Skip to content

🎯 Use Cases & Personas

This page provides clear, step-by-step workflows for both coders and business users, covering all major use-cases for sktime-mcp.

👤 User Personas

1. Coder

  • Goal: Wants code files as outputs.

Use-Case 1: Benchmarking Time Series Models

  • Task: Use a model (e.g., ARIMA) and benchmark it on a dataset (local or web).
  • What happens:
    • Your dataset is automatically loaded (see Loading Data).
    • The selected model (e.g., ARIMA, Prophet) is instantiated and fitted.
    • Benchmarking is performed behind the scenes.
    • Forecast results and code files are generated for you. You just need to submit your query and wait for the results/status.

Use-Case 2: Composing Pipelines

  • Task: Build and benchmark a pipeline (e.g., Deseasonalizer → Detrender → Estimator).

  • What happens:

    • The pipeline components you specify are validated and composed automatically.
    • The pipeline is run on your dataset in the background.
    • Code files and benchmark results are produced for you. You only need to submit your pipeline query and monitor the status.

  • Designing Custom Pipelines:

    • You can specify any sequence (e.g., deseasonalizer → detrender → estimator).
    • If unsure, see pipeline design guide.

2. Business Person

  • Goal: Wants forecasts, analysis, and reports as outputs.

Use-Case: Get Forecasts & Reports

  • What happens:
    • Your data (CSV or demo) is loaded and processed automatically.
    • The requested analysis (forecast, report) is performed in the background.
    • Results (forecasts, charts, summary reports) are made available for you to view or download. Just submit your request and wait for completion.

📂 Loading Data

1. Demo Datasets

  • Use built-in demo datasets for quick testing.

2. User-Defined Data

  • Local CSV:
    • Place your CSV in the project directory.
    • The system will automatically load your file when you reference it in your query—no manual code needed.
  • Web URL (Planned):
    • Soon, you’ll be able to load CSVs directly from a web URL by simply providing the link in your query.
    • Action: Add web URL support and update docs when ready.

📝 Outputs

  • Code Files:
    • Scripts and notebooks for reproducibility.
  • Forecasts:
    • CSV, plots, and summary tables.

🔗 See Also