{
  "id": 73002,
  "title": "How to deal with historical_transactions.csv",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/73002",
  "author_name": "",
  "post_date": "2018-11-29T00:34:36.267235400Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello, everyone. I'm wondering how should we deal with historical_transactions.csv and new_merchant_transactions.csv. The features in train.csv are very little. I guess we need to use data in historical_transactions.csv and new_merchant_transactions.csv. But how should we use them?</p>",
  "messages": [
    {
      "id": "429510",
      "postDate": "11/29/2018 00:34:36",
      "content": "<p>Hello, everyone. I'm wondering how should we deal with historical_transactions.csv and new_merchant_transactions.csv. The features in train.csv are very little. I guess we need to use data in historical_transactions.csv and new_merchant_transactions.csv. But how should we use them?</p>",
      "rawMarkdown": "Hello, everyone. I'm wondering how should we deal with historical_transactions.csv and new_merchant_transactions.csv. The features in train.csv are very little. I guess we need to use data in historical_transactions.csv and new_merchant_transactions.csv. But how should we use them?",
      "votes": null
    },
    {
      "id": "429878",
      "postDate": "11/29/2018 13:48:01",
      "content": "<p>The common thread between the train/test data and the historical and new transaction data is the <em>card_id</em>.  So for each <em>card_id</em> you want to look at buying patterns (how many transactions, total amount of all transactions, counts and amounts for different <em>city_id</em>, <em>state_id</em>, <em>subsector_id</em>, etc., characteristics of the most frequented merchant) or changes in buying patterns between the new transactions and the historical transactions.</p>",
      "rawMarkdown": "The common thread between the train/test data and the historical and new transaction data is the *card\\_id*.  So for each *card\\_id* you want to look at buying patterns (how many transactions, total amount of all transactions, counts and amounts for different *city\\_id*, *state\\_id*, *subsector\\_id*, etc., characteristics of the most frequented merchant) or changes in buying patterns between the new transactions and the historical transactions.",
      "votes": null
    },
    {
      "id": "430073",
      "postDate": "11/29/2018 19:29:45",
      "content": "<p>Very nice explanation, very helpful. Thanks Norm!</p>",
      "rawMarkdown": "Very nice explanation, very helpful. Thanks Norm!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 429878,
      "author_name": "nsecord",
      "author_url": "",
      "post_date": "11/29/2018 13:48:01",
      "content": "<p>The common thread between the train/test data and the historical and new transaction data is the <em>card_id</em>.  So for each <em>card_id</em> you want to look at buying patterns (how many transactions, total amount of all transactions, counts and amounts for different <em>city_id</em>, <em>state_id</em>, <em>subsector_id</em>, etc., characteristics of the most frequented merchant) or changes in buying patterns between the new transactions and the historical transactions.</p>",
      "votes": null,
      "replies": [
        {
          "id": 430073,
          "author_name": "jinjunding",
          "author_url": "",
          "post_date": "11/29/2018 19:29:45",
          "content": "<p>Very nice explanation, very helpful. Thanks Norm!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "429510": "Hello, everyone. I'm wondering how should we deal with historical_transactions.csv and new_merchant_transactions.csv. The features in train.csv are very little. I guess we need to use data in historical_transactions.csv and new_merchant_transactions.csv. But how should we use them?",
    "429878": "The common thread between the train/test data and the historical and new transaction data is the *card\\_id*.  So for each *card\\_id* you want to look at buying patterns (how many transactions, total amount of all transactions, counts and amounts for different *city\\_id*, *state\\_id*, *subsector\\_id*, etc., characteristics of the most frequented merchant) or changes in buying patterns between the new transactions and the historical transactions.",
    "430073": "Very nice explanation, very helpful. Thanks Norm!"
  },
  "source": "meta"
}