I am using Pulse sensor to record BPM of persons when they are awake and drowsy. To collect data, finger was attached to sensor connected to Arduino, then using USB cable data was transferred to Arduino IDE. Then I stored data from Arduino IDE to Excel files. So, I have collected the data, trained SVM model on it.
Now, normally we divide data into training set and testing set. Then train ML model on training set and then test on testing set. Now, instead of testing on stored data, I want to send data from sensor (or Arduino IDE) to ML model and make predictions continuously whether the person is drowsy or awake.
I am beginner in ML. So, how should I proceed with this? How should I deploy my ML model and then data to it?
What you are trying to achieve is one of the hard things in ML deployment – getting your model to infer at scale.
There is no one-size-fits all recipe for this. Perhaps Google for ‘streaming machine learning’.
Another opportunity may be to perform a minibatch approach: say, you temporarily store 500 measurements upon receiving them – and then you feed them to the model for prediction just at once.