btw, @addisonhoward and @sohier thanks for this ML comp!
\n\nCheck the information about the previous competitions:
\n\n\n\nSimilar competitions\n- VSB Power Line Fault Detection\n- LANL Earthquake Prediction funny time series :)
\n\nPosts/Papers\n- Application of Neural Networks for the Prediction of the Energy Consumption in a Supermarket\n- Short-Term Memory Solar Energy Forecasting at University of Illinois\n> LSTM networks performed the best for the prediction of the solar energy due to their abilities to connect recent features to perform present day predicitions.
\n\nData driven prediction models of energy use of appliances in a low-energy house\n> This paper presents and discusses data-driven predictive models for the energy use of appliances.
More buildings produce more generalizable modeling frameworks: Benchmarking open-source prediction methods on public building electrical meter data\n> This paper outlines a library of open source regression techniques from the Scikit-Learn Python library and describes the process of applying them to open hourly electrical meter data from 482 non-residential buildings\n
Some thoughts about Energy Consumption and Similar Data by @robikscube (kaggler and electrical engineer).
Important conclusion by @dmitriyguller
\n\n\n\n\nIf meter_reading= 0.0 means that a reading wasn't taken, and then the next positive reading \"catches up\" on the incremental changes since the last positive reading, then it's actually a problem with the RMSLE metric.\n With a mean-preserving metric, this bunching would just create noise, but on average you would still get the right incremental value; if you were supposed to have readings of \"100\" and \"200\" for the two hours, but you get \"0\" and \"300\" instead, either way you get 150 on average. However, (log(101) + log(201)) / 2 amounts to a different number than (log(1) + log(301)) / 2; it's 4.96 vs 2.85, in this example. The implication is that your prediction of the building's energy use should take into account the likelihood of the meter not updating every hour, implicitly or explicitly.\n Is the prediction for the building low because it's energy efficient, or because its meters get stuck more often? The organizers won't know, so there will be a big question mark over any conclusion that they reach from the models trained to optimize the RMSLE metric. This is the kind of nonsense that happens when you use a kludge metric, and don't think through what side effects you're introducing by using it.
\n
I'll keep adding stuff :) 👍
", "messages": [ { "id": "650312", "postDate": "10/16/2019 09:05:08", "content": "btw, @addisonhoward and @sohier thanks for this ML comp!
\n\nCheck the information about the previous competitions:
\n\n\n\nSimilar competitions\n- VSB Power Line Fault Detection\n- LANL Earthquake Prediction funny time series :)
\n\nPosts/Papers\n- Application of Neural Networks for the Prediction of the Energy Consumption in a Supermarket\n- Short-Term Memory Solar Energy Forecasting at University of Illinois\n> LSTM networks performed the best for the prediction of the solar energy due to their abilities to connect recent features to perform present day predicitions.
\n\nData driven prediction models of energy use of appliances in a low-energy house\n> This paper presents and discusses data-driven predictive models for the energy use of appliances.
More buildings produce more generalizable modeling frameworks: Benchmarking open-source prediction methods on public building electrical meter data\n> This paper outlines a library of open source regression techniques from the Scikit-Learn Python library and describes the process of applying them to open hourly electrical meter data from 482 non-residential buildings\n
Some thoughts about Energy Consumption and Similar Data by @robikscube (kaggler and electrical engineer).
Important conclusion by @dmitriyguller
\n\n\n\n\nIf meter_reading= 0.0 means that a reading wasn't taken, and then the next positive reading \"catches up\" on the incremental changes since the last positive reading, then it's actually a problem with the RMSLE metric.\n With a mean-preserving metric, this bunching would just create noise, but on average you would still get the right incremental value; if you were supposed to have readings of \"100\" and \"200\" for the two hours, but you get \"0\" and \"300\" instead, either way you get 150 on average. However, (log(101) + log(201)) / 2 amounts to a different number than (log(1) + log(301)) / 2; it's 4.96 vs 2.85, in this example. The implication is that your prediction of the building's energy use should take into account the likelihood of the meter not updating every hour, implicitly or explicitly.\n Is the prediction for the building low because it's energy efficient, or because its meters get stuck more often? The organizers won't know, so there will be a big question mark over any conclusion that they reach from the models trained to optimize the RMSLE metric. This is the kind of nonsense that happens when you use a kludge metric, and don't think through what side effects you're introducing by using it.
