Idea for colors prediction model

Hi everyone,

Could someone help me about an idea for a color prediction system using a Bayesian approch ?
I have a vehicles maintenance system where the people decide some actions to execute based on 2 flag colors: blue and red.
I´m thinking about to use a bayesian inference to tell the probabilities for the next flag be blue or red based on some simple past data.
The data I have are: A sequencial index (a integer) just to sorting records, a begin and end timestamp for the last N flags and the flag color. For example:

Idx, Begin, End, Color
1,07.01.2019 09:30,07.01.2019 09:37;red
2,07.01.2019 09:40,07.01.2019 09:50;red
3,07.01.2019 09:53,07.01.2019 10:20;blue

The objective is to return the probability for the next color flag.

Does anyone have an idea for a good model for this ?

Thank you all.

This sounds to me like a CTHMM. @RJTK has done some work in this area (Multivariate time series inference). An alternative that PyMC3 handles well is a lagged regression (an example of which is also in that link)

1 Like

Thanks chartl. I´ll check it !