Machine Learning in R & Predictive Models | 3 Courses in 1

install R and RStudio on your computer, expand the R interface, and quickly get started with R.

Explore the motivation for using R and RStudio, learn that R is an open-source language for statistical computing and graphics, and see how RStudio enhances coding and reproducible research.

Install R from cran and RStudio by following platform-specific steps for Windows, Mac, and Linux, using the cran web page and RStudio download; the free RStudio version suffices.

Install R and RStudio on Windows using CRAN and RStudio downloads, run the installers, and accept default settings. Explore manuals and community resources for help.

Explore the RStudio interface by learning the source, console, environment, and file panes, and master setting a working directory, executing code, and accessing help and shortcuts.

Explore the RStudio interface, set up a working directory via menu, script, or console, and learn how to access help, run code, and view plots and packages.

Get an essential crash course in R basics: install and manage packages, explore data types, vectors, factors, and data frames, use functions, and read data into R.

Install and manage R packages in the studio via the package menu or install.packages, then load with library and document steps with comments, including bedfast for remote sensing time series.

Learn how to create and use variables in R with the assign operator. View results in the environment pane and perform arithmetic with named variables.

Assign X as the sum of eight and seven, and agric as the product of four and two in R; print results in the console and name variables meaningfully.

Explore data types and data structures in R by examining objects, vectors, and atomic types such as numeric, integer, logical, and character, plus factors with levels and attributes.

Explore how to assign and convert data types in R, including numeric, integer, and character, using as.integer and L suffix, and work with factors and ordered levels.

Explore vector creation and operations in R, including six vector types, declaring factors, combining vectors, type coercion to character, indexing, arithmetic, recycling rules, and built-in max and var functions.

Learn how factors represent categorical data in R, with integer levels assigned by alphabetical order; specify level order using the levels argument for controlled data input and analysis.

Explore data frames as the two-dimensional, tabular data type in R, with columns as variables and rows as observations, and use structure and summary to inspect types and statistics.

Explore control structures in R, including if and if else, to govern script flow. Apply vectorized logic to a numeric vector and use print for output.

Learn how to use for loops in R to repeat code, iterate over vector elements, and print or index results across a defined or vector-based range.

Learn to import data into R from Excel, CSV, and text files by setting the working directory, choosing the right read functions, and adjusting headers and separators.

Learn the prediction process in machine learning, from defining the problem and assembling training and test data to building a prediction function, selecting features, and evaluating accuracy in spam emails.

Identify the key components of prediction models—question, data, feature design, and algorithm—and show how data quality and feature selection drive interpretability, speed, scalability, and accuracy.

Compare in-sample and out-of-sample errors to understand generalization, and see how overfitting to training data amplifies noise, reducing performance on new data, with a spam example included.

Explore overfitting and sample errors in machine learning modeling in R through a hands-on lab, using spam data to test sample in/out, and tune rule-based predictions for performance.

Design predictive models by properly splitting data into training, testing, and validation sets, using cross-validation to select features and estimate parameters, then applying a hold-out test for out-of-sample error.

Explore common error types in binary, multiclass, and continuous predictions, and measure them with sensitivity, specificity, accuracy, positive/negative predictive values, and MSE/RMSE.

Use cross-validation to estimate out-of-sample accuracy by partitioning the training set, while preserving the independent test set, noting time-series constraints and bootstrap biases, with carrot package in R.

Use data that directly relates to the target variable to improve predictions, avoid unrelated data, and beware confounding factors and misleading correlations, including lag data considerations.

Explore unsupervised learning and clustering to reveal structure in unlabeled high-dimensional data, using distance metrics and hierarchical clustering to group and visualize results.

Discover unsupervised learning with hierarchical clustering in R, using a simulated dataset to build a distance matrix and visualize how points merge into clusters, illustrated by a clustering diagram.

Simulate and plot a 12-point data set, customize visuals, compute pairwise distances, and perform hierarchical clustering in R, exploring methods and arguments through hands-on practice.

Explore hierarchical clustering and how to merge points. Compare complete linkage and average linkage by cluster distance, and learn to test both approaches for better results.

Explore heat maps and cluster heat maps to visualize magnitude across two dimensions using color scales, and learn how hierarchical clustering orders rows and columns to reveal patterns.

Learn to build a data matrix and create heat maps in R from the spam dataset, then interpret color patterns to distinguish spam from not spam.

Run kmeans clustering in R on a data frame, view centers and cluster assignments, and visualize results; experiment with different cluster counts to observe reassignments.

Apply k-means clustering to a spam dataset using two features (capital letters and punctuation), visualize the results, and assess cluster quality against known spam labels.

Apply kmeans clustering to a spam data example and visualize the results with heat maps, ordering columns by cluster to reveal patterns in high-dimensional data.

