{
  "id": 75441,
  "title": "Feature Engineering Intuition",
  "url": "/competitions/vsb-power-line-fault-detection/discussion/75441",
  "author_name": "",
  "post_date": "2018-12-21T17:06:22.486090100Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I think for feature engineering we should go with discrete fourier transform and wavelet transform to extract insights  from the signal's data:</p>\n\n<p>Its a signal processing use case where looking at the signal, we should be able to tell power discharges, power discharges lead to transients and transients means high frequency components in the signal. Thus applying fourier transform and taking out the transients frequency will help predict the output.</p>",
  "messages": [
    {
      "id": "443450",
      "postDate": "12/21/2018 17:06:22",
      "content": "<p>I think for feature engineering we should go with discrete fourier transform and wavelet transform to extract insights  from the signal's data:</p>\n\n<p>Its a signal processing use case where looking at the signal, we should be able to tell power discharges, power discharges lead to transients and transients means high frequency components in the signal. Thus applying fourier transform and taking out the transients frequency will help predict the output.</p>",
      "rawMarkdown": "I think for feature engineering we should go with discrete fourier transform and wavelet transform to extract insights  from the signal's data:\n\nIts a signal processing use case where looking at the signal, we should be able to tell power discharges, power discharges lead to transients and transients means high frequency components in the signal. Thus applying fourier transform and taking out the transients frequency will help predict the output.",
      "votes": null
    },
    {
      "id": "443571",
      "postDate": "12/21/2018 22:11:36",
      "content": "<p>I agree and I'm looking into these features. Conveniently <code>scipy.signal</code> and <code>scipy.fftpack</code> has quite a few of these functions implemented already.</p>\n\n<p><a href=\"https://docs.scipy.org/doc/scipy/reference/signal.html\">https://docs.scipy.org/doc/scipy/reference/signal.html</a></p>\n\n<p><a href=\"https://docs.scipy.org/doc/scipy/reference/fftpack.html\">https://docs.scipy.org/doc/scipy/reference/fftpack.html</a></p>",
      "rawMarkdown": "I agree and I'm looking into these features. Conveniently `scipy.signal` and `scipy.fftpack` has quite a few of these functions implemented already.\n\nhttps://docs.scipy.org/doc/scipy/reference/signal.html\n\nhttps://docs.scipy.org/doc/scipy/reference/fftpack.html",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 443571,
      "author_name": "timothycwillard",
      "author_url": "",
      "post_date": "12/21/2018 22:11:36",
      "content": "<p>I agree and I'm looking into these features. Conveniently <code>scipy.signal</code> and <code>scipy.fftpack</code> has quite a few of these functions implemented already.</p>\n\n<p><a href=\"https://docs.scipy.org/doc/scipy/reference/signal.html\">https://docs.scipy.org/doc/scipy/reference/signal.html</a></p>\n\n<p><a href=\"https://docs.scipy.org/doc/scipy/reference/fftpack.html\">https://docs.scipy.org/doc/scipy/reference/fftpack.html</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "443450": "I think for feature engineering we should go with discrete fourier transform and wavelet transform to extract insights  from the signal's data:\n\nIts a signal processing use case where looking at the signal, we should be able to tell power discharges, power discharges lead to transients and transients means high frequency components in the signal. Thus applying fourier transform and taking out the transients frequency will help predict the output.",
    "443571": "I agree and I'm looking into these features. Conveniently `scipy.signal` and `scipy.fftpack` has quite a few of these functions implemented already.\n\nhttps://docs.scipy.org/doc/scipy/reference/signal.html\n\nhttps://docs.scipy.org/doc/scipy/reference/fftpack.html"
  },
  "source": "meta"
}