Machine Learning Algorithm Performance Analysis

This project investigates the effectiveness of various machine learning algorithms in classifying the Iris Dataset, comparing multiple Neural Network architectures and a K-Means Clustering model. The analysis explores how different nonlinear activation functions (Sigmoid, ReLU, Leaky ReLU, Swish) and learning algorithms (Gradient Descent, RMSprop, Adam) impact classification accuracy. Additionally, a dropout layer is implemented to mitigate overfitting. The project includes a detailed PDF report summarizing findings and a Jupyter Notebook containing the full implementation, offering a comprehensive study for those interested in classification-based machine learning.

The project is implemented in Python using Jupyter Notebook for an interactive and reproducible analysis. TensorFlow and Keras are leveraged for constructing and training neural networks, while Scikit-Learn is used for data preprocessing and K-Means Clustering. The dataset is loaded from a CSV file, and Matplotlib and Seaborn are utilized for visualizing performance metrics. The combination of these technologies enables an in-depth evaluation of different machine learning approaches to classification problems.

• See Project on Github