2
$\begingroup$

I am planning to use neural networks for time series analysis and forecasting and i was looking for a reference or any textbook i can use for that purpose. It would be great if it has practical R or python code with it.

$\endgroup$
1
  • $\begingroup$ I've re-added the neural network tag because your question asks about neural networks. $\endgroup$
    – Sycorax
    Jun 16, 2018 at 22:06

1 Answer 1

4
$\begingroup$

Well you can use neural networks for time series forecasting. I am not sure about the analysis part. You can use recurrent neural network (RNN) , LSTM (long short term memory) a type of RNN, for your problem. Recurrent neural networks are Artificial neural networks with cycles. They are the natural architecture for learning over sequences.

Few points that might be helpful for you to start with:

1) Consult a book on RNN and LSTM. The book discusses in detail about the problems that RNN and LSTM are being used to solve.

2) ALso have a look at some great blogs on the web. This blog is one of the most popular for understanding the recurrent neural network and LSTM in most simplistic ways. This is highly recommended.

3) Go through research papers that compare traditional time series approaches like ARIMA, VAR, ARIMAX etc to LSTM, RNN, GRU etc.

Ex- 1) Comparison of multivariate forecasting technique VAR (Vectpor Autoregression) with RNN. This paper will give you a perspective of how traditional statistical model like VAR compare to neural network.

2) This paper is about forecasting economics and financial time series. It compares ARIMA to LSTM.

4) Last but not the least. This tutorial gives a fair amount of idea on how to prepare time series data for forecasting and also familiarizes with using LSTM using keras. It contains reproducible example with code. This will be very useful if you quickly want to get your hands dirty with implementing LSTM.

$\endgroup$
8
  • $\begingroup$ Welcome to CV! Please edit your question so that the academic articles have complete citation information AND summarize the key points. Links can go dead, but we want to have answers which are useful in the long-term. $\endgroup$
    – Sycorax
    Jun 15, 2018 at 18:55
  • $\begingroup$ @naive thank you so much for this answer but How could you search for papers that compare traditional methods with NN? Or you have already read those papers? $\endgroup$
    – floyd
    Jun 15, 2018 at 19:32
  • 1
    $\begingroup$ @floyd Welcome! Have read the first one that i mentioned and am planning to read the second. Just some creative googling will land you up at the right place! $\endgroup$
    – naive
    Jun 15, 2018 at 19:41
  • 1
    $\begingroup$ @floyd you can search for papers relating to neural networks at the arxiv(arxiv.org) library. Just type VAR RNN in the search id. You will land up with the first article that i mentioned. Hope this helps! $\endgroup$
    – naive
    Jun 15, 2018 at 20:17
  • 1
    $\begingroup$ @floyd if you found the answer satisfactory and if its meeting your requirements please accept it! Thanks! $\endgroup$
    – naive
    Jun 15, 2018 at 20:50

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