# Time series prediction using ARIMA vs LSTM

The problem that I am dealing with is predicting time series values. I am looking at one time series at a time and based on for example 15% of the input data, I would like to predict its future values. So far I have come across two models:

• LSTM (long short term memory; a class of recurrent neural networks)
• ARIMA

I have tried both and read some articles on them. Now I am trying to get a better sense on how to compare the two. What I have found so far:

1. LSTM works better if we are dealing with huge amount of data and enough training data is available, while ARIMA is better for smaller datasets (is this correct?)
2. ARIMA requires a series of parameters (p,q,d) which must be calculated based on data, while LSTM does not require setting such parameters. However, there are some hyperparameters we need to tune for LSTM.

Other than the above-mentioned properties, I could not find any other points or facts which could help me toward selecting the best model. I would be really grateful if someone could help me finding articles, papers or other stuff (had no luck so far, only some general opinions here and there and nothing based on experiments.)

I have to mention that originally I am dealing with streaming data, however for now I am using NAB datasets which includes 50 datasets with the maximum size of 20k data points.

• Why don't you just try the two models on part of your data, see which one is better at forecasting, and choose it. Or use both models and combine their forecasts. Forecast combinations often outperform the individual forecasts. – Richard Hardy Jul 11 '16 at 17:58
• @RichardHardy I already have done that and am aware of their performance on my datasets. I am trying to get better understanding of both , specially their drawbacks to see which one might be the best candidate to handle the upcoming data samples. – ahajib Jul 11 '16 at 18:04
• – Franck Dernoncourt Aug 3 '17 at 23:13
• Please read the help center -- in particular the third-last paragraph which says "Please note, however, that cross-posting is not encouraged on SE sites. Choose one best location to post your question. Later, if it proves better suited on another site, it can be migrated." – Glen_b Aug 4 '17 at 2:41

## 1 Answer

A comparison of artificial neural network and time series models for forecasting commodity prices compares the performance of ANN and ARIMA in predicting financial time series. I think it is a good starting point for your literature review.

In many cases, neural networks tend to outperform AR-based models. However, I think that one major drawback (which is not discussed that much in the academic litterature) with more advanced machine learning methods is that they use black boxes. This is a big problem if you had to explain how the model works to someone who doesn't know that much of these models (for example in a corporation). But if you are doing this analysis just as a school work, I don't think this is going to be an issue.

But like the previous commentator said usually the best way is to form an ensemble estimator in which you combine two or more models.

• The ref you cited deals with simple feedforward neural nets and it's too old to be useful (1990s is a century ago). The OP question asks about recurrent neural net with the LSTM architecture and this paper doesn't cover that. – horaceT Jul 11 '16 at 18:58
• As @horaceT mentioned this paper is a bit outdated and if you could suggest a more recent paper which includes info on LSTMs would be awesome. Thanks – ahajib Jul 11 '16 at 20:24