MNIST digit classification in python using scikit-learn- Stochastique gradient descent and random forest

Table of contents

Summary

Import the relevant libraries

Loading the MNIST data

Exploratory data analysis

Preprocess the data set

Cleaning the data set

Separate Features and Labels

Plotting the data set

Data Splitting Process

Training Process

Training a binary classifier

STOCHASTIQUE GRADIENT DESCENT

RANDOM FOREST ALGORITHM

Performance Measures

Cross Validation

Confusion Matrix

Precision

Recall

F1

Precision/Recall Trade-off

The Test set

SUMMARY

The goal of this notebook is to analyze a classification model with the MNIST data, so in this notebook, i will detect one number from MNIST data set using binary classifiers (A classifier is an algorithm of machine learning that will determine the class to which the input data belongs to, based on a set of features). And then i will evaluate the measures of performance, and choose the model that have a great accuracy.

In This notebook, i will use the most popular machine learning approches to solve this classification problem by using the sklearn library.

The Data set

In this notebook I am using the MNIST Digits dataset is an open dataset and is available in Kaggle which can be accessed from the link below.

[](https://www.kaggle.com/c/digit-recognizer)

The dataset consists of 10 classes of handwritten Images pictures each with a number between 0-9, and each image has 784 features, this is beacuse each image is 28*28 pixels.

Project Setup

Installation

A very easy way to install these packages is to download and install the Conda distribution. This distribution of conda is available on all platforms (Windows, Linux and Mac OSX).

Dependencies

The main libraries involved in this notebook are:

-NumPy : for multidimensional array computing

-Pandas : for data manipulation

-SciKit-Learn : for machine learning Techniques

-Matplotlib : for data visualization

Key Concepts

Workflow steps:

• Acquire the dataset
• Explore the dataset
• Preprocess the dataset
• Visualize the dataset
• Split the data
• Model training
• Measure the performance
• Submit the results

How the project is organized

The Project consist of four files:

• Explore and Preprocess the data set.ipynb

• Visualizing, Spliting the data set and Training the model .ipynb

• Model Evaluation.ipynb

• Submit the results.ipynb