Machine Learning Bootcamp: Build ML models using GenAI

Install Python via Anaconda distribution to get Python, Jupyter Notebook, data science libraries, and VSCode, plus a brief Python crash course using the 3.7 version for mac or Windows.

Learn three ways to open Jupyter Notebook—from Anaconda Navigator, Anaconda Prompt, or Command Prompt—and manage your default directory and notebook files.

Celebrate reaching this milestone in the machine learning bootcamp using GenAI, as you join the top 50% of learners and access Q&A, AI assistant, subtitles, offline viewing, and your certificate.

Explore how to create and run code in Jupyter notebooks using Anaconda Navigator, including code, markdown, and raw cells, with handy shortcuts like Ctrl+Enter and Alt+Enter.

Learn Python basics by performing arithmetic operations like addition, subtraction, multiplication, division, exponentiation, and modulus with the board mass rule and parentheses. Assign values to variables and use comparison operators.

Master strings in Python basics: quote with double or single quotes, assign to variables, and print with or without format. Learn Len, replace, indexing and slicing, and type checks.

Explore Python basics with lists, tuples, and dictionaries, including list creation, indexing, slicing, mutability, and nested lists, plus key-value access and common methods.

Explore the NumPy library in Python, learning to create and manipulate arrays and matrices with ndarray, import conventions, shapes, dtypes, and key functions like arange and linspace.

Learn to manipulate data with pandas by importing csv data, inspecting with head and describe, and indexing with loc and iloc to access rows and set a custom index.

Learn Seaborn for data visualization: import seaborn as sns, plot histograms with distplot, adjust kde and color, and use the iris dataset for joint plot and pair plot.

Learn to classify data types for effective analysis by distinguishing qualitative (nominal and ordinal) from quantitative (discrete and continuous) data, with practical examples.

Explore descriptive and inferential statistics, including measures of center and dispersion, frequency distributions, charts such as bar graphs and histograms, and regression concepts.

Describe data using frequency distributions for qualitative and quantitative data, convert to bar charts and histograms, and explore descriptive statistics and the normal distribution.

Explore mean, median, mode, and mid-range as descriptive measures of center, including population versus sample mean, handling outliers, and applying mode to qualitative data.

Explore measures of dispersion: range, standard deviation, and variance, and learn how outliers affect range and how variance is the square of the standard deviation.

Learn how machine learning uses past data to optimize performance, with supervised and unsupervised learning, classification and regression, parametric and non-parametric models, and real-world applications.

Formulate business problems as statistical problems, identify dependent and independent variables, tidy data, preprocess data, split into train and test sets, train and validate the model, and deploy predictions.

Define the business context and key factors affecting the variables of interest to identify data needs, and gather data via primary or secondary research.

Identify data needed from business knowledge, request internal and external data, and perform quality checks. Use cart abandonment insights: source channels, cart steps, and cart value to guide data preparation.

Assemble a housing dataset with 506 observations and 19 columns, where price is the dependent variable and factors are independent; create a data dictionary with variable definitions and primary key.

Import data in Python with pandas and read_csv in a Jupyter notebook, guided by ChatGPT prompts; set file paths, create a dataframe, and view the first rows with df.head.

Inspect descriptive statistics for each variable, including mean, median, mode, range, quartiles, and counts for categorical data, and understand how the extended data dictionary supports univariate analysis and detects outliers.

Explores univariate analysis in Python by using df.describe, identifying categorical and numerical columns, and visualizing distributions with histograms and count plots.

Identify outliers with box plots, scatter plots, and histograms, then cap at 99th percentile three times the value and lower percentile, or apply the sigma approach to reduce variance.

Identify and cap outliers in a python workflow using box plots and the 99th percentile, demonstrating how to compute upper limits with numpy and apply capping across numerical columns.

Learn missing value imputation for machine learning: decide when to drop rows, impute with mean, median, or mode, or use segment or neighboring means guided by business knowledge.

Learn missing value imputation in Python by replacing numerical nulls with median and creating a none category for water body, updating the data frame in place.

Identify seasonality in time-based data as recurring patterns driven by weather and holidays. Multiply by the seasonal correction factor to normalize data and improve model fit.

Explore bivariate analysis with scatter plots and correlation matrices to identify linear relationships, handle multicollinearity, and transform variables for regression.

Transform variables to normalize data and simplify features by averaging distance columns, dropping bustier, treating missing values, capping outliers, and applying square root or log transforms to skewed crime rate.

Apply univariate and bivariate analysis to prune unusable variables, removing constants like bus terminal and low-fill or sensitive data, and rely on business knowledge and scatterplots to iteratively select features.

Learn to convert categorical variables into dummy variables for regression using 0/1 coding, with n minus one dummies for nominal data, and interpret the results.

Encode categorical features with dummy variables using pandas get_dummies to convert airport and water body into numeric booleans, aligning text data for machine learning input and reducing dimensionality.

explains correlation analysis with scatter plots and a correlation matrix, defines positive, negative, and zero correlations, notes the -1 to 1 correlation coefficient, and discusses multicollinearity, causation, and variable selection.

Learn to build a correlation matrix to reduce data dimensionality by identifying and removing redundant features, using a heat map to compare price with crime rate, air quality, and parks.

