A company manages GPS data for a fleet of buses, and has the following table.
Table:
['fecha', 'latitud', 'longitud', 'altitud', 'velocidad', 'cog', 'nsat',
'tipo', 'utc_hora', 'fix_fecha', 'imei', 'registro', 'input1', 'input2',
'input3', 'input4', 'hdop', 'adc', 'ignicion', 'adc2', 'power',
'driverid', 'ibutton2', 'ibutton3', 'ibutton4', 'trailerid', 'adc3',
'adc4', 'horometro', 'odometro', 'panico', 'bateria', 'bateriaint']
The variable ignicion is binary, it means if the engine is on (1) or off (0). The idea is to find the devices that are failing, because if ignicion is on it sends information every 40 seconds and if it is off every 10 min. What happens is that sometimes the GPS behaves abnormally and the goal is to identify those GPS .
I think it's not a problem of machine learning because there is no variable to predict, so I think it's a problem of descriptive analysis. It could also be a problem of time series.