NetflixOC
Analysis of Netflix's Original content with data from February 2013 to May 2017
Author: Hunter Kempf
Installations
This project requires installation of the following python packages:
- Pandas
- Numpy
- Sklearn
- Matplotlib
Project Motivation
Netflix is a company that I am interested in especially because they are a leader in the streaming media business that many large content companies are trying to emulate. Despite this they have a bit of a reputation for producing almost anything they can get and seem to at least in the eyes of the public take a Quantity over Quality approach. I am interested in diving into data about the shows they fund to get more insights into their approach to content creation.
Executive Summary
Monthly Analysis
Netflix is stepping up their original content production and from 2013 to mid 2017 has gone from one show premier every other month
Yearly Analysis
The average show rating per year premiered is declining as netflix is making more shows.
High Rating Analysis
There aren't any clear trends in show length or number of episodes for highly ranked shows.
Genre Analysis
There are clear associations of genre's with IMDB ratings. This shows the areas that netflix is succeeding in with their original content. this seems to mostly be around Drama's, Marvel shows and Docu-Series. It would be interesting to see how this compares with other major content producers (Disney, Warner Media etc)
Status Analysis
IMDB rating seems to have a correlation with the Status of the show and at a high level aligns with the various levels of Status from Ongoing to Ended.
Status Predictions
Despite not having much data to work with and using a simple model the accuracy in predicting the Status of a show is pretty decent. It would be interesting to get more general data on other companies and content to see how it can compare.
Evaluating Analysis Results
Questions:
- Just how much more Original Content is Netflix making in compared to when they first started making it in 2013?
- What are the rating trends for Netflix’s Original Content? What types of Content are successful?
- And finally can we predict if Netflix will renew or cancel a show?
Question 1 Results
Netflix is making significantly more content in 2017 from when it first started making original content in 2013. From the monthly rates in 2013 to the monthly premier rates in 2017 there was a ~750% increase in the rate at which netflix has been premiering new original content
Question 2 Results
As a whole the mean rating for Original Content shows that Netflix has been producing has been falling. Having said this the drop in IMDB Rating is because Netflix is producing more low quality content but not because Netflix is producing less high rated content. This means that users may have to sift through some more poor quality content but in general there is more high quality content available for them to watch.
As far as the types of content that have high ratings Netflix is most proficient at making Drama, Marvel, Foreign Language and Docu-Series. These genres have produced the best ratings of netflix shows. Interestingly all of these categories except foreign language seem to have average runtimes between 45 min and 70 min.
Question 3 Results
Yes we can sucessfully model our data! For how few observations and features we have and for the simple linear regression model I chose the f1 score results in the .67 range is actually pretty impressive. Furthermore this modeling can be applied to shows that we have in the pending status to give us some insights into which shows may be cancelled or renewed.
Next Steps / Areas for Improvement
Next Steps
- I would like to extend analysis like this to other content creation companies like Disney or Warner Media
- This would likely require building my own dataset through webscraping
- This could allow me to pull ratings from other sources than IMDB as well as bring in new features like cost of production or Director/Cast information
- I would like to see how my simple model performs on more recent data
- This would require me to extend the dataset through to 2019 which would probably have to be done through webscraping
- Since this analysis only covers series I would like to extend it to Movies as well since Netflix especially recently has been making more of those
Areas for Improvement
- This Analysis really lacks the ammount of Data that I would like to have (only 109 observations)
- This Data has pretty limited feature information about a show
File Descriptions
The file structure in this project is very simple:
- Netflix_Originals_EDA.ipynb
- This is the jupyter notebook that I wrote my python code in for my analysis of the Netflix Dataset
- Netflix_Originals_EDA.html
- This is an html file that was compiled from the jupyter notebook