![\"\"](\"https://www.mdpi.com/make/make-01-00056/article_deploy/html/images/make-01-00056-g001-550.jpg\")
Hello Fellow Kagglers:
\n\nJust a short note to introduce you to the ASHRAE \"Contest Administration\" Team, led by myself. As the current chair of ASHRAE Technical Committee 4.7 (Energy Calculations), it gives me great pleasure to see this contest launch - the next step in supporting ASHRAE's mission - \"To serve humanity by advancing the arts and sciences of heating, ventilation, air conditioning, refrigeration and their allied fields.\"
\n\nOur team is supported two of my good colleagues and friends - Dr. Jeff Haberl and Krishnan Gowri.
\n\nWe represent a small, diverse and dedicated group of volunteers who have been working together over the past 16 months to create the relationships and pull together the necessary resources to launch a successful Kaggle data competition.
\n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain. During the contest, the 'Administration Team\" will be monitoring the discussion board to respond to your queries, as appropriate. Contest winners and participants should expect more interaction with the Administration Team as we work together to find creative ways to disseminate the results to our community at large.
\n\nGood Luck!
\n\nAll the Best,
\n\n_Chris
\n\nChris Balbach, PE, CEM CMVP, BESA\nVice President of Research and Development\nPerformance Systems Development of NY, LLC\n124 Brinndley Street, Suite 4, Ithaca, NY 14850\nhttp://www.psdconsulting.com\nph: (607)-327-1647
", "messages": [ { "id": "650429", "postDate": "10/16/2019 11:23:42", "content": "Hello Fellow Kagglers:
\n\nJust a short note to introduce you to the ASHRAE \"Contest Administration\" Team, led by myself. As the current chair of ASHRAE Technical Committee 4.7 (Energy Calculations), it gives me great pleasure to see this contest launch - the next step in supporting ASHRAE's mission - \"To serve humanity by advancing the arts and sciences of heating, ventilation, air conditioning, refrigeration and their allied fields.\"
\n\nOur team is supported two of my good colleagues and friends - Dr. Jeff Haberl and Krishnan Gowri.
\n\nWe represent a small, diverse and dedicated group of volunteers who have been working together over the past 16 months to create the relationships and pull together the necessary resources to launch a successful Kaggle data competition.
\n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain. During the contest, the 'Administration Team\" will be monitoring the discussion board to respond to your queries, as appropriate. Contest winners and participants should expect more interaction with the Administration Team as we work together to find creative ways to disseminate the results to our community at large.
\n\nGood Luck!
\n\nAll the Best,
\n\n_Chris
\n\nChris Balbach, PE, CEM CMVP, BESA\nVice President of Research and Development\nPerformance Systems Development of NY, LLC\n124 Brinndley Street, Suite 4, Ithaca, NY 14850\nhttp://www.psdconsulting.com\nph: (607)-327-1647
", "rawMarkdown": "Hello Fellow Kagglers:\n\nJust a short note to introduce you to the ASHRAE \"Contest Administration\" Team, led by myself. As the current chair of ASHRAE Technical Committee 4.7 (Energy Calculations), it gives me great pleasure to see this contest launch - the next step in supporting ASHRAE's mission - \"To serve humanity by advancing the arts and sciences of heating, ventilation, air conditioning, refrigeration and their allied fields.\"\n\nOur team is supported two of my good colleagues and friends - [Dr. Jeff Haberl](https://www.linkedin.com/in/jeff-haberl-fashrae-fibpsa-44a4b946/) and [Krishnan Gowri](https://www.linkedin.com/in/krishnangowri/). \n\nWe represent a small, diverse and dedicated group of volunteers who have been working together over the past 16 months to create the relationships and pull together the necessary resources to launch a successful Kaggle data competition. \n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain. During the contest, the 'Administration Team\" will be monitoring the discussion board to respond to your queries, as appropriate. Contest winners and participants should expect more interaction with the Administration Team as we work together to find creative ways to disseminate the results to our community at large. \n\nGood Luck!\n\nAll the Best,\n\n_Chris\n\nChris Balbach, PE, CEM CMVP, BESA\nVice President of Research and Development\nPerformance Systems Development of NY, LLC\n124 Brinndley Street, Suite 4, Ithaca, NY 14850\nhttp://www.psdconsulting.com\nph: (607)-327-1647", "votes": null }, { "id": "650506", "postDate": "10/16/2019 12:59:25", "content": "Hi Chris! Thanks for hosting this competition. I think it's going to be a lot of fun.
