Machine Learning Regression Masterclass in Python

Download and install the Anaconda distribution on Windows, Mac, or Linux, choose Python 3.7, and access built-in Jupyter notebooks for Python data science.

Discover how Jupyter Notebooks, an open source web app, lets you write live code and narrative text in one document, enabling data cleaning, visualizations, and machine learning.

Master simple linear regression, predicting y from x with the least squares method, using Python and scikit-learn, and train and test your models for sales, volume, and fuel economy.

Explore the intuition of simple linear regression, linking temperature to revenue. Derive the best-fit line y = mx + b and predict unseen revenue.

learn how least squares yields the best fit line for simple linear regression by minimizing squared residuals between actual and predicted values. apply training and testing splits to assess generalization.

Apply simple linear regression to predict ice cream stand revenue from outside temperature using a dataset of 500 samples, following six steps: import libraries, load data, visualize, split, train, test.

Visualize the ice cream data with pandas read_csv and seaborn, exploring temperature and revenue through joint plots, pair plots, and lm plots to identify distributions and a potential linear relationship.

Split the dataset into training and testing sets using train_test_split, with X as temperature and Y as revenue. Shuffle data for generalization and use an 80/20 (training/testing) split.

Train a linear regression model in scikit-learn by importing LinearRegression, instantiating a regressor, and fitting it on X_train and y_train to obtain slope and intercept; compare fit_intercept true vs false.

Develop and test a simple linear regression model in Python, using train-test data, fit intercept, and interpret the slope and intercept. Visualize results and generate predictions with the regressor.

Develop a simple linear regression model to predict vehicle fuel economy from horsepower, and learn to explore data, visualize it, perform training-testing splits, and estimate the slope and intercept.

Apply simple linear regression to predict fuel economy from horsepower, building a model of mileage per gallon using the fuel economy dataset; visualize data and inspect statistics.

Visualize the fuel economy dataset with seaborn plots: joint plot, pair plot, and lm plot, revealing the inverse relationship between horsepower and mpg while inspecting data with head.

Train a linear regression model on a 70/30 train-test split with scikit-learn, noting a negative slope and intercept around 35–37. Experiment with fit_intercept false to compare coefficients now.

Train and test a linear regression model to predict miles per gallon from horsepower, comparing predictions to y_test. Visualize training and testing data and examine coefficients and intercept settings.

Assess regression model performance by comparing predictions to actual data. Explore metrics like mean absolute error, mean squared error, rmse, mape, mpe, and understand training–testing splits for unseen data.

Learn R square and adjusted R square as metrics for regression, quantifying explained variance and goodness of fit, with examples showing how added predictors affect them.

Explore the bias-variance trade-off in regression, comparing linear and high-order polynomial models, and learn how training versus testing performance reveals underfitting, overfitting, and the need for generalization.

Explore the theory and intuition of polynomial regression, build two practical projects from scratch, and apply nth-degree polynomial models with scikit-learn to predict salaries and commodity prices.

Explore polynomial regression as an extension of simple linear regression, modeling y as an nth degree polynomial in x to capture curvature, with salary versus experience as an example.

Learn to predict employee salary from years of experience with polynomial regression, explore the employee_salary dataset, visualize data, and compare linear and polynomial models.

Visualize the relationship using seaborn joint plot, lm plot, and pair plot; identify that a simple linear model fits poorly, then apply polynomial regression with the full training data.

Train a linear regression model on X train and Y train, extracting the slope and intercept with fit intercept true. Compare to fit intercept false and visualize the fit line.

Learn to implement polynomial regression in Python using scikit-learn's PolynomialFeatures and fit_transform to convert a single feature into polynomial columns, then fit a linear regressor on the transformed data.

Train polynomial regression models in Python using a poly regressor to transform features and visualize second and fifth order fits, showing higher orders yield better salary predictions from experience.

apply polynomial regression to model cost per unit against production volume, compare linear and polynomial fits, visualize results, and train/test with economies of scale context.

apply polynomial regression to estimate cost of goods sold from production units, and explore data with joint, pair, and lm plots in python using pandas, numpy, seaborn, and matplotlib.

Train a linear regression model with sklearn, inspect slope and intercept, visualize data with the fit line, and introduce polynomial regression to improve accuracy.

Explore polynomial regression in python by applying polynomial features to transform data, fitting a linear model with different degrees, and visualizing improved accuracy over linear regression.

Explore the theory and intuition of multiple linear regression and build two practical projects from scratch while applying scikit-learn in Python to predict stock prices and university acceptance rate.

Extend simple linear regression to multiple variables and learn the regression equation y = b0 + b1 x1 + b2 x2 + ... + bn xn, including implementation in scikit-learn.

Explore a simple multiple linear regression project that predicts the S&P 500 price from interest rates and employment using synthetic data, with data import, cleaning, training, evaluation, and visualization.

Visualizes the data set with seaborn plots and teaches creating training and testing splits. Demonstrates data cleaning to prepare for multiple linear regression in Python using scikit-learn.

