Timeline for Regression model is doing exceptionally very well on time series
Current License: CC BY-SA 4.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Nov 28 at 3:04 | vote | accept | S. M. | ||
| Nov 23 at 8:59 | answer | added | cinch | timeline score: 1 | |
| Nov 23 at 8:41 | comment | added | cinch | Sure. Depending on the sense of 'bad', it may be statistical flux without significance, generally with 1800 features you need some regularization. | |
| Nov 23 at 8:36 | comment | added | S. M. | @cinch thank you for your valuable comments. I have one little confusion, without shuffle ridge regression is working bad than linear but after shuffle I get the reverse, what would you comments on this? | |
| Nov 23 at 7:40 | comment | added | cinch | This question is about lr model not lstm. lstm starts working better than lr for all 3-day window test samples doesn't necessary mean lstm is not working as expected since it may still learn local patterns that persist even in the absence of strict time-based structure, though lstm generally needs much more data and certainly is not recommended for your 3-day prediction problem. The behavior you're observing after shuffling, if the new MSE is much larger than the original MSE, suggests that shuffling exposes the intrinsic low noise temporally-dependent regular structure of your data. | |
| Nov 23 at 6:30 | comment | added | S. M. | @cinch would I share with you complete code with lstm? | |
| Nov 23 at 4:12 | comment | added | S. M. | @cinch but after shuffling lr model is working worst as expected, but lstm model is started working well than lr model after shuffling. what you comment on this scenario? since we know that in low amount of data( in my case 3-day window), the sophisticated model like lstm is often work worst, but after your suggesting shuffle, lr model is working as expected but lstm model not working as expected. what would you comment on this? | |
| Nov 23 at 1:43 | comment | added | S. M. | @cinch for any day we have only 15 mins data available, this data is 24×25 csv filé. On which day the data is available, it's 15 mins data. | |
| Nov 23 at 1:23 | history | edited | S. M. | CC BY-SA 4.0 |
added 18 characters in body
|
| Nov 23 at 0:53 | comment | added | S. M. | @cinch it's tec data, each data points are pixels. So, would you write the code review? | |
| Nov 22 at 19:24 | history | edited | S. M. |
edited tags
|
|
| S Nov 22 at 19:15 | review | First questions | |||
| Nov 28 at 3:10 | |||||
| S Nov 22 at 19:15 | history | asked | S. M. | CC BY-SA 4.0 |