Classify driver based on time-series sensor data



I want to build a model that can detect which driver is driving now the car based on a dataset that contains 20 driver records for 3600s each driver ( the dataset contains all the car sensors values every second ) .

So , now i have that dataset that contain the drivers records . How can i train a model that can identify the driver based on 60 seconds ( or more for example).

Means , i want to make predictions with a dataframe of rows and not a single row . because we can't identify a driver with a single row .


John Karimov

Posted 2018-04-11T11:43:24.513

Reputation: 41

1This is a very interesting problem. First of all I am looking forward to hearing what others have to say. It is a multivariate time series. If we look at it as a supervised learning, should not LSTM-RNN be able to help for multiclass-classification? My question is, how each driver is labeled? An exact driven name/id or a it is rather behavioral labeling? This is very much people at automotive industry are trying to look at. – TwinPenguins – 2018-04-11T12:34:18.737

Yes , my first idea was to use LSTM-RNN and i'm looking for how to make data reliable for it ( convert dataframe to time sequences ) . Every driver has it's own ID ( or letter ) as label . – John Karimov – 2018-04-11T12:42:48.513


Have you seen the answer and suggested links in this post? I think I pretty much have the very same dataset like yours. I think I wrote a function to sample efficiently (because the sensory data was collected every milliseconds) over the 3-min period; I can share this if you want. But I am not sure this is what you want at this stage. Is not your data already with timestamp? I would actually like to try these methods myself. Happy to collaborate as well if you like. ;-)

– TwinPenguins – 2018-04-11T12:49:46.773

Sure , it will be great to collaborate with you :) . My data has a timestamp ( every second ) . So , your function must be helpful for me :) . – John Karimov – 2018-04-11T13:21:09.493

No answers