Statistics for AI Data Science and Business Analysis - 2026

Explore exploratory data analysis (eda) to uncover patterns and outliers using descriptive statistics, summary statistics, and visualization, illustrated via an Ask.com dataset case study.

Define data analysis and explore exploratory data analysis (eda) to reveal patterns in datasets. Classify eda into univariate, bivariate, and multivariate approaches using graphical and non graphical techniques.

Explore univariate non graphical exploratory data analysis (eda), using measures of center, dispersion, position, and shape, and examine outliers' impact on mean, std, and model assumptions like regression and anova.

Explore univariate graphical eda with hands-on visualization of single-column data using pie charts, histograms, line graphs, box plots, and dot plots to reveal distribution and outliers.

Explore multivariate non graphical EDA using covariance and correlation, examine association and confounding variables, and apply cross tabulation and a correlation matrix with hands-on data analysis.

Explore multivariate data using graphical methods, including scatterplots, heatmaps, and scatterplot matrices (pair plots) in seaborn, to reveal relationships, strength, and distributions across variables with hue by gender.

Identify data types, perform univariate, bivariate, and multivariate analyses, and address missing values and outliers to conduct structured exploratory data analysis.

Learn graphical exploratory data analysis techniques by variable type, using pie and bar charts for categorical data, histograms and line plots for univariate data, and scatterplots and heatmaps for relationships.

Perform hands-on exploratory data analysis on the Titanic data set in google Colab, covering univariate, bivariate, and multivariate analyses with visualizations of survival by gender, class, and embarkation.

Explore the Boston crime dataset via exploratory data analysis, visualizing serious crimes by year and by day of the week, using incident, offense, and location details.

Load the Boston crime dataset in Google Colab from Dropbox and filter to 2018–2019. Convert day of week to an ordered category and use count plots to compare offenses.

Analyze the PUBG game dataset to map player skills, visualize skill counts, and explore correlations between walking distance and win rate, healing and boosting, and vehicle destruction and win rate.

Analyze PUBG game data in Google Colab, compute average kills, visualize kill distributions, and reveal top correlates with win place percentage using heatmaps.

Explore FIFA player data with age, potential, spring speed, and earnings to visualize how age affects potential and spring speed, compare overall scores by nationality and club.

Explore the FIFA game analysis dataset to study how age influences potential and sprint speed, using joint plots and scatter plots to link ball control and dribbling.

Download and unzip the COVID-19 dataset in Google Colab, load with pandas and numpy, and analyze the maximum infection rate with happiness data via heatmaps and scatter plots.

Explore probability fundamentals, its use in data science and business risk, and quantify events with sample space, distributions, and simple coin-toss examples.

Explore key probability terminologies, including random experiments, sample spaces, and events; distinguish discrete versus continuous random variables with practical examples and outcome mapping to real numbers.

Learn the basic rules of probability—zero and one bounds, the complement, and a sample space that sums to one—through ball and coin examples, and introduces marginal and joint probability concepts.

Examine marginal and joint probabilities, plus conditional probability, through a Netflix subscriber case. Learn to convert frequency to probability distributions and read tables for gender and show preferences.

Explore disjoint (mutually exclusive) and non disjoint events, learn their unions, and apply probability formulas with examples like coin toss, pass/fail, and jack or red cards.

Explore dependent and independent events in probability, learn their key differences, and see practical examples that illustrate how one event affects or does not affect another.

Master the product rule of probability for independent and dependent events, including the specific and general multiplication rules, and learn to compute joint probabilities.

Explore a manifold bank use case to compute and compare probabilities using a credit card dataset. Practice calculating joint, marginal, and conditional probabilities in a hands-on Colab workflow.

Learn Bayes theorem and its conditional probability foundation, derive the formula P(B)=P(B|A)P(A)/P(B), and explore practical examples such as drawing aces and disease testing in data science.

Explore how Bayes theorem drives data science by testing hypotheses, enabling classifiers, and optimizing hyperparameters through Bayesian optimization for improved model accuracy.

Apply Bayes theorem to rainfall and flood data in Python, map June rainfall over 500 mm to binary, and compute joint, marginal, and conditional probabilities to assess flood risk.

Explore how a random variable maps sample space outcomes to real numbers, distinguishing discrete and continuous types and describing probabilities with pmf, cdf, and pdf.