\n\nI was wondering if you could clear up some confusion about using actualized weather data in a time series competition.\nAs you said:
\n\n\n\n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain.
\n
But typically the best practice is to not actualized data from the future when trying to predict the future - and is considered to be \"Data leakage\"
\n\nIs there something unique about this competition that makes using actualized weather data in the test set different? Or is it more so because that in a competitive setting there is no possible way to restrict the use of weather data?
", "rawMarkdown": "Hi Chris! Thanks for hosting this competition. I think it's going to be a lot of fun.\n\nI was wondering if you could clear up some confusion about using actualized weather data in a time series competition.\nAs you said:\n> Our shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain.\n\n[But typically the best practice is to not actualized data from the future when trying to predict the future](https://towardsdatascience.com/data-leakage-part-i-think-you-have-a-great-machine-learning-model-think-again-ad44921fbf34) - and is considered to be \"Data leakage\"\n\nIs there something unique about this competition that makes using actualized weather data in the test set different? Or is it more so because that in a competitive setting there is no possible way to restrict the use of weather data?", "votes": null }, { "id": "650657", "postDate": "10/16/2019 15:19:01", "content": "Thanks for the hosting.
", "rawMarkdown": "Thanks for the hosting.", "votes": null }, { "id": "650932", "postDate": "10/16/2019 21:07:16", "content": "Hi Rob
\n\nOne of the 'real world' use cases for the types of prediction models that this competition is targeting are Energy Efficiency Investments to our built environment that are financed using \"Energy Performance Contracting\" (EPC) mechanisms.
\n\nIn an effort to increase the standardization (thus lower the costs) of EPC project delivery, the Efficiency Valuation Organization (EVO) has created and distributes \"best practice\" documents for the Measurement and Verification procedures incorporated into Energy Performance Contracts. These documents and procedures are referred to as the International Performance Measurement and Verification Protocol (IPMVP).
\n\nThe details of IPMVP \"best-practice\" documents are behind a pay-wall, but the premise of how models that forecast energy usage can be seen here.
\n\nIn short, the green area of the time series cartoon, between the forecast energy and the \"actual' (i.e. measured), energy, represents 'savings' related to a specific set of improvements financed under a Performance Contract agreement.
\n\nHope that helps explain some of the ways we hope results from this competition will be able to be used - to improve the techniques currently used for forecasting energy usage, after improvements to a building have occurred.
\n\n_Chris
", "rawMarkdown": "Hi Rob\n\nOne of the 'real world' use cases for the types of prediction models that this competition is targeting are Energy Efficiency Investments to our built environment that are financed using \"[Energy Performance Contracting](https://www.energy.gov/eere/slsc/energy-savings-performance-contracting)\" (EPC) mechanisms. \n\nIn an effort to increase the standardization (thus lower the costs) of EPC project delivery, the Efficiency Valuation Organization (EVO) has created and distributes \"best practice\" documents for the Measurement and Verification procedures incorporated into Energy Performance Contracts. These documents and procedures are referred to as the International Performance Measurement and Verification Protocol (IPMVP). \n\nThe details of IPMVP \"best-practice\" documents are behind a pay-wall, but the premise of how models that forecast energy usage can be seen [here](https://evo-world.org/en/products-services-mainmenu-en/protocols/ipmvp).\n\nIn short, the green area of the time series cartoon, between the forecast energy and the \"actual' (i.e. measured), energy, represents 'savings' related to a specific set of improvements financed under a Performance Contract agreement.\n\nHope that helps explain some of the ways we hope results from this competition will be able to be used - to improve the techniques currently used for forecasting energy usage, after improvements to a building have occurred. \n\n_Chris", "votes": null }, { "id": "651212", "postDate": "10/17/2019 07:19:08", "content": "Hi @chrisbalbach \nThanks!\nCan you say whether the dataset has buildings that have\\don't have these improvements that are supposed to save energy?\nor whether some of the buildings have improvements installed during the train\\test period?\nMaybe a specific building has improvements that affect warm water and not electricity within the dataset?
\n\nThanks!
