{
  "id": 79117,
  "title": "Single model result about 3.670, waiting for you guy to join us.",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/79117",
  "author_name": "Jianfei Wang",
  "post_date": "2019-01-31T05:20:36.056000",
  "votes": 3,
  "comment_count": 28,
  "views": 0,
  "content": "<p>Hi guys.\nCurrently, out team has three people, me and two creative guys. our best single model is about 3.670. with many other process, we got 3.655 model result. Many days don't submit, just for the total submit limits.\nLooking for some guy who are creative and has a good single model result. Comment next, i will contract you if ok.\nThank you very much! \nUpdated:\nOur total submit times are about 290 times, there are some limit to merge</p>",
  "messages": [
    {
      "id": 464035,
      "postDate": "2019-01-31T05:20:36.057Z",
      "content": "<p>Hi guys.\nCurrently, out team has three people, me and two creative guys. our best single model is about 3.670. with many other process, we got 3.655 model result. Many days don't submit, just for the total submit limits.\nLooking for some guy who are creative and has a good single model result. Comment next, i will contract you if ok.\nThank you very much! \nUpdated:\nOur total submit times are about 290 times, there are some limit to merge</p>",
      "rawMarkdown": "Hi guys.\nCurrently, out team has three people, me and two creative guys. our best single model is about 3.670. with many other process, we got 3.655 model result. Many days don't submit, just for the total submit limits.\nLooking for some guy who are creative and has a good single model result. Comment next, i will contract you if ok.\nThank you very much! \nUpdated:\nOur total submit times are about 290 times, there are some limit to merge",
      "votes": 3
    },
    {
      "id": 464037,
      "postDate": "2019-01-31T05:30:45.743Z",
      "content": "<p>really single 3.670...? wow... </p>",
      "rawMarkdown": "really single 3.670...? wow... ",
      "votes": 1,
      "replies": [
        {
          "id": 464038,
          "postDate": "2019-01-31T05:32:51.550Z",
          "content": "<p>about this, yes guy</p>",
          "rawMarkdown": "about this, yes guy"
        },
        {
          "id": 464045,
          "postDate": "2019-01-31T05:51:05.970Z",
          "content": "<p>wow. really awesome. my single 3.683 hahaha </p>",
          "rawMarkdown": "wow. really awesome. my single 3.683 hahaha ",
          "votes": 1
        },
        {
          "id": 464056,
          "postDate": "2019-01-31T06:05:09.800Z",
          "content": "<p>I add some features from the public kernel such as a kernel 3.691 recently, i think some basic featues are meanfull, but not work for me.</p>",
          "rawMarkdown": "I add some features from the public kernel such as a kernel 3.691 recently, i think some basic featues are meanfull, but not work for me.",
          "votes": 1
        },
        {
          "id": 464104,
          "postDate": "2019-01-31T07:44:02.417Z",
          "content": "<p>Thanks. i also try that features but score decrease. i try more. Fighting :D</p>",
          "rawMarkdown": "Thanks. i also try that features but score decrease. i try more. Fighting :D"
        }
      ]
    },
    {
      "id": 472618,
      "postDate": "2019-02-16T10:10:32.173Z",
      "content": "<p>Are you guys still looking for teammates? We have two people with best single model of 3.680 </p>",
      "rawMarkdown": "Are you guys still looking for teammates? We have two people with best single model of 3.680 "
    },
    {
      "id": 466422,
      "postDate": "2019-02-05T11:02:53.647Z",
      "content": "<p>Is the single model referring to the result of the K-fold LGM, and is it not using the ffm model?</p>",
      "rawMarkdown": "Is the single model referring to the result of the K-fold LGM, and is it not using the ffm model?"
    },
    {
      "id": 464489,
      "postDate": "2019-02-01T00:59:44.967Z",
      "content": "<p><a href=\"/codlife\">@codlife</a> Thats an awesome score. One question - are you using merchants.csv?</p>",
      "rawMarkdown": "@codlife Thats an awesome score. One question - are you using merchants.csv?",
      "replies": [
        {
          "id": 464500,
          "postDate": "2019-02-01T01:36:47.840Z",
          "content": "<p>Yes,guy, but with very little help,</p>",
          "rawMarkdown": "Yes,guy, but with very little help,",
          "votes": 1
        },
        {
          "id": 464502,
          "postDate": "2019-02-01T01:38:57.987Z",
          "content": "<p>Why you call everyone \"guy\"?)))\nNo offence</p>",
          "rawMarkdown": "Why you call everyone \"guy\"?)))\nNo offence"
        },
        {
          "id": 464509,
          "postDate": "2019-02-01T02:25:59.897Z",
          "content": "<p>No reason, haha, sorry if i'm wrong.thank you</p>",
          "rawMarkdown": "No reason, haha, sorry if i'm wrong.thank you"
        },
        {
          "id": 466031,
          "postDate": "2019-02-04T14:54:56.793Z",
          "content": "<p>I guess it's bro, hey bro, yes bro, thanks bro...</p>",
          "rawMarkdown": "I guess it's bro, hey bro, yes bro, thanks bro..."
