from voting_classifier import predict
import numpy as np

fil = "siwoco-swellex-topic-bigram-compact-nh.csv"

data = []
labels = []

with open(fil,"r") as f:
    for l in f:
        if not l.strip():
            continue
        values,target = l.rstrip().rsplit(",",1)
        values = np.array(values.split(","), dtype="float64").reshape(1,-1)
        data.append(values)
        labels.append(target)

def cmap(pred):
    if pred == "A1":
        return 0
    if pred == "A2":
        return 1
    if pred == "B1":
        return 2
    if pred == "B2":
        return 3
    if pred == "C1":
        return 4
    return -1

def f1(p,r):
    return (2*p*r)/(p+r)

def mean(l):
    return float(sum(l))/max(len(l),1)

matrix = {}
matrix["A1"] = [0,0,0,0,0]
matrix["A2"] = [0,0,0,0,0]
matrix["B1"] = [0,0,0,0,0]
matrix["B2"] = [0,0,0,0,0]
matrix["C1"] = [0,0,0,0,0]

for i in range(len(data)):
    pred = predict(data[i])
    target = labels[i]
    matrix[target][cmap(pred)] += 1

r_a1 = float(matrix["A1"][0]) / sum(matrix["A1"])
r_a2 = float(matrix["A2"][1]) / sum(matrix["A2"])
r_b1 = float(matrix["B1"][2]) / sum(matrix["B1"])
r_b2 = float(matrix["B2"][3]) / sum(matrix["B2"])
r_c1 = float(matrix["C1"][4]) / sum(matrix["C1"])       
    
p_a1 = float(matrix["A1"][0]) / (matrix["A1"][0] + matrix["A2"][0] + matrix["B1"][0] + matrix["B2"][0] + matrix["C1"][0])
p_a2 = float(matrix["A2"][1]) / (matrix["A1"][1] + matrix["A2"][1] + matrix["B1"][1] + matrix["B2"][1] + matrix["C1"][1])
p_b1 = float(matrix["B1"][2]) / (matrix["A1"][2] + matrix["A2"][2] + matrix["B1"][2] + matrix["B2"][2] + matrix["C1"][2])
p_b2 = float(matrix["B2"][3]) / (matrix["A1"][3] + matrix["A2"][3] + matrix["B1"][3] + matrix["B2"][3] + matrix["C1"][3])
p_c1 = float(matrix["C1"][4]) / (matrix["A1"][4] + matrix["A2"][4] + matrix["B1"][4] + matrix["B2"][4] + matrix["C1"][4])

a1_f1 = f1(p_a1,r_a1)
a2_f1 = f1(p_a2,r_a2)
b1_f1 = f1(p_b1,r_b1)
b2_f1 = f1(p_b2,r_b2)
c1_f1 = f1(p_c1,r_c1)
    
avg_f1 = mean([a1_f1,a2_f1,b1_f1,b2_f1,c1_f1])

print("Confusion matrix")
print("\t".join([str(x) for x in matrix["A1"]]))
print("\t".join([str(x) for x in matrix["A2"]]))
print("\t".join([str(x) for x in matrix["B1"]]))
print("\t".join([str(x) for x in matrix["B2"]]))
print("\t".join([str(x) for x in matrix["C1"]]))
print("*"*60)
        
print("\tRecall\t\tPrecision\tF1")
print("A1\t{} {} {}".format(r_a1,p_a1,a1_f1))
print("A2\t{} {} {}".format(r_a2,p_a2,a2_f1))
print("B1\t{} {} {}".format(r_b1,p_b1,b1_f1))
print("B2\t{} {} {}".format(r_b2,p_b2,b2_f1))
print("C1\t{} {} {}".format(r_c1,p_c1,c1_f1))
print("-"*60)
print("Average F1: {}".format(avg_f1))