Feature Extraction for telematics time series data problem

I have telematics data basically containing dateTime, latitude, longitude, speed, ignition status, main battery on off status, reason for datageneration,GPS direction, GPS fix etc.

Using this i need to build model to perform the following

1. Identify loose connections that is if i get data with ignition status value 1 and 0 alternating that means there is loose connection.
2. identify repeated same reason
3. Identify GPS faults--> That is lat and lon remains unchanged but speed is >0 which means vehicle is moving or GPS fix keeps toggling between 0-1.

This is some of the examples of what i need to achieve. So here i am under the impression that one model wont be enough to solve all the above 3 stated problem and also which model do i need to consider. Since its time series data and need to remember old state, i thought of considering rnn models. Now my problem is for supervised learning i need to label them and also create features out of them. Can someone guide me with above feature extraction and labelling?