        },
        {
          "id": 466059,
          "postDate": "2019-02-04T15:54:38.060Z",
          "content": "<p>Stupid question...but what is the total submit limit?</p>",
          "rawMarkdown": "Stupid question...but what is the total submit limit?"
        },
        {
          "id": 467451,
          "postDate": "2019-02-07T06:43:08.287Z",
          "content": "<p>r u using fm?</p>",
          "rawMarkdown": "r u using fm?"
        }
      ]
    },
    {
      "id": 464239,
      "postDate": "2019-01-31T12:49:03.237Z",
      "content": "<p>Hello,</p>\n\n<p>Depends what you call single model.\nI simply have two sub models. One model without outliers and a model predicting outliers. And a ridge regression to combine this two outputs. Nothing manual and therefore I guess I'm not overfitting leaderboard given my low submission count. </p>\n\n<p>I'm using default model parameters (the one shared by Konrad's not the one tuned from other Kernels). Implemented a script to optimize hyper-parameters for each of my models, but haven't yet ran my models with optimized paramaters. </p>\n\n<p>But I feel like I might be missing some clues.</p>",
      "rawMarkdown": "Hello,\n\nDepends what you call single model.\nI simply have two sub models. One model without outliers and a model predicting outliers. And a ridge regression to combine this two outputs. Nothing manual and therefore I guess I'm not overfitting leaderboard given my low submission count. \n\nI'm using default model parameters (the one shared by Konrad's not the one tuned from other Kernels). Implemented a script to optimize hyper-parameters for each of my models, but haven't yet ran my models with optimized paramaters. \n\nBut I feel like I might be missing some clues.",
      "replies": [
        {
          "id": 464254,
          "postDate": "2019-01-31T13:42:39.683Z",
          "content": "<p>Single model mean just one model result</p>",
          "rawMarkdown": "Single model mean just one model result"
        }
      ]
    },
    {
      "id": 464205,
      "postDate": "2019-01-31T11:27:55.543Z",
      "content": "<p>lb3.70 is great! By the way,How many features ？</p>",
      "rawMarkdown": "lb3.70 is great! By the way,How many features ？",
      "replies": [
        {
          "id": 464255,
          "postDate": "2019-01-31T13:42:55.360Z",
          "content": "<p>about 250 features</p>",
          "rawMarkdown": "about 250 features",
          "votes": 3
        },
        {
          "id": 464261,
          "postDate": "2019-01-31T13:50:35.663Z",
          "content": "<p>great work！</p>",
          "rawMarkdown": "great work！"
        }
      ]
    },
    {
      "id": 464161,
      "postDate": "2019-01-31T09:24:59.660Z",
      "content": "<p>Hi!\nI admit i have a weak single model. \n3.646 cv and 3.695 LB.\nBut it is simple, only 36 features. \nThe main thing I've worked on and still working on is a strong and interesting blending pipeline. \nCurrently I'm on 3.634 cv and 3.679 LB.\nNo post processing. And there's still much to work on.</p>\n\n<p>I'm sure i can add to the team by utilizing the models and their features to the pipeline.</p>\n\n<p>Let me know if you wanna Skype to discuss things. You can pm me your Skype handle.</p>",
      "rawMarkdown": "Hi!\nI admit i have a weak single model. \n3.646 cv and 3.695 LB.\nBut it is simple, only 36 features. \nThe main thing I've worked on and still working on is a strong and interesting blending pipeline. \nCurrently I'm on 3.634 cv and 3.679 LB.\nNo post processing. And there's still much to work on.\n\nI'm sure i can add to the team by utilizing the models and their features to the pipeline.\n\nLet me know if you wanna Skype to discuss things. You can pm me your Skype handle.",
      "replies": [
        {
          "id": 464258,
          "postDate": "2019-01-31T13:46:31.977Z",
          "content": "<p>Guy, thank you very much, i will talk to my teammates, not sure to merge mainly for the submit times limit, you can looking for other teammates at the same time.</p>",
          "rawMarkdown": "Guy, thank you very much, i will talk to my teammates, not sure to merge mainly for the submit times limit, you can looking for other teammates at the same time.",
          "votes": 1
        },
        {
          "id": 464279,
          "postDate": "2019-01-31T14:43:14.273Z",
          "content": "<p>No Problem, Thanks</p>",
          "rawMarkdown": "No Problem, Thanks"
        }
      ]
    },
    {
      "id": 464123,
      "postDate": "2019-01-31T08:30:16.860Z",
      "content": "<p>with outlier classification , its 3.673 for me. Let me know.</p>",
      "rawMarkdown": "with outlier classification , its 3.673 for me. Let me know.",
      "replies": [
        {
          "id": 464147,
          "postDate": "2019-01-31T08:58:26.603Z",
          "content": "<p>Guy, I‘m happy to see your result, but out total submit times can't merge, thank you very much!</p>",
          "rawMarkdown": "Guy, I‘m happy to see your result, but out total submit times can't merge, thank you very much!"
