The popular IRIS dataset is used for training of linear and non-linear SVM models. The hyper-parameters are fine-tuned of the models are fine-tuned using K-Fold Cross Validation and GridSearch to improve model performance.

The dataset is made of examples of irises, each represented with a feature vector of dimension 4. The examples belong to one of 3 categories (setosa, versicolor, and virginica). The feature vectors contain the width and length of the sepal and of the petals.

• Number of Instances: 150 (50 in each of three classes)
• Number of Attributes: 4 numeric, predictive attributes and the class
• Attribute Information:
  • sepal length in cm
  • sepal width in cm
  • petal length in cm
  • petal width in cm
  • class:
    • Iris-Setosa
    • Iris-Versicolour
    • Iris-Virginica

• The Scikit-learn implementation of linear and non-linear SVM has been used.
• SVM have hyper-parameters which need to be fined tuned for better model performance. The fine-tuning was done using:
  • K-Fold Cross Validation
  • Grid Search

The Fig. below shows the decision boundary obtained using Linear SVM.

The Fig. below shows the cross-validation accuracy of Linear SVM.

The Fig. below shows the decision boundary obtained using Linear SVM.

Detailed results have been provided in the Report.pdf file.

The Fig. below shows the cross-validation accuracy of Linear SVM.

• Scikit-learn

📦IRIS-classification-using-SVMs
┣ 📂Documents
┃ ┣ 📜Report.pdf
┃ ┗ 📜Task Description.docx
┣ 📂Resources
┃ ┣ 📜linear-svm-acc.PNG
┃ ┣ 📜linear-svm-db.PNG
┃ ┣ 📜non-linear-svm-acc.PNG
┃ ┗ 📜non-linear-svm-db.PNG
┣ 📜Linear_SVM.ipynb
┣ 📜Non-Linear_SVC.ipynb
┗ 📜README.md