Handwritten_Digits_Classification

Identifying the handwritten digits using classification algorithms and also implementing GridSearchCV for Hyper parameters tuning.

INTRODUCTION

The dataset is of 0-9 digits converted from image to input features, every row represents a specific digit image along with labels for every row. We can view the image using Matplotlib. Every digit is represented as 28x28 pixels, giving 784 features to work with.

Data Set

OBJECTIVE

To apply classification algorithms and recognise the digit using the pixel intensities of the handwritten image as features.

APPROACH

Import the modules and dependencies.
Load the dataset.
Analyse the dataset, get shape and info and check for null values and balance of the dataset.
Plot the images of labels.
Slice the data into features and label.
Split the data into training dataset and testing dataset in 3:1 ratio.
Apply feature scaling and normalization on features of training and testing datasets.
Apply classification algorithms.
Tune the hyperparameters using GridSearchCV.
Compare the scores of the models and fit the best model.

ACCURACY

Logistic Regression : 92.11
SVM Classifier Linear : 94.0
SVM Classifier RBF : 96.92
K Nearest Neighbors : 93.90
Decision Tree Classifier : 83.75
Random Forest Classifier : 95.21
GridSearchCV model : 97.18

CONCLUSION

Applied different classification algorithms along with the implementation of hyper parameters tuning using GridSearchCV to predict the digit. GridSearchCV gave pretty high accuracy – 97.18 %