Algorithm 1: Sensor Selection

Input: Sn×m matrix of time series for n sensors with m samples each, a vector r = {r1,…, rm} of a reference

time series, a correlation coefficient limit ρ ∈ [—1,1], and a set of validity rules = {υ1,…, υl}.

Output: A set of reliable sensors.

n ← rows(S);

m ← cols(S);

← ∅;

// For each sensor entry in the time series matrix S.

for i ← 1 to n do

// Adjust the valid readings among the reference and the sensor.

s* ← { si,j (si,j ≠ λ) and (rj ≠ λ), 1 ≤ j ≤ m };

r* ← { rj (si,j ≠ λ) and (rj ≠ λ),1 ≤ j ≤ m };

// Compute the correlation coefficient between s* and r*.

ρ* ← cor(s*, r*);

// The correlation satisfies the given limit?

if ρ* ≥ ρ;

then

// The sensor entry satisfies all validity rules?

if ${\wedge }_{k=1}^l{\mathit{\upsilon }}_k\left(\text{s}\ast \right)$;

then

← ∪ i;

end

end

end

return