ARIMA Modeling Application
A Rhino R/Shiny Application
By Andrew Disher in R R/Shiny Time Series Modeling
What is it?
The ARIMA Modeling application is a side project of mine that allows a user to practice and hone their skills with ARIMA time series modeling. The user can specify a numnber of model parameters and observe the model output, as well as view a plethora of statistical diagnostic plots. Like in any real mdoeling scenario, the user is motivated to change and refine their model using these diagnostics until they achieve a model that they are content with.
Motivation for the Project
I have always been extremely interested in time series modeling and it remains one of my greatest passions. I have explored many time series models and used them in a variety of contexts, but the Autoregressive Integrated Moving Average model has proven time and time again to be one of the most reliable methods of understanding and modeling time series data.
ARIMA is considered the gold standard, and nine times out of ten, a Data Scientist should at least consider this model, in addition to any other models they think worthwhile. It is considered the gold standard of time series modeling, and will always prove to be a good baseline when comparing other time series models. Therefore, it is one of the most valuable tools in the data science toolkit one can have and understanding its use cases is fundamental.
This app helps to acquaint analysts with the iterative modeling process by providing frequently used diagnostic plots that aid in this process.
How it was Built
The application was built using the R/Shiny framework, developed by Posit.
Like many of my R Shiny projects, I like to include various other web technologies, like SASS, which greatly improves the functionality and appearance of the app so it can be used more easily and provide a better user experience. No one wants to use an ugly app!
To deploy it, I used AWS EC2 alongside with Docker, which helped to create a reproducible environment on the AWS EC2 instance. renv took care of tracking my R package dependencies, like in so many of my other projects.
Picture Preview
