{
  "id": 5969,
  "title": "Ensenble in this competition",
  "url": "/competitions/ams-2014-solar-energy-prediction-contest/discussion/5969",
  "author_name": "",
  "post_date": "2013-10-07T02:01:23.807Z",
  "votes": null,
  "comment_count": 27,
  "views": 8470,
  "content": "<p>I'm a bit curious. I tried doing model ensenbling in this competition but without much success. So far, my best model uses only one algorithm.</p>\n<p>Anybody had any luck with ensenbling in this competition?</p>",
  "messages": [
    {
      "id": "32070",
      "postDate": "10/07/2013 02:01:23",
      "content": "<p>I'm a bit curious. I tried doing model ensenbling in this competition but without much success. So far, my best model uses only one algorithm.</p>\n<p>Anybody had any luck with ensenbling in this competition?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32083",
      "postDate": "10/07/2013 13:31:37",
      "content": "<p>I have much success with 98 single models rather forecasting in long format</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32086",
      "postDate": "10/07/2013 14:01:53",
      "content": "<p>Hello. &nbsp; I tried ensembling extra trees regressor and random forest regressor,&nbsp;&nbsp; but did not improve on MAE error. &nbsp; For instance,&nbsp; Gaussian Process Regressor requires that&nbsp; trainX and trainY have the same number of rows.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32094",
      "postDate": "10/07/2013 18:50:12",
      "content": "<p>I tried Random Forest Regressor in several setups and the results performance were far worse than using Ridge, Lasso and ElasticNet.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32096",
      "postDate": "10/07/2013 19:21:56",
      "content": "<p>I think its something related to the metric. Maybe MAE is just more difficult to ensenble.</p>\n<p>Of course, in thi task its possible to do some kind of combination with the many forecasts, but i couldnt manage to use another method and combine with my current model.</p>\n<p>So for this task i think single model will have to do.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32097",
      "postDate": "10/07/2013 19:50:31",
      "content": "<p>Sounds right,</p>\n<p><br>It makes sense that MAE punishes high variance much less than MSE and I've typically seen ensembling presented/explained primarily as a way to control model variance.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32115",
      "postDate": "10/08/2013 14:46:30",
      "content": "<p>I've seen some improvement, but the feature engineering had much more impact.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32143",
      "postDate": "10/08/2013 21:08:10",
      "content": "<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32146",
      "postDate": "10/08/2013 21:22:50",
      "content": "<p>Yes I think you're right Leustagos. I tried some featuring engineering in several ways, and did not obtained any better performance.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32151",
      "postDate": "10/08/2013 22:06:11",
      "content": "<p>im my case it only made it worse...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32163",
      "postDate": "10/09/2013 05:22:28",
      "content": "<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32173",
      "postDate": "10/09/2013 12:07:31",
      "content": "<p>[quote=Herimanitra;32163]</p>\n<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>\n<p>[/quote]</p>\n<p>What do you mean by numerical model?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32174",
      "postDate": "10/09/2013 13:52:52",
      "content": "<p>[quote=Leustagos;32173]</p>\n<p>[quote=Herimanitra;32163]</p>\n<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>\n<p>[/quote]</p>\n<p>What do you mean by numerical model?</p>\n<p>[/quote]</p>\n<p>like Spline interp... not standard ML models</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32176",
      "postDate": "10/09/2013 14:04:00",
      "content": "<p>Thats a secret! To be revealed after the competition...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32239",
      "postDate": "10/11/2013 02:55:08",
      "content": "<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32249",
      "postDate": "10/11/2013 10:28:20",
      "content": "<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32255",
      "postDate": "10/11/2013 15:51:54",
      "content": "<p>[quote=Leustagos;32249]</p>\n<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Just to clarify, are you saying that you are using the same set of features for all stations, either through a muti-target estimator, or running the estimator once per station but with the identical set of features, and still obtaining an MAE of less than 1900k?&nbsp; This is the same as Kevin's question, but stated more explicitly ...</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32256",
