In this project, we are going to explore the main classification algorithms and learn how to find and train the best possible model for the classification of handwritten digits. We will need to process a dataset that includes images of handwritten numbers from 0 to 9. The ultimate goal is to train the model to identify a digit on the picture.
Get hands-on experience with the Keras dataset, train a variety of classification algorithms, and find the best one using scikit-learn tools.
Stage 1 : Figure out how to load the dataset from the external resource, preprocess arrays, and provide general information about the data.
Stage 2 : Split the data into train and test sets and verify the class distribution.
Stage 3 : Explore sklearn classification algorithms, train the models, and compare the results.
Stage 4 : Normalize data, train the models, and explore if that improves the results.
Stage 5 : Search through classifier parameter values and find a set of parameters that yields the best result.
To learn more about this project, please visit HyperSkill Website - Classification of Handwritten Digits.
This project's difficulty has been labelled as Challenging where this is how HyperSkill describes each of its four available difficulty levels:
- Easy Projects - if you're just starting
- Medium Projects - to build upon the basics
- Hard Projects - to practice all the basic concepts and learn new ones
- Challenging Projects - to perfect your knowledge with challenging tasks
This Repository contains one .py file:
analysis.py - Contains the code used to complete the project's requirements
Project was built using python version 3.11.3
Download the analysis.py file to your local repository and open the project in your choice IDE and run the project. The performance of four models - Logistic Regression, K Nearest Neighbors, Decision Tree, and Random Forest - were compared to each other and evaluated using the accuracy metric. Please read each Stage's docstring to know the requirements of each stage and understand the output of the program.