Explore probability distributions for discrete and continuous random variables, including binomial and geometric distributions, and uniform, exponential, and normal distributions with mean and standard deviation.

Generate random numbers and visualize distributions for discrete and continuous variables, using Bernoulli, binomial, and geometric distributions alongside uniform, exponential, and normal distributions in Google Colab.

Explore the central limit theorem through a hands-on Google Colab workflow, show how the sampling distribution of the mean of independent variables becomes approximately normal with larger samples.

Explore the applications of probability distributions, including discrete types: binomial, multinomial, geometric, Poisson, and continuous types such as normal and uniform, used in forecasting, risk analysis, and customer service.

Learn to assess data normality using skewness, histograms, and Q-Q plots, then transform non-normal data to normal with log, reciprocal, square root, exponential, and Box-Cox methods.

Discover basic probability through practical examples, calculating winning chances from sample spaces and events, wheel spins and food orders, using the addition rule.

Analyze a car dataset to identify the most sold body type, calculate marginal and joint probabilities, and explore conditional probability with fuel types using cross-tab analysis.

Apply Bayes theorem to a casino data problem, using conditional probability to compute the probability a player was lying given they won.

This hands-on session loads the students start csv, analyzes the height column for normality, and uses z scores with the cdf to compute the probability between 175 and 190 cm.

Explore inferential statistics by linking population and sample, parameters and statistics, and applying hypothesis testing with null hypotheses and regression analysis, including linear, multiple linear, logistic, and ordinal regression.

Take a hands on look at population and sample by generating a synthetic dataset, computing population mean and variance, and sampling 500 observations to compare and discuss representation.

Explore the three forms of statistical inference: point estimation, interval estimation, and hypothesis testing, and learn how sample statistics estimate population parameters, including bias, confidence levels, and interval bounds.

Explore the concept of confidence intervals, including confidence level, margin of error, and how to estimate a population parameter using sample statistics and the standard formula.

Demonstrate computing the margin of error and confidence interval from a 200-sample of the Boston housing data, using the z critical value and the sample mean and standard deviation.

Explore hypothesis testing, including defining null and alternate hypotheses, choosing a test statistic, and interpreting p values and critical values to decide whether to reject the null.

Explore z test for comparing sample means to an assumed mean with known standard deviation, and apply a 40-observation passport processing example at 0.05 to test null and alternative hypotheses.

Perform a z-test on passport processing time data in Google Colab using pandas to compute the z statistic and p-value, interpret the null hypothesis, and preview the one-sample t-test.

Learn the one sample t-test for a hypothesis that an unknown mean differs from a specified value, using the t statistic, p-values, and alpha 0.05.

Apply a one-sample t-test to Tollywood movie production costs in a Colab workflow, testing against a 50 crore mean and interpreting the p-value and null versus alternative hypotheses.

Learn to perform a two-sample t-test comparing means of two independent normal populations with equal variances, using a height-change example with Horlicks and hypothesis testing at alpha 0.05.

Perform a two-sample t-test on independent groups to test the null and alternate hypotheses about health drink effects on children's height, using alpha 0.05 and p-value decision, in Google Colab.

learn how to apply a paired sample t-test to compare alcohol consumption before and after breakup, using a lawbreakers csv dataset with alpha 0.05 to assess significant change.

Conduct a hands-on paired sample t-test on pre-breakup and post-breakup alcohol consumption to assess impact of breakup. Interpret the p-value, noting p>0.05 indicates no significant relationship.

Apply a chi square goodness of fit nonparametric test to compare observed and expected food preferences of passengers, evaluating Captain Cook's beliefs about vegetarian, non-vegetarian, low calorie, and diabetic options.

Perform a chi squared goodness of fit test in Google Colab using Python, comparing observed and expected frequencies, interpreting the p-value to support the null hypothesis.

Explore the one-way ANOVA to compare mean detergent sales across 0%, 10%, and 20% discounts, illustrating how hypothesis testing assesses discount impact on quantity sold.

Load the dataset in pandas, visualize discount groups (0%, 10%, 20%) with seaborn, and perform a one-way ANOVA to test if discounts have a significant impact on quantity sold (p<0.05).

analyze car data to compute the population mean and standard deviation of peak rpm, draw a sample of 30 with pandas, and determine the 95% confidence interval for peak rpm.

