Artificial Intelligence and Machine Learning: Complete Guide

Learn intelligent search algorithms by defining problems and representations, using a city map to find the best routes with greedy search and A star search, and explore related data structures.

Explore the core components of a problem for search algorithms: initial state, end state, state space, actions, and the solution, with the Arad to Bucharest example.

Learn how heuristics guide search by prioritizing promising moves to solve problems faster, illustrated with puzzle states, distance-based costs, and using straight-line distance to find the shortest path.

Explore the intuition of ordered arrays as a data structure for ascending values. See how linear search guides insertion and how deletions shift elements to maintain order.

Explore implementing a sorted array in Python with a class named Ordered Vector, managing capacity and inserting elements in order by shifting values, demonstrated in a Google Colab notebook.

Build a city map for intelligent search by modeling cities as vertices with adjacency and heuristic distances, then use an ordered array to explore with greedy search.

Explore the intuition of greedy search, using straight-line distance as a heuristic to find the shortest path from Arad to Bucharest by expanding the closest city first.

Implement grid search from scratch to solve city graph problem, using a grid class, the objective Bucharest, visited tracking, and an ordered array with recursive expansion of adjacent cities.

Discover how the A* algorithm blends heuristic and real road distances to find the shortest path from Arad to Bucharest, demonstrated with route calculations and a greedy comparison.

Implement a star search in a Google Colab notebook, using a star distance (heuristic plus road cost) and an ordered adjacency structure to navigate from Arad to Bucharest.

Explore a south Brazil city map with greedy search and a star search using straight line distance; show a star yields 20 km shorter route (228 km vs 248 km).

Explore optimization algorithms and metaheuristics, including climb, simulated annealing, and genetic algorithms, through practical case studies on flight schedules and product transport.

Apply optimization algorithms to a Europe case study, selecting the cheapest round-trip flights from Lisbon, Madrid, Paris, Brussels, London, and Dublin to Rome using a flight dataset.

Represent the problem for optimization by converting flight data into a Python dictionary that captures origin, destination, depart, arrival, and price.

Learn to implement and print flight schedules from a data structure, calculate total price, and explore optimization algorithms to find the cheapest flights among many possibilities.

Define a fitness function to minimize flight costs by evaluating schedules, use mlrose to test discrete optimization with algorithms like hill climbing, simulated annealing, and genetic algorithm.

Introduces hill climb algorithm, starting from a random solution and moving to the best neighbor to minimize cost, and demonstrates local vs global optima, applicable to minimization and maximization.

Implement hill climb algorithm with the Melrose Library to optimize flight schedules, revealing the best solution and price, and visualize results from Lisbon to Rome at 99 and 129.

Learn simulated annealing, optimization algorithm inspired by physics. Start from a random solution, use a decreasing temperature to sometimes accept worse moves and escape local minima toward the global minimum.

Implement simulated annealing with adjustable temperature and decay, and compare results to hill climb. Tune initial temperature and max attempts to improve the best solution and price.

Learn the basics of genetic algorithms, including initial population, individuals, chromosomes, and genes, and how evaluation and stopping criteria guide selection, crossover, and mutation to improve solutions.

implement a genetic algorithm by configuring population size and mutation probability, visualize the best solution and price, and compare with hill climb and simulated annealing across settings.

Apply optimization algorithms to select a 0/1 loading vector for 14 products to maximize revenue within a three cubic meter truck capacity, balancing space and price.

Install and import the library; model products by name, space, and price; load into 3 m3 truck with a 0/1 vector; maximize revenue via a fitness function using hill climb.

Explore fuzzy logic within artificial intelligence, examine its intuition and applications, and implement a tipping case study using the S-k Fuzzy Library to model service and food quality.

Explore fuzzy logic applications in devices like washing machines, vacuum cleaners, air conditioning, dishwashers, and microwaves, where sensors guide controllers to adjust cycles, water, power, and detergent.

Explore the basics of fuzzy logic, mapping cost and benefit to non-binary categories through rule-based reasoning. Learn about membership functions: triangle, trapezoid, shoulder trapezoid, and sigmoid that create overlapping ranges.

Begin implementing fuzzy logic for a restaurant tipping system in a Google Colab notebook, linking service and food quality to tip, and install and import scikit-fuzzy for future membership functions.

Map numbers to quality and service categories using three membership functions, range 0-10, visualize their intersections, then map outcomes to tip categories (0-20) and build rules.

Implement a fuzzy control system with variables, rules, and a simulation to compute tips. Visualize results and explore membership functions and intersections for food and service quality.

Explore fuzzy systems by creating membership functions for quality and service using sigmoid, Gaussian, and Pi functions, and visualize robust intersections in a Google Colab notebook.

Discover machine learning fundamentals, including classification, regression, clustering, and reinforcement learning, with intuition for algorithms like naive Bayes, decision trees, SVM, and linear regression, using Orange 2.

Explore how machine learning and data science fit inside artificial intelligence, covering classification, regression, clustering, and real-world applications from face detection to self-driving cars and streaming recommendations.

Learn how classification uses a training data set with features like credit history and income to predict loan risk. It also classifies product purchases and image data.

Explore the intuition of naive Bayes for risk, using data and a probability table to classify customers as high, moderate, or low by credit history, debts, properties, and income.

Implement a Naïve Bayes classifier in Orange by loading the Zoo dataset, training with an 80/20 split, evaluating with cross-validation, and interpreting the confusion matrix and accuracy.

explains how decision trees classify loan risk using income and credit history, yielding leaves labeled high, moderate, or low; contrasts with naive bayes and notes entropy and information gain.

