It seems that feature engineering is the key for making model more robust for the task, however , I have a question with respect to feature engineering , how does one go on to create do the same . I have seen that since it is time-series data , a lot of features might be related to that . Are there any resources for learning how to deal with time-series data as well ? Any advices would be appreciated .
", "messages": [ { "id": "1102889", "postDate": "12/05/2020 12:55:09", "content": "It seems that feature engineering is the key for making model more robust for the task, however , I have a question with respect to feature engineering , how does one go on to create do the same . I have seen that since it is time-series data , a lot of features might be related to that . Are there any resources for learning how to deal with time-series data as well ? Any advices would be appreciated .
", "rawMarkdown": "It seems that feature engineering is the key for making model more robust for the task, however , I have a question with respect to feature engineering , how does one go on to create do the same . I have seen that since it is time-series data , a lot of features might be related to that . Are there any resources for learning how to deal with time-series data as well ? Any advices would be appreciated .", "votes": null }, { "id": "1102894", "postDate": "12/05/2020 13:03:22", "content": "Solutions of past competitions (e.g. 2019 Data Science Bowl) might be valuable.
", "rawMarkdown": "Solutions of past competitions (e.g. 2019 Data Science Bowl) might be valuable.", "votes": null }, { "id": "1102938", "postDate": "12/05/2020 13:50:14", "content": "Hi , I would try to check the same . But I also had a question with regards to code , as to how we create the code for the same .
", "rawMarkdown": "Hi , I would try to check the same . But I also had a question with regards to code , as to how we create the code for the same .", "votes": null }, { "id": "1102967", "postDate": "12/05/2020 14:21:31", "content": "I think the documentation of pandas, Q & A websites (e.g. stackoverflow) and kaggle notebook will help you.
", "rawMarkdown": "I think the documentation of pandas, Q & A websites (e.g. stackoverflow) and kaggle notebook will help you.", "votes": null }, { "id": "1103263", "postDate": "12/05/2020 19:12:23", "content": "Has anybody come up with a feature that has gain close to that of overall question difficulty?
", "rawMarkdown": "Has anybody come up with a feature that has gain close to that of overall question difficulty?", "votes": null }, { "id": "1103626", "postDate": "12/06/2020 04:47:20", "content": "Hi caleb , I have question/suggestion can we use the amount of time taken to answer , to relate with overall question difficulty? Though very different concepts ,we can a guess that harder question would require more time to solve.
", "rawMarkdown": "Hi caleb , I have question/suggestion can we use the amount of time taken to answer , to relate with overall question difficulty? Though very different concepts ,we can a guess that harder question would require more time to solve.", "votes": null }, { "id": "1103630", "postDate": "12/06/2020 04:51:33", "content": "Do you use nn or lgb? Your score is awosome!!!!
", "rawMarkdown": "Do you use nn or lgb? Your score is awosome!!!!", "votes": null }, { "id": "1109302", "postDate": "12/11/2020 14:17:23", "content": "@sahilmaheshwari For Time Series Data Related to Feature Engineering , I use following resources !
\nMachine Learning Mastery blog for feature generation here , however it is very basic in nature ! .
There was competition called M5 Forecasting Accuracy , in that Konstantin Yakolvev Published a very great kernel on Time Series competition ! It can be found here , What he has done is he has demonstrated how to make rolling and lags features based on this !
You might find this helpful !
", "rawMarkdown": "sahilmaheshwari For Time Series Data Related to Feature Engineering , I use following resources !\n\n1. Machine Learning Mastery blog for feature generation [here ](https://machinelearningmastery.com/basic-feature-engineering-time-series-data-python/) , however it is very basic in nature ! .\n\n2. There was competition called **M5 Forecasting Accuracy** , in that Konstantin Yakolvev Published a very great kernel on Time Series competition ! It can be found [here ](https://www.kaggle.com/kyakovlev/m5-lags-features) , What he has done is he has demonstrated how to make rolling and lags features based on this ! \n\nYou might find this helpful !", "votes": null }, { "id": "1111536", "postDate": "12/13/2020 19:25:02", "content": "That seems like an interesting idea. There is another discussion going on with ideas on how to calculate here. I was thinking that duration could go both ways - yes longer for more difficult questions - one point here is that there are inherent differences depending on part, i.e, it takes longer to read a passage or watch a video and answer questions than it does to fill in a missing word - but also could indicate lack of effort and hence be correlated with lower accuracy.
", "rawMarkdown": "That seems like an interesting idea. There is another discussion going on with ideas on how to calculate [here](https://www.kaggle.com/yanamal/explanation-and-elapsed-time-for-current-question/comments). I was thinking that duration could go both ways - yes longer for more difficult questions - one point here is that there are inherent differences depending on part, i.e, it takes longer to read a passage or watch a video and answer questions than it does to fill in a missing word - but also could indicate lack of effort and hence be correlated with lower accuracy.", "votes": null }, { "id": "1111958", "postDate": "12/14/2020 06:53:59", "content": "Thanks Caleb, I will check that out !