", "rawMarkdown": "Hi @chrisbalbach \nThanks!\nCan you say whether the dataset has buildings that have\\don't have these improvements that are supposed to save energy?\nor whether some of the buildings have improvements installed during the train\\test period?\nMaybe a specific building has improvements that affect warm water and not electricity within the dataset?\n\nThanks!", "votes": null }, { "id": "651686", "postDate": "10/17/2019 19:51:17", "content": "Consider the scenario laid out in this diagram. This competition simulates the modeling challenge presented at the end of the timeline when measured energy and weather conditions are known and the adjusted baseline energy must be calculated. \n
For further details, I recommend reading this paper by @claytonmiller: https://www.mdpi.com/2504-4990/
\n\nEdit: Updated for clarity
", "rawMarkdown": "Consider the scenario laid out in this diagram. This competition simulates the modeling challenge presented at the end of the timeline when measured energy and weather conditions are known and the adjusted baseline energy must be calculated. \n\n\nFor further details, I recommend reading [this paper](https://www.mdpi.com/2504-4990/1/3/56) by @claytonmiller: https://www.mdpi.com/2504-4990/\n\nEdit: Updated for clarity", "votes": null }, { "id": "652093", "postDate": "10/18/2019 10:39:51", "content": "Thanks for hosting this competition @chrisbalbach \nCould you inform us the location where this data was collected?\nAs holidays are massive outliers on the consumption and could greatly impact the predictability of the models (as your linked paper also points out).\nIs the location break down similar to the one in the paper?\nThank you again already in advance.
", "rawMarkdown": "Thanks for hosting this competition @chrisbalbach \nCould you inform us the location where this data was collected?\nAs holidays are massive outliers on the consumption and could greatly impact the predictability of the models (as your linked paper also points out).\nIs the location break down similar to the one in the paper?\nThank you again already in advance.", "votes": null }, { "id": "653455", "postDate": "10/20/2019 13:42:47", "content": "Thanks for hosting this interesting competition! In the description of the previous competition in 1994 it's stated that:
\n\n\n\n\nSix winning teams were formally recognized, all participants were asked to write an ASHRAE paper to document their efforts and to present at ASHRAE conferences. As a part of these efforts, over a dozen peer-reviewed papers were published and several software vendors incorporated the algorithms described in these papers.
\n
I think I found what is the summarizing paper here: https://www.osti.gov/biblio/33315, but is there anything publicly available?
", "rawMarkdown": "Thanks for hosting this interesting competition! In the description of the previous competition in 1994 it's stated that:\n\n> Six winning teams were formally recognized, all participants were asked to write an ASHRAE paper to document their efforts and to present at ASHRAE conferences. As a part of these efforts, over a dozen peer-reviewed papers were published and several software vendors incorporated the algorithms described in these papers.\n\nI think I found what is the summarizing paper here: https://www.osti.gov/biblio/33315, but is there anything publicly available?", "votes": null }, { "id": "654201", "postDate": "10/21/2019 15:32:17", "content": "We posted several of the papers written about the prior competitions with our rehost of the prior competition: https://www.kaggle.com/c/great-energy-predictor-shootout-i/data
", "rawMarkdown": "We posted several of the papers written about the prior competitions with our rehost of the prior competition: https://www.kaggle.com/c/great-energy-predictor-shootout-i/data", "votes": null }, { "id": "655055", "postDate": "10/22/2019 16:37:10", "content": "The second link isn't quite right! As it is, the one ending in just 2504-4990 links to a paper on an anti-tumor agent, titled \"Design, Synthesis and Applications of Hyaluronic Acid-Paclitaxel Bioconjugates\" by Francesca Leonelli, Angela La Bella, Luisa Maria Migneco and Rinaldo Marini Bettolo. \nThe first link is correct, and the link address is: https://www.mdpi.com/2504-4990/1/3/56
", "rawMarkdown": "The second link isn't quite right! As it is, the one ending in just 2504-4990 links to a paper on an anti-tumor agent, titled \"Design, Synthesis and Applications of Hyaluronic Acid-Paclitaxel Bioconjugates\" by Francesca Leonelli, Angela La Bella, Luisa Maria Migneco and Rinaldo Marini Bettolo. \nThe first link is correct, and the link address is: https://www.mdpi.com/2504-4990/1/3/56", "votes": null }, { "id": "655993", "postDate": "10/23/2019 19:01:29", "content": "Ahh, didn't notice. Cheers!
