How to predict an event (or action) based on a window of time-series measurements?

I have an input vector $$X$$, which contains a series of measurements within a period, e.g. 100 measurements in 1 sec. The goal is to predict an event, let's say, moving forward, backward or static.

I don't want to predict the output just by looking at one series of measurements, but by looking at a window of $$n$$ vectors $$X$$ of measurements, making it dependant on the previous measurements, because of the noise in the measurements.

Is there a way RNN can help me with this? Many to one architecture? LSTM? CNN of 1D + LSTM + dense?

• Hi and welcome to AI SE. So, to clarify, you want to feed to the model at once $n$ vectors $X_1, X_2, \dots, X_n$, each of which contains e.g. $M=100$ measurements, i.e. $|X_i| = 100$, for all $i$? – nbro Mar 20 '20 at 2:13
• Hi, basically I want to predict the class based on these X vector with 100 measuments, but also depend on the previous X inputs I fed. Is there a model which takes into account previous states? Or do I need to feed 'n' number of X in a certain window of states? I heard about RNN and Lstm. – Erick Medina Mar 20 '20 at 2:46
• Yes, RNN may be helpful for your problem. Please, edit your post to answer my questions above. Also, don't just ask "Is there a way RNN can help me with this?", because people could say either "yes" or "no". You could ask "How specifically could I design an LSTM to solve this problem?" What should the input of this LSTM be? What should be the output? And so on. These are more specific questions. – nbro Mar 20 '20 at 22:38