{ "id": 118685, "title": "What is leaked data?", "url": "/competitions/ashrae-energy-prediction/discussion/118685", "author_name": "", "post_date": "2019-11-23T15:12:04.489324Z", "votes": 3, "comment_count": 2, "views": 0, "content": "

Is this data that I should delete? I just graduated from college and doing this for fun. But I'm getting very low at 1.46.

", "messages": [ { "id": "679870", "postDate": "11/23/2019 15:12:04", "content": "

Is this data that I should delete? I just graduated from college and doing this for fun. But I'm getting very low at 1.46.

", "rawMarkdown": "Is this data that I should delete? I just graduated from college and doing this for fun. But I'm getting very low at 1.46.", "votes": null }, { "id": "679941", "postDate": "11/23/2019 17:46:30", "content": "
\n

Data Leakage is the presence of unexpected additional information in the training data, allowing a model or machine learning algorithm to make unrealistically good predictions.

\n \n

Leakage is a pervasive challenge in applied machine learning, causing models to over-represent their generalization error and often rendering them useless in the real world. It can be caused by human or mechanical error, and can be intentional or unintentional in both cases.

\n
\n\n

You can find out more info in the competition documentation here:\nhttps://www.kaggle.com/docs/competitions#what-is-leakage

", "rawMarkdown": "> Data Leakage is the presence of unexpected additional information in the training data, allowing a model or machine learning algorithm to make unrealistically good predictions.\n\n> Leakage is a pervasive challenge in applied machine learning, causing models to over-represent their generalization error and often rendering them useless in the real world. It can be caused by human or mechanical error, and can be intentional or unintentional in both cases.\n\nYou can find out more info in the competition documentation here:\nhttps://www.kaggle.com/docs/competitions#what-is-leakage", "votes": null }, { "id": "680195", "postDate": "11/24/2019 07:46:50", "content": "

In terms of competitions, it means that the target values of the test set have been found, or some very strong indicator like the mean of the meter reading per house, per hour or something like that. It is not data that is given by the organizers. In fact, they try to stop participants from finding this data. :) To improve your score, you should check out other high scoring kernels and see what they have done and you haven't. This way, over time you will learn more concepts which can be applied in other competitions as well.

", "rawMarkdown": "In terms of competitions, it means that the target values of the test set have been found, or some very strong indicator like the mean of the meter reading per house, per hour or something like that. It is not data that is given by the organizers. In fact, they try to stop participants from finding this data. :) To improve your score, you should check out other high scoring kernels and see what they have done and you haven't. This way, over time you will learn more concepts which can be applied in other competitions as well.", "votes": null } ], "comments": [ { "id": 679941, "author_name": "andreagarritano", "author_url": "", "post_date": "11/23/2019 17:46:30", "content": "
\n

Data Leakage is the presence of unexpected additional information in the training data, allowing a model or machine learning algorithm to make unrealistically good predictions.

\n \n

Leakage is a pervasive challenge in applied machine learning, causing models to over-represent their generalization error and often rendering them useless in the real world. It can be caused by human or mechanical error, and can be intentional or unintentional in both cases.

\n
\n\n

You can find out more info in the competition documentation here:\nhttps://www.kaggle.com/docs/competitions#what-is-leakage

", "votes": null, "replies": [] }, { "id": 680195, "author_name": "yushg123", "author_url": "", "post_date": "11/24/2019 07:46:50", "content": "

In terms of competitions, it means that the target values of the test set have been found, or some very strong indicator like the mean of the meter reading per house, per hour or something like that. It is not data that is given by the organizers. In fact, they try to stop participants from finding this data. :) To improve your score, you should check out other high scoring kernels and see what they have done and you haven't. This way, over time you will learn more concepts which can be applied in other competitions as well.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "679870": "Is this data that I should delete? I just graduated from college and doing this for fun. But I'm getting very low at 1.46.", "679941": "> Data Leakage is the presence of unexpected additional information in the training data, allowing a model or machine learning algorithm to make unrealistically good predictions.\n\n> Leakage is a pervasive challenge in applied machine learning, causing models to over-represent their generalization error and often rendering them useless in the real world. It can be caused by human or mechanical error, and can be intentional or unintentional in both cases.\n\nYou can find out more info in the competition documentation here:\nhttps://www.kaggle.com/docs/competitions#what-is-leakage", "680195": "In terms of competitions, it means that the target values of the test set have been found, or some very strong indicator like the mean of the meter reading per house, per hour or something like that. It is not data that is given by the organizers. In fact, they try to stop participants from finding this data. :) To improve your score, you should check out other high scoring kernels and see what they have done and you haven't. This way, over time you will learn more concepts which can be applied in other competitions as well." }, "source": "meta" }