        }
      ]
    },
    {
      "id": 464094,
      "postDate": "2019-01-31T07:31:57.633Z",
      "content": "<p>May I join your team? My best 'single model' is 3.672 which combined a Normal value model and a Outlier model.</p>",
      "rawMarkdown": "May I join your team? My best 'single model' is 3.672 which combined a Normal value model and a Outlier model.",
      "replies": [
        {
          "id": 464146,
          "postDate": "2019-01-31T08:56:55.483Z",
          "content": "<p>Good result, i will talk with my friends, thank you! but our submit times are many</p>",
          "rawMarkdown": "Good result, i will talk with my friends, thank you! but our submit times are many"
        },
        {
          "id": 464259,
          "postDate": "2019-01-31T13:48:21.647Z",
          "content": "<p>Thanks. I'm interested in 'many other process'.</p>",
          "rawMarkdown": "Thanks. I'm interested in 'many other process'."
        }
      ]
    },
    {
      "id": 472578,
      "postDate": "2019-02-16T07:50:28.410Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 464037,
      "author_name": "choco",
      "author_url": "",
      "post_date": "2019-01-31T05:30:45.743000",
      "content": "<p>really single 3.670...? wow... </p>",
      "votes": 1,
      "replies": [
        {
          "id": 464038,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T05:32:51.550000",
          "content": "<p>about this, yes guy</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 464045,
          "author_name": "choco",
          "author_url": "",
          "post_date": "2019-01-31T05:51:05.970000",
          "content": "<p>wow. really awesome. my single 3.683 hahaha </p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 464056,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T06:05:09.800000",
          "content": "<p>I add some features from the public kernel such as a kernel 3.691 recently, i think some basic featues are meanfull, but not work for me.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 464104,
          "author_name": "choco",
          "author_url": "",
          "post_date": "2019-01-31T07:44:02.417000",
          "content": "<p>Thanks. i also try that features but score decrease. i try more. Fighting :D</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 472618,
      "author_name": "Tung",
      "author_url": "",
      "post_date": "2019-02-16T10:10:32.173000",
      "content": "<p>Are you guys still looking for teammates? We have two people with best single model of 3.680 </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 466422,
      "author_name": "kwx_007",
      "author_url": "",
      "post_date": "2019-02-05T11:02:53.647000",
      "content": "<p>Is the single model referring to the result of the K-fold LGM, and is it not using the ffm model?</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 464489,
      "author_name": "Aditya Vyas",
      "author_url": "",
      "post_date": "2019-02-01T00:59:44.967000",
      "content": "<p><a href=\"/codlife\">@codlife</a> Thats an awesome score. One question - are you using merchants.csv?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464500,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-02-01T01:36:47.840000",
          "content": "<p>Yes,guy, but with very little help,</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 464502,
          "author_name": "Konstantin Yakovlev",
          "author_url": "",
          "post_date": "2019-02-01T01:38:57.987000",
          "content": "<p>Why you call everyone \"guy\"?)))\nNo offence</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 464509,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-02-01T02:25:59.897000",
          "content": "<p>No reason, haha, sorry if i'm wrong.thank you</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 466031,
          "author_name": "Endi Niu",
          "author_url": "",
          "post_date": "2019-02-04T14:54:56.793000",
          "content": "<p>I guess it's bro, hey bro, yes bro, thanks bro...</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 466059,
          "author_name": "lennart",
          "author_url": "",
          "post_date": "2019-02-04T15:54:38.060000",
          "content": "<p>Stupid question...but what is the total submit limit?</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 467451,
          "author_name": "SubikashPal",
          "author_url": "",
          "post_date": "2019-02-07T06:43:08.287000",
          "content": "<p>r u using fm?</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 464239,
      "author_name": "Jacques Peeters",
      "author_url": "",
      "post_date": "2019-01-31T12:49:03.237000",
      "content": "<p>Hello,</p>\n\n<p>Depends what you call single model.\nI simply have two sub models. One model without outliers and a model predicting outliers. And a ridge regression to combine this two outputs. Nothing manual and therefore I guess I'm not overfitting leaderboard given my low submission count. </p>\n\n<p>I'm using default model parameters (the one shared by Konrad's not the one tuned from other Kernels). Implemented a script to optimize hyper-parameters for each of my models, but haven't yet ran my models with optimized paramaters. </p>\n\n<p>But I feel like I might be missing some clues.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464254,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T13:42:39.683000",
          "content": "<p>Single model mean just one model result</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 464205,
