I have made measurements on sensors (light,humidity etc.) and I result in statistical curves/graphs. When I do the same experiments, I get a curve that looks like the previous in general, not the same of course. What I want is to model the curve and result to an equation so that when I run the experiment again and take a similar curve(graph) to say this is light sensor, or this is humidity sensor..etc. The problem is that I do not know whether this is feasible, and where to start from.. Do I need Machine Learning? Something else? Thanks...
You can use simple neural network which would learn how to determine type of sensor given measurement. To train the neural net you need data which means you would need to gather several dozens or hundreds of measurements and label them (the more data the more accurate predictions from neural network)
Hovewer, if the measurements for given sensor are very similar and from specified range, you don't really need machine learning. You just need to compute to which type of sensor your new measurement is the most similar to.
One possible approach would be to :
Take a few measures for each class of sensor
For each class create a vector of fixed length that would contain averaged values of measurement, for example if your light sensor measurements from 3 experiments look like this:
[1,4,5,3,8]
[1,3,4,3,7]
[1,3,5,3,6]
Then you average it to single vector:
[1, 3.33, 4.66, 3, 7]
When you take a new measure and want to determine it's class, you compute Mean Absolute Error of the new measurement for avereged vector of each class. The class with the lowest error is the sensor that the measurement was taken with