I'm hesitating between these two techniques for business data (activity logs, purchases) for classification:

  • I take all the data and consider it as a multidimensional time serie and use a deep learning model for time series

  • I calculate mean, sum, number of records, variance for each feature instead of using the whole time serie and then I use xgboost

Anyone knows the pros and cons ?

They are about 70 different features and a million of lines in the dataset


1 Answer 1


In my personal opinion, I wouldn´t use DL models for time series regression no matter how much data is available. They are hard to train in this type of task and you´ll be requiring a huge amount of data and time to get great results. Consider a DL model would require more features than just the plain time serie.

XGBoost is more friendly and requires less time and data for training. A disavantage could be the amount of features, as you give more to the model, it takes longer to train, I think 70 features would not be a big deal for it.

Finally, I suggest doing a feature selection just to ensure data quality and speed up the training of whatever model you´ll be working on.

Hope it helps!

  • $\begingroup$ It helps thanks. I've already used xgboost but I was curious about the DL method. I keep in mind it would be usefull with other data $\endgroup$
    – Jiayme C
    Aug 13, 2020 at 10:26

Your Answer

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

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