      "author_name": "process531",
      "author_url": "",
      "post_date": "2019-01-31T11:27:55.543000",
      "content": "<p>lb3.70 is great! By the way,How many features ？</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464255,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T13:42:55.360000",
          "content": "<p>about 250 features</p>",
          "votes": 3,
          "replies": []
        },
        {
          "id": 464261,
          "author_name": "process531",
          "author_url": "",
          "post_date": "2019-01-31T13:50:35.663000",
          "content": "<p>great work！</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 464161,
      "author_name": "AmirH",
      "author_url": "",
      "post_date": "2019-01-31T09:24:59.660000",
      "content": "<p>Hi!\nI admit i have a weak single model. \n3.646 cv and 3.695 LB.\nBut it is simple, only 36 features. \nThe main thing I've worked on and still working on is a strong and interesting blending pipeline. \nCurrently I'm on 3.634 cv and 3.679 LB.\nNo post processing. And there's still much to work on.</p>\n\n<p>I'm sure i can add to the team by utilizing the models and their features to the pipeline.</p>\n\n<p>Let me know if you wanna Skype to discuss things. You can pm me your Skype handle.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464258,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T13:46:31.977000",
          "content": "<p>Guy, thank you very much, i will talk to my teammates, not sure to merge mainly for the submit times limit, you can looking for other teammates at the same time.</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 464279,
          "author_name": "AmirH",
          "author_url": "",
          "post_date": "2019-01-31T14:43:14.273000",
          "content": "<p>No Problem, Thanks</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 464123,
      "author_name": "SubikashPal",
      "author_url": "",
      "post_date": "2019-01-31T08:30:16.860000",
      "content": "<p>with outlier classification , its 3.673 for me. Let me know.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464147,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T08:58:26.603000",
          "content": "<p>Guy, I‘m happy to see your result, but out total submit times can't merge, thank you very much!</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 464094,
      "author_name": "Correlation",
      "author_url": "",
      "post_date": "2019-01-31T07:31:57.633000",
      "content": "<p>May I join your team? My best 'single model' is 3.672 which combined a Normal value model and a Outlier model.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 464146,
          "author_name": "Jianfei Wang",
          "author_url": "",
          "post_date": "2019-01-31T08:56:55.483000",
          "content": "<p>Good result, i will talk with my friends, thank you! but our submit times are many</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 464259,
          "author_name": "Correlation",
          "author_url": "",
          "post_date": "2019-01-31T13:48:21.647000",
          "content": "<p>Thanks. I'm interested in 'many other process'.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 472578,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-02-16T07:50:28.410000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "464035": "Hi guys.\nCurrently, out team has three people, me and two creative guys. our best single model is about 3.670. with many other process, we got 3.655 model result. Many days don't submit, just for the total submit limits.\nLooking for some guy who are creative and has a good single model result. Comment next, i will contract you if ok.\nThank you very much! \nUpdated:\nOur total submit times are about 290 times, there are some limit to merge",
    "464037": "really single 3.670...? wow... ",
    "472618": "Are you guys still looking for teammates? We have two people with best single model of 3.680 ",
    "466422": "Is the single model referring to the result of the K-fold LGM, and is it not using the ffm model?",
    "464489": "@codlife Thats an awesome score. One question - are you using merchants.csv?",
    "464239": "Hello,\n\nDepends what you call single model.\nI simply have two sub models. One model without outliers and a model predicting outliers. And a ridge regression to combine this two outputs. Nothing manual and therefore I guess I'm not overfitting leaderboard given my low submission count. \n\nI'm using default model parameters (the one shared by Konrad's not the one tuned from other Kernels). Implemented a script to optimize hyper-parameters for each of my models, but haven't yet ran my models with optimized paramaters. \n\nBut I feel like I might be missing some clues.",
    "464205": "lb3.70 is great! By the way,How many features ？",
    "464161": "Hi!\nI admit i have a weak single model. \n3.646 cv and 3.695 LB.\nBut it is simple, only 36 features. \nThe main thing I've worked on and still working on is a strong and interesting blending pipeline. \nCurrently I'm on 3.634 cv and 3.679 LB.\nNo post processing. And there's still much to work on.\n\nI'm sure i can add to the team by utilizing the models and their features to the pipeline.\n\nLet me know if you wanna Skype to discuss things. You can pm me your Skype handle.",
    "464123": "with outlier classification , its 3.673 for me. Let me know.",
    "464094": "May I join your team? My best 'single model' is 3.672 which combined a Normal value model and a Outlier model.",
    "472578": ""
  }
}