Walk through project 11 in colab; load cars data with pandas; compute population and sample means and std dev; and calculate margin of error and confidence interval for peak rpm.

Analyze a five-column, 120-row blood pressure dataset with patient id, sex, age, bp pre, and bp post; perform a statistical test to obtain a p-value and decide null or alternative.

This hands-on project 12 solution demonstrates testing pre- and post-treatment blood pressure data in Colab, formulating null and alternative hypotheses, computing p-values, and rejecting null to show drug effect.

Analyze a four-column, 400-row student admissions dataset (GPA, rank, GI codes, I admit) and perform an independent sample d test to assess if GPA affects rank, with p-value and conclusion.

Download and load the dataset, test whether GPA affects student rank using a hypothesis test, and reject the null with p<0.05, confirming GPA's impact before moving to regression.

Explore regression concepts, distinguishing linear and logistic regression, and predict continuous versus categorical outcomes using simple and multiple linear models with independent and dependent variables, and evaluate errors.

Explore linear regression assumptions: linearity, no multicollinearity, and constant error variance with normally distributed residuals. Learn checks via VIF, correlation, Q-Q plot, one-hot encoding, and at least 20 records.

Explore the intuition of linear regression by modeling data with a straight line and minimizing error, using the normal equation to find the intercept and coefficients for the model.

Explore linear regression using the normal equation, derive theta with an augmented features matrix (including bias), and visualize the fit with hands-on Python code.

Apply linear regression with scikit-learn on an advertising dataset in Google Colab, performing simple and multiple regression and evaluating with r2 score and mean squared error and mean absolute error.

Master hands-on checks for linear regression assumptions, including linearity, additivity, residual behavior, and autocorrelation. Evaluate homoscedasticity, multicollinearity with VIF, normality of errors, using regression plots and Goldfeld-Quandt tests for validation.

Evaluate linear regression fit using the R-squared value, the coefficient of determination, and explained variance. Apply Python to compute R-squared and interpret values such as 0.8972 (89.72%).

Learn gradient descent to train a linear regression model by iteratively updating beta naught and beta one to minimize the mean squared error loss using a learning rate.

Apply gradient descent to a Python regression on advertising data, standardizing TV, radio, and newspaper features to predict sales, and update parameters using mean squared error.

This hands-on lecture implements gradient descent from scratch in Python, using Google Colab, with initialization, update, prediction, and cost functions to minimize loss.

Apply linear regression with the stats model library to predict the value from rooms and distance in a three-column dataset, loaded into a Google Colab notebook.

Apply linear regression with statsmodels on the Boston housing data to predict home values from rooms and distance, using simple and multiple regression and interpretation of the regression table.

Load the salary dataset with year of experience and salary, perform null-value checks, split data, and build a simple linear regression to predict salary from experience.

Predict salaries from years of experience with a simple linear regression without library, using a 10% test split; evaluate with r2 score around 0.89 and mse 48.5 million.

Analyze a house price prediction project with the dataset and median house value as target, applying multiple linear regression, loading data, checking shape, nulls, splitting, training, evaluating, and extracting coefficients.

in project 16, predict median house value from California housing data in Google Colab, split data, and train a regression model evaluated with r2 score and mean absolute error.

Explore medical cost prediction by building a linear regression model on a dataset with age, sex, bmi, children, smoker, and region; preprocess categoricals, train, predict, and evaluate on test data.

Preprocess the medical cost dataset with one hot encoding, fit a linear regression model, and report r-squared 0.766 and mean squared error 32275850.

Analyze a dataset (r&d spend, administration, marketing, state, profit) by loading, checking shape and nulls, encoding state, applying linear regression with scaling and cross-validation, and plotting predicted vs actual profit.

Explore project 18: company profit prediction with a hands-on Google Colab workflow, using label encoding, train-test split, linear regression with k-fold cross-validation, and RMSE and R2 evaluation against R&D spend.

Logistic regression is a supervised classification model that outputs probabilities between 0 and 1 via a sigmoid curve, applying a threshold to assign classes from continuous or categorical predictors.

Explore logistic regression with the sigmoid function applied to a linear model, showing how weights and bias shape the 0 to 1 probability curve for classification.