\n
I'll keep adding stuff :) 👍
", "rawMarkdown": "btw, @addisonhoward and @sohier thanks for this ML comp!\n\nCheck the information about the previous competitions:\n\n1. [Great Energy Predictor Shootout I](https://www.kaggle.com/c/great-energy-predictor-shootout-i/overview)\n1. [Great Building Energy Predictor Shootout II (1994)](https://oaktrust.library.tamu.edu/bitstream/handle/1969.1/2137/ESL-TR-94-07-01.pdf?sequence=1&isAllowed=y)\n\n**Similar competitions**\n- [VSB Power Line Fault Detection](https://www.kaggle.com/c/vsb-power-line-fault-detection)\n- [LANL Earthquake Prediction](https://www.kaggle.com/c/LANL-Earthquake-Prediction) funny time series :)\n\n**Posts/Papers**\n- [Application of Neural Networks for the Prediction of the Energy Consumption in a Supermarket](https://pdfs.semanticscholar.org/57ff/6819a860ee00a23b4b3360d3a124c6cdb426.pdf)\n- [Short-Term Memory Solar Energy Forecasting at University of Illinois](https://github.com/adelekuzmiakova/CS229-machine-learning-solar-energy-predictions/blob/master/cs229-final-report.pdf)\n> LSTM networks performed the best for the prediction of the solar energy due to their abilities to connect recent features to perform present day predicitions.\n\n- [Data driven prediction models of energy use of appliances in a low-energy house](https://github.com/LuisM78/Appliances-energy-prediction-data)\n> This paper presents and discusses data-driven predictive models for the energy use of appliances.\n\n- **[More buildings produce more generalizable modeling frameworks: Benchmarking open-source prediction methods on public building electrical meter data](https://github.com/buds-lab/building-prediction-benchmarking)**\n> This paper outlines a library of open source regression techniques from the Scikit-Learn Python library and describes the process of applying them to open hourly electrical meter data from 482 non-residential buildings\n\n\n- [Some thoughts about Energy Consumption and Similar Data](https://www.kaggle.com/c/ashrae-energy-prediction/discussion/112887#latest-650277) by @robikscube (kaggler and electrical engineer).\n\n---\nImportant conclusion by @dmitriyguller \n\n>If meter_reading= 0.0 means that a reading wasn't taken, and then the next positive reading \"catches up\" on the incremental changes since the last positive reading, then it's actually a problem with the RMSLE metric.\nWith a mean-preserving metric, this bunching would just create noise, but on average you would still get the right incremental value; if you were supposed to have readings of \"100\" and \"200\" for the two hours, but you get \"0\" and \"300\" instead, either way you get 150 on average. However, (log(101) + log(201)) / 2 amounts to a different number than (log(1) + log(301)) / 2; it's 4.96 vs 2.85, in this example. The implication is that your prediction of the building's energy use should take into account the likelihood of the meter not updating every hour, implicitly or explicitly.\nIs the prediction for the building low because it's energy efficient, or because its meters get stuck more often? The organizers won't know, so there will be a big question mark over any conclusion that they reach from the models trained to optimize the RMSLE metric. This is the kind of nonsense that happens when you use a kludge metric, and don't think through what side effects you're introducing by using it.\n\n---\n\n**I'll keep adding stuff :) 👍**", "votes": null }, { "id": "650382", "postDate": "10/16/2019 10:02:48", "content": "Thanks @jesucristo ! This is really useful :smile:
", "rawMarkdown": "Thanks @jesucristo ! This is really useful :smile:", "votes": null }, { "id": "651303", "postDate": "10/17/2019 09:53:32", "content": "👍
", "rawMarkdown": "👍", "votes": null }, { "id": "651312", "postDate": "10/17/2019 10:07:20", "content": "Thanks, you have my upvote.
