Learn Machine Learning in 21 Days

Learn how to use NumPy to create vectors and matrices, perform indexing, slicing, reshaping, and basic operations like mean, min, max, and matrix multiplication.

Explore data manipulation with Pandas: create and inspect dataframes, import datasets like Titanic, use head and info, and perform grouping and aggregation with mean, count, and percentages.

Explore simple linear regression, modeling the relationship between one independent variable and one dependent variable using a straight line, with examples like salary by years of experience.

Discover how multiple linear regression uses multiple independent variables to predict a single dependent variable, with an intuitive comparison to simple regression and its advantages and limitations.

Learn why polynomial regression better captures nonlinear data than linear models, create polynomial features, train a linear model on them, and tune the degree to balance fit and overfitting.

Explore support vector regression (svr) that uses kernels to fit linear and nonlinear data with a boundary defined by epsilon and support vectors, plus its advantages and drawbacks.

Understand decision tree regression, a flowchart-like model that splits data to leaves using regional averages. Note advantages like minimal preprocessing and no scaling.

Explore how random forest regression uses a bagging ensemble of decision trees to improve accuracy for both classification and regression, while handling missing values and outliers.

Build a regression model to predict car prices from the dataset using encoded features and train-test split. Deploy a Flask web app to input data and display predictions.

Explore how logistic regression converts linear outputs to probabilities for binary and multiclass classification using sigmoid and softmax. Learn its simplicity, limitations with nonlinear data and outliers, and implementation steps.

Explore the k-nearest neighbor algorithm for supervised classification, using Euclidean distance to assign new points by the five nearest neighbors, with applications in intrusion detection and data mining.

Explore how support vector machines classify data by finding the optimal hyperplane with maximum margin, handling linear and non-linear separation using kernels and outlier awareness.

Explore kernel SVM concepts, including linear, polynomial, and radial basis function kernels, and learn how mapping to higher dimensions yields a maximum-margin hyperplane for non-linearly separable data.

Explore the Naive Bayes classification algorithm, its independence assumption, and how to implement it in code, with real-time prediction, multiclass classification, and text-based applications like spam filtering and sentiment analysis.

Learn the basics of decision tree classification and its differences from regression, with a weather forecast example and real-world applications in biomedical engineering, medicine, astronomy, financial analysis, and manufacturing.

Explore random forest classification by building multiple decision trees from subsamples, voting to produce the final prediction, and training with fit and predict methods using a specified number of trees.

Explore an end-to-end classification mini project using the iris dataset, encoding categorical labels, splitting data, training a random forest, and evaluating with a confusion matrix and accuracy metrics.

Discover how to build a good model that generalizes to unseen data, avoiding underfitting and overfitting.

Learn to evaluate and fine-tune models using cross-validation and grid search, including tenfold cross-validation and parameter grids for algorithms like random forest, to identify best settings.

Learn supervised, unsupervised, and reinforcement learning basics with regression, classification, clustering, and association, plus deploying a Python web app to Heroku.