      "postDate": "10/11/2013 16:18:13",
      "content": "<p>[quote=KK Surugucchi;32255]</p>\n<p>[quote=Leustagos;32249]</p>\n<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Just to clarify, are you saying that you are using the same set of features for all stations, either through a muti-target estimator, or running the estimator once per station but with the identical set of features, and still obtaining an MAE of less than 1900k?&nbsp; This is the same as Kevin's question, but stated more explicitly ...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Yes. Same everything.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32258",
      "postDate": "10/11/2013 16:40:57",
      "content": "<p>Thanks. I think I went in the wrong direction...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32272",
      "postDate": "10/11/2013 20:24:09",
      "content": "<p>[quote=Kevin Hwang;32258]</p>\n<p>Thanks. I think I went in the wrong direction...</p>\n<p>[/quote]</p>\n<p>Maybe not, maybe it is just my style of creating a model. What if i could get an improvment doing specific things for some stations?</p>\n<p>Maybe i'll try it later, but up until now i didnt. This is just what it means.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "32277",
      "postDate": "10/12/2013 02:55:03",
      "content": "<p>well, the best I could achieve was 205k without any feature selection and can't improve further. That's when I start focus on feature selection(not feature engineering like Amazon Employee one). Maybe or maybe not you can further improve with feature selection. But I definitely should be able to improve a lot by selecting the right model and parameter setting.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33304",
      "postDate": "11/06/2013 10:47:25",
      "content": "<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33305",
      "postDate": "11/06/2013 10:50:48",
      "content": "<p>I don think so. Most RF implementations targets MSE, and the metric of this competition is MAE, so i don't think it will give good results.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33309",
      "postDate": "11/06/2013 12:12:32",
      "content": "<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33310",
      "postDate": "11/06/2013 12:14:02",
      "content": "<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>\n<p>[/quote]</p>\n<p>hmm... Im currently running RF regression, its been on for last 12 hours lets see. btw, this score was achieved after feature engineering or on the basic features?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33311",
      "postDate": "11/06/2013 12:15:23",
      "content": "<p>[quote=Abhishek;33310]</p>\n<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>\n<p>[/quote]</p>\n<p>hmm... Im currently running RF regression, its been on for last 12 hours lets see. btw, this score was achieved after feature engineering or on the basic features?</p>\n<p>[/quote]</p>\n<p>after feature engineering...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33332",
      "postDate": "11/06/2013 18:28:01",
      "content": "<p>[quote=Herimanitra;33311]</p>\n<p>[quote=Abhishek;33310]</p>\n<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]2036k[/quote]hmm...<br> Im currently running RF regression, its been on for last 12 hours lets <br>see. btw, this score was achieved after feature engineering or on the <br>basic features?[/quote]after feature engineering...</p>\n<p>[/quote]</p>\n<p>Is this just feature selection or more involved feature engineering?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "33782",
      "postDate": "11/16/2013 00:50:53",
      "content": "<p>We got good ensembling (averaging?) of a stacked model (nearest station 5113 * 98 rows) 196 LB with models based on single stations (203 LB and 204 LB) This was 192 on LB</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 32083,
      "author_name": "herimanitra",
      "author_url": "",
      "post_date": "10/07/2013 13:31:37",
      "content": "<p>I have much success with 98 single models rather forecasting in long format</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32086,
      "author_name": "ijaganjac",
      "author_url": "",
      "post_date": "10/07/2013 14:01:53",
      "content": "<p>Hello. &nbsp; I tried ensembling extra trees regressor and random forest regressor,&nbsp;&nbsp; but did not improve on MAE error. &nbsp; For instance,&nbsp; Gaussian Process Regressor requires that&nbsp; trainX and trainY have the same number of rows.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32094,
      "author_name": "euclides",
      "author_url": "",
      "post_date": "10/07/2013 18:50:12",
