Comparative Analysis of Machine Learning Models on the Iris Dataset

Program Project Program Identification Program ID: T5 Program Title: Machine Learning Module Program Information: Project Title: Comparative Analysis of Machine Learning Models on the Iris Dataset Description Project Overview: The aim of this project is to provide hands-on experience to trainees in both supervised and unsupervised machine learning tasks using the famous Iris dataset. The project involves two main tasks: 1.Unsupervised Learning: Clustering and outlier detection on the dataset. 2.Supervised Learning: Building baseline models, comparing multiple algorithms, tuning the best-performing model, and evaluating it against an ensemble of algorithms. Dataset: Iris Dataset • The Iris dataset is a well-known dataset in the machine learning community, containing 150 samples of iris flowers, each with four features (sepal length, sepal width, petal length, and petal width) and a target variable specifying the iris species (setosa, versicolor, or virginica). • Dataset Link: Iris Dataset: https://archive.ics.uci.edu/ml/datasets/iris Tasks and Mark Distribution: 1. Data Preprocessing (10 marks): 1 – V 1.0/05032024 • Load the Iris dataset. • Perform data exploration and visualization. • Check for missing values and handle them if any. • Split the dataset into features and target variables. 2. Unsupervised Learning: Clustering and Outlier Detection (20 marks): • Apply K-means clustering algorithm to cluster the data. • Visualize the clusters. • Detect outliers using appropriate techniques such as isolation forest or DBSCAN. • Evaluate the clustering results. 3. Supervised Learning: Baseline Model (10 marks): • Choose an appropriate evaluation metric based on the problem (classification). • Split the dataset into training and testing sets. • Build a baseline model (e.g., logistic regression or decision tree) using default parameters. • Evaluate the baseline model’s performance. 4. Model Comparison (30 marks): • Select 3-4 machine learning algorithms (e.g., SVM, Random Forest, Gradient Boosting) suitable for the problem. • Implement each algorithm and evaluate its performance using cross-validation. • Compare the performance of algorithms based on evaluation metrics. • Select the best-performing algorithm. 5. Model Tuning and Ensemble (20 marks): • Perform hyperparameter tuning on the best-performing algorithm using Grid Search or Random Search. • Evaluate the tuned model’s performance. • Implement an ensemble of the top-performing algorithms and compare its performance with the tuned model. 6. Documentation and Presentation (10 marks): • Provide a clear and concise report documenting the project process, including data preprocessing, model implementation, evaluation, and conclusions. • Prepare a presentation summarizing the key findings and insights from the project. Total Marks: 100 Note: • Trainees are encouraged to seek guidance from instructors, conduct additional research, and experiment with different approaches to enhance their understanding and skills in machine learning. The project aims to provide a comprehensive learning experience in both supervised and unsupervised learning tasks using a real-world dataset. • Maximum team count is three 2 – V 1.0/05032024 Project Outcomes By the end of this project trainee will deliver: A. Python code in a Notebook B. Presentation for the project showing their results 3 – V 1.0/05032024