Explore the assumptions of logistic regression: binary outcomes, linear relation to log odds, minimal multicollinearity, and adequate sample size, with multiclass tasks via multiple binary classifiers.

Implement logistic regression from scratch using gradient descent, exploring the sigmoid function, binary cross-entropy loss, and plotting decision boundaries on linear and nonlinear data with accuracy evaluation.

Explore the diabetes prediction project by loading the numerical dataset, displaying and training, predicting, and evaluating a model against outcome with features like pregnancies, glucose, blood pressure, bmi, and age.

Explore diabetes prediction using the Pima Indians Diabetes database, preprocess data with train-test split and normalization, and apply logistic regression to achieve about 76.6% accuracy.

Explore heart disease prediction with logistic regression on a dataset including sex, age, smoking, hypertension, diabetes, cholesterol, and stroke history; preprocess, train, and evaluate the model.

Explore heart disease prediction by downloading dataset from Google Drive, cleaning by dropping education and missing values, performing 80-20 split, training a model, and evaluating accuracy and roc auc score.

Explore the Titanic survival dataset and apply logistic regression using passenger class, sex, age, siblings/spouses aboard, and parents/children aboard; prepare data and evaluate on test data.

Develop a Titanic survival predictor by cleaning the data, dropping cabin and passenger id, encoding features, and training a logistic regression model to achieve 80% accuracy and auc near 0.78.

Explore project 22 with the nursery student dataset, including columns like parent's occupation, housing, and health conditions, to classify nursery school applications using the idea tree model.

Learn to solve project 22 on the nurseries to rent dataset using Google Colab, from loading a headerless csv to one-hot encoding, logistic regression, and 92.43% accuracy.

Explore resampling techniques for machine learning, including cross-validation (k-fold and leave-one-out) and bootstrapping, to assess model performance and reduce bias in training data.

Explore cross validation techniques to estimate model performance and generalization. Implement holdout, k-fold, stratified k-fold, leave-p-out, Monte Carlo (shuffle split), and time series cross validation with practical demonstrations.

Explore flight price prediction by loading the dataset, surveying columns like airline, journey date, route, times, and price. Train a regression model to predict prices and evaluate performance.

Load the flight price prediction dataset, preprocess dates and times, apply one-hot encoding, and train a random forest model to predict prices.

Explore modeling concrete compressive strength with a data-driven approach, training a predictive model on cement, slag, fly ash, coarse and fine aggregates to optimize mix design and reduce test time.

Demonstrates solving concrete strength prediction (project 24) with a UCI dataset in Google Colab, from data setup to evaluating linear regression, lasso, and elastic net with correlation analyses.

Explore the us baseball salary prediction project by analyzing a baseball dataset with at bats, hits, home runs, runs, and runs batted in; train and evaluate model on test data.

Explore how to perform model selection for a machine learning task, from data gathering and preprocessing to evaluating models like linear, logistic, SVM, tree-based methods, and neural networks.

Portable deployment prep of a beginner-friendly end-to-end machine learning project using iris data, preprocessing with standard scalar, train-test split, and a decision tree, saved with joblib.

Learn how to deploy a machine learning model as a Flask web app by loading a saved standard scaler and classifier with joblib, handling input, applying transformation, and returning predictions.

Learn to build and deploy a shareable data science web app with the Streamlit library, from writing a Python script to running it and deploying for free.

Explore Docker containers for data science and machine learning projects, focusing on packaging, portability, and lightweight deployment. Learn Docker concepts like images, containers, Dockerfile, Docker Hub, and orchestration with Kubernetes.

Explore how generative AI drives art, music, writing, and personalized user experiences. Learn prompt engineering with ChatGPT to craft precise prompts that steer large language models toward targeted content.

Understand how transformer architecture replaces sequential RNNs with parallel self-attention. Explore encoder-decoder models, multi-head attention, positional encoding, and residual connections that enable efficient sequence learning.

Learn practical mlops for data scientists and devops engineers, building end-to-end pipelines from experimentation to production use cases with hands-on code and cloud and docker workflows.

Discover how enterprise ml ops bridges data science, software engineering, and business stakeholders, automates the ml lifecycle with ci/cd, docker, kubernetes, mlflow, and cloud platforms.