Machine Learning & Data Science A-Z: Hands-on Python

Build two 2x2 arrays and compare matrix multiplication with element-wise multiplication, using the dot-style matrix multiply and the star operator to show different results.

Learn to divide arrays with numpy, convert results to integers using floor, and apply numpy’s math functions; finish with random numbers and normal and uniform distributions.

Explore generating normal and uniform distributions with NumPy, create random arrays of specified size, and filter data with logical conditions to select values in a range.

Learn pandas basics by comparing series and data frames, label a series with age indices, create and filter data, and understand indexing changes for handling student age data.

Master pandas for data preprocessing by dropping columns, replacing values, and using head, tail, and sort_values, then read data from csv or excel files into a dataframe.

Learn to visualize climate data with matplotlib by plotting a line graph of year versus temperature index, then add x and y labels and a title.

Learn to visualize two customers’ monthly electricity consumption with Matplotlib by creating lists, plotting with colors and markers, labeling axes and legends, and arranging two subplots side by side.

Learn how to create and compare bar charts in Matplotlib, adjust bar width, place bars side by side, and customize x-axis labels for monthly data.

Compute the mean, variance, and standard deviation from data with simple formulas. Explore how these statistics relate to normal and uniform distributions through practical examples and Python analysis.

Explore covariance to understand how variables relate, compute the covariance matrix and correlation, and visualize these relationships with seaborn heatmaps.

Identify missing values by distinguishing null and NaN, use info to spot gaps, replace entries that are not numbers with NaN, and apply numeric conversion for preprocessing.

Identify and locate missing values during preprocessing using is not and NaN detection. Drop or fill these values using forward or backward fill, or mean imputation with a simple imputer.

Define outliers as data points far from the main range, and introduce quartiles (Q1, Q3), the IQR, and box plots, plus mild and extreme outlier calculations.

Learn to concatenate data with pandas by attaching a new column or rows, using axis=1 for column-side joins and axis=0 for row-side joins, including read_csv usage.

Explore dummy coding and one-hot encoding with Pandas get_dummies to convert categorical features, such as peak and off-peak, into numeric inputs for machine learning models.

Understand normalization to scale features to a common range, and apply min-max scaling and distance-based norms like Euclidean and Manhattan in sklearn for modeling.

Explore the types of machine learning models: supervised, unsupervised, and reinforcement learning, using cat and dog labeling examples. Understand how labeled data trains models and how reinforcement learning builds policies.

Explore supervised learning models, focusing on classification and regression. Learn to prepare the iris dataset for binary and multiclass classification with four features across 150 samples, labeling data for training.

Develop a k-nearest neighbors classifier by setting six neighbors and using Minkowski distance with p=1. Train on iris data with X and Y, then predict new samples.

Learn to create training and test sets for k-NN by using sklearn train_test_split with test_size, random_state, shuffle, and stratify, then train, predict, and evaluate accuracy.

Learn decision tree concepts as a classifier with root nodes, branch nodes, and leaves guiding end decisions. See a house-purchase scenario and iris data sets to illustrate how decisions unfold.

Develop a decision tree classifier on the iris dataset using a sklearn-like library, training with a train/test split, fitting the model, predicting, and achieving about 97 percent accuracy; next, cross-validation.

This session demonstrates developing a naive bayes classifier using scikit-learn's gaussian nb on the iris dataset, achieving about 97% accuracy and applying ten-fold cross-validation to report scores.

Explore logistic regression as a binary classifier for cancer versus not cancer, using the logistic function to separate two classes and predict cancer probability.

Develop a logistic regression model for breast cancer classification by loading the dataset, preparing X and y, and performing a train-test split.

Explore evaluating a logistic regression classifier using confusion matrices, classification reports, and ROC curves to assess precision, recall, and AUC on breast cancer data.

Develop your first multiple linear regression model using the linear model package, train-test split, apply inverse transform to predictions, and compare with actual values for evaluation metrics.

Explore evaluation metrics for regression, including mean absolute error, mean squared error, RMSE, absolute percentage error (MAPE), and R-squared, using actual versus predicted values and residuals to assess model performance.

Explore polynomial regression concepts, showing how one feature with powers creates a curve that fits data better than simple linear or multiple regression.

Develop a polynomial linear regression model in Python using scikit-learn's PolynomialFeatures to transform the Boston dataset's number of rooms feature, apply a train-test split, fit, and evaluate with r-squared.

Explore random forest regression as an ensemble learning method that combines multiple decision trees to produce robust regression predictions, using parameters like the number of trees and maximum depth.

Understand SVR using an epsilon tube to ignore inside errors, minimize outside errors via slack variable C, and apply kernels: polynomial, Gaussian (RBF), sigmoid, and hyperbolic tangent in Python.

Develop a support vector regression model, preprocess and scale data, train and predict, inverse-transform results, and evaluate with R^2, RMSE, and MAE, comparing with random forest and MLR.

Evaluate a k-means clustering model using inertia and the elbow method to choose the optimal number of clusters, and implement your first clustering model in Python.

Build a first k-means clustering model on the iris dataset without labels, using sepal length, sepal width, petal length, and petal width to group 150 samples into three species.

Explore evaluating k-means models by measuring inertia across varying cluster counts, plotting inertia versus k, and using the elbow method to identify three as the optimal number of clusters.

Explore hierarchical clustering as a bottom-up approach, using the linkage methods—complete, single, and average—to build a tree-like graph that groups samples by distance.

Explore hyperparameters and model tuning to boost performance across classification, regression, and clustering. Use for loops and grid search to optimize kernel choices and the number of clusters.

Learn how to tune a support vector regression model with grid search, testing kernels like rbf and linear, gamma values, and mean squared error scoring, after normalizing the data.

Explore overfitting and underfeeding, with examples of a model that trains perfectly but fails on the test set, and a simpler model that generalizes better and yields higher test accuracy.