", "rawMarkdown": "Ahh, didn't notice. Cheers!", "votes": null }, { "id": "656054", "postDate": "10/23/2019 21:23:30", "content": "Nice work! Thank you.
", "rawMarkdown": "Nice work! Thank you.", "votes": null }, { "id": "656084", "postDate": "10/23/2019 22:16:23", "content": "@chrisbalbach - A related question to Henrique's - also related to location - are the timestamps in local time or GMT? Also, is there an hour shift for DST in the timestamps? I suppose I can derive the time zone by fitting to known consumption patterns, but if you will provide that information, I can save the time..
", "rawMarkdown": "chrisbalbach - A related question to Henrique's - also related to location - are the timestamps in local time or GMT? Also, is there an hour shift for DST in the timestamps? I suppose I can derive the time zone by fitting to known consumption patterns, but if you will provide that information, I can save the time..", "votes": null }, { "id": "661746", "postDate": "10/30/2019 16:56:38", "content": "I am noticed that there are two competitions of great energy predictor. Which data makes more sense? And Competition III data seems has some reading errors on meter_reading. How could we know the result of this meter_reading is not correct, which should not be included when training the model. In other words, any criteria or hinds that we could get?\nBTW, when I loaded competition III data, I could not find any data dictionary to describe the columns. However, when I reviewed the other competitor's discussion, some of them seems have files for describing the data. For example, meter column is code: 0 means electricity, 1 is chilledwater, and so on.
\n\nThanks.
", "rawMarkdown": "I am noticed that there are two competitions of great energy predictor. Which data makes more sense? And Competition III data seems has some reading errors on meter_reading. How could we know the result of this meter_reading is not correct, which should not be included when training the model. In other words, any criteria or hinds that we could get?\nBTW, when I loaded competition III data, I could not find any data dictionary to describe the columns. However, when I reviewed the other competitor's discussion, some of them seems have files for describing the data. For example, meter column is code: 0 means electricity, 1 is chilledwater, and so on. \n\nThanks.", "votes": null }, { "id": "673058", "postDate": "11/14/2019 13:23:00", "content": "The data dictionary is available in Data tab once you scroll down the full description
", "rawMarkdown": "The data dictionary is available in Data tab once you scroll down the full description", "votes": null }, { "id": "674459", "postDate": "11/16/2019 14:38:35", "content": "Thanks for hosting! I'm happy to see this competition on Kaggle.
", "rawMarkdown": "Thanks for hosting! I'm happy to see this competition on Kaggle.", "votes": null }, { "id": "675794", "postDate": "11/18/2019 15:05:53", "content": "Awesome
", "rawMarkdown": "Awesome", "votes": null }, { "id": "688944", "postDate": "12/06/2019 09:19:01", "content": "great，thanks！
", "rawMarkdown": "great，thanks！", "votes": null }, { "id": "697482", "postDate": "12/18/2019 01:34:47", "content": "sir how can be the row 41697600
", "rawMarkdown": "sir how can be the row 41697600", "votes": null } ], "comments": [ { "id": 650506, "author_name": "robikscube", "author_url": "", "post_date": "10/16/2019 12:59:25", "content": "Hi Chris! Thanks for hosting this competition. I think it's going to be a lot of fun.
\n\nI was wondering if you could clear up some confusion about using actualized weather data in a time series competition.\nAs you said:
\n\n\n\n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain.
\n
But typically the best practice is to not actualized data from the future when trying to predict the future - and is considered to be \"Data leakage\"
\n\nIs there something unique about this competition that makes using actualized weather data in the test set different? Or is it more so because that in a competitive setting there is no possible way to restrict the use of weather data?
", "votes": null, "replies": [] }, { "id": 650657, "author_name": "grapestone5321", "author_url": "", "post_date": "10/16/2019 15:19:01", "content": "Thanks for the hosting.
", "votes": null, "replies": [] }, { "id": 650932, "author_name": "chrisbalbach", "author_url": "", "post_date": "10/16/2019 21:07:16", "content": "Hi Rob
\n\nOne of the 'real world' use cases for the types of prediction models that this competition is targeting are Energy Efficiency Investments to our built environment that are financed using \"Energy Performance Contracting\" (EPC) mechanisms.
