Feature Extraction or Feature Selection ?

Hi All,

I am hoping someone can advise me. I am using an accelerometer to gather raw data. After some filtering stages I am left with nice looking waveform, which I then want to use do use to create a fall detection device. I have been exploring the machine learning app. But I am now stuck on this topic.. should I be using feature extraction techniques or feature selection techniques to build my model.

I understand that extraction involves extracting certain properties such as peak acceleration or distance between peeks etc... but I'm getting lost with feature selection methods.

Thanks in advance.

Answer 1

[Image Analyst]

I'm not sure what you mean by feature selection. I mean, you could select dozens of features to measure, but some of them might be meaningless, therefore you select only the features you want. For example you may measure acceleration in the three directions, temperature, time of day, heart rate, GPS coordinates, and instantaneous stock market index. But perhaps only the 3 accelerations are the features you're interested in to gauge the severity or occurrence of a fall. So you select only those to measure, and you don't need to bother measuring the others.