Train and evaluate a multilinear regression model with scikit-learn, fit intercept options, predict on test data, and visualize predictions versus ground truth while assessing rmse and r-squared.

Visualize training data and a three-dimensional model surface with a mesh grid and linspace. Reveal how interest rates and employment predict the S&P 500 and expose the model coefficients.

Predict graduate school admission probability using a real dataset with a multi regression model, using GRE, TOEFL, university rating, SOP strength, GPA, and research experience.

Visualize and clean the admission dataset, create training and testing sets, and explore distributions, correlations, and box plots to support a multilinear regression model predicting admission probability.

Split the data into training and testing, verify no nulls, train a multiple linear regression model to predict admission chance, and review coefficients and intercept.

Evaluate the model on the test data, plot predictions against ground truth, and compute RMSE, MSE, MAE, R squared, adjusted R squared, and MAPE.

Retrain a regression model using features: GRE score and TOEFL score, to predict admission chance, compare rmse, mae, r-squared, and adjusted r-squared against using all parameters, and visualize 3D plane.

Explore the intuition and theory of logistic regression, turning linear predictions into probabilities with a sigmoid, and use pandas, seaborn, and scikit-learn to train, evaluate, and classify data.

Learn how logistic regression converts a linear equation into a probability via the sigmoid, enabling binary classification through a threshold, with examples using hours of study and pass/fail.

Explore how a confusion matrix evaluates a classification model, detailing true positives, true negatives, false positives, and false negatives, and measuring precision, recall, and accuracy.

Learn to import a comma-separated value dataset with encoding handling, inspect a binary target for ad clicks, and set up features for a logistic regression classifier in Python.

Load the training dataset, 499 samples, split into clicked and not clicked, and visualize time spent on site vs salary with scatter plots, box plots, and histograms by click status.

Clean the data by dropping emails, names, and country, define x from time on site and salary and y from clicked, and apply standard scaling before training.

Train a logistic regression model in Python to predict clicked from time spent and salary, using train-test split and achieving about 86% test accuracy via confusion matrix and classification report.

Visualize logistic classifier results by plotting training and testing data, model predictions, and decision boundaries. Compare true labels with misclassifications via a confusion matrix interpretation to assess performance.

Build multi-layered feedforward artificial neural networks for regression using Keras, and optimize with Adam; load data from a CSV with pandas, visualize with seaborn, and evaluate over epochs.

Explain the basics of artificial neural networks for regression using a temperature and revenue example, introduce a simple linear model, and show how a multi-layer network maps inputs to predictions.

Develop a regression model using an artificial neural network to predict house prices in King County, Washington, based on features like bedrooms, living area, lot size, and year built.

Learn the basics of artificial neural networks, including neurons, weights, bias, and activation, and how a fully connected feedforward network learns from data by example and minimizes error.

Take a single neuron model with inputs x1, x2, x3, weights, and bias; apply an activation function to produce a binary classification, then assemble mini neural networks in TensorFlow playground.

Explore how neural networks train via forward propagation, error calculation, and back propagation to adjust weights with gradients, using the Adam optimizer for practical one-line training in a Jupyter notebook.

Construct a multilayer perceptron by combining neurons, using matrices of inputs, weights, and biases with activation functions to minimize error.

Predict house prices with artificial neural networks for regression using King County data. Load and explore the Casey_house_data.csv dataset, noting 21k samples, 21 columns, 20 features, and one target.

Visualize the dataset with scatter plots, histograms, heatmaps, and pair plots to reveal how square footage, bathrooms, and year built relate to house prices in King County.

Train the model on minmax-scaled data with a train-test split, using a seven-input dense network in Keras and a linear output, and assess performance with Adam and mean squared error.

Evaluate the train model by analyzing training and validation loss, testing with scaled inputs, predicting house prices, and assessing performance with RMSE, MSE, MAE, and R²/adjusted R².

Apply multiple linear regression to predict house prices using scikit-learn on the same training data, compare results with an artificial neural network, and explore coefficients, intercept, and evaluation metrics.

Increase model performance by adding more features, applying normalization with the min-max scaler, and comparing neural networks with linear regression while evaluating RMSE, MAE, and R-squared.

Explore ridge and lasso regression as regularization techniques to combat overfitting by penalizing model complexity, increasing bias to improve generalization, and balancing training and testing performance.

Explore ridge and lasso regression as regularization tools using scikit-learn. Learn the bias-variance trade-off and how regularization improves generalization across training and testing.

Ridge regression introduces a penalty term to reduce variance and improve generalization, balancing bias and overfitting for better training and testing performance.

Explore how lasso regression uses alpha times the absolute value of slope as a penalty to generalize better, unlike ridge. It can drive some coefficients to zero, enabling feature selection.

Apply ridge and lasso regression to predict house prices in King County using common features. Adjust the alpha parameter to improve generalization via regularization and compare results with least squares.