", "rawMarkdown": "Thanks Caleb, I will check that out !", "votes": null } ], "comments": [ { "id": 1102894, "author_name": "mamasinkgs", "author_url": "", "post_date": "12/05/2020 13:03:22", "content": "Solutions of past competitions (e.g. 2019 Data Science Bowl) might be valuable.
", "votes": null, "replies": [ { "id": 1102938, "author_name": "sahilmaheshwari", "author_url": "", "post_date": "12/05/2020 13:50:14", "content": "Hi , I would try to check the same . But I also had a question with regards to code , as to how we create the code for the same .
", "votes": null, "replies": [] }, { "id": 1102967, "author_name": "mamasinkgs", "author_url": "", "post_date": "12/05/2020 14:21:31", "content": "I think the documentation of pandas, Q & A websites (e.g. stackoverflow) and kaggle notebook will help you.
", "votes": null, "replies": [] }, { "id": 1103630, "author_name": "shichenyang", "author_url": "", "post_date": "12/06/2020 04:51:33", "content": "Do you use nn or lgb? Your score is awosome!!!!
", "votes": null, "replies": [] } ] }, { "id": 1103263, "author_name": "calebeverett", "author_url": "", "post_date": "12/05/2020 19:12:23", "content": "Has anybody come up with a feature that has gain close to that of overall question difficulty?
", "votes": null, "replies": [ { "id": 1103626, "author_name": "sahilmaheshwari", "author_url": "", "post_date": "12/06/2020 04:47:20", "content": "Hi caleb , I have question/suggestion can we use the amount of time taken to answer , to relate with overall question difficulty? Though very different concepts ,we can a guess that harder question would require more time to solve.
", "votes": null, "replies": [] }, { "id": 1111536, "author_name": "calebeverett", "author_url": "", "post_date": "12/13/2020 19:25:02", "content": "That seems like an interesting idea. There is another discussion going on with ideas on how to calculate here. I was thinking that duration could go both ways - yes longer for more difficult questions - one point here is that there are inherent differences depending on part, i.e, it takes longer to read a passage or watch a video and answer questions than it does to fill in a missing word - but also could indicate lack of effort and hence be correlated with lower accuracy.
", "votes": null, "replies": [] }, { "id": 1111958, "author_name": "sahilmaheshwari", "author_url": "", "post_date": "12/14/2020 06:53:59", "content": "Thanks Caleb, I will check that out !
", "votes": null, "replies": [] } ] }, { "id": 1109302, "author_name": "sayedathar11", "author_url": "", "post_date": "12/11/2020 14:17:23", "content": "@sahilmaheshwari For Time Series Data Related to Feature Engineering , I use following resources !
\nMachine Learning Mastery blog for feature generation here , however it is very basic in nature ! .
There was competition called M5 Forecasting Accuracy , in that Konstantin Yakolvev Published a very great kernel on Time Series competition ! It can be found here , What he has done is he has demonstrated how to make rolling and lags features based on this !
You might find this helpful !
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1102889": "It seems that feature engineering is the key for making model more robust for the task, however , I have a question with respect to feature engineering , how does one go on to create do the same . I have seen that since it is time-series data , a lot of features might be related to that . Are there any resources for learning how to deal with time-series data as well ? Any advices would be appreciated .", "1102894": "Solutions of past competitions (e.g. 2019 Data Science Bowl) might be valuable.", "1102938": "Hi , I would try to check the same . But I also had a question with regards to code , as to how we create the code for the same .", "1102967": "I think the documentation of pandas, Q & A websites (e.g. stackoverflow) and kaggle notebook will help you.", "1103263": "Has anybody come up with a feature that has gain close to that of overall question difficulty?", "1103626": "Hi caleb , I have question/suggestion can we use the amount of time taken to answer , to relate with overall question difficulty? Though very different concepts ,we can a guess that harder question would require more time to solve.", "1103630": "Do you use nn or lgb? Your score is awosome!!!!", "1109302": "sahilmaheshwari For Time Series Data Related to Feature Engineering , I use following resources !\n\n1. Machine Learning Mastery blog for feature generation [here ](https://machinelearningmastery.com/basic-feature-engineering-time-series-data-python/) , however it is very basic in nature ! .\n\n2. There was competition called **M5 Forecasting Accuracy** , in that Konstantin Yakolvev Published a very great kernel on Time Series competition ! It can be found [here ](https://www.kaggle.com/kyakovlev/m5-lags-features) , What he has done is he has demonstrated how to make rolling and lags features based on this ! \n\nYou might find this helpful !", "1111536": "That seems like an interesting idea. There is another discussion going on with ideas on how to calculate [here](https://www.kaggle.com/yanamal/explanation-and-elapsed-time-for-current-question/comments). I was thinking that duration could go both ways - yes longer for more difficult questions - one point here is that there are inherent differences depending on part, i.e, it takes longer to read a passage or watch a video and answer questions than it does to fill in a missing word - but also could indicate lack of effort and hence be correlated with lower accuracy.", "1111958": "Thanks Caleb, I will check that out !" }, "source": "meta" }