0
$\begingroup$

I have traffic volume data (Surrey City, CA) like this

the data are arterial traffic volume count from February 2013 - 2014 in Surrey

I wish to use Artificial neural network (Deep Learning) or ARIMA to predict traffic flow/volume of the urban area with the use of previous traffic count data. I want to know whether it is a good technique to predict traffic condition. I want to make the data 2013 as training set and the data 2014 as a testing. Is it possible to do traffic flow/volume prediction in next time period if only use this dataset or we need other data to support it?

Did I choose the right method? Or which method is suitable for this case? I really appreciate if somebody could help me with how to do that... I need guidance, I am new in neural networks.

Improve this question
$\endgroup$
1
$\begingroup$

You wont benefit from deep learning in this problem, this isnt a classification task, and deep learning is much better suited for those. I would apply an kalman filter or give seasonal arima (sarima) a try, you want to stick with a simple model for this one.

Improve this answer
$\endgroup$
2
  • $\begingroup$ Thank you very much.. so Deep Learning is suitable for the big data only, isn't it? $\endgroup$
    – Hendrik Tampubolon
    May 21 '15 at 21:39
  • $\begingroup$ What do you mean by big data? No. Its suited for problems where complex, non linear representations are seeked, for example images or speech. In summary in problems complex enough for the model easily ovrerfit. Your data arent even i.d.d, there is some kind of short term memory. (both arima and kalman exploit it). Also, deep learning is a broad term, here i am assuming you mean multi layered networks trained with dropout, but the same applies to cnn's, tough rnn's could work. (But they are too much trouble to train) $\endgroup$
    – Ramalho
    May 21 '15 at 21:58
0
$\begingroup$

Papers on traffic volume prediction have applied Kalman filter. A very cited one is the following by Okutani and Stephanedes.

Improve this answer
$\endgroup$
1
  • $\begingroup$ thank you so much for your suggestion.. anyway why i couldn't found the pdf file of that paper? $\endgroup$
    – Hendrik Tampubolon
    May 21 '15 at 21:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.