Complete Predictive Analysis & Machine Learning Bootcamp

Address class imbalance in predictive analytics by balancing data and tuning ensemble methods. Load data, split into training and testing, and evaluate with a classification report.

grid search helps identify the best parameter combinations for classifiers by exhaustively evaluating a grid of values, using cross-validation and metrics such as precision and recall.

Learn how Adaboost regressors compute feature importance to identify key features, enable dimensionality reduction, and boost regression performance through iterative emphasis on difficult cases.

Apply an extremely random forest regressor to predict the number of vehicles using a preprocessing pipeline on Dodgers Stadium game-day traffic data with day, time, and game indicators.

Apply AI-powered traffic prediction using label encoding, train-test split, and an extremely random forest regressor, evaluated by mean absolute error.

Explore unsupervised learning and clustering to identify subgroups in unlabeled data using similarity metrics. See how centroids and iterations update cluster assignments to reveal intrinsic properties of datapoints.

Learn to perform clustering with a gamins clustering model: load data, choose five clusters, initialize centroids, train, predict labels on a grid, and visualize cluster boundaries and centroids.

Explore clustering with mean shift, including loading two-dimensional data, estimating the bandwidth for density estimation, and seeding and visualizing the resulting clusters.

Discover affinity propagation, a clustering algorithm that does not require predefining number of clusters. It identifies exemplars via message passing with responsibility and availability updates, guided by a preference parameter.

Apply affinity propagation to stock data, forming clusters of stocks by opening and closing costs, within a graph-based model, after loading symbol mappings and normalizing data.

Estimate clustering quality with a silhouette-like metric that compares intra-cluster and nearest-cluster distances, using Euclidean distance, to determine the optimal number of clusters.

Evaluate clustering quality with a Kamins clustering model, test various cluster counts, plot scores versus clusters, and identify the optimal number of clusters.

Explore Gaussian mixture models as a parametric approach that represents data as a combination of multiple Gaussian components, enabling modeling of region-specific behaviors and applications in finance, biometrics, and databases.

Explore implementing a gaussian mixture model, loading and splitting data, configuring components and training iterations, and evaluating training and testing accuracy with plots of three ellipses representing distinct distributions.

Classify data into predefined classes using a training data set and labels to build a classification model for face recognition, spam identification, and recommendation engine, emphasizing generalization through ample samples.

Explore how logistic regression uses a logistic function to model relationships between independent inputs and dependent outputs for classification, enabling probability-based class prediction.

Explore logistic regression in Python using scikit-learn, train a classifier on sample data, and visualize decision boundaries with a mesh grid and plotting utilities.

Explore the Naive Bayes classifier, its independence assumption, and Bayes' theorem for estimating class probabilities, illustrated with a cheetah example and training, cross-validation, and accuracy and precision metrics.

Learn to build and visualize confusion metrics and the classification report by comparing ground truth and predicted labels, plotting a greyscale confusion matrix with labeled axes for five classes.

Explore how support vector machines define a separating hyperplane to classify data in binary and multi-class problems, maximizing the margin between classes and identifying support vectors.

Explore SVM classifier examples through practical preprocessing, one-vs-one modeling, and cross-validation, including data loading, label encoding, handling string and numeric attributes, and predicting new points.

Explore the logic programming paradigm, where programs express facts and rules to compute outcomes by deduction and unification. See how declarative reasoning solves problems with examples like sums and chess.

Match mathematical expressions using logic programming in Python, installing a logic programming package to compare expressions and deduce variable values via commutative and associative properties.

Explore how logic programming parses a family tree to infer relationships such as parent, grandparent, siblings, uncles, and grandchildren from a dataset.

Analyze geography through logic programming by modeling US states with facts and rules, querying coastal and bordering states to uncover location-based insights.

Explore logic programming with a puzzle solver example that uses facts and rules to assign four people, pets, and countries, and identify who owns the rabbit.

Explore how heuristic search speeds up problem solving by eliminating obviously wrong options and scoring nodes with a heuristic function, offering fast, practical yet not always optimal solutions.

Investigate local search techniques for constraint satisfaction problems, including hill climbing and simulated annealing, guided by an objective function to update states toward the goal.

solve a constraint satisfaction problem with a best-first heuristic search. combine path cost and Manhattan-distance heuristic to efficiently reach the goal in a sliding puzzle.

Explore how heuristics solve the region coloring problem as a constraint satisfaction task, using Python to assign four colors so adjacent regions never share a color.

Utilize the A* algorithm to compute the shortest path in a maze from the start to the goal, navigating obstacles with a Euclidean distance heuristic and movement costs.

Explore heuristic search for puzzle solving by combining a cost function and heuristic estimates, using best-first search and Manhattan distance to efficiently find the optimal solution.