Learn linear regression as a simple, supervised learning tool and a foundation for machine learning, using least squares to predict house prices and estimate variable effects.

Learn simple linear regression for predicting house price from rooms, estimating beta zero and beta one (intercept and slope) via least squares on 506 observations, and interpret residuals and predictions.

Assess the accuracy of regression coefficients from a sample by comparing sample and population lines, using standard errors and confidence intervals to test if house price relates to room count.

Assess model accuracy by examining residual standard error and r square to gauge how much of the variability in Y the model explains, with adjusted r squared accounting for predictors.

Perform simple linear regression in Python to predict price from room number using OLS and a linear regression model, including coefficients, R-squared, and regression plots.

Explore multiple linear regression with 16 predictors, learn how beta coefficients reflect variable effects on price, and interpret the residual sum of squares, R-squared, and p-values for model assessment.

Use the f statistic to test if the model's predictors collectively relate to the response, adjusting for the number of variables, and then inspect p-values and t-values for individual predictors.

Interpret the effects of categorical variables by converting to dummies (airport yes; water body lake/river) and interpreting betas and p-values; airport increases price by 1.13, water body effects are inconclusive.

Master multiple linear regression in Python with ChatGPT assisted workflows, using statsmodels and sklearn to predict price from multiple independent variables, with model fitting, intercept, and coefficient interpretation.

Compare training and test mean squared error to assess model performance on unseen data, and apply validation set, leave-one-out, or k-fold cross validation to select the best model.

Explore the bias variance trade-off in machine learning, showing how model flexibility affects variance and bias, and how to minimize the total test error amid overfitting concerns.

Demonstrates ChatGPT assisted Python code to split data with train_test_split, fit a linear regression model, and compute train and test r-squared to gauge performance and overfitting.

Explore shrinkage methods for all predictors, focusing on ridge regression and lasso, where a lambda penalty shrinks coefficients toward zero to reduce variance and improve interpretability.

Demonstrates ridge and lasso regression in Python by standardizing data with StandardScaler, tuning alpha via validation curves, and evaluating train and test R-squared to select the best lambda.

Learn about heteroscedasticity, where error variance grows with y, shown by funnel-shaped residuals. Transform y with log or sqrt to stabilize variance and achieve near-constant residual variance.

Explore three classification models—logistic regression, KNN, and LDA—and predict property sold within three months using price as a feature in a pre-processed 506-transaction dataset.

Load classification data in Python with ChatGPT assistance in a Jupyter notebook, ensure clean, ready-to-use features like price, residential area, air quality, and room number, and predict sold status.

Explore two business questions: prediction and inference, using house data to predict whether a listing sells within three months and to estimate each independent variable's effect on response, via classifiers.

Explain why linear regression cannot be used for classification, detailing issues with multi-level responses, probability interpretation, and outlier sensitivity, and introduce logistic regression as an alternative.

Learn to classify with logistic regression by modeling probabilities with the sigmoid function, set decision boundaries, handle outliers, and estimate coefficients via maximum likelihood on credit default data.

Learn to build a simple logistic regression in Python with sklearn and statsmodels, guided by chat-assisted code, and interpret the intercept and price coefficient to explain sold outcomes.

Shows how simple logistic regression uses beta zero and beta one to estimate y=1 probability, and how standard error, z value, and p value test price impact against 5% threshold.

Explore logistic regression with multiple predictors, estimating beta coefficients by maximum likelihood to compute probabilities and classify with a 0.5 threshold, and consider multi-class extensions via linear discriminant analysis.

Explore multiple logistic regression in python with ChatGPT assisted guidance, fitting a model on multiple variables, comparing sklearn and statsmodels, examining coefficients, intercept, standard errors, and r-squared.

Explore the confusion matrix to evaluate how well predictions match true values, and learn about type I and type II errors (false positives and false negatives) and adjusting decision boundary.

Demonstrates logistic regression probabilities for house sold versus not sold, converts them with thresholds into binary predictions, and uses a confusion matrix to tune true and false positives and negatives.

Compare confusion matrix terms: true negative, false positive, false negative, and true positive to evaluate classifier performance. Use precision, sensitivity, specificity, and ROC AUC to compare models.

Evaluate model performance in Python using confusion matrix, accuracy, precision, recall, F1, and ROC AUC. Understand how true positives and negatives, false positives, and false negatives shape these metrics.

Explore linear discriminant analysis, its Bayes-based conditional probabilities, and its use for multi-class problems, comparing it with logistic regression and quadratic discriminant analysis using a practical example.

Train an lda model in Python using sklearn, generate predictions, and visualize the confusion matrix and classification report to compare with logistic regression.

Learn how to split classification data into train and test sets in Python using scikit-learn's train_test_split, define X and y, set random_state, and verify shapes (404x16 train, 102x16 test).

Explore the intuition behind the k nearest neighbors classifier, a non-parametric Bayes-inspired method. Learn how to choose an optimal k, understand boundary behavior, and standardize variables for knn.

Build a k-nearest neighbors model in python by standardizing features with StandardScaler, training on train/test data, and evaluating with a confusion matrix, classification report, and accuracy score.

Explore k-NN in Python using grid search cross-validation to test k values from 1 to 10, identify the best estimator (k=4), and evaluate with accuracy and a confusion matrix.