
This video talks about a brief course description and tells you what to expect from this course.
Build a solid foundation in R programming and related concepts, gain a grounding in statistics and basic linear algebra, and cultivate an interest in machine learning and data modeling.
This video tells an outline about what you will learn in subsequent sessions of the course.
In this lecture you will learn about the various techniques of Machine Learning.
Explore cross validation as a robust method to evaluate and compare learning algorithms by splitting data into training and test (validation) sets, then estimate prediction errors to assess model performance.
Explore cross-validation by random subsampling data into training and testing sets, fitting models, and evaluating accuracy across iterations. Estimate the average error rate using out-of-sample estimation and leave-one-out cross-validation.
Learn to use the caret package in R for pre-processing, feature selection, model training, and evaluation with cross-validation and confusion metrics on datasets like spam.
Learn how classification uses training data to assign documents to categories and enable natural language processing tasks like spam filtering and market segmentation.
Explains k-nearest neighbors classification using majority vote among neighbors, with Euclidean and Minkowski distances for numeric data and Hamming distance for categorical data, and notes standardization to handle mixed scales.
Load the Wisconsin breast cancer data, drop the id column, encode diagnosis as a factor, scale features, and apply knn with a selected number of neighbors; evaluate with a cross-table.
Understand the naive bayes classifier and how prior, likelihood, and posterior probabilities guide classification in high-dimensional data, with practical examples of gender and color classification.
Implement a naive-bayes classifier in R for spam detection using text mining: import data, clean text, build corpus, create document-term matrix, convert to presence/absence, split data, train, and evaluate.
Explore linear discriminant analysis, a simple method that searches for a linear combination of variables to best separate two classes and minimize misclassification.
Load the forensic glass dataset from a package, perform linear discriminant analysis, view probabilities, group means, and discriminant coefficients, and predict classifications for new data.
Explore how artificial neural networks, a statistical learning model inspired by biological neurons, estimate functions from many inputs using interconnected neurons, weights, activation functions, and hidden layers.
Explore conceptualizing neural networks and backpropagation, with examples like predicting football plays, brain activity patterns triggering robotic arms, self-driving cars from road images, and chewing pattern analysis.
Implement a neural network in R by preparing training and test partitions, training with neural net functions, and validating predictions on an infertility dataset.
Explore how back propagation, the backward propagation of errors, trains artificial neural networks by propagating error signals from outputs back to hidden layers to adjust weights via gradient descent.
Compare neural networks with back propagation, demonstrating an implementation on an inverter dataset with two hidden layers, and contrasting error rates and step counts.
Learn how support vector machines, a supervised learning method, use a hyperplane to separate data into two classes, maximize the margin, and operate in higher dimensional feature spaces.
Apply support vector machines in R by loading a dataset, splitting training and test sets, and experimenting with kernels such as RBF to evaluate model performance.
Discover how a decision tree classifies data from root node to terminal nodes using splitting, pruning, and grafting, highlighting interpretability and handling missing values, and risks of too many splits.
Learn how to implement a decision tree in R using a flowers dataset, splitting into training and testing sets, building and evaluating the model, and predicting new values.
Bootstrap aggregation, or bagging, improves accuracy by drawing multiple random training samples, training separate models on each, and averaging their predictions.
Boosting combines many predictors to reduce training error by iteratively focusing on misclassified points and summing weighted classifiers for improved predictions, using GBM with trees.
Explore how random forest builds an ensemble of decision trees for classification and regression, using random data subsets and majority voting to improve accuracy and handle missing data.
Explore the implementation of random forest in R, including training and testing, inspecting trees with getTree, predicting species, and evaluating accuracy and standard deviation.
Explore how clustering groups objects by similarity, forming clusters that are more alike within than between. Compare two main approaches—k-means unsupervised learning and hierarchical clustering—and review their implementation.
Explore how k-means clustering groups data into k clusters by assigning points to the nearest seed and recalculating centroids, using distance measures and iteration to achieve stable, well-separated clusters.
Implement k-means clustering in R by loading the required package, performing a 70/30 training-testing split, and evaluating three clusters with their means and label comparisons.
This lecture covers hierarchical clustering in R, using dist to compute Euclidean distances for continuous data and binary distances, then builds clusters with hclust and plots a dendrogram.
Explore predicting with linear regression by modeling how the dependent variable changes when independent variables vary, using the regression equation and an error term.
Learn how to implement linear regression in R using the faithful dataset, fit a model with waiting time as the predictor for eruptions, and generate predictions and prediction intervals.
Apply a training and test split to a multiple covariates regression using a data subset. Fit a model with ten predictors, assess residuals, and explore diagnostic plots colored by race.
Predict binary, ordered, and multinomial outcomes using logistic regression, blending continuous and discrete predictors to model whether an event occurs rather than when.
Explains logistic regression on the S&P 500 daily returns (2001–2005) using lag1, lag2, and lag3 as predictors, with training before 2005, test on 2005, binomial link function, and 0.5 threshold.
Forecast time series data by analyzing past and present data at regular intervals, identifying trend, seasonality, and random fluctuations. Explore additive and multiplicative decomposition for accurate forecasts.
Explore forecasting implementation by loading sales data into a time series, decomposing it into seasonal, trend, and irregular components, and applying exponential smoothing for forecasts.
Inspired by the field of Machine Learning? Then this course is for you!
This course is intended for both freshers and experienced hoping to make the bounce to Data Science.
R is a statistical programming language which provides tools to analyze data and for creating high-level graphics.
The topic of Machine Learning is getting exceptionally hot these days in light of the fact that these learning algorithms can be utilized as a part of a few fields from software engineering to venture managing an account. Students, at the end of this course, will be technically sound in the basics and the advanced concepts of Machine Learning.