{
  "id": 612956,
  "title": "genetic algo feature selection ->local search feature selection->xgboost bo",
  "content": "<p>here is my training pipe..it is not very performant but demonstrate how to increase the features space and let ga nature select the right set of features</p>\n<p>genetic algo feature selection -&gt;local search feature selection-&gt;xgboost bayesian optimisation ….walk forward cross validation..\n.from local smape 57 to 43.</p>\n<p>bonus to come the generic algorithm and xgboost in pure JAX that runs on the TPU session 100 times faster</p>",
  "comments": [],
  "comment_count": 0,
  "messages": [
    {
      "id": 3305661,
      "postDate": "2025-10-23T07:49:52.057Z",
      "content": "<p>here is my training pipe..it is not very performant but demonstrate how to increase the features space and let ga nature select the right set of features</p>\n<p>genetic algo feature selection -&gt;local search feature selection-&gt;xgboost bayesian optimisation ….walk forward cross validation..\n.from local smape 57 to 43.</p>\n<p>bonus to come the generic algorithm and xgboost in pure JAX that runs on the TPU session 100 times faster</p>",
      "rawMarkdown": "here is my training pipe..it is not very performant but demonstrate how to increase the features space and let ga nature select the right set of features\n\ngenetic algo feature selection ->local search feature selection->xgboost bayesian optimisation ....walk forward cross validation..\n.from local smape 57 to 43.\n\nbonus to come the generic algorithm and xgboost in pure JAX that runs on the TPU session 100 times faster"
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-20T00:52:21Z",
  "competition_id": 111876,
  "competition_slug": "china-real-estate-demand-prediction",
  "team_id": 14348387,
  "solution_url": ""
}