\n\nIn an effort to increase the standardization (thus lower the costs) of EPC project delivery, the Efficiency Valuation Organization (EVO) has created and distributes \"best practice\" documents for the Measurement and Verification procedures incorporated into Energy Performance Contracts. These documents and procedures are referred to as the International Performance Measurement and Verification Protocol (IPMVP).
\n\nThe details of IPMVP \"best-practice\" documents are behind a pay-wall, but the premise of how models that forecast energy usage can be seen here.
\n\nIn short, the green area of the time series cartoon, between the forecast energy and the \"actual' (i.e. measured), energy, represents 'savings' related to a specific set of improvements financed under a Performance Contract agreement.
\n\nHope that helps explain some of the ways we hope results from this competition will be able to be used - to improve the techniques currently used for forecasting energy usage, after improvements to a building have occurred.
\n\n_Chris
", "votes": null, "replies": [] }, { "id": 651212, "author_name": "tpthegreat", "author_url": "", "post_date": "10/17/2019 07:19:08", "content": "Hi @chrisbalbach \nThanks!\nCan you say whether the dataset has buildings that have\\don't have these improvements that are supposed to save energy?\nor whether some of the buildings have improvements installed during the train\\test period?\nMaybe a specific building has improvements that affect warm water and not electricity within the dataset?
\n\nThanks!
", "votes": null, "replies": [] }, { "id": 651686, "author_name": "chrisbalbach", "author_url": "", "post_date": "10/17/2019 19:51:17", "content": "Consider the scenario laid out in this diagram. This competition simulates the modeling challenge presented at the end of the timeline when measured energy and weather conditions are known and the adjusted baseline energy must be calculated. \n
For further details, I recommend reading this paper by @claytonmiller: https://www.mdpi.com/2504-4990/
\n\nEdit: Updated for clarity
", "votes": null, "replies": [ { "id": 655055, "author_name": "lindsberlin", "author_url": "", "post_date": "10/22/2019 16:37:10", "content": "The second link isn't quite right! As it is, the one ending in just 2504-4990 links to a paper on an anti-tumor agent, titled \"Design, Synthesis and Applications of Hyaluronic Acid-Paclitaxel Bioconjugates\" by Francesca Leonelli, Angela La Bella, Luisa Maria Migneco and Rinaldo Marini Bettolo. \nThe first link is correct, and the link address is: https://www.mdpi.com/2504-4990/1/3/56
", "votes": null, "replies": [] } ] }, { "id": 652093, "author_name": "hmendonca", "author_url": "", "post_date": "10/18/2019 10:39:51", "content": "Thanks for hosting this competition @chrisbalbach \nCould you inform us the location where this data was collected?\nAs holidays are massive outliers on the consumption and could greatly impact the predictability of the models (as your linked paper also points out).\nIs the location break down similar to the one in the paper?\nThank you again already in advance.
", "votes": null, "replies": [ { "id": 656084, "author_name": "wjholst", "author_url": "", "post_date": "10/23/2019 22:16:23", "content": "@chrisbalbach - A related question to Henrique's - also related to location - are the timestamps in local time or GMT? Also, is there an hour shift for DST in the timestamps? I suppose I can derive the time zone by fitting to known consumption patterns, but if you will provide that information, I can save the time..
", "votes": null, "replies": [] } ] }, { "id": 653455, "author_name": "tahaum", "author_url": "", "post_date": "10/20/2019 13:42:47", "content": "Thanks for hosting this interesting competition! In the description of the previous competition in 1994 it's stated that:
\n\n\n\n\nSix winning teams were formally recognized, all participants were asked to write an ASHRAE paper to document their efforts and to present at ASHRAE conferences. As a part of these efforts, over a dozen peer-reviewed papers were published and several software vendors incorporated the algorithms described in these papers.
\n
I think I found what is the summarizing paper here: https://www.osti.gov/biblio/33315, but is there anything publicly available?
", "votes": null, "replies": [ { "id": 654201, "author_name": "sohier", "author_url": "", "post_date": "10/21/2019 15:32:17", "content": "We posted several of the papers written about the prior competitions with our rehost of the prior competition: https://www.kaggle.com/c/great-energy-predictor-shootout-i/data
", "votes": null, "replies": [] }, { "id": 655993, "author_name": "tahaum", "author_url": "", "post_date": "10/23/2019 19:01:29", "content": "Ahh, didn't notice. Cheers!
", "votes": null, "replies": [] }, { "id": 661746, "author_name": "xiaoyingyu06", "author_url": "", "post_date": "10/30/2019 16:56:38", "content": "I am noticed that there are two competitions of great energy predictor. Which data makes more sense? And Competition III data seems has some reading errors on meter_reading. How could we know the result of this meter_reading is not correct, which should not be included when training the model. In other words, any criteria or hinds that we could get?\nBTW, when I loaded competition III data, I could not find any data dictionary to describe the columns. However, when I reviewed the other competitor's discussion, some of them seems have files for describing the data. For example, meter column is code: 0 means electricity, 1 is chilledwater, and so on.
\n\nThanks.
", "votes": null, "replies": [] }, { "id": 673058, "author_name": "aravindankrishnan", "author_url": "", "post_date": "11/14/2019 13:23:00", "content": "The data dictionary is available in Data tab once you scroll down the full description
", "votes": null, "replies": [] } ] }, { "id": 656054, "author_name": "agastya124", "author_url": "", "post_date": "10/23/2019 21:23:30", "content": "Nice work! Thank you.
", "votes": null, "replies": [] }, { "id": 674459, "author_name": "dc108349", "author_url": "", "post_date": "11/16/2019 14:38:35", "content": "Thanks for hosting! I'm happy to see this competition on Kaggle.
", "votes": null, "replies": [] }, { "id": 675794, "author_name": "rjchauhan", "author_url": "", "post_date": "11/18/2019 15:05:53", "content": "Awesome
", "votes": null, "replies": [] }, { "id": 688944, "author_name": "qkmichael", "author_url": "", "post_date": "12/06/2019 09:19:01", "content": "great，thanks！
", "votes": null, "replies": [] }, { "id": 697482, "author_name": "amitsorde", "author_url": "", "post_date": "12/18/2019 01:34:47", "content": "sir how can be the row 41697600
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "650429": "Hello Fellow Kagglers:\n\nJust a short note to introduce you to the ASHRAE \"Contest Administration\" Team, led by myself. As the current chair of ASHRAE Technical Committee 4.7 (Energy Calculations), it gives me great pleasure to see this contest launch - the next step in supporting ASHRAE's mission - \"To serve humanity by advancing the arts and sciences of heating, ventilation, air conditioning, refrigeration and their allied fields.\"\n\nOur team is supported two of my good colleagues and friends - [Dr. Jeff Haberl](https://www.linkedin.com/in/jeff-haberl-fashrae-fibpsa-44a4b946/) and [Krishnan Gowri](https://www.linkedin.com/in/krishnangowri/). \n\nWe represent a small, diverse and dedicated group of volunteers who have been working together over the past 16 months to create the relationships and pull together the necessary resources to launch a successful Kaggle data competition. \n\nOur shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain. During the contest, the 'Administration Team\" will be monitoring the discussion board to respond to your queries, as appropriate. Contest winners and participants should expect more interaction with the Administration Team as we work together to find creative ways to disseminate the results to our community at large. \n\nGood Luck!\n\nAll the Best,\n\n_Chris\n\nChris Balbach, PE, CEM CMVP, BESA\nVice President of Research and Development\nPerformance Systems Development of NY, LLC\n124 Brinndley Street, Suite 4, Ithaca, NY 14850\nhttp://www.psdconsulting.com\nph: (607)-327-1647", "650506": "Hi Chris! Thanks for hosting this competition. I think it's going to be a lot of fun.\n\nI was wondering if you could clear up some confusion about using actualized weather data in a time series competition.\nAs you said:\n> Our shared goals are to uncover (and disseminate to the public) current \"best practices\" for methods of time series predictions that can be useful in our domain.\n\n[But typically the best practice is to not actualized data from the future when trying to predict the future](https://towardsdatascience.com/data-leakage-part-i-think-you-have-a-great-machine-learning-model-think-again-ad44921fbf34) - and is considered to be \"Data leakage\"\n\nIs there something unique about this competition that makes using actualized weather data in the test set different? Or is it more so because that in a competitive setting there is no possible way to restrict the use of weather data?", "650657": "Thanks for the hosting.", "650932": "Hi Rob\n\nOne of the 'real world' use cases for the types of prediction models that this competition is targeting are Energy Efficiency Investments to our built environment that are financed using \"[Energy Performance Contracting](https://www.energy.gov/eere/slsc/energy-savings-performance-contracting)\" (EPC) mechanisms. \n\nIn an effort to increase the standardization (thus lower the costs) of EPC project delivery, the Efficiency Valuation Organization (EVO) has created and distributes \"best practice\" documents for the Measurement and Verification procedures incorporated into Energy Performance Contracts. These documents and procedures are referred to as the International Performance Measurement and Verification Protocol (IPMVP). \n\nThe details of IPMVP \"best-practice\" documents are behind a pay-wall, but the premise of how models that forecast energy usage can be seen [here](https://evo-world.org/en/products-services-mainmenu-en/protocols/ipmvp).