Optimising FOR LOOP or alternative to it - python-3.x

Question

I am using a FOR LOOP to calculate a simple probability on a dataset with approximately 500K rows of data.
For loop
class_ = 4
class_freq = Counter(list_[-1] for list_ in train_list) # Counter({5: 1476, 1: 1531, 4: 1562, 3: 1430, 2: 1498, 7: 1517, 6: 1486})
def cp(x, class_, freq_): # x is column index passed from another function
for row in train_list:
pos = 0
neg = 0
if row[x] == 1 and row[54] == class_:
pos+=1
else:
neg+=1
cal_0 = (neg + 0.1) / (class_freq[class_value] + 0.2)
cal_1 = (pos + 0.1) / (class_freq[class_value] + 0.2)
if prob_1 > prob_0:
return prob_1
else:
return prob_0
Train_list sample
[3050, 180, 4, 277, -3, 5782, 221, 242, 156, 2721, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1]
[2818, 119, 19, 30, 10, 5213, 248, 220, 92, 4497, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]
[3182, 115, 10, 553, 10, 4605, 237, 231, 124, 1768, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5]
[3024, 312, 18, 474, 177, 5785, 169, 224, 194, 4961, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2]
[3067, 32, 4, 30, -2, 6679, 219, 230, 147, 2947, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4]
[2716, 1, 10, 234, 27, 2100, 206, 222, 153, 5581, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 4]
...
The FOR LOOP works well on small dataset (few hundred rows) as expected. Unfortunately, when I try to use it on 20K rows of data, the processing time take ages. I cannot imagine how long it will take to run 500K rows of data.
FOR LOOP is grossly bad in performance for large dataset. What is an alternative to this? Will Lambda improve processing speed? I appreciate advice and assistance here, thanks.
Edited:
Thanks to everyone comments, I have tried to work on another algorithm to replace the FOR LOOP.
def cp(x, class_, class_):
filtered_list = [t for t in train_list if t[54] == class_]
count_binary = Counter(binary[col] for binary in filtered_list)
binary_1 = count_binary[1]
binary_0 = count_binary[0]
cal_0 = (binary_0 + 0.1) / (class_freq[class_value] + 0.2)
cal_1 = (binary_1 + 0.1) / (class_freq[class_value] + 0.2)
if prob_1 > prob_0:
return prob_1
else:
return prob_0
I am still running the above code in my program and the process is not done yet - so can't tell if it is much efficient. I will appreciate if someone can provide their opinion on this new block of code.
FYI, if this is indeed a better and more efficient code, then the issue of processing speed is most likely on other parts of my code.