I calculated an entropy level of user's behaviour for its possible states of occurence (H:=Home, w:=Work or E:=Elsewhere) for a day. Say a user A has its possible states for each hour of a day as {H,H,H,H,H,H,H,H,H,W,W,W,W,W,W,W,W,W,E,E,H,H,H,H}
P(H)=13/24=0.54
P(W)=9/24=0.38
P(E)=2/24=0.08
Using shannon's entropy, I wanted to calculate the entropy level of a user to figure out how predictable the user is.
H(A)=(-P(H)*log2(H)) + (-P(W)*log2(W)) + (-P(E)*log2(E))
=0.48 + 0.53 + 0.29
=1.3
How to distinguish this information in bits as low entropy or high entropy?