User Guide

Purpose

This guide explains how to use the final three-module tourism analytics app.

Step 1: Explore the Series

Open the Time Series Visual Analysis tab to:

  1. Choose a monthly series such as Visitor Arrivals: China or Hotel Room Occupancy Rate.
  2. Inspect the recent trend with a line chart.
  3. Confirm the unit and source metadata before moving to deeper analysis.

Step 2: Compare Country Recovery Patterns

Open the Time Series Clustering tab to:

  1. Select a focused set of country-level visitor-arrivals series.
  2. Set the year window and choose the normalization mode.
  3. Pick a working k and run the clustering.
  4. Read the module in order: Dashboard, Pattern Explorer, Focus Market in Context, Assignments.

The clustering unit is a country trajectory, not an individual month. The module is designed to answer which markets recovered in similar ways, not to assign each month into a market state.

Step 3: Forecast Future Values

Open the Forecasting tab to:

  1. Select a country-level visitor-arrivals series.
  2. Compare that arrivals series with hotel occupancy, average length of stay, and room revenue in the context chart.
  3. Set the holdout horizon, choose the model set, and click Run Forecasting.
  4. Compare the seasonal-naive baseline with ETS, ARIMA, or both.
  5. Review the forecast plot, decomposition views, split summary, and accuracy table.

The forecasting module prefers the full modeltime workflow, but it can fall back to a lighter forecast implementation when that dependency chain is unavailable. The model labels and interpretation stay consistent across both paths.

Interpretation Tips

  1. Use the explorer first to understand seasonality and structural breaks.
  2. Use clustering second to identify which country-arrival trajectories behave similarly.
  3. Use forecasting last to project one country-level arrivals series while interpreting it against hotel and stay indicators.

Data Scope

The current app uses two coordinated inputs:

  • Shared arrivals backbone: data/raw/visitor_arrivals_full_dataset.xlsx
  • Shared clustering artifacts: data/processed/clustering_country_wide.csv, data/processed/clustering_country_long.csv, data/processed/clustering_series_metadata.csv
  • Optional supporting tourism context: data/raw/tourism_update.xlsx

Country-level visitor arrivals remain the common analytical target across explorer, clustering, and forecasting. Hotel occupancy, average length of stay, number of hotels, and total room revenue are used only as optional supporting context for interpretation.