Recognizing Handwritten Digits

I am Zeeshan Darediya, a Data Science enthusiast. This is my first blog post ever. I will try to be most explainable I could be.

So, I got this project idea while exploring www.suvenconsultants.com. This is a code where we use SVM to recognize the handwritten digits from given dataset of images (imported from sklearn library)

This is how I have imported the above dataset and declared the model.

Using matplotlib library, I have visualized the images.

Now we will try to use 3 different validation criterias.

In the first case, we used just 5 values for test while the rest of 1791 of images were training data and we achieved 100% accuracy which states the model is overfitted.

In the next case we used 105 values as test data while the rest of 1691 images for training the model. We achieved an accuracy of 97% which is still a bit too high.

In the third case, we used 150 values as test data and trained the model with 1646 values to train the model. Now the accuracy was somewhere close to 94% which looks just perfect.

I hope I was clear with my explanations through this blog .

At last, I am thankful to mentors at https://internship.suvenconsultants.com for providing awesome problem statements and giving many of us a Coding Internship Exprience. Thank you www.suvenconsultants.com