Introduction:
-
This program applies basic machine learning (classification) concepts on Fisher's Iris Data to predict the species of a new sample of Iris flower.
-
The data set consists of 50 samples from each of three species of Iris (Iris setosa, Iris virginica and Iris versicolor).
-
Four features were measured from each sample (in centimetres): - Length of the sepals - Width of the sepals - Length of the petals - Width of the petals
-
The central goal here is to design a model that makes useful classifications for new flowers or, in other words, one which exhibits good generalization.
Software and Libraries:
- Python (Scikit-learn, Pandas, Matplotlib, NumPy)
- Jupyter Notebook (Anaconda3)
Summary of this project:
-
The loaded data is summarized to have basic idea about the data and then visualised with the following plots to get the idea of distribution:
- Box and whisker plots - Histograms - Scatter Matrix plot (to spot structured relationships between input variables) -
The data is splitted into Training set(80%) and Validation set(20%).
-
Following 6 different models are built: (2 linear and last 4 are non-linear algorithms)
- Logistic Regression (LR) - Linear Discriminant Analysis (LDA) - K-Nearest Neighbors (KNN) - Classification and Regression Trees (CART) - Gaussian Naive Bayes (NB) - Support Vector Machines (SVM) -
Stratified 10-fold cross validation technique is used to estimate accuracy of the 6 models.
-
From the results, it is observed that Support Vector Machines (SVM) has the largest estimated accuracy score at about 0.98 or 98%.
-
SVM model is fitted on the entire training set and is used to make predictions on the Validation set.
-
Finally, the predictions are evaluated by calculating Classification Accuracy, Confusion Matrix and Classification Report.