Practical Machine Learning using Python

Explore random variables and the normal (Gaussian) distribution, illustrated by sprint timings, bell curves, and practical uses in HR appraisal and quality control.

Explore Python variables and conditions, mastering built-in data types (integers, floats, booleans, strings), dynamic typing, using the type function for checks, and conditional logic with if/elif/else and proper indentation.

Explore how Python and NumPy store numeric data as multi-dimensional arrays and create arrays with zeros, ones, full, eye, random, arange, reshape, and perform transpose and dot product.

Explore Python lambda functions as anonymous, single-expression tools for fast inline operations. Apply map and reduce to series and dataframes, including cubing and a life expectancy derivation.

Explore loan status distribution (charged-off, fully paid, current) with a bar plot, revealing a 14.2% charged-off rate, then use a heatmap to visualize correlations and null values.

Explore how exploratory data analysis reveals correlations between loan grade and default ratio using cross tabulations and derived metrics, and examine effects of tenure, employer name, and verification status.

Analyze debt-to-income ratio, public records, and income against loan default using dti bins and cross-tab insights to reveal small business, interest rate, and income effects.

Explore linear regression as a simple yet powerful machine learning algorithm for prediction problems, covering cost functions, assumptions, and the transition from simple to multiple regression with predictor variables.

Learn to predict car prices with linear regression using historical sales data, and preprocess both numeric and categorical features by extracting brand, handling outliers, and cleaning anomalies.

Learn to model binary outcomes with the logit model, linking log-odds to predictor variables through the sigmoid function, and optimize coefficients via maximum likelihood and gradient descent.

Build a logistic regression model to predict telecom churn from merged churn, customer, and internet data. Clean total charges and analyze churn patterns.

Compare logistic regression evaluation using AUC-ROC, interpret confusion matrices and accuracy metrics, and understand thresholds and ROC curves for model performance.

Demonstrates how Naive Bayes, including multinomial and Gaussian variants, computes posterior probabilities for text classification with bag-of-words and idf, enabling fast spam detection and sentiment analysis.

Explore how Gini index and entropy quantify impurity in decision trees, illustrated with the iris dataset, showing how splits yield pure leaves and when to stop growing.

Explore how random forests build multiple decision trees with bagging, feature subsampling, and voting to improve accuracy; learn pruning, hyperparameters, out-of-bag evaluation, and feature importance for practical credit default prediction.

Master how support vector machines classify data using linear and nonlinear approaches. Learn how the C parameter controls hard versus soft margins to improve generalization.

Explore practical support vector machines through three examples: linearly separable data with a linear SVM, nonlinear moon data with polynomial features, and kernel trick classification, using a pipeline with standardization.