{
  "id": 41258,
  "title": "does anyone try lstm to model the user behavior ?",
  "url": "/competitions/kkbox-churn-prediction-challenge/discussion/41258",
  "author_name": "",
  "post_date": "2017-10-15T12:37:31.375389400Z",
  "votes": 3,
  "comment_count": 7,
  "views": 0,
  "content": "",
  "messages": [
    {
      "id": "231598",
      "postDate": "10/15/2017 12:37:31",
      "content": "",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "231643",
      "postDate": "10/15/2017 15:02:58",
      "content": "<p>yes i use such model but does not gain better performance than xgboost</p>",
      "rawMarkdown": "yes i use such model but does not gain better performance than xgboost",
      "votes": null
    },
    {
      "id": "233104",
      "postDate": "10/19/2017 08:10:08",
      "content": "<p>I am working on it ...</p>",
      "rawMarkdown": "I am working on it ...",
      "votes": null
    },
    {
      "id": "233709",
      "postDate": "10/20/2017 22:22:52",
      "content": "<p>Hey I'm going to try to implement lstm or dilated lstm, using tensorflow and protocol buffers.  Let me know if you want to team up,- I'm new and this is quite a task.</p>",
      "rawMarkdown": "Hey I'm going to try to implement lstm or dilated lstm, using tensorflow and protocol buffers.  Let me know if you want to team up,- I'm new and this is quite a task.",
      "votes": null
    },
    {
      "id": "233789",
      "postDate": "10/21/2017 08:35:27",
      "content": "<p>Hey, sounds great. I implemented an LSTM iterator for users_log.csv in dl4j for testing the results, but if you are more familiar with tensorflow, we can do it there again. Tensorflow provides more flexibility, so TF would be the better choice in my opinion. Let us work together :-)</p>",
      "rawMarkdown": "Hey, sounds great. I implemented an LSTM iterator for users_log.csv in dl4j for testing the results, but if you are more familiar with tensorflow, we can do it there again. Tensorflow provides more flexibility, so TF would be the better choice in my opinion. Let us work together :-)",
      "votes": null
    },
    {
      "id": "234001",
      "postDate": "10/22/2017 02:17:26",
      "content": "<p>How exciting!  I've been fishing for a partner for weeks :)  I hope you noticed I said 'Im new', because no, I do not have any substantial Tensorflow experience, however, I have quite a bit of faith that Tensorflow has a very promising ecology and future.  Give me two weeks and I will try to figure out how to get the data into protobuf form, then let's get in touch.  Being new, I'm not sure how effective or quick I'll be...  working with me may bore you to tears lol.</p>",
      "rawMarkdown": "How exciting!  I've been fishing for a partner for weeks :)  I hope you noticed I said 'Im new', because no, I do not have any substantial Tensorflow experience, however, I have quite a bit of faith that Tensorflow has a very promising ecology and future.  Give me two weeks and I will try to figure out how to get the data into protobuf form, then let's get in touch.  Being new, I'm not sure how effective or quick I'll be...  working with me may bore you to tears lol.",
      "votes": null
    },
    {
      "id": "234349",
      "postDate": "10/23/2017 03:27:43",
      "content": "<p>hi, guys, i have implemented a lstm version with keras, and got 0.20468, maybe we can add facebook friends to discuss the problems ~ my facebook page is <a href=\"https://www.facebook.com/kadeswang\">https://www.facebook.com/kadeswang</a> ~</p>",
      "rawMarkdown": "hi, guys, i have implemented a lstm version with keras, and got 0.20468, maybe we can add facebook friends to discuss the problems ~ my facebook page is https://www.facebook.com/kadeswang ~",
      "votes": null
    },
    {
      "id": "234414",
      "postDate": "10/23/2017 08:37:44",
      "content": "<p>I am going to contact you later today ...</p>",
      "rawMarkdown": "I am going to contact you later today ...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 231643,
      "author_name": "chenzelin",
      "author_url": "",
      "post_date": "10/15/2017 15:02:58",
      "content": "<p>yes i use such model but does not gain better performance than xgboost</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 233104,
      "author_name": "patrickklein",
      "author_url": "",
      "post_date": "10/19/2017 08:10:08",
      "content": "<p>I am working on it ...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 233709,
      "author_name": "stravinsky",
      "author_url": "",
      "post_date": "10/20/2017 22:22:52",
      "content": "<p>Hey I'm going to try to implement lstm or dilated lstm, using tensorflow and protocol buffers.  Let me know if you want to team up,- I'm new and this is quite a task.</p>",
      "votes": null,
      "replies": [
        {
          "id": 233789,
          "author_name": "patrickklein",
          "author_url": "",
          "post_date": "10/21/2017 08:35:27",
          "content": "<p>Hey, sounds great. I implemented an LSTM iterator for users_log.csv in dl4j for testing the results, but if you are more familiar with tensorflow, we can do it there again. Tensorflow provides more flexibility, so TF would be the better choice in my opinion. Let us work together :-)</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 234001,
          "author_name": "stravinsky",
          "author_url": "",
          "post_date": "10/22/2017 02:17:26",
          "content": "<p>How exciting!  I've been fishing for a partner for weeks :)  I hope you noticed I said 'Im new', because no, I do not have any substantial Tensorflow experience, however, I have quite a bit of faith that Tensorflow has a very promising ecology and future.  Give me two weeks and I will try to figure out how to get the data into protobuf form, then let's get in touch.  Being new, I'm not sure how effective or quick I'll be...  working with me may bore you to tears lol.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 234349,
          "author_name": "aronml",
          "author_url": "",
          "post_date": "10/23/2017 03:27:43",
          "content": "<p>hi, guys, i have implemented a lstm version with keras, and got 0.20468, maybe we can add facebook friends to discuss the problems ~ my facebook page is <a href=\"https://www.facebook.com/kadeswang\">https://www.facebook.com/kadeswang</a> ~</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 234414,
          "author_name": "patrickklein",
          "author_url": "",
          "post_date": "10/23/2017 08:37:44",
          "content": "<p>I am going to contact you later today ...</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "231598": "",
    "231643": "yes i use such model but does not gain better performance than xgboost",
    "233104": "I am working on it ...",
    "233709": "Hey I'm going to try to implement lstm or dilated lstm, using tensorflow and protocol buffers.  Let me know if you want to team up,- I'm new and this is quite a task.",
    "233789": "Hey, sounds great. I implemented an LSTM iterator for users_log.csv in dl4j for testing the results, but if you are more familiar with tensorflow, we can do it there again. Tensorflow provides more flexibility, so TF would be the better choice in my opinion. Let us work together :-)",
    "234001": "How exciting!  I've been fishing for a partner for weeks :)  I hope you noticed I said 'Im new', because no, I do not have any substantial Tensorflow experience, however, I have quite a bit of faith that Tensorflow has a very promising ecology and future.  Give me two weeks and I will try to figure out how to get the data into protobuf form, then let's get in touch.  Being new, I'm not sure how effective or quick I'll be...  working with me may bore you to tears lol.",
    "234349": "hi, guys, i have implemented a lstm version with keras, and got 0.20468, maybe we can add facebook friends to discuss the problems ~ my facebook page is https://www.facebook.com/kadeswang ~",
    "234414": "I am going to contact you later today ..."
  },
  "source": "meta"
}