F_Score

ml_F_Score.py

# An approach to do F_Score without np, however, using np is quite effective.
def F_Score(pred_list, true_list):
  def bitwise(a, b, core): 
    return [int(core(pred, true)) for pred, true in zip(pred_list, true_list)]

  true_positive = bitwise(pred_list, true_list, lambda pred, true: pred==1 and true==1) 
  true_positive = sum(true_positive)

  false_positive = bitwise(pred_list, true_list, lambda pred, true: pred==1 and true==0) 
  false_positive = sum(false_positive)

  false_negative = bitwise(pred_list, true_list, lambda pred, true: pred==0 and true==1)  
  false_negative = sum(false_negative)
 
  score = 0
  if (true_positive+false_positive) * (true_positive+false_negative) > 0: 
    precision = true_positive/(true_positive+false_positive);
    recall    = true_positive/(true_positive+false_negative);
    score     = (2*precision*recall)/(precision+recall);

  return score 

pred_list = [1, 1, 0, 0, 0, 1, 0]   # prediction
true_list = [1, 0, 1, 0, 1, 1, 1]   # true

print(F_Score(pred_list, 
         true_list))

pred_list = [0, 0, 0, 0, 0, 0, 0]   # prediction
true_list = [0, 0, 0, 0, 0, 0, 0]   # true

print(F_Score(pred_list, 
         true_list))