      "content": "<p>I tried Random Forest Regressor in several setups and the results performance were far worse than using Ridge, Lasso and ElasticNet.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32096,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/07/2013 19:21:56",
      "content": "<p>I think its something related to the metric. Maybe MAE is just more difficult to ensenble.</p>\n<p>Of course, in thi task its possible to do some kind of combination with the many forecasts, but i couldnt manage to use another method and combine with my current model.</p>\n<p>So for this task i think single model will have to do.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32097,
      "author_name": "alecradford",
      "author_url": "",
      "post_date": "10/07/2013 19:50:31",
      "content": "<p>Sounds right,</p>\n<p><br>It makes sense that MAE punishes high variance much less than MSE and I've typically seen ensembling presented/explained primarily as a way to control model variance.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32115,
      "author_name": "konradb",
      "author_url": "",
      "post_date": "10/08/2013 14:46:30",
      "content": "<p>I've seen some improvement, but the feature engineering had much more impact.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32143,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/08/2013 21:08:10",
      "content": "<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32146,
      "author_name": "euclides",
      "author_url": "",
      "post_date": "10/08/2013 21:22:50",
      "content": "<p>Yes I think you're right Leustagos. I tried some featuring engineering in several ways, and did not obtained any better performance.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32151,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/08/2013 22:06:11",
      "content": "<p>im my case it only made it worse...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32163,
      "author_name": "herimanitra",
      "author_url": "",
      "post_date": "10/09/2013 05:22:28",
      "content": "<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32173,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/09/2013 12:07:31",
      "content": "<p>[quote=Herimanitra;32163]</p>\n<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>\n<p>[/quote]</p>\n<p>What do you mean by numerical model?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32174,
      "author_name": "herimanitra",
      "author_url": "",
      "post_date": "10/09/2013 13:52:52",
      "content": "<p>[quote=Leustagos;32173]</p>\n<p>[quote=Herimanitra;32163]</p>\n<p>[quote=Leustagos;32143]</p>\n<p>funny. i'm going in the direction of doing less and less feature engineering. In fact i know of a model that can achieve 1980k (current 10th place) that almost dont use feature engineering.</p>\n<p>[/quote]</p>\n<p>Waouh! It's surely a numerical model</p>\n<p>[/quote]</p>\n<p>What do you mean by numerical model?</p>\n<p>[/quote]</p>\n<p>like Spline interp... not standard ML models</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32176,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/09/2013 14:04:00",
      "content": "<p>Thats a secret! To be revealed after the competition...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32239,
      "author_name": "kevinhwang",
      "author_url": "",
      "post_date": "10/11/2013 02:55:08",
      "content": "<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32249,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/11/2013 10:28:20",
      "content": "<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32255,
      "author_name": "kksurugucchi",
      "author_url": "",
      "post_date": "10/11/2013 15:51:54",
      "content": "<p>[quote=Leustagos;32249]</p>\n<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Just to clarify, are you saying that you are using the same set of features for all stations, either through a muti-target estimator, or running the estimator once per station but with the identical set of features, and still obtaining an MAE of less than 1900k?&nbsp; This is the same as Kevin's question, but stated more explicitly ...</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32256,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/11/2013 16:18:13",
      "content": "<p>[quote=KK Surugucchi;32255]</p>\n<p>[quote=Leustagos;32249]</p>\n<p>[quote=Kevin Hwang;32239]</p>\n<p>Does feature engineering include feature selection?</p>\n<p>Are you saying there exists one algorithm which using all features(w/o feature selection per station), can achieve average MAE of 1980?&nbsp;</p>\n<p>Thanks.</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>MAE of 1980k or 1980000...&nbsp;</p>\n<p>As of now, i'm not doing any special treatment per station, and i managed to beat 1900k like that...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Just to clarify, are you saying that you are using the same set of features for all stations, either through a muti-target estimator, or running the estimator once per station but with the identical set of features, and still obtaining an MAE of less than 1900k?&nbsp; This is the same as Kevin's question, but stated more explicitly ...</p>\n<p>&nbsp;</p>\n<p>[/quote]</p>\n<p>Yes. Same everything.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32258,