\n\nIn short, the green area of the time series cartoon, between the forecast energy and the \"actual' (i.e. measured), energy, represents 'savings' related to a specific set of improvements financed under a Performance Contract agreement.\n\nHope that helps explain some of the ways we hope results from this competition will be able to be used - to improve the techniques currently used for forecasting energy usage, after improvements to a building have occurred. \n\n_Chris", "651212": "Hi @chrisbalbach \nThanks!\nCan you say whether the dataset has buildings that have\\don't have these improvements that are supposed to save energy?\nor whether some of the buildings have improvements installed during the train\\test period?\nMaybe a specific building has improvements that affect warm water and not electricity within the dataset?\n\nThanks!", "651686": "Consider the scenario laid out in this diagram. This competition simulates the modeling challenge presented at the end of the timeline when measured energy and weather conditions are known and the adjusted baseline energy must be calculated. \n\n\nFor further details, I recommend reading [this paper](https://www.mdpi.com/2504-4990/1/3/56) by @claytonmiller: https://www.mdpi.com/2504-4990/\n\nEdit: Updated for clarity", "652093": "Thanks for hosting this competition @chrisbalbach \nCould you inform us the location where this data was collected?\nAs holidays are massive outliers on the consumption and could greatly impact the predictability of the models (as your linked paper also points out).\nIs the location break down similar to the one in the paper?\nThank you again already in advance.", "653455": "Thanks for hosting this interesting competition! In the description of the previous competition in 1994 it's stated that:\n\n> Six winning teams were formally recognized, all participants were asked to write an ASHRAE paper to document their efforts and to present at ASHRAE conferences. As a part of these efforts, over a dozen peer-reviewed papers were published and several software vendors incorporated the algorithms described in these papers.\n\nI think I found what is the summarizing paper here: https://www.osti.gov/biblio/33315, but is there anything publicly available?", "654201": "We posted several of the papers written about the prior competitions with our rehost of the prior competition: https://www.kaggle.com/c/great-energy-predictor-shootout-i/data", "655055": "The second link isn't quite right! As it is, the one ending in just 2504-4990 links to a paper on an anti-tumor agent, titled \"Design, Synthesis and Applications of Hyaluronic Acid-Paclitaxel Bioconjugates\" by Francesca Leonelli, Angela La Bella, Luisa Maria Migneco and Rinaldo Marini Bettolo. \nThe first link is correct, and the link address is: https://www.mdpi.com/2504-4990/1/3/56", "655993": "Ahh, didn't notice. Cheers!", "656054": "Nice work! Thank you.", "656084": "chrisbalbach - A related question to Henrique's - also related to location - are the timestamps in local time or GMT? Also, is there an hour shift for DST in the timestamps? I suppose I can derive the time zone by fitting to known consumption patterns, but if you will provide that information, I can save the time..", "661746": "I am noticed that there are two competitions of great energy predictor. Which data makes more sense? And Competition III data seems has some reading errors on meter_reading. How could we know the result of this meter_reading is not correct, which should not be included when training the model. In other words, any criteria or hinds that we could get?\nBTW, when I loaded competition III data, I could not find any data dictionary to describe the columns. However, when I reviewed the other competitor's discussion, some of them seems have files for describing the data. For example, meter column is code: 0 means electricity, 1 is chilledwater, and so on. \n\nThanks.", "673058": "The data dictionary is available in Data tab once you scroll down the full description", "674459": "Thanks for hosting! I'm happy to see this competition on Kaggle.", "675794": "Awesome", "688944": "great，thanks！", "697482": "sir how can be the row 41697600" }, "source": "meta" }