Introduction

This is again a note to myself. This code has been prepared following different tutorials and SO answers. I have tried to cover many of them in reference below. The notebook is commented to the best of my ability. Starting with 11% in first trial, the final model is able to achieve accuracy of 95%.

If you are new, you might want to read this first:

Introduction To Machine Learning

A tutorial To Find Best Scikit classifiers For Sentiment Analysis

Steps

1. Import the data in a dataframe
2. Make data series of pixel data and the label
3. Split the data in test and train samples
4. Use simple SVC without any tuning. (Got 11% accuracy)
5. Modify the data. Take out the gradient of 0-255 and replace it with simple binary 0 and 1
6. Again use SVC (Got 90% accuracy)
7. Use GridSearchCV to find better value of parameters
8. Again use SCV with the optimum value of parameters (Got 95% accuracy)

Code

Sorry, something went wrong. Reload?

Sorry, we cannot display this file.

Sorry, this file is invalid so it cannot be displayed.

view raw
mnist-dataset-scikit.ipynb
hosted with ❤ by GitHub

https://gist.github.com/rohitsingh20122992/cf2bd168a5dbad794808610e209d3c6e

Reference:

http://scikit-learn.org/stable/modules/generated/sklearn.model_selection.GridSearchCV.html#sklearn.model_selection.GridSearchCV

https://stackoverflow.com/questions/46330329/finding-the-values-of-c-and-gamma-to-optimise-svm

https://www.kaggle.com/archaeocharlie/a-beginner-s-approach-to-classification

https://gist.github.com/rohitsingh20122992/cf2bd168a5dbad794808610e209d3c6e