", "rawMarkdown": "Thanks, you have my upvote.", "votes": null }, { "id": "658506", "postDate": "10/26/2019 04:39:20", "content": "[update 26Oct]\nCompetition by DataDriven Power Laws: Forecasting Energy Consumption \n@willkoehrsen was top3 there, you can top3 solutions in the official github.
", "rawMarkdown": "[update 26Oct]\nCompetition by DataDriven [Power Laws: Forecasting Energy Consumption ](https://www.drivendata.org/competitions/51/electricity-prediction-machine-learning/) \n@willkoehrsen was top3 there, you can top3 solutions [in the official github.](https://github.com/drivendataorg/power-laws-forecasting)", "votes": null }, { "id": "658545", "postDate": "10/26/2019 06:20:00", "content": "Do you know how to get the data of ‘ Power Laws: Forecasting Energy Consumption’？
", "rawMarkdown": "Do you know how to get the data of ‘ Power Laws: Forecasting Energy Consumption’？", "votes": null }, { "id": "658613", "postDate": "10/26/2019 08:58:02", "content": "I don't think it's possible. Generally in DrivenData, data remains available only during the competition.
", "rawMarkdown": "I don't think it's possible. Generally in DrivenData, data remains available only during the competition.", "votes": null }, { "id": "662958", "postDate": "11/01/2019 09:10:49", "content": "@jesucristo Thanks for this useful resource!!
", "rawMarkdown": "jesucristo Thanks for this useful resource!!", "votes": null } ], "comments": [ { "id": 650382, "author_name": "raimonds1993", "author_url": "", "post_date": "10/16/2019 10:02:48", "content": "Thanks @jesucristo ! This is really useful :smile:
", "votes": null, "replies": [] }, { "id": 651303, "author_name": "", "author_url": "", "post_date": "10/17/2019 09:53:32", "content": "👍
", "votes": null, "replies": [] }, { "id": 651312, "author_name": "xceptions", "author_url": "", "post_date": "10/17/2019 10:07:20", "content": "Thanks, you have my upvote.
", "votes": null, "replies": [] }, { "id": 658506, "author_name": "jesucristo", "author_url": "", "post_date": "10/26/2019 04:39:20", "content": "[update 26Oct]\nCompetition by DataDriven Power Laws: Forecasting Energy Consumption \n@willkoehrsen was top3 there, you can top3 solutions in the official github.
", "votes": null, "replies": [ { "id": 658545, "author_name": "", "author_url": "", "post_date": "10/26/2019 06:20:00", "content": "Do you know how to get the data of ‘ Power Laws: Forecasting Energy Consumption’？
", "votes": null, "replies": [] }, { "id": 658613, "author_name": "arnabbiswas1", "author_url": "", "post_date": "10/26/2019 08:58:02", "content": "I don't think it's possible. Generally in DrivenData, data remains available only during the competition.
