Supervised Machine Learning for beginners

Explore binary classification, where observations with features are labeled as two classes, using the breast cancer dataset to illustrate features, targets, and decision boundaries in supervised learning with scikit-learn.

Explore multi class classification by extending binary classification to three iris classes using the iris dataset, with features like petal length and width, visualized via scatter plots.

Learn how regression predicts continuous values using the Boston housing data to estimate property prices, exploring features like crime, rooms, and age, and fitting a linear model that minimizes error.

Learn how to handle categorical data in supervised learning by converting non-numerical attributes into numerical representations using one-hot encoding, pandas get_dummies, and the abalone dataset.

Apply min max scaling to normalize numerical attributes, transforming values to the 0-1 range with (x - min) / (max - min) and using fit transform.

Standardization centers data around zero by subtracting the mean and dividing by the standard deviation, creating a comparable scale for features. It avoids min max scaling's outlier sensitivity.

Split the prepared data into training and test sets to evaluate how well a model generalizes, using an 80/20 shuffled split with a fixed random state for reproducibility.

Assess model performance after training by splitting data into training and test sets, then compute accuracy to measure how often predictions match the ground truth.

Explore how the confusion matrix goes beyond accuracy by revealing class performance on iris dataset. Read a 3x3 matrix of actual vs predicted classes to identify misclassifications and guide improvements.

Explore precision and recall as concise metrics beyond accuracy. Relate them to confusion matrix terms: true positives, false positives, false negatives, and true negatives.

This lecture demonstrates supervised learning with linear regression, training and evaluating models on real data, computing coefficients and intercept, predicting prices, and improving accuracy through data preparation and handling outliers.

Learn logistic regression, a binary classification extension of linear regression using the logistic function to output probabilities and separate classes with a hyperplane, including one-vs-rest for multiclass.

Explore k-nearest neighbors classification by storing training data and predicting new labels from the majority of the k closest points, with two-feature demonstrations and multiclass extensions.

Explores using KNN for regression by predicting dollar amounts with nearest neighbors, averaging labels for k neighbors, and evaluating with RMSE, including standardization and hyperparameter tuning of k.

Explore decision trees and their variants, gaining explainable, threshold-based classification that handles nonlinear data and categorical features, demonstrated on iris data with a shallow, visualizable model.

Apply decision trees to regression tasks by predicting leaf averages, handle categorical features with one-hot encoding, and evaluate performance using RMSE on a train-test split.

Learn how random forests derive from decision trees, train multiple trees on data subsets, and use majority votes to improve binary classification accuracy.

Explore supervised learning with Python and Jupyter, cover data preparation, symbol learning, model evaluation, and test algorithms like logistic regression, cane neighbors, and decision trees, plus hyper parameter optimization.

Install Python 3.7, upgrade pip, install required libraries, and verify the setup by launching Jupyter notebook and running a quick hello world.