{
  "id": 92990,
  "title": "Data slicing",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/92990",
  "author_name": "",
  "post_date": "2019-05-22T07:12:13.861097300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>How do you cut the data for the training sample? Just successively at 150,000 points without overlapping or at 150k with some kind of overlap (for example, an offset of 10%)? Try someone who can calculate the forecast for shorter pieces and then get a final prediction as a combination of forecasts for short fragments?</p>",
  "messages": [
    {
      "id": "535040",
      "postDate": "05/22/2019 07:12:13",
      "content": "<p>How do you cut the data for the training sample? Just successively at 150,000 points without overlapping or at 150k with some kind of overlap (for example, an offset of 10%)? Try someone who can calculate the forecast for shorter pieces and then get a final prediction as a combination of forecasts for short fragments?</p>",
      "rawMarkdown": "How do you cut the data for the training sample? Just successively at 150,000 points without overlapping or at 150k with some kind of overlap (for example, an offset of 10%)? Try someone who can calculate the forecast for shorter pieces and then get a final prediction as a combination of forecasts for short fragments?",
      "votes": null
    },
    {
      "id": "535211",
      "postDate": "05/22/2019 13:48:19",
      "content": "<p>If you are talking about slicing for list, plz refer to the following:\n<a href=\"https://stackoverflow.com/questions/509211/understanding-slice-notation\">https://stackoverflow.com/questions/509211/understanding-slice-notation</a></p>\n\n<p>For numpy array slicing, plz refer to the following:\n<a href=\"https://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm\">https://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm</a>\n<a href=\"https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html\">https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html</a></p>\n\n<p>And for the overlapping segment generation, it shouldn't be that difficult to implement with for loop. Give it a try!</p>",
      "rawMarkdown": "If you are talking about slicing for list, plz refer to the following:\nhttps://stackoverflow.com/questions/509211/understanding-slice-notation\n\nFor numpy array slicing, plz refer to the following:\nhttps://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm\nhttps://docs.scipy.org/doc/numpy/reference/arrays.indexing.html\n\nAnd for the overlapping segment generation, it shouldn't be that difficult to implement with for loop. Give it a try!",
      "votes": null
    },
    {
      "id": "535499",
      "postDate": "05/23/2019 03:31:52",
      "content": "<p>I tried all of those techniques above:\noverlapping - I got overfitted results\n2nd level features - 150k split into 15 parts, calculated features and then for each feature calculated mean of, std of etc. - also LB score went up</p>",
      "rawMarkdown": "I tried all of those techniques above:\noverlapping - I got overfitted results\n2nd level features - 150k split into 15 parts, calculated features and then for each feature calculated mean of, std of etc. - also LB score went up",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 535211,
      "author_name": "pukkinming",
      "author_url": "",
      "post_date": "05/22/2019 13:48:19",
      "content": "<p>If you are talking about slicing for list, plz refer to the following:\n<a href=\"https://stackoverflow.com/questions/509211/understanding-slice-notation\">https://stackoverflow.com/questions/509211/understanding-slice-notation</a></p>\n\n<p>For numpy array slicing, plz refer to the following:\n<a href=\"https://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm\">https://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm</a>\n<a href=\"https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html\">https://docs.scipy.org/doc/numpy/reference/arrays.indexing.html</a></p>\n\n<p>And for the overlapping segment generation, it shouldn't be that difficult to implement with for loop. Give it a try!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 535499,
      "author_name": "pavelvod",
      "author_url": "",
      "post_date": "05/23/2019 03:31:52",
      "content": "<p>I tried all of those techniques above:\noverlapping - I got overfitted results\n2nd level features - 150k split into 15 parts, calculated features and then for each feature calculated mean of, std of etc. - also LB score went up</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "535040": "How do you cut the data for the training sample? Just successively at 150,000 points without overlapping or at 150k with some kind of overlap (for example, an offset of 10%)? Try someone who can calculate the forecast for shorter pieces and then get a final prediction as a combination of forecasts for short fragments?",
    "535211": "If you are talking about slicing for list, plz refer to the following:\nhttps://stackoverflow.com/questions/509211/understanding-slice-notation\n\nFor numpy array slicing, plz refer to the following:\nhttps://www.tutorialspoint.com/numpy/numpy_indexing_and_slicing.htm\nhttps://docs.scipy.org/doc/numpy/reference/arrays.indexing.html\n\nAnd for the overlapping segment generation, it shouldn't be that difficult to implement with for loop. Give it a try!",
    "535499": "I tried all of those techniques above:\noverlapping - I got overfitted results\n2nd level features - 150k split into 15 parts, calculated features and then for each feature calculated mean of, std of etc. - also LB score went up"
  },
  "source": "meta"
}