Automated Time Series Analysis Dashboard Guide - bri-thetechie/TSA GitHub Wiki

Overview

The Automated App is designed for quick and easy exploration of time series data. It automatically performs several key analyses with minimal user input.

How to Use

  1. Upload Data: Upload your time series data file (supported formats: .csv and .xlsx) using the "Browse" button in the sidebar.
  2. Select Column: Choose the numeric column containing your time series data from the "Select Numeric Column" dropdown.
  3. Set Frequency: Specify the frequency of your time series (e.g., 12 for monthly, 252 for daily stock data) and click "Set Frequency".
  4. Forecast Horizon: Optionally, enter the number of periods you want to forecast in the "Forecast Horizon" field.
  5. Run Analysis: Click the "Run Automated Analysis" button.

Output Tabs

The main panel will display the results in the following tabs:

  • Time Series Plot: A plot of your time series data over time.
  • Stationarity Tests: Results of the Augmented Dickey-Fuller (ADF) and HEGY tests (HEGY is run if frequency > 1).
  • Decomposition: Plots showing the decomposed time series (trend, seasonal, and random components). The app attempts both additive and multiplicative decomposition.
  • Detrending: Plots showing the original series and the moving average detrended series.
  • ACF/PACF: Autocorrelation Function (ACF) and Partial Autocorrelation Function (PACF) plots to help identify potential model orders.
  • Periodogram: A plot showing the spectral density of the time series at different frequencies.
  • ARIMA/SARIMA Model: Summary of the automatically fitted ARIMA or SARIMA model (using auto.arima()).
  • GARCH Model: Summary of the GARCH(1,1) model fitted to the residuals if heteroscedasticity is detected.
  • Forecasting: A plot of the forecast along with the forecast results. If a GARCH model was fitted, a forecast of the conditional variance might also be shown.
  • Residual Analysis: Results of various residual diagnostic tests (Shapiro-Wilk, Durbin-Watson, Box-Pierce, ARCH LM) and plots (residual plot, QQ-plot, ACF of residuals) to assess the model fit.