Proper algorithm for short term predictions
up vote
0
down vote
favorite
I am currently working on my final project in university and I have to do some Machine Learning. I have to say I am not experienced with ML. I have data with a timestamp, zone number (6 zones) and number of calls. I need to predict the number of calls and initially i decided to use Multilinear regression. However, while researching i found about time series analysis and I am wondering now, which one would be better for making predicitons in my case.
From what I understood time series analysis is good for forecast, but is it good for short term predictions. Like predicting number of calls tomorrow or next week? I want to make short term predictions. Maximum in the next month.
I just have read so much that I got confused!
I would very much appreciate, if you could advice me, what is better.
machine-learning time-series linear-regression
asked Nov 22 at 0:28
Kosi
41
add a comment |
up vote
0
down vote
favorite
I am currently working on my final project in university and I have to do some Machine Learning. I have to say I am not experienced with ML. I have data with a timestamp, zone number (6 zones) and number of calls. I need to predict the number of calls and initially i decided to use Multilinear regression. However, while researching i found about time series analysis and I am wondering now, which one would be better for making predicitons in my case.
From what I understood time series analysis is good for forecast, but is it good for short term predictions. Like predicting number of calls tomorrow or next week? I want to make short term predictions. Maximum in the next month.
I just have read so much that I got confused!
I would very much appreciate, if you could advice me, what is better.
machine-learning time-series linear-regression
asked Nov 22 at 0:28
Kosi
41
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39
add a comment |
up vote
0
down vote
favorite
up vote
0
down vote
favorite
I am currently working on my final project in university and I have to do some Machine Learning. I have to say I am not experienced with ML. I have data with a timestamp, zone number (6 zones) and number of calls. I need to predict the number of calls and initially i decided to use Multilinear regression. However, while researching i found about time series analysis and I am wondering now, which one would be better for making predicitons in my case.
From what I understood time series analysis is good for forecast, but is it good for short term predictions. Like predicting number of calls tomorrow or next week? I want to make short term predictions. Maximum in the next month.
I just have read so much that I got confused!
I would very much appreciate, if you could advice me, what is better.
machine-learning time-series linear-regression
asked Nov 22 at 0:28
Kosi
41
I am currently working on my final project in university and I have to do some Machine Learning. I have to say I am not experienced with ML. I have data with a timestamp, zone number (6 zones) and number of calls. I need to predict the number of calls and initially i decided to use Multilinear regression. However, while researching i found about time series analysis and I am wondering now, which one would be better for making predicitons in my case.
From what I understood time series analysis is good for forecast, but is it good for short term predictions. Like predicting number of calls tomorrow or next week? I want to make short term predictions. Maximum in the next month.
I just have read so much that I got confused!
I would very much appreciate, if you could advice me, what is better.
machine-learning time-series linear-regression
machine-learning time-series linear-regression
asked Nov 22 at 0:28
Kosi
41
asked Nov 22 at 0:28
Kosi
41
asked Nov 22 at 0:28
Kosi
41
asked Nov 22 at 0:28
Kosi
41
asked Nov 22 at 0:28
Kosi
41
41
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39
add a comment |
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39
add a comment |
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
active
oldest
votes
Thanks for contributing an answer to Stack Overflow!
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
Some of your past answers have not been well-received, and you're in danger of being blocked from answering.
Please pay close attention to the following guidance:
- Please be sure to answer the question. Provide details and share your research!
But avoid …
- Asking for help, clarification, or responding to other answers.
- Making statements based on opinion; back them up with references or personal experience.
To learn more, see our tips on writing great answers.
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
StackExchange.ready(
function () {
StackExchange.openid.initPostLogin('.new-post-login', 'https%3a%2f%2fstackoverflow.com%2fquestions%2f53422304%2fproper-algorithm-for-short-term-predictions%23new-answer', 'question_page');
}
);
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Sign up or log in
StackExchange.ready(function () {
StackExchange.helpers.onClickDraftSave('#login-link');
});
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Sign up using Google
Sign up using Facebook
Sign up using Email and Password
Post as a guest
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
Required, but never shown
If you have 24 months of data, you can model the first 21 months, and use the last 3 months to test short-term predictions. For example, model 21 months and then test on month 22. Then model 22 months and test on month 23. Finally, model 23 months and test using month 24. Another possible technique is to convert timestamps to days such as "day 100 at 8:00 AM" converts to 100.25 to make the date information numeric. You may want to code "normal business day" as a 1 and other days as 0 as a rough way to account for weekends and holidays, that might be useful.
– James Phillips
Nov 22 at 1:41
@JamesPhillips For the linear regression that I wanted to use at first i converted my time stamp can I use Multilinear regression or i should necessary use Time series since i have a timestamp. That is the confusing part, because i am not sure my data has some seasonality a.k.a i am not sure that my dependant variable depends on the time of the year. Or is it that if you have time in our data there for sure must be some pattern?
– Kosi
Nov 22 at 14:26
If you inspect scatterplots of month versus density for a the different species, you should be able to see whether or not seasonal variation exists - it might be obvious from the plots, or it might not seem large enough to be a concern.
– James Phillips
Nov 22 at 15:07
@JamesPhillips Thanks! So if there is no obvious seasonality or trend in the data, and it is just random peaks and drops can i use Multiple linear regression or the fact that we have peaks and drops just means that the data is non stationary over time and regressions it not the best fir model? Thank you in advance and sorry for the probably not so smarty questions :))
– Kosi
Nov 22 at 15:36
If there is a large and obvious annual cyclical variation, accounting for this would improve the model, true - and if it is small, then accounting for this small variation may not yield significant improvement. Random noise is expected and cannot be directly modeled.
– James Phillips
Nov 22 at 17:39