{ "id": 176094, "title": "I tried random forest, neural network and linear regression.", "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/176094", "author_name": "", "post_date": "2020-08-20T13:37:39.478249200Z", "votes": 3, "comment_count": 11, "views": 0, "content": "

Hi all,

\n

I tried some standard algorithms- linear regression, neural network and random forest.

\n

I found random forest to be the best of the above but there's a lot to improve as the score is only -9.10 . Let me know if you have any suggestions:
\nhttps://www.kaggle.com/srikanthpotukuchi/random-forest-submission

\n

Regards,
\nSrikanth

", "messages": [ { "id": "978902", "postDate": "08/20/2020 13:37:39", "content": "

Hi all,

\n

I tried some standard algorithms- linear regression, neural network and random forest.

\n

I found random forest to be the best of the above but there's a lot to improve as the score is only -9.10 . Let me know if you have any suggestions:
\nhttps://www.kaggle.com/srikanthpotukuchi/random-forest-submission

\n

Regards,
\nSrikanth

", "rawMarkdown": "Hi all,\n\nI tried some standard algorithms- linear regression, neural network and random forest. \n\nI found random forest to be the best of the above but there's a lot to improve as the score is only -9.10 . Let me know if you have any suggestions: \nhttps://www.kaggle.com/srikanthpotukuchi/random-forest-submission \n\nRegards,\nSrikanth", "votes": null }, { "id": "978959", "postDate": "08/20/2020 14:31:28", "content": "

Hi! @srikanthpotukuchi, based on my observations of the dataset (EDA and Visualizations), seems logical that any ML that separates the spaces in some planes will work better. I'm planning to use support vector machine (SVM), plus feature creation based on K-mean

", "rawMarkdown": "Hi! @srikanthpotukuchi, based on my observations of the dataset (EDA and Visualizations), seems logical that any ML that separates the spaces in some planes will work better. I'm planning to use support vector machine (SVM), plus feature creation based on K-mean", "votes": null }, { "id": "978965", "postDate": "08/20/2020 14:35:43", "content": "

Interesting, I'll try SVM. I didn't even think of feature creation based on k-means, will try it!

", "rawMarkdown": "Interesting, I'll try SVM. I didn't even think of feature creation based on k-means, will try it!", "votes": null }, { "id": "981220", "postDate": "08/22/2020 09:34:04", "content": "

Did you try neural networks so far?

", "rawMarkdown": "Did you try neural networks so far?", "votes": null }, { "id": "984161", "postDate": "08/24/2020 22:28:31", "content": "

I tried a NN and got CV 6.715 and PL of 6.89, and this was the first try so it's very doable.

", "rawMarkdown": "I tried a NN and got CV 6.715 and PL of 6.89, and this was the first try so it's very doable.", "votes": null }, { "id": "984261", "postDate": "08/25/2020 02:07:13", "content": "

@srikanthpotukuchi here is more information, this is quite advanced techniques but provide the difference in a competition.
\nhttps://towardsdatascience.com/how-to-create-new-features-using-clustering-4ae772387290

", "rawMarkdown": "srikanthpotukuchi here is more information, this is quite advanced techniques but provide the difference in a competition.\n[https://towardsdatascience.com/how-to-create-new-features-using-clustering-4ae772387290](url)", "votes": null }, { "id": "984265", "postDate": "08/25/2020 02:11:39", "content": "

I did! https://www.kaggle.com/srikanthpotukuchi/neural-network-model-submission-score

", "rawMarkdown": "I did! https://www.kaggle.com/srikanthpotukuchi/neural-network-model-submission-score", "votes": null }, { "id": "984268", "postDate": "08/25/2020 02:14:39", "content": "

Interesting, will try it. I need to do some feature engineering.

", "rawMarkdown": "Interesting, will try it. I need to do some feature engineering.", "votes": null }, { "id": "986099", "postDate": "08/26/2020 07:50:38", "content": "

I'm a newbie in this competition, could you guy tell me what CV and PL is. Thanks

", "rawMarkdown": "I'm a newbie in this competition, could you guy tell me what CV and PL is. Thanks", "votes": null }, { "id": "986221", "postDate": "08/26/2020 09:49:59", "content": "

PL - Public score
\nCV - score on the validation (local)

", "rawMarkdown": "PL - Public score\nCV - score on the validation (local)", "votes": null }, { "id": "986243", "postDate": "08/26/2020 10:06:27", "content": "

What's your recommendation on how to use the IMG -Data? I am still reading and trying to figure things out. I am starting with EDA again: How many images per record, how many records per patient ..

