![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F4001300%2F2300ce6da43d4e6dbe376b5ed993078d%2FClipboard04.jpg?generation=1729218053439456&alt=media\")
In this competition, unlike previous ones, the time series API has been updated. I struggled to understand it, but I finally started to grasp how to use lag, so I’d like to share my insights.
\nhttps://www.kaggle.com/code/chumajin/janestreet-easy-to-understand-new-time-series-api
\nI think there are 3 steps.
\nStep 1 The data is split by day using group_by.
\nStep 2 The data made by Step1 is split by time_id using group_by, and the lag is generated (for time_id == 0).
\nStep 3 After inputting the predicted values, the submission file is generated.
\nThe following image is the comparing the new time series API and my understanding
\n(Detail is in above my public notebook! )
It can be understood that the official API code omits the part where the date_id and time_id using a for loop are assigned. By understanding this overall structure, I believe you will be able to participate in the competition in easy.
\nEnjoy !
", "messages": [ { "id": "3020893", "postDate": "10/18/2024 02:28:32", "content": "In this competition, unlike previous ones, the time series API has been updated. I struggled to understand it, but I finally started to grasp how to use lag, so I’d like to share my insights.
\nhttps://www.kaggle.com/code/chumajin/janestreet-easy-to-understand-new-time-series-api
\nI think there are 3 steps.
\nStep 1 The data is split by day using group_by.
\nStep 2 The data made by Step1 is split by time_id using group_by, and the lag is generated (for time_id == 0).
\nStep 3 After inputting the predicted values, the submission file is generated.
\nThe following image is the comparing the new time series API and my understanding
\n(Detail is in above my public notebook! )
It can be understood that the official API code omits the part where the date_id and time_id using a for loop are assigned. By understanding this overall structure, I believe you will be able to participate in the competition in easy.
\nEnjoy !
", "rawMarkdown": "In this competition, unlike previous ones, the time series API has been updated. I struggled to understand it, but I finally started to grasp how to use lag, so I’d like to share my insights.\n\nhttps://www.kaggle.com/code/chumajin/janestreet-easy-to-understand-new-time-series-api\n\n# Summary\n\nI think there are 3 steps.\n\nStep 1 The data is split by day using group_by.\n\nStep 2 The data made by Step1 is split by time_id using group_by, and the lag is generated (for time_id == 0).\n\nStep 3 After inputting the predicted values, the submission file is generated.\n\nThe following image is the comparing the new time series API and my understanding\n(Detail is in above my public notebook! )\n\n\n\nIt can be understood that the official API code omits the part where the date_id and time_id using a for loop are assigned. By understanding this overall structure, I believe you will be able to participate in the competition in easy.\n\nEnjoy !", "votes": null }, { "id": "3021100", "postDate": "10/18/2024 07:35:33", "content": "You can also check my work related to the submission API here: https://www.kaggle.com/code/shiyili/js24-rmf-generate-synthetic-test-data
", "rawMarkdown": "You can also check my work related to the submission API here: https://www.kaggle.com/code/shiyili/js24-rmf-generate-synthetic-test-data", "votes": null }, { "id": "3021129", "postDate": "10/18/2024 08:02:23", "content": "@shiyili OK. I will check it. Thank you very much!
", "rawMarkdown": "shiyili OK. I will check it. Thank you very much!", "votes": null }, { "id": "3054531", "postDate": "11/24/2024 19:18:25", "content": "Hi,
\nThanks for that.
\nShall we understand the lag as a previous day close equivalent, or previous day change in the asset price (which ever responder_6 is meant to represent ?)
\nTherefore, to get this info on the train set, it's possible to use the responders values at the last time_id of each date_id of the previous day to get those lags ? i.e for date_id 2, we use the last time_id data of each responders for each symbol present in date_id 1 ?
Thanks!
", "rawMarkdown": "Hi,\nThanks for that.\nShall we understand the lag as a previous day close equivalent, or previous day change in the asset price (which ever responder_6 is meant to represent ?) \nTherefore, to get this info on the train set, it's possible to use the responders values at the last time_id of each date_id of the previous day to get those lags ? i.e for date_id 2, we use the last time_id data of each responders for each symbol present in date_id 1 ?\n\nThanks!", "votes": null }, { "id": "3078527", "postDate": "12/22/2024 12:41:02", "content": "dude this is really helpful, insanely good saved my day big thanks
", "rawMarkdown": "dude this is really helpful, insanely good saved my day big thanks", "votes": null } ], "comments": [ { "id": 3021100, "author_name": "shiyili", "author_url": "", "post_date": "10/18/2024 07:35:33", "content": "You can also check my work related to the submission API here: https://www.kaggle.com/code/shiyili/js24-rmf-generate-synthetic-test-data
", "votes": null, "replies": [ { "id": 3021129, "author_name": "chumajin", "author_url": "", "post_date": "10/18/2024 08:02:23", "content": "@shiyili OK. I will check it. Thank you very much!
", "votes": null, "replies": [] }, { "id": 3078527, "author_name": "zhangyue199", "author_url": "", "post_date": "12/22/2024 12:41:02", "content": "dude this is really helpful, insanely good saved my day big thanks
", "votes": null, "replies": [] } ] }, { "id": 3054531, "author_name": "julienfron", "author_url": "", "post_date": "11/24/2024 19:18:25", "content": "Hi,
\nThanks for that.
\nShall we understand the lag as a previous day close equivalent, or previous day change in the asset price (which ever responder_6 is meant to represent ?)
\nTherefore, to get this info on the train set, it's possible to use the responders values at the last time_id of each date_id of the previous day to get those lags ? i.e for date_id 2, we use the last time_id data of each responders for each symbol present in date_id 1 ?
Thanks!
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "3020893": "In this competition, unlike previous ones, the time series API has been updated. I struggled to understand it, but I finally started to grasp how to use lag, so I’d like to share my insights.\n\nhttps://www.kaggle.com/code/chumajin/janestreet-easy-to-understand-new-time-series-api\n\n# Summary\n\nI think there are 3 steps.\n\nStep 1 The data is split by day using group_by.\n\nStep 2 The data made by Step1 is split by time_id using group_by, and the lag is generated (for time_id == 0).\n\nStep 3 After inputting the predicted values, the submission file is generated.\n\nThe following image is the comparing the new time series API and my understanding\n(Detail is in above my public notebook! )\n\n\n\nIt can be understood that the official API code omits the part where the date_id and time_id using a for loop are assigned. By understanding this overall structure, I believe you will be able to participate in the competition in easy.\n\nEnjoy !", "3021100": "You can also check my work related to the submission API here: https://www.kaggle.com/code/shiyili/js24-rmf-generate-synthetic-test-data", "3021129": "shiyili OK. I will check it. Thank you very much!", "3054531": "Hi,\nThanks for that.\nShall we understand the lag as a previous day close equivalent, or previous day change in the asset price (which ever responder_6 is meant to represent ?) \nTherefore, to get this info on the train set, it's possible to use the responders values at the last time_id of each date_id of the previous day to get those lags ? i.e for date_id 2, we use the last time_id data of each responders for each symbol present in date_id 1 ?\n\nThanks!", "3078527": "dude this is really helpful, insanely good saved my day big thanks" }, "source": "meta" }