1

I am pretty new to most algorithms and with there being so many out there, it is hard to choose as a beginner which one to use. Therefore, I am looking for some recommendations here. The problem is as follows: assume different series of treatments [T1, T2, T3] for a person. Now, every person can have a different set of treatments. My goal is to use the given data (so all the sets of completed treatment arrays) and predict the next treatment sequence for a new patient, which let's say is halfway through with its pathway.

I was thinking about using an LSTM model, but frankly am curious what models could be of use and especially why. If someone can explain this clearly, it would be much much appreciated.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
Jacky Chu
  • 11
  • 1
  • Is $T_i$ a sequence itself as well? – gunes Apr 19 '21 at 09:10
  • Yes, it is. The length of the sequence varies per patient and runs as far as 25. The idea is to use all the data available to make a prediction on what the 'next best treatment' is for the current patient (where for example the current sequence consists of [T1,T3,T4, x, y] and we would want to predict x and y. – Jacky Chu Apr 19 '21 at 10:43
  • So, if $T_i$ is a sequence, that makes $[T_1,T_2,T_3]$ a sequence of sequences, but I couldn't get the same feeling from your example in your comment. – gunes Apr 19 '21 at 10:45
  • I totally misread your comment. It is indeed not a sequence in itself. It can be seen as a string containing information about the treatment. Sorry for the confusion. – Jacky Chu Apr 19 '21 at 10:55
  • Please add new information as an edit to post, not only as comments. Ahd use the opportunity to tell us more specifics, as of now the Q is to abstract – kjetil b halvorsen Apr 19 '21 at 12:09

0 Answers0