Apply a decision tree to the zoo dataset, training with 80% of data and testing with 20%, achieving about 89% accuracy and visualizing the confusion matrix.

Explore rule based learning in machine learning, where data yields rules to classify customer risk (high, moderate, low) using income and credit history, with a default ruling for unmatched cases.

Explore rule-based learning in Orange using the zoo dataset to induce rules. Evaluate with 80/20 train/test split and confusion matrix, achieving 94% accuracy and outperforming naive Bayes and decision trees.

Learn how the k nearest neighbors algorithm classifies a new point by euclidean distance to labeled points, choosing k and using majority class among the closest neighbors.

Load the dataset, configure the kNN algorithm in Orange, and predict animal types from features using a 80/20 train/test split and euclidean distance; compare 92% to 94% accuracy.

Explore how support vector machines find the best straight line to classify data and maximize margins with support vectors.

Explore svm testing on the zoo dataset, visualize data, adjust cost and kernel parameters, train on 80% and test on 20%, achieving 93% accuracy and examining the confusion matrix.

Learn logistic regression as a classification method that finds the best boundary to separate pay yes from pay no data, shown as an s-shaped separator.

Learn to apply logistic regression on the zoo data set to predict animal type, compare algorithms (knn, svm, naive bayes, CN2 rule, decision tree), and evaluate with test and score.

Learn how k-fold cross-validation evaluates algorithms by partitioning data into k folds, training on k-1 folds, testing on the remaining fold, and averaging results for robust accuracy.

Investigate wine quality prediction with the wine dataset using multiple algorithms such as SVM, naive Bayes, and decision trees. Apply 80/20 training testing and cross-validation to compare accuracies.

Learn to classify cat and dog images in Orange by transforming pixels into embeddings and applying tree and other algorithms, then evaluate with ten-fold cross-validation achieving 92 percent accuracy.

Learn how regression predicts numerical values from x variables to forecast y, with examples like sales from advertising costs and wind speed from weather factors.

Explore the intuition of linear regression to predict health plan cost from age by fitting a line with coefficients b0 and b1, showing training estimates for simple and multiple regression.

Explore linear regression in orange by loading a car price dataset, examining numeric targets, and evaluating predictions using MSE, RMSE, MAE, and R2 with cross-validation.

Build a regression model to predict house prices from 21 features (21,613 instances) using a train-test split, removing id and date, and evaluating mean absolute error.

Learn how clustering groups data to reveal patterns across markets, texts, and preferences. See examples from market segmentation, politics and sports topics, Netflix recommendations, and Covid-19 paper clustering.

Explore how k-means clusters data by age and salary, with three initial centroids and iterative distance-based assignments. Update centroids by averaging points in each cluster and repeat until no changes.

Learn to cluster data sets with the k-means algorithm in orange, using the iris data set and visualizing results to compare clusters with original classes.

Explore the credit card clients data set by creating a bill amount feature from six months, then cluster with k-means to reveal patterns between limit and spending, guiding targeted marketing.

Cluster images in orange by converting pixels to numeric embeddings, measuring euclidean distances, and applying hierarchical clustering to visualize cat and dog image groups.

Explore association rules to uncover data patterns, enabling market basket analysis, promotions, and shelf placement strategies, with examples like beer and diapers and dropout risk.

Learn the apriori algorithm for mining association rules from market basket data, using support and confidence thresholds to select items and generate rules.

Implement the a priori algorithm in orange to mine frequent itemsets and generate association rules from a market basket dataset, adjusting minimum support and confidence to discover meaningful rules.

Generate association rules from the adults dataset using only categorical features, after removing numeric attributes and preprocessing when needed, then tune support and confidence to uncover meaningful rules.

Apply preprocessing to handle missing values and normalize numerical features before running a knn classifier, improving accuracy from 75% to 82% on the adults dataset.

Discretization converts continuous variables into categories using equal frequency discretization, enabling association rule mining on the adults dataset. Explore adjusting attributes, support, and confidence to discover income-related rules.

Feature selection demonstrates pre-processing by selecting top features from the adults dataset, compares Naive Bayes and random forest, and shows how accuracy and training time vary with fewer features.

Learn how to reduce data dimensions with PCA on the adult dataset and compare to feature selection, testing naive Bayes accuracy of 82% before and 81% after eight components.

Apply principal component analysis to reduce the adult dataset to two components, then cluster with K-means and visualize on a scatterplot of the first two components for income classification.

Detect outliers in datasets using scatter plots and unsupervised methods like one-class SVM and local outlier factor, with examples from credit data and fraud detection.

Install the time series package, load the dataset, select date and open, visualize with data table and charts, and train an algorithm to predict the next 30 prices.

Forecast time series with Arima and Var models, set open as target, tune p, D, q, and forecast 30 steps ahead; evaluate with cross-validation and rmse, mae, and mape.

Explore basic charts in orange to visualize the iris dataset, a flower classification task, using distributions, scatter plots, three V's, mosaic displays, and circle-based visualizations.

Explore covid-19 time series data with interactive visualizations, including heatmaps, line plots, and scatter plots, using data tables and country filters from January to April 2020.

Explore covid-19 time series data and visualize it with data tables, then merge with the HDI dataset to analyze life expectancy, mean years of schooling, and physicians per 10,000.

Explore visualizing the Covid dataset on a geographic map using the Orange Tree geo package, mapping latitudes and longitudes, coloring by physicians, and tracking date-driven spread.