I am new with anything related to Machine Learning and would like to know how, and if there is a way to predict simple outcomes from patterns, for example, I have a dataset of [1, 2, 3, 1, 3, 2, 1]. The user does action 3, and the dataset turns to [1, 2, 3, 1, 3, 2, 1, 3]. How can I predict what the user would do next?
I am planning to use this with CoreML. I found this on Apple's developer website. However, I believe this is predicting based on data provided on the model. (I'm assuming it would be with CreateML).
What would be the recommended approach to my problem? (Sorry for my English, I will try to clarify more if needed). Thank you :)
This looks like a classifier that work on a sequence of input data. A typical model used for this sort of thing is an LSTM. Create ML currently does not let you train sequence models such as LSTMs, so you'd need to use a training package like Keras (or Turi Create for specific applications).
Here's a blog post I wrote a while back that explains how LSTMs work and how to implement them on iOS: http://machinethink.net/blog/recurrent-neural-networks-with-swift/ Since then, Core ML has appeared, which also supports LSTM models and is a little easier to use.