This project focuses on the development of an advanced machine-learning model. It leverages the principles of natural selection to optimize rule bundles for the classification of data—in this case, the Iris dataset. The algorithm iterates through generations of rule bundles, employing mechanisms such as crossover and mutation to evolve increasingly accurate predictions. Each rule bundle comprises a set of rules that assign weights to class attributes, allowing for a two-tiered prediction process at both the rule and bundle levels. Performance is meticulously evaluated based on accuracy, with the best-performing bundles carried forward to subsequent generations. This methodical approach aims to refine the predictive capabilities of the model, contributing to the overarching goals of the parent repository.
This project is a component of a broader project aimed at enhancing road safety through image recognition. It houses a custom version of the GoogLeNet architecture, fine-tuned for the Drive Assistant system to detect potholes with 88.46% accuracy. Notable features include a modified GoogLeNet for better detection, a curated dataset with labeled road images, and the use of data augmentation and regularization to boost model efficacy and mitigate overfitting. For a comprehensive understanding of the model’s design, effectiveness, and application, users are directed to the README file in this sub-repository.
The HazardousAsteroid_Detection project utilizes machine learning to predict potentially dangerous asteroids with a Random Forest classifier achieving a notable 99% accuracy. Feature selection was also conducted using the Random Forest algorithm, highlighting its versatility within the project.
This project utilizes the R language to offer a detailed analysis of global COVID-19 vaccination data. We explore vaccination rates considering factors like population, vaccine types, regions, and income, applying diverse sampling methods to gauge their impact on data interpretation. Key undertakings include validating the Central Limit Theorem through histograms and examining the bootstrap bias estimate in relation to sample size and deviation.
IBM ML 💼
The project contains code that was implemented to successfully complete the IBM AI Engineer professional certificate on Coursera. It reflects a comprehensive educational journey through various aspects of AI and Machine Learning.
This repository hosts a Text Preprocessing Toolkit, tailored for individuals working on Natural Language Processing (NLP) or text analytics projects.