|
| 1 | +--- |
| 2 | +import MainLayout from '../layouts/MainLayout.astro'; |
| 3 | +import Navbar from '../components/Navbar.astro'; |
| 4 | +--- |
| 5 | +<MainLayout title="M5 Benchmark"> |
| 6 | + <Navbar /> |
| 7 | + <article class="mt-4 rounded-xl border border-slate-700 bg-slate-900/70 p-5 space-y-4"> |
| 8 | + <h1 class="text-2xl font-bold">M5: Why it matters for time-series forecasting</h1> |
| 9 | + <p class="text-slate-300"> |
| 10 | + The <strong>M5 Forecasting Competition</strong> is one of the most-used modern benchmarks for practical demand forecasting. |
| 11 | + It uses real Walmart retail sales data with rich hierarchy (item, department, store, state) and calendar/event effects. |
| 12 | + </p> |
| 13 | + |
| 14 | + <section> |
| 15 | + <h2 class="text-lg font-semibold mb-2">Why M5 is a strong standard</h2> |
| 16 | + <ul class="list-disc ml-5 space-y-1 text-slate-300"> |
| 17 | + <li><strong>Real-world complexity:</strong> noisy retail demand, intermittent series, promotions, and seasonality.</li> |
| 18 | + <li><strong>Hierarchical structure:</strong> forecasts must make sense across aggregation levels, not just one series.</li> |
| 19 | + <li><strong>Scale:</strong> thousands of related series test both model quality and operational reliability.</li> |
| 20 | + <li><strong>Community adoption:</strong> many papers and libraries report M5-style performance, so comparisons are meaningful.</li> |
| 21 | + </ul> |
| 22 | + </section> |
| 23 | + |
| 24 | + <section> |
| 25 | + <h2 class="text-lg font-semibold mb-2">What Weatherman does with M5</h2> |
| 26 | + <ul class="list-disc ml-5 space-y-1 text-slate-300"> |
| 27 | + <li>Lets you run a quick <strong>M5 demo mode</strong> from the landing page.</li> |
| 28 | + <li>Samples a configurable number of M5 series via <code>n_series</code>.</li> |
| 29 | + <li>Runs rolling backtests and reports SMAPE so you can compare models on historical windows.</li> |
| 30 | + <li>Then generates forward forecasts for each series.</li> |
| 31 | + </ul> |
| 32 | + </section> |
| 33 | + |
| 34 | + <section class="rounded-lg border border-slate-700 bg-slate-950 p-3"> |
| 35 | + <h3 class="font-semibold mb-2">When to use M5 demo vs your own payload</h3> |
| 36 | + <ul class="list-disc ml-5 space-y-1 text-slate-300"> |
| 37 | + <li><strong>M5 demo:</strong> great for sanity checks, regressions, and model-comparison smoke tests.</li> |
| 38 | + <li><strong>Your payload:</strong> best for business-specific signal, seasonality, and decision support.</li> |
| 39 | + </ul> |
| 40 | + </section> |
| 41 | + |
| 42 | + <p class="text-sm text-slate-400">Note: M5 is a benchmark, not your business. Treat it as a quality baseline, then validate on your own data.</p> |
| 43 | + <p><a href="/" class="text-sky-300 underline">← Back to Forecast form</a></p> |
| 44 | + </article> |
| 45 | +</MainLayout> |
0 commit comments