This app is intended to provide different instruments to professional who work with time series; it is composed by three main panels:
- Data Cleaning: contains different functions that help clean the dataset (remove outliers, fill gaps and normalize sampling frequency).
- Model - Work in Progress: help the user find the best model to fit the time-series.
- Forecast - Work in Progress: provide the users forecasts based on the selected model and allows the download of the model configuration.
A tutorial for the first panel is available here
The app can be launched by installing the Shiny library on R typing in your console the following comands:
install.packages("shiny")
library(shiny)
runGitHub(repo = "bigliolimatteo/Time_Series_modeling_app", ref='main')
Every needed library will be automatically installed and loaded in your system without any further notice (the list of the needed packages can be found in sessionInfo.txt).
We developed two optimizations of the code using the RCpp libraries that can be found in the file data_engineering.R.
At the end of the file you can find also the microbenchmark comparisons with the corresponding R functions.
data_engineering.R: Here you can find the functions that manipulate the dataset.
app.R: This is the core of the app containing the ui and server applications.
tutorial.md: A brief tutorial that goes through the main functionalities of the app.
default_datasets: This is a folder containing three default time-series datasets to help the user play with the app even if he/she doesn't have a custom dataset at hand.
old_dependencies: This folder contains app dependencies that were removed by CRAN but are useful for the app itself; they are automatically installed.
tutorial_images: This is a folder containing images used in the tutorial.
www: This is a folder containing images used in the app.
sessionInfo.txt: information about loaded libraries and system specs.