", "votes": null, "replies": [] } ] }, { "id": 662958, "author_name": "pradeepmuniasamy", "author_url": "", "post_date": "11/01/2019 09:10:49", "content": "@jesucristo Thanks for this useful resource!!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "650312": "btw, @addisonhoward and @sohier thanks for this ML comp!\n\nCheck the information about the previous competitions:\n\n1. [Great Energy Predictor Shootout I](https://www.kaggle.com/c/great-energy-predictor-shootout-i/overview)\n1. [Great Building Energy Predictor Shootout II (1994)](https://oaktrust.library.tamu.edu/bitstream/handle/1969.1/2137/ESL-TR-94-07-01.pdf?sequence=1&isAllowed=y)\n\n**Similar competitions**\n- [VSB Power Line Fault Detection](https://www.kaggle.com/c/vsb-power-line-fault-detection)\n- [LANL Earthquake Prediction](https://www.kaggle.com/c/LANL-Earthquake-Prediction) funny time series :)\n\n**Posts/Papers**\n- [Application of Neural Networks for the Prediction of the Energy Consumption in a Supermarket](https://pdfs.semanticscholar.org/57ff/6819a860ee00a23b4b3360d3a124c6cdb426.pdf)\n- [Short-Term Memory Solar Energy Forecasting at University of Illinois](https://github.com/adelekuzmiakova/CS229-machine-learning-solar-energy-predictions/blob/master/cs229-final-report.pdf)\n> LSTM networks performed the best for the prediction of the solar energy due to their abilities to connect recent features to perform present day predicitions.\n\n- [Data driven prediction models of energy use of appliances in a low-energy house](https://github.com/LuisM78/Appliances-energy-prediction-data)\n> This paper presents and discusses data-driven predictive models for the energy use of appliances.\n\n- **[More buildings produce more generalizable modeling frameworks: Benchmarking open-source prediction methods on public building electrical meter data](https://github.com/buds-lab/building-prediction-benchmarking)**\n> This paper outlines a library of open source regression techniques from the Scikit-Learn Python library and describes the process of applying them to open hourly electrical meter data from 482 non-residential buildings\n\n\n- [Some thoughts about Energy Consumption and Similar Data](https://www.kaggle.com/c/ashrae-energy-prediction/discussion/112887#latest-650277) by @robikscube (kaggler and electrical engineer).\n\n---\nImportant conclusion by @dmitriyguller \n\n>If meter_reading= 0.0 means that a reading wasn't taken, and then the next positive reading \"catches up\" on the incremental changes since the last positive reading, then it's actually a problem with the RMSLE metric.\nWith a mean-preserving metric, this bunching would just create noise, but on average you would still get the right incremental value; if you were supposed to have readings of \"100\" and \"200\" for the two hours, but you get \"0\" and \"300\" instead, either way you get 150 on average. However, (log(101) + log(201)) / 2 amounts to a different number than (log(1) + log(301)) / 2; it's 4.96 vs 2.85, in this example. The implication is that your prediction of the building's energy use should take into account the likelihood of the meter not updating every hour, implicitly or explicitly.\nIs the prediction for the building low because it's energy efficient, or because its meters get stuck more often? The organizers won't know, so there will be a big question mark over any conclusion that they reach from the models trained to optimize the RMSLE metric. This is the kind of nonsense that happens when you use a kludge metric, and don't think through what side effects you're introducing by using it.\n\n---\n\n**I'll keep adding stuff :) 👍**", "650382": "Thanks @jesucristo ! This is really useful :smile:", "651303": "👍", "651312": "Thanks, you have my upvote.", "658506": "[update 26Oct]\nCompetition by DataDriven [Power Laws: Forecasting Energy Consumption ](https://www.drivendata.org/competitions/51/electricity-prediction-machine-learning/) \n@willkoehrsen was top3 there, you can top3 solutions [in the official github.](https://github.com/drivendataorg/power-laws-forecasting)", "658545": "Do you know how to get the data of ‘ Power Laws: Forecasting Energy Consumption’？", "658613": "I don't think it's possible. Generally in DrivenData, data remains available only during the competition.", "662958": "jesucristo Thanks for this useful resource!!" }, "source": "meta" }![https://raw.githubusercontent.com/buds-lab/building-prediction-benchmarking/master/visualization_notebooks/example-boxplots.png](\"https://raw.githubusercontent.com/buds-lab/building-prediction-benchmarking/master/visualization_notebooks/example-boxplots.png\"){kind=link}