https://github.com/estevesx10/ml1-adaboost-analysis-optimization

AdaBoost Analysis and Optimization [Machine Learning I UC Project]
https://github.com/estevesx10/ml1-adaboost-analysis-optimization

adaboostclassifier boosting-ensemble weak-learners

Last synced: about 1 month ago
JSON representation

AdaBoost Analysis and Optimization [Machine Learning I UC Project]

• Host: GitHub
• URL: https://github.com/estevesx10/ml1-adaboost-analysis-optimization
• Owner: EstevesX10
• License: mit
• Created: 2024-04-03T20:31:37.000Z (9 months ago)
• Default Branch: main
• Last Pushed: 2024-06-23T22:00:57.000Z (6 months ago)
• Last Synced: 2024-06-24T20:40:33.611Z (6 months ago)
• Topics: adaboostclassifier, boosting-ensemble, weak-learners
• Language: Jupyter Notebook
• Homepage:
• Size: 23.3 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        




# ML1 | AdaBoost Classifier Analysis & Optimization





    





    

        

    








    

        

    

    

        

    

    

        

    

    

         

    



## Project Overview

In recent years, the rise of **Artificial Intelligence** and the widespread use of **Machine Learning** have revolutionized the way we tackle complex real-world challenges. However, due to the **diverse nature of data involved**, choosing the right algorithm is crucial to achieve efficient and effective solutions. Therefore, understanding the **strengths** and **weaknesses** behind different Machine Learning algorithms, and knowing how to **adapt them** to meet specific challenges, can become a fulcral skill to develop.

Furthermore, since the **choice of algorithm** greatly depends on the specific task and data involved, it's clear that there is no **"Master Algorithm"** (No algorithm can solve every problem). For example, while Linear Discriminants effectively delineate boundaries in data that is linearly separable, they struggle to capture relationships in more complex, higher-dimensional spaces.

This Project focuses on the following topic:



> With no Master Algorithm, is it possible to improve a existing Machine Learning Algorithm in characteristics it struggles the most?



Therefore, after choosing a **Machine Learning Algorithm** and gaining a thorough understanding of its theoretical and empirical aspects, we aim to **refine it**, specifically **targeting its weaknesses** in solving classification problems.



    



## Classifier Selection

Nowadays, since **singular Machine Learning Algorithms** can fall short to predict the whole data given, we decided to study an **Ensemble Algorithm**. Since these Algorithms can combine outputs of multiple models it makes them more prone to **better address more complex problems** and **provide better solutions**.

Consequently, after careful consideration, we decided to focus on enhancing the **AdaBoost Algorithm M1**, which is employed in **binary classification problems**.

  

    

        


        AdaBoost (Adaptive Boosting) is a type of ensemble learning technique used in machine learning to solve both classification and regression problems. It consists on training a series of weak classifiers on the dataset. Therefore, with each iteration, the algorithm increases the focus on data points that were previously predicted incorrectly.

        

    

    

        


        

    

  


As a result, the AdaBoost algorithm builds a model by considering all the individual **weak classifiers** which are **weighted based on their performance**. Consequently, classifiers with **higher predictive accuracy contribute more to the final decision** which **reduces the influence of less accurate ones** in the final prediction. 

## Authorship

- **Authors** → [Gonçalo Esteves](https://github.com/EstevesX10) and [Nuno Gomes](https://github.com/NightF0x26)

- **Course** → Machine Learning I [CC2008]

- **University** → Faculty of Sciences, University of Porto





`README.md by Gonçalo Esteves`