I have a small dataset from 2006 to 2023, I would like to predict monthly sales for the next year. This is my data:

enter image description here

I already tried Prophet and NeuralProphet, but unfortunately they don't work well after many trials with different parameters, I am doing my internship, so I don't have so much experience in this field, it would be great if experts give me some recommendation which algorithms can give me better results ? I was thinking about LSTM, but I think it needs many sample data for training. I am using python for programming.

  • $\begingroup$ You can start even more basic with an autocorrelation and then autoregression models, especially if you have a smaller dataset before jumping to LSTMs. $\endgroup$
    – m13op22
    Feb 8 at 20:05
  • $\begingroup$ Just found an even better comparison between ARIMA, Prophet, and LSTM $\endgroup$
    – m13op22
    Feb 8 at 20:09

1 Answer 1


A few years ago a blog post criticizing Prophet became very popular: https://www.microprediction.com/blog/prophet. One of the Prophet authors commented on the issue, basically agreeing with the post in this Twitter thread

The author of the post talks a lot about time series forecasting on his blog (https://www.microprediction.com/blog), and maintains a ranking with comparisons between different approaches (https://www.microprediction.com/ blog/fast). I think you can get ideas from there.

In general, Python people seem to tend to use pmdarima or other packages that have an auto-ARIMA implementation, while R people tend to use auto.arima from the forecast package.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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