Data Science in R: Regression & Classification Analysis

Install and set up R and RStudio, customize the R interface, and quickly start using R for data science tasks.

Explore why data scientists use R and RStudio for statistical computing, graphics, and reproducible research, highlighting open-source, cross-platform benefits and GUI options like Rattle.

Install R and RStudio on a Windows PC in two steps: download and run the R installer from the project web page, then download and install RStudio.

Navigate the RStudio interface, source, console, environment/history, and files, set up a working directory via menu or script, and learn to access help and run code with outputs and plots.

Master the basics of R by installing and managing packages, exploring data types, vectors, factors, data frames, and functions, and reading data into R.

Install and manage R packages in RStudio by loading from repository or package archive data, using install.packages and library, and selectively call specific functions with comments.

Learn how to assign variables in R with simple arithmetic (8 + 7 and 4 * 2), print results in the console, and use meaningful names.

Explore core data types and structures in R, including vectors and atomic types (numeric, integer, logical, complex), characters, and factors, plus attributes and class for object manipulation.

Discover how to assign and coerce data types in R, including numeric, integer, and character variables, perform logical operations, and work with factors, levels, and ordered factors for analyses.

Explore how vectors work in R, including vector types, type conversion, indexing, and arithmetic operations, with recycling rules and built-in functions like max and var.

Explore data frames as the core two-dimensional data structure in R, with columns as variables and rows as observations; inspect them with dim, str, and summary for analysis.

Explore control structures in R, focusing on if and else statements, true/false blocks, and vectorized forms with practical examples and a print output.

Learn to import data into R from Excel, text, and CSV formats by setting the working directory, using read.table/read.csv and read_excel, and viewing previews and summary statistics.

Learn how regression analysis predicts a continuous outcome from one or more predictors, using simple and multiple regression, cross-sectional and time-series data, and techniques like principal components and cross-validation.

Explore graphical analysis in regression models, a supervised learning approach, visualize relationships with scatterplots and density plots, and identify outliers and normality assumptions.

Explore linear regression in R using the diamonds dataset: analyze carat vs price with scatterplots, assess outliers and normality, and fit a model to predict price from multiple variables.

Assess linear relationships with correlation, interpreting values from -1 to 1 to gauge strong or weak associations, illustrated by a 0.924 strong positive link between carat and price.

Assess if the model fits your data by comparing R-squared, adjusted R-squared, F statistics, and p-values, using information criteria and absolute percentage error to choose the best model.

Explore linear regression diagnostics by assessing model significance, coefficient significance, and R-squared measures while predicting diamond price from carat, and preparing for multiple regression.

Use AIC and BIC to assess model fit and guide selection across regression models, including linear and logistic, with the Diamond Center data modeling price by size.

Evaluate model performance by using root mean square error for regression and accuracy for classification, and assess generalization with out-of-sample evaluation via confusion matrices.

Predict diamond prices with a full linear regression model and compute RMSE to evaluate in-sample error, then prepare for training-validation splits.

split the diamonds dataset into an 80/20 train-test set with a reproducible seed, train on 80%, and evaluate out-of-sample rmse on the 20% test data.

Learn to estimate a multiple linear regression predicting sales from YouTube, Facebook, and newspaper spend, assess coefficient significance, and use an 80/20 train-test split with lm in R.

Predict sales with a multiple linear regression model using Facebook, YouTube, and newspaper advertising, and assess predictions with mean squared error.

Explore nonlinear regression essentials in R, covering polynomial, spline, and generalized additive models, and learn to compare models using RMSE and R-squared to select the best fit.

Explore non-linear relationships with polynomial regression in R on the Boston dataset, predicting median house value from the lower-status population, using an 80/20 train-test split and fifth-order models.

Apply logarithmic transformations to predictive variables to reduce skewness and improve model validity when data deviate from normal distribution, demonstrated with Boston housing regression.

Execute spline regression in R using cubic splines and quartile knots to predict median Boston house value from lower status, with an 80/20 train-test split and visualizing the fit.

Learn how generalized additive models automatically fit nonlinear relationships in data by using spline bases within the GAM framework in R, illustrated with Boston housing price prediction and model interpretation.

Explore model selection essentials in R to identify the best predictors for a simpler, high-performing linear model. Compare stepwise and best subset approaches for high-dimensional data.

Learn supervised learning with labeled data to train and predict outputs. Explore binary and multiclass classification with logistic regression, decision trees, support vector machines, neural networks, and k-nearest neighbors.

Learn how the caret package unifies many prediction algorithms under a single interface, enabling preprocessing, data splitting, cross-validation, training and predicting, and model comparison with confusion matrices.

Train a supervised k-nearest neighbours classifier in R on the iris data, then predict species for unseen samples and evaluate overall accuracy, experimenting with different k values.

Explore how confusion metrics measure classification accuracy by comparing predicted and actual labels, detailing true/false positives and negatives, and applying to a two-by-two table with covid examples.

Train a logistic regression classifier on sonar data with 60 features to distinguish rocks from mines, then assess accuracy using a confusion matrix and a 0.5 threshold.

Explore how ROC curves and AUC assess logistic regression classification performance across thresholds, balancing true positive and false positive rates, and interpreting area under the curve.

Explore cart decision trees for classification and regression, using divide-and-conquer splits to form leaves and nodes, and pruning to prevent overfitting; bootstrap aggregation and random forests help produce accurate models.

Apply decision trees in R to classify sonar data into mines or rocks using the airport package and regression trees, and compare performance to a logistic regression model.

Explore random forest theory as an accurate, nonparametric ensemble classifier of classification and regression trees, trained on random data subsets, with out-of-bag validation and grid search.

Apply random forest classification in R with ranger on sonar data to distinguish mines from rocks, using a 60/40 train-test split and variable importance insights.

Compare and select the best predictive model for churn using a unified caret framework with train-control and 5-fold resampling, evaluating linear models, random forest, and SVM.

Complete your final project by applying regression and classification to churn data, testing at least five models from the caret library, comparing performance, and selecting the best model.

Follow Geo World on YouTube and explore additional courses on remote sensing, data science, and machine learning, while engaging with quizzes and promotions in this exciting community.