Decision Trees, Random Forests, Bagging & XGBoost: R Studio

Install R from the official R-project site and set up RStudio to run scripts, using the script window and output window for Windows users, with a quick statistics crash course.

Celebrate reaching this milestone in the decision trees, random forests, bagging, and XGBoost course on R Studio, and access resources and your certificate to continue.

Master the basics of R and R Studio: run code with control enter, use comments, create variables with <-, perform vector operations, and manage the workspace with ls and rm.

learn how to install, load, and manage packages in R, including using library and require, installing from CRAN repositories, and scripting installations for reproducible analysis.

Add data in R by building datasets, or entering manually, or importing from CSP file. Explore the iris dataset with help and str, then load it with data(Iris).

Learn to input data manually in R by assigning values, using concatenation, and generating sequences (multiples of five from five to fifty) with the sequence function.

Import tab-delimited product data and comma-delimited customer data into the workspace, create data frames with headers, and inspect structure (1862 observations, four variables; 793 observations, nine variables).

Create a frequency distribution of regions from customer data and visualize it with a bar plot in R, adjusting color, orientation, borders, and labels, then export.

Learn to create histograms in R to visualize age distributions by binning into categories with breaks, display frequencies, customize color and labels, and export the chart.

Explore how machine learning uses past data to optimize performance, distinguish supervised and unsupervised learning, and apply classification and regression to real-world problems.

Learn the seven-step process to build a machine learning model—from problem formulation and data preparation to train-test split, model training, validation, and deployment for prediction and monitoring.

Explore the basics of decision trees, including root and leaf nodes, splitting to form regions, and regression and classification trees, illustrated with study hours and scores.

Understand how a regression tree partitions data into regions and predicts each region's mean. See how greedy binary splitting selects variables and splits to minimize the sum of squared errors.

Control tree growth by setting stopping criteria such as minimum observations to split, minimum observations at leaf nodes, and maximum depth, to prevent overfitting.

Explore a simulated movie dataset with 18 columns, where 17 predictors estimate the collection, the dependent variable, using a regression tree on budget, marketing, and genre.

Learn how to import a data set into R, inspect headers and variables, perform mean imputation for missing values, and prepare the data for training a decision tree model.

Learn how to split data into training and testing sets in R using an 80/20 split, set seed for reproducibility, and evaluate model performance on unseen data.

Build a regression tree in r using rpart and rpart.plot packages to train on a movie dataset, predict box office on the test set, and evaluate with mean squared error.

Prune large decision trees to balance interpretability and performance with cost complexity pruning using an alpha parameter that minimizes RSS plus terminal nodes, selected by cross-validation.

Learn how to prune a regression tree in R using the rpart package, selecting the cp value via cross-validated error to create a simpler, more accurate model.

Analyze classification trees that assign the most frequent class in each region, and compare splitting criteria such as classification error rate, Gini index, and cross entropy.

Use a 506-movie dataset to build a classification model predicting Oscar wins from its variables. Split the data into training and testing sets to train and evaluate performance.

Build a classification decision tree in R using regression template, impute missing values, split data into train and test, fit an rpart model with classification, plot tree, and evaluate accuracy.

Decision trees are easy to explain, graphically representable, and handle qualitative predictors without dummy variables. Yet a single tree may have lower accuracy, but ensembles can significantly improve performance.

Explore ensemble methods like bagging, random forest, and boosting to reduce variance in decision-tree predictions. See how bootstrapping and averaging multiple trees improve regression and classification accuracy.

Learn bagging in R with the randomForest package, using bootstrap samples and all predictors, compare its MSE to pruned trees, and understand the trade-off between prediction accuracy and interpretability.

Explore how random forest improves over bagging by reducing correlated tree results through random predictor subsets, and apply the M rule of thumb for variable selection.

Build a random forest model in R with the randomForest package, using a formula and predictors from the green data, and tune mtry for improved mse versus bagging.

Explore boosting techniques in ensemble learning, including gradient boosting, AdaBoost, and XGBoost, using sequential trees, residuals, shrinkage, depth, and regularization to improve performance.

Install and load gbm package, use gradient boosting in R to tune n.trees, interaction.depth, and shrinkage, predict on test data, and compare mean squared error with bagging and random forest.

Learn to implement ada boosting for classification in R with the adabag package, train boosted models, and evaluate with a confusion matrix, tuning mfinal to improve accuracy.

Explore xgboosting in R by preparing data in ab matrix format, converting categorical variables to dummy variables, and training a multi-class classifier with tunable learning rate, max depth, and iterations.

Identify the business context and key variables through primary and secondary research, then gather data to model factors like cart abandonment along the customer journey.

Identify data needed, request internal and external data, perform data receipt quality check, and study cart abandonment by marketing channels and the three buying steps: cart review, address entry, payment.

Identify the price as the dependent variable and uncover independent factors, standardize variable names with underscores, merge sources, and build a data dictionary with primary keys and definitions.

Import the dataset from a csv file into RStudio with read.csv(header=TRUE), creating data frame B, then use str to show 506 observations and 19 variables.

Explore univariate analysis by examining descriptive statistics for each variable, including mean, median, mode, range, quartiles, and standard deviations, while identifying outliers and missing values using the extended data dictionary.

Perform exploratory data analysis in r by examining the data dictionary, distributions, histograms, and scatter plots to identify outliers, missing values, and categorical variables that affect price and crime rate.

Identify and treat outliers using box plots, scatter plots, and histograms; apply imputation methods such as capping at 99th percentile, lower limits, and sigma-based replacement to preserve model accuracy.

Apply capping to outliers in hard rooms and rainfall by setting upper bounds at three times 99th percentile and lower bounds at 0.3 times first quartile, improving mean–median alignment.

Learn to handle missing values by imputation with mean, median, or mode, or zero when sensible, and apply segment means for groups, guided by business knowledge.

Impute missing values in R by replacing with the mean, handling NA with na.rm, and identifying NA entries using is.na and which, followed by assignment to update the dataset.

Explore seasonality in data, such as summer sales and tourism fluctuations, and learn to normalize by multiplying observations with a correction factor using M = mean year over mean month.

Analyze two-variable relationships with scatter plots and correlation matrices, decide to keep, discard, or transform variables, and apply transformations to achieve linearity for regression.

Transform the crime rate with log of one plus crime rate to linearize price, shown by scatter plots. Create an average distance variable from four distances and remove unused columns.

Identify and remove non-informative variables, including single-value and missing-value issues, iteratively refine features with business and regulatory knowledge for decision trees, random forests, and XGBoost.

Create dummy variables to convert categorical data into numeric inputs for regression models by coding each category as 0 or 1, using n minus one variables for n categories.

Create dummy variables in R using the dummy's package to convert the airport and water body categories into numeric columns, drop redundant columns, and prepare a numerical dataset for regression.

Learn to interpret positive, negative, and zero correlations with scatter plots and correlation coefficients, distinguish correlation from causation, and use a correlation matrix to manage multicollinearity in modeling.

Compute and round a correlation matrix in R to relate variables to price. Identify high correlations and remove one variable, such as deleting box and keeping air quality.

Celebrate reaching the final milestone by completing all lectures to obtain your certificate of completion and download it from your email or course site; share a review to motivate others.