", "rawMarkdown": "What's your recommendation on how to use the IMG -Data? I am still reading and trying to figure things out. I am starting with EDA again: How many images per record, how many records per patient ..", "votes": null }, { "id": "986354", "postDate": "08/26/2020 12:10:43", "content": "

Nice work. Keep it up

", "rawMarkdown": "Nice work. Keep it up", "votes": null } ], "comments": [ { "id": 978959, "author_name": "cv13j0", "author_url": "", "post_date": "08/20/2020 14:31:28", "content": "

Hi! @srikanthpotukuchi, based on my observations of the dataset (EDA and Visualizations), seems logical that any ML that separates the spaces in some planes will work better. I'm planning to use support vector machine (SVM), plus feature creation based on K-mean

", "votes": null, "replies": [ { "id": 978965, "author_name": "srikanthpotukuchi", "author_url": "", "post_date": "08/20/2020 14:35:43", "content": "

Interesting, I'll try SVM. I didn't even think of feature creation based on k-means, will try it!

", "votes": null, "replies": [] }, { "id": 984261, "author_name": "cv13j0", "author_url": "", "post_date": "08/25/2020 02:07:13", "content": "

@srikanthpotukuchi here is more information, this is quite advanced techniques but provide the difference in a competition.
\nhttps://towardsdatascience.com/how-to-create-new-features-using-clustering-4ae772387290

", "votes": null, "replies": [] } ] }, { "id": 981220, "author_name": "fireheart7", "author_url": "", "post_date": "08/22/2020 09:34:04", "content": "

Did you try neural networks so far?

", "votes": null, "replies": [ { "id": 984265, "author_name": "srikanthpotukuchi", "author_url": "", "post_date": "08/25/2020 02:11:39", "content": "

I did! https://www.kaggle.com/srikanthpotukuchi/neural-network-model-submission-score

", "votes": null, "replies": [] } ] }, { "id": 984161, "author_name": "lukereijnen", "author_url": "", "post_date": "08/24/2020 22:28:31", "content": "

I tried a NN and got CV 6.715 and PL of 6.89, and this was the first try so it's very doable.

", "votes": null, "replies": [ { "id": 984268, "author_name": "srikanthpotukuchi", "author_url": "", "post_date": "08/25/2020 02:14:39", "content": "

Interesting, will try it. I need to do some feature engineering.

", "votes": null, "replies": [] }, { "id": 986099, "author_name": "hajau1", "author_url": "", "post_date": "08/26/2020 07:50:38", "content": "

I'm a newbie in this competition, could you guy tell me what CV and PL is. Thanks

", "votes": null, "replies": [] }, { "id": 986221, "author_name": "lukereijnen", "author_url": "", "post_date": "08/26/2020 09:49:59", "content": "

PL - Public score
\nCV - score on the validation (local)

", "votes": null, "replies": [] } ] }, { "id": 986243, "author_name": "ChristianDenich", "author_url": "", "post_date": "08/26/2020 10:06:27", "content": "

What's your recommendation on how to use the IMG -Data? I am still reading and trying to figure things out. I am starting with EDA again: How many images per record, how many records per patient ..

", "votes": null, "replies": [] }, { "id": 986354, "author_name": "sukanthen", "author_url": "", "post_date": "08/26/2020 12:10:43", "content": "

Nice work. Keep it up

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "978902": "Hi all,\n\nI tried some standard algorithms- linear regression, neural network and random forest. \n\nI found random forest to be the best of the above but there's a lot to improve as the score is only -9.10 . Let me know if you have any suggestions: \nhttps://www.kaggle.com/srikanthpotukuchi/random-forest-submission \n\nRegards,\nSrikanth", "978959": "Hi! @srikanthpotukuchi, based on my observations of the dataset (EDA and Visualizations), seems logical that any ML that separates the spaces in some planes will work better. I'm planning to use support vector machine (SVM), plus feature creation based on K-mean", "978965": "Interesting, I'll try SVM. I didn't even think of feature creation based on k-means, will try it!", "981220": "Did you try neural networks so far?", "984161": "I tried a NN and got CV 6.715 and PL of 6.89, and this was the first try so it's very doable.", "984261": "srikanthpotukuchi here is more information, this is quite advanced techniques but provide the difference in a competition.\n[https://towardsdatascience.com/how-to-create-new-features-using-clustering-4ae772387290](url)", "984265": "I did! https://www.kaggle.com/srikanthpotukuchi/neural-network-model-submission-score", "984268": "Interesting, will try it. I need to do some feature engineering.", "986099": "I'm a newbie in this competition, could you guy tell me what CV and PL is. Thanks", "986221": "PL - Public score\nCV - score on the validation (local)", "986243": "What's your recommendation on how to use the IMG -Data? I am still reading and trying to figure things out. I am starting with EDA again: How many images per record, how many records per patient ..", "986354": "Nice work. Keep it up" }, "source": "meta" }