A Collection of Data Science Take-Home Challenges
The company of this challenge allows users to upload videos online, just like YouTube.
This company is interested in knowing whether a video is "hot" (i.e. trending up in terms of popularity), stable or going down. Understanding this would allow to optimize the videos promoted on the home-page and, therefore, maximize ads revenue.
Company XYZ is an online video streaming company, just like YouTube or Dailymotion.
The Head of Product has identified as a major problem for the site a very high home page drop-off rate. That is, users come to the home-page and then leave the site without taking any action or watching any video. Since customer acquisition costs are very high, this is a huge problem: the company is spending a lot of money to acquire users who don't generate any revenue.
Currently, the videos shown on the home page to new users are manually chosen. The Head of Product had this idea of creating a new recommended video section on the home page.
She asked you the following:
Classify each video into one these 3 categories:
"Hot" - means trending up. These videos are candidates to be shown.
"Stable and Popular" - video view counts are flat, but very high. These videos are candidates to be shown too.
"Everything else" - these videos won't be shown
What are the main characteristics of the "hot videos"?
After having identified the characteristics of the hot videos, how would you use this information from a product standpoint?
We have 2 tables downloadable by clicking here.
The 2 tables are:
video_count - provides information about how many times each video was seen by day
video_features - characteristics of the video.
Let's check one video: how many times was it seen on a given day?
head (video_count, 1)
|video_id||2303||id of the video|
|count||22||it was seen just 22 times.|
|date||2015-01-07||on Jan, 7th.|
Let's now check the characteristics of that video 2303:
subset(video_features, video_id == 2303)
|video_id||2303||It is the video we care about. Same as above|
|video_length||1071||the video lasts almost 18 min (1071 seconds)|
|video_language||Cn||the video is in Chinese|
|video_upload_date||2014-12-10||was uploaded on Dec, 10|
|video_quality||1080p||video quality is 1080p, i.e. very high|