{
  "id": 119190,
  "title": "I would like to ask: How to determine the newly established features are useful",
  "url": "/competitions/ashrae-energy-prediction/discussion/119190",
  "author_name": "",
  "post_date": "2019-11-27T05:27:53.170528100Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>For example, in the competition, I created 20 columns of new features. \nWhat method should I use to test the validity of the features? Or is there a similar share？\nThanks</p>",
  "messages": [
    {
      "id": "682236",
      "postDate": "11/27/2019 05:27:53",
      "content": "<p>For example, in the competition, I created 20 columns of new features. \nWhat method should I use to test the validity of the features? Or is there a similar share？\nThanks</p>",
      "rawMarkdown": "For example, in the competition, I created 20 columns of new features. \nWhat method should I use to test the validity of the features? Or is there a similar share？\nThanks",
      "votes": null
    },
    {
      "id": "682275",
      "postDate": "11/27/2019 07:03:14",
      "content": "<p>In general, there are two different approaches for feature selection: wrapper and filter methods. </p>\n\n<p>Roughly speaking, wrapper methods create a lot of models with different subsets of features, and then select the features which result in the best performing model. As an example, you can take a look at Recursive Feature Elimination (RFE) here: <a href=\"https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html\">https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html</a>  </p>\n\n<p>On the other hand, filter methods use statistics to investigate the relationship between each predictor and target variable. For example, correlation statistical measures are widely used as a basis for filter feature selection methods.  </p>",
      "rawMarkdown": "In general, there are two different approaches for feature selection: wrapper and filter methods. \n\nRoughly speaking, wrapper methods create a lot of models with different subsets of features, and then select the features which result in the best performing model. As an example, you can take a look at Recursive Feature Elimination (RFE) here: https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html  \n\nOn the other hand, filter methods use statistics to investigate the relationship between each predictor and target variable. For example, correlation statistical measures are widely used as a basis for filter feature selection methods.",
      "votes": null
    },
    {
      "id": "682327",
      "postDate": "11/27/2019 09:16:34",
      "content": "<p>If you have stable and reliable CV, your CV should improve along side your LB score. So finding reliable CV is your first priority before attempting feature engineering.</p>",
      "rawMarkdown": "If you have stable and reliable CV, your CV should improve along side your LB score. So finding reliable CV is your first priority before attempting feature engineering.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 682275,
      "author_name": "starl1ght",
      "author_url": "",
      "post_date": "11/27/2019 07:03:14",
      "content": "<p>In general, there are two different approaches for feature selection: wrapper and filter methods. </p>\n\n<p>Roughly speaking, wrapper methods create a lot of models with different subsets of features, and then select the features which result in the best performing model. As an example, you can take a look at Recursive Feature Elimination (RFE) here: <a href=\"https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html\">https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html</a>  </p>\n\n<p>On the other hand, filter methods use statistics to investigate the relationship between each predictor and target variable. For example, correlation statistical measures are widely used as a basis for filter feature selection methods.  </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 682327,
      "author_name": "teeyee314",
      "author_url": "",
      "post_date": "11/27/2019 09:16:34",
      "content": "<p>If you have stable and reliable CV, your CV should improve along side your LB score. So finding reliable CV is your first priority before attempting feature engineering.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "682236": "For example, in the competition, I created 20 columns of new features. \nWhat method should I use to test the validity of the features? Or is there a similar share？\nThanks",
    "682275": "In general, there are two different approaches for feature selection: wrapper and filter methods. \n\nRoughly speaking, wrapper methods create a lot of models with different subsets of features, and then select the features which result in the best performing model. As an example, you can take a look at Recursive Feature Elimination (RFE) here: https://scikit-learn.org/stable/modules/generated/sklearn.feature_selection.RFE.html  \n\nOn the other hand, filter methods use statistics to investigate the relationship between each predictor and target variable. For example, correlation statistical measures are widely used as a basis for filter feature selection methods.",
    "682327": "If you have stable and reliable CV, your CV should improve along side your LB score. So finding reliable CV is your first priority before attempting feature engineering."
  },
  "source": "meta"
}