances of “I”
* **is\_edited**: whether or not the comment has been edited
* **is\_quoted**: whether or not comment quotes another
* **no\_quoted**: number of quotes in the comment
* **senti\_neg**: negative sentiment score
* **senti\_neu**: neutral sentiment score
* **senti\_pos**: positive sentiment score
* **senti\_comp**: compound sentiment score
* **word\_count**: number of words in the comment
### Parent level features
* **time\_since\_parent**: the time in seconds between comment and the parent comment
* **parent\_score:** score of parent comment (NaN if the comment doesn’t have a parent)
* **parent\_cos\_angle**: cosine similarity between comment and its parent comment’s embeddings ()
### Comment tree root features
* **is\_root**: whether or the comment is a root
* **time\_since\_comment\_tree\_root**: the time in seconds between comment and the comment tree root
* **comment\_tree\_root\_score**: score of comment tree root
### Thread level features
* **link\_score**: upvotes of on thread comment is on
* **upvote\_ratio**: the percentage of upvotes from all votes on thread comment is on
* **link\_ups**: number of upvotes on thread
* **time\_since\_link**: time in seconds since the thread was created
* **no\_past\_comments**: number of comments on thread before comment was posted
* **score\_till\_now**: score of thread at the time this comment was posted
* **title\_cos\_angle**: cosine similarity between comment and its thread’s title’s embeddings
* **is\_selftext**: whether or not thread had selftext
## Thanks and best of luck !!!