search

armut / Occupancy-Detection

Occupancy detection in a room using supervised machine learning algorithms.
4 stars 10 forks source link

Questions about the scatter matrix #1

Open fresh1994 opened 5 years ago

fresh1994 commented 5 years ago

Thank you for sharing your code and report. I am confused with the scattering matrix with respect to occupancy. Do you have a detailed description of that? I can't understand the correlation between every two variables. Thanks a lot.

armut commented 5 years ago

image

This one right? By this matrix we can scrutinize correlations. Elements of the correlations are shown in different colors to make them distinguishable. For example let's take Humidity-HumidityRatio tuple. The graph at the intersection of these two features seems rather tangled right? You can't neatly take apart the colors which means they are highly correlated and you probably won't want to take both of them as features because they almost mean the same thing. On the other hand, the graph at the intersection of light and CO2 seems somewhat neatly slicable which means that they are not much correlated. By selecting features that are most uncorrelated between themselves and most correlated with the target, you get more healthy results in terms of feature extraction.