Shiny: a data scientist’s best friend
One of the most important skill for data scientists to have is the ability to clearly communicate results to a general audience. The impact of data scientists’ work depends on how well others can understand their insights to take further actions. But most data science projects are hard to digest and deep in the math and the computation! Therefore, to clearly communicate data science findings can be very hard, especially when you have an audience with very diverse backgrounds. This is why tools like Shiny become a data scientist best friend.
Shiny is an RStudio package to develop interactive web apps using the R programming language. Here are a few benefits of Shiny:
- Great to communicate results via interactive charts, visualizations, text and tables.
- Easy to use. If you know R, there’s not a lot more to learn in order to develop a cool shiny app rapidly. I follow this excellent tutorial to quickly learn the core concepts of shiny.
- Easy to share with colleagues and friends.
- Shiny apps look very cool!
Car accidents predictions based on weather
Some of my data science colleagues at IBM built a very cool model to predict car accidents in New York City. The model was trained using historical data of car accidents and IBM weather’s data. Weather conditions per zip code were used as features to train a logistic regression with Spark that predicts the probability of car accidents. The math and code behind this project takes a little bit of time (days!) to follow and understand. But, I can clearly show you the results of the trained model in action in a shiny app that shows the probability of car accidents per zip code. Here is a screenshot of the shiny app that shows the predicted probabilities of car accidents on an interactive map. Note that the day and time can be chosen interactively.
Planning your next vacation
There is nothing more annoying than getting flight delays. This is why I built a shiny app for you and me to use next time we are planning a trip. The web app can be used to explore the average flight arrival delays (in minutes) for each airport in the US. Users’ can interact with the app by choosing the month and the year to be explored. In addition, users’ can click on airports to get the airport name, code, state, city and average delay. The size of the airport bubbles depends on the volume of flights for each airport. Negative average delays are early arrival flights.
Would you like to build your own shiny?
Originally published at datascience.ibm.com on October 14, 2016.