Explore selecting the number of clusters in k-means on spam data, using two predictors, plotting accuracy across 2–15 clusters, and identifying eight clusters for high explained variability.

Assess whether clustering makes sense for a dataset by visualizing distance matrices and using Hopkins statistics; the Boston housing data shows a clustering tendency, supporting unsupervised learning.

Assess clustering performance using k-means and silkroad values, visualize results on Boston data with five clusters, and compare to supervised cross-tabulation and confusion metrics on labeled data.

Explore the carrot package, a unified framework for preprocessing data, cross-validation, training and predicting with multiple algorithms, and model comparison via a confusion matrix.

Delve into supervised learning with labeled data. Explore binary, multiclass, and multilabel tasks and algorithms like logistic regression, decision trees, SVMs, neural networks, random forests, and KNN.

Demonstrates supervised classification using the k-nearest neighbors algorithm in R, training on 50 iris samples and predicting 50 others, evaluating accuracy and tuning the k parameter.

Evaluate classification predictions with confusion matrices, distinguishing true positives and true negatives from false positives and false negatives, and compute accuracy, specificity, and positive predictive value and negative predictive value.

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

Explore the roc curve and area under the curve (auc) to evaluate classification with varying thresholds, balancing true positive rate and false positive rate on sonar data using logistic regression.

Explore regression analysis to predict a continuous outcome from one or more predictors, covering simple and multiple regression, linear and non-linear relationships, cross-validation, and validation metrics.

Explore graphical analysis for regression in R, using scatter plots, density plots, and box plots to assess relationships, outliers, and normality before fitting linear regression.

Explore linear regression in R to predict diamond price from carat and other features, using scatterplots, box plots, normality checks, and training-validation with predictions.

Explore how correlation measures linear dependence between variables, and how carat and price show a very strong positive linear relationship with r = 0.924.

Learn how to assess model fit using R squared, adjusted R squared, F statistics, p value, AIC, and BIC, and compare absolute percentage error to select the best predictive model.

Explore AIC and BIC as criteria for model selection, measuring goodness of fit across logistic and linear regression, where lower values indicate better models.

Evaluate prediction model performance in supervised learning using RMSE for regression and accuracy for classification, addressing in-sample versus out-of-sample error and using a confusion matrix.

Predict diamond prices with a full linear regression model and evaluate in-sample error using root mean square error (RMSE); learn to split data into validation and training sets for assessment.

Learn to evaluate prediction models with an 80/20 train-test split, set a seed for reproducibility, and assess out-of-sample RMSE on the test set, with cross-validation for other data sets.

Apply multiple linear regression to predict sales from YouTube, Facebook, and newspaper ad spend, estimate coefficients and intercept, assess significance, and validate with an 80/20 train‑test split.

Predict future sales with a multiple linear regression model using Facebook, YouTube, and newspaper, validate with mean squared error, and note potential interaction effects.

Explore non-linear regression in R with polynomial, spline, and generalized additive models, learn to compute and compare models using RMSE and R-squared to select the best fit.

Explore polynomial regression in R with Boston dataset to predict median house value from lower-status population percentage, using quadratic to fifth-order models and 80/20 train-test split, visualizing fits with ggplot.

Explore how a logarithmic transformation of predictive variables reduces skewness and aligns data with normal distribution to improve model validity in R, demonstrated with the Boston house price example.

Learn spline regression in R as an alternative to polynomial regression, using knots and cubic splines on the Boston data to model median home value from lower status.

Explore generalized additive models in R, a flexible approach that automatically fits splines to reveal non-linear effects. Apply to the Boston housing data and compare GAM performance using RMSE.

Learn how decision trees perform classification and regression through if-then splits, creating leaves and nodes, and how pruning and bagging with random forests prevent overfitting.

Apply decision trees in R to classify sonar data with 60 features into mines or rocks using the airport and rpart packages. Visualize and assess accuracy with a confusion matrix.

Explore random forest theory as an ensemble classifier built from multiple decision trees using bagging. Learn tuning with grid search and apply to classification and regression tasks.

Apply random forest in R using ranger and caret to classify mines versus rocks with the sonar dataset, training on 60/40 splits and evaluating accuracy, sensitivity, and specificity.

Compare predictive models—linear, random forest, and support vector machine—using a unified caret framework on a churn dataset, and select the best model based on resampled roc, sensitivity, and specificity.

Understand model selection essentials in R by choosing predictive variables to build simpler, high-performing models. Compare best subset regression, stepwise regression, and other high-dimensional techniques for optimal linear models.

Tackle the final project by testing at least five diverse models on the Charn data, compare performance, and select the best model; optional predictions with your own data are recommended.

Celebrate learner progress and invite students to explore the instructor’s Udemy page, YouTube channel Jio World, and social channels for guided courses in remote sensing, data science, and machine learning.