      "author_name": "kevinhwang",
      "author_url": "",
      "post_date": "10/11/2013 16:40:57",
      "content": "<p>Thanks. I think I went in the wrong direction...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32272,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "10/11/2013 20:24:09",
      "content": "<p>[quote=Kevin Hwang;32258]</p>\n<p>Thanks. I think I went in the wrong direction...</p>\n<p>[/quote]</p>\n<p>Maybe not, maybe it is just my style of creating a model. What if i could get an improvment doing specific things for some stations?</p>\n<p>Maybe i'll try it later, but up until now i didnt. This is just what it means.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 32277,
      "author_name": "kevinhwang",
      "author_url": "",
      "post_date": "10/12/2013 02:55:03",
      "content": "<p>well, the best I could achieve was 205k without any feature selection and can't improve further. That's when I start focus on feature selection(not feature engineering like Amazon Employee one). Maybe or maybe not you can further improve with feature selection. But I definitely should be able to improve a lot by selecting the right model and parameter setting.</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33304,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "11/06/2013 10:47:25",
      "content": "<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33305,
      "author_name": "leustagos",
      "author_url": "",
      "post_date": "11/06/2013 10:50:48",
      "content": "<p>I don think so. Most RF implementations targets MSE, and the metric of this competition is MAE, so i don't think it will give good results.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33309,
      "author_name": "herimanitra",
      "author_url": "",
      "post_date": "11/06/2013 12:12:32",
      "content": "<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33310,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "11/06/2013 12:14:02",
      "content": "<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>\n<p>[/quote]</p>\n<p>hmm... Im currently running RF regression, its been on for last 12 hours lets see. btw, this score was achieved after feature engineering or on the basic features?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33311,
      "author_name": "herimanitra",
      "author_url": "",
      "post_date": "11/06/2013 12:15:23",
      "content": "<p>[quote=Abhishek;33310]</p>\n<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]</p>\n<p>2036k</p>\n<p>[/quote]</p>\n<p>hmm... Im currently running RF regression, its been on for last 12 hours lets see. btw, this score was achieved after feature engineering or on the basic features?</p>\n<p>[/quote]</p>\n<p>after feature engineering...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33332,
      "author_name": "kevinhwang",
      "author_url": "",
      "post_date": "11/06/2013 18:28:01",
      "content": "<p>[quote=Herimanitra;33311]</p>\n<p>[quote=Abhishek;33310]</p>\n<p>[quote=Herimanitra;33309]</p>\n<p>[quote=Abhishek;33304]</p>\n<p>Did anyone achieve a &quot;good&quot; score using RandomForest?</p>\n<p>[/quote]2036k[/quote]hmm...<br> Im currently running RF regression, its been on for last 12 hours lets <br>see. btw, this score was achieved after feature engineering or on the <br>basic features?[/quote]after feature engineering...</p>\n<p>[/quote]</p>\n<p>Is this just feature selection or more involved feature engineering?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 33782,
      "author_name": "domcastro",
      "author_url": "",
      "post_date": "11/16/2013 00:50:53",
      "content": "<p>We got good ensembling (averaging?) of a stacked model (nearest station 5113 * 98 rows) 196 LB with models based on single stations (203 LB and 204 LB) This was 192 on LB</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "32070": "",
    "32083": "",
    "32086": "",
    "32094": "",
    "32096": "",
    "32097": "",
    "32115": "",
    "32143": "",
    "32146": "",
    "32151": "",
    "32163": "",
    "32173": "",
    "32174": "",
    "32176": "",
    "32239": "",
    "32249": "",
    "32255": "",
    "32256": "",
    "32258": "",
    "32272": "",
    "32277": "",
    "33304": "",
    "33305": "",
    "33309": "",
    "33310": "",
    "33311": "",
    "33332": "",
    "33782": ""
  },
  "source": "meta"
}