Skip to content

Instantly share code, notes, and snippets.

@morganmcg1
Created November 29, 2017 17:58
Show Gist options
  • Save morganmcg1/ad401d3bf8b3054ba13099f42aa7c253 to your computer and use it in GitHub Desktop.
Save morganmcg1/ad401d3bf8b3054ba13099f42aa7c253 to your computer and use it in GitHub Desktop.
For multi class problems decide how the model performed in a binary situation
# for multi class problems decide whether the model was correct going long or short
def correct_quadrant_classifications(cls_labels, cm):
#Find nuumber of correct classifications
correct_clf = 0
if len(cls_labels) % 2 == 0:
for i in range(0, len(cls_labels)):
correct_clf = correct_clf + cm[i][i]
# calculate the rate of correct predictions
correct_clf_rate = correct_clf/cm.sum()
#print('\n{:.0f} out of {:.0f} predictions were correct'.format(correct_clf, cm.sum()))
# Find how many predictions were in the correct quadrant
correct_quad = 0
if len(cls_labels) % 2 == 0:
for i in range(0, int(len(cls_labels)/2)):
for j in range(0, int(len(cls_labels)/2)):
correct_quad = correct_quad + cm[i][j]
for i in range(int(len(cls_labels)/2) ,len(cls_labels)):
for j in range(int(len(cls_labels)/2) ,len(cls_labels)):
correct_quad = correct_quad + cm[i][j]
print('''\nTotal number of predictions in the correct quadrant was {:.0f} out of {:.0f},
this corresponds to a {:.1f}% classification accuracy rate\n'''.format(
correct_quad, cm.sum(), correct_quad/cm.sum()*100))
# calculate the rate of predictions in the correct quadrant, e.g. long or short
correct_quad_rate = correct_quad/cm.sum()
else:
print('cannot calculate correct quadrant classification on odd number of classes')
return correct_clf_rate, correct_quad_rate
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment