ML & MLOps Masters 2026 - Build, Train, Evaluate, Deployment

Master Python installation and environment setup with options like python.org downloads, Anaconda Navigator, Jupyter Notebook, VS Code or Spyder, and Google Colab for practice.

Learn to use Google Colab for python practice, avoiding setup issues with ipynb notebooks. Create notebooks, run code with shift-enter, text blocks, and upload files while exploring lists and tuples.

Learn to debug without an instructor using ChatGPT and other AI tools to diagnose and fix common Python errors in data analytics and data science.

Explore Python, open source language that is object oriented and scripting, with easy learning, large standard library, cross platform support, and a strong community for data science and backend development.

Explore Python basics by understanding variables and keywords, learn installation options via python.org, Anaconda Navigator, or Colab, and explore types with print and type.

Master data types, operators, and operands in Python, including numeric, dictionaries, booleans, and sequences; explore type casting, input, string immutability, and operator precedence.

Learn Python data structures—lists, tuples, sets, dictionaries—with emphasis on lists mutability, indexing, slicing, and nested lists, plus extend, append, pop, remove, sort, and shallow copy.

Master tuples in Python: immutable sequences, unlike lists, defined with round brackets or comma notation. Learn indexing, slicing, concatenation, nesting, and practical uses, including length, access, and sorting workarounds.

Explore sets in Python as unordered collections of unique elements, learn to initialize with curly braces, use the in keyword for membership, and perform unions, intersections, differences, and symmetric differences.

Explore Python dictionaries as unordered key-value stores; learn to access, update, add, delete, sort, and understand immutable keys with values using keys(), values(), get(), and update() for real-world data.

Master Python loops and iterations, covering for and while loops, iterables (lists, strings, dictionaries, sets), iterators, and list comprehensions for practical examples from odd/even checks to dictionary iteration.

Master Python functions, from built-ins to user-defined and lambda forms, and apply map, filter, and reduce with practical examples like BMI calculations and even/odd checks.

Explore map, reduce, and filter in Python as functional programming paradigms, using lambda expressions to simplify area calculations and data transforms with practical examples like Celsius to Fahrenheit conversion.

Master file handling in Python by opening, reading, writing, and appending text files, including reading lines and closing files, with a nod to pandas read_csv for csv data.

Explore python control structures and operators, apply if-else decision making, and build simple programs such as even/odd checks, age-based rules, discounts, and range-based iterations.

Explore the fundamentals of object oriented programming in Python, including classes, objects, constructors, and instance and class variables. Learn about inheritance, polymorphism, encapsulation, and practical examples with rectangles and circles.

Explore NumPy, the numerical Python library for fast multi-dimensional ndarrays, including creating zeros, ones, full, identity, and random arrays, with indexing, shaping, and basic arithmetic using the np alias.

Explore pandas basics for Python data analysis: create and inspect dataframes, read files, use head, info, describe, select columns, filter with conditions, handle nulls, and perform joins and merges.

Understand data visualization in Python to turn data into insights with matplotlib, seaborn, and plotly. Discover chart types, correlations, patterns, and how visual analytics support data-driven decisions.

Learn to plot diverse charts with matplotlib in Python, using pandas data frames, from histograms and bar charts to pie charts and scatter plots, with code practices.

Explore seaborn to visualize data with kd plots, dist plots, pair plots, and heat maps; install and import, load iris dataset, and compare with matplotlib for clearer insights.

Explore the basics of statistics and its role in data analytics and data science, using charts and scatter plots to compare age and salary, and distinguish population from sample.

Explore descriptive and inferential statistics and identify data types, including categorical and numerical (qualitative and quantitative) data.

Explore descriptive statistics in the ML & MLOps masters 2026 course, focusing on central tendency and variability, including mean, median, mode, variance, standard deviation, and frequency distribution for data sets.

Explore inferential statistics, including estimating population parameters and hypothesis testing, by drawing inferences about a population from sample data, illustrated with a country’s average age example.

Explore qualitative (categorical) and quantitative (numerical) data, including nominal and ordinal types, with examples like gender and location, and visualize them using pie and bar charts.

Explore quantitative data, defined as numerical values, and distinguish discrete data from continuous data with examples like class size and height, including plotting approaches.

Explore population versus samples and learn how data from a subset estimates the whole, using random and nonprobability sampling techniques.

Define population as the entire group and a sample as a subset; show why researchers use samples to estimate the population mean from data chunks.

Use sampling to analyze large populations when full data collection is impractical, inferring population metrics from a representative sample and speeding analysis while reducing data collection errors.

Explore probability and non-probability sampling, learn how to create samples from a population, and examine techniques like simple random, stratified, cluster, systematic, purposive, voluntary response, snowball, and convenience-based methods.

Explore cluster random sampling, dividing the population into clusters and randomly selecting some, ensuring clusters reflect the population and cover it, with less statistical certainty than simple random sampling.

Explore probability sampling techniques such as random, systematic, stratified, and cluster, and compare with non-probability methods like convenient, purposive, voluntary response, and snowball sampling.

Explore non-probability sampling techniques, including convenience, purposive, voluntary response, and snowball sampling, and contrast them with probability-based sampling to understand sample selection in research.

Learn population sampling, the use of a representative sample, and how population variance, sample variance, and standard deviation (sigma) are calculated from mu (mean).

Explore why sample variance uses n minus one and population variance uses n, via Bessel's correction. This bias correction yields a more unbiased variance estimate and impacts standard deviation.

Explore descriptive analytics and statistics by detailing central tendency measures—mean, median, and mode—and dispersion measures like range, interquartile range, variance, standard deviation, and mean deviation.

Examine measures of central tendency, including mean, median, and mode, to identify the center of data distributions with descriptive statistics guidance and examples of when to use each.

Compute the mean, or average, as the sum of observations divided by their count. Use mean to impute missing values and identify central tendency for discrete and continuous data.

Understand the median as the middle value in ordered data, or average of the two middle values for odd and even sets, and why it resists outliers in skewed data.

Identify the mode as the most frequent value in data, including unimodal, bimodal, and multimodal cases, and learn practical imputation strategies for categorical data.

Explore measures of dispersion to understand data spread around the mean. Learn range, variance, standard deviation, interquartile range, and the roles of Q1, Q3, mu, and n.

Identify the range as the difference between the highest and lowest values, a simple dispersion measure. Use it to highlight extreme values and guide the quality control chart.

Calculate the interquartile range to measure the middle 50% of data, using Q3 minus Q1 to assess dispersion and detect outliers.

Learn how variance and standard deviation measure dispersion around the mean, and how standard deviation is square root of variance, with population (n) and sample (n-1) using sigma notation.

Explore mean deviation as the average of absolute deviations from the mean, and note the median minimizes this value, illustrated by a 0.88 deviation.

Explore the fundamentals of probability, its theories and formulas, and why we use it. Learn the additional rule, independent events, cumulative probability, conditional probability, and Bayes theorem in upcoming videos.

Learn probability as the measure of how likely an event is, using favorable outcomes over total outcomes with coin toss and die examples, and explore trials, events, and sample space.

Explore the addition rule of probability for mutually exclusive and non mutually exclusive events. Apply P(A or B) = P(A) + P(B) − P(A ∩ B) with dice examples.

Explore independent events in probability, where one event does not affect another, and apply P(A and B) = P(A)P(B) with examples like rolling a die and tossing a coin.

Explore cumulative probability, defined as the probability that a random variable falls within a given range, illustrated with two-coin and two-die examples.

Learn conditional probability through practical examples: compute probability of B given A using intersection, with and without replacement, including urn and sports drinks scenarios.

Explore Bayes theorem as an extension of conditional probability, derive p(a|b) and p(b|a) from p(a), p(b), and p(a∩b), and solve example problems.

Apply Bayes theorem to compare the probability a black ball came from bag x versus bag y, using conditional probabilities and two-bag scenarios.

Explore probability distributions, from uniform to normal, and learn how to describe outcomes, apply scaling techniques, and convert to standard normal.

Explore uniform distribution, including discrete and continuous cases, with dice examples. Learn that discrete probability is 1/n, area represents probability, and the mean is (a+b)/2 with std dev sqrt((b-a)^2/2).

Learn the binomial distribution in statistics, a probability model for independent trials with fixed success probability. Use the formula nCx p^x (1-p)^{n-x} on coin-toss examples.

Master Poisson distribution, a discrete probability model for number of events in a time period, using lambda as the average rate. Include a call center example with formula P(X=x)=(lambda^x e^{-lambda})/x!.

Explore the normal distribution, also known as the Gaussian or bell curve, and apply the empirical rule to interpret data within mu plus minus one, two, and three sigma.

We estimate mu and compute the mean and standard deviation to assess normal distribution. We note that data rarely follow normal distribution and discuss skewness, kurtosis, and transformations.

Explore skewness and how it reveals data asymmetry, with right and left skew versus symmetric distributions. Learn logarithmic transformation, square root transformation, cube root transformation, and reciprocal to approach normality.

Kurtosis measures the thickness of data distribution and its deviation from normality, distinguishing mesokurtic and platykurtic shapes, handling skewness, and guiding transformations like log, square root, and cube root.

Explore normal and standard normal distributions and how z-scores measure deviation from the mean using mu and sigma. Learn feature scaling with standardization and normalization for machine learning readiness.

Calculate a z score using x, mu, and sigma to standardize a test score. Consult the z table to determine how the score ranks among peers.

Compute the z-score for 1.5 m in a class height distribution and use the z-table to find 4.75% of students are shorter than 1.5 m, illustrating normal distribution.

Explore covariance and correlation, and examine how these two statistics relate to each other in this module.

Learn how covariance reveals the directional relationship between two variables, with positive or negative covariance, using the formula and a stock example to illustrate means and deviations.

Explore correlation as how variables move together linearly, measured by Pearson's coefficient from -1 to 1, with negative and zero correlation, differentiating it from covariance and using heatmaps and python.

Compare covariance and correlation, noting how they signal direction and the magnitude of relationships in data. Covariance is unbounded, while correlation lies between -1 and 1.

Explore hypothesis testing, a method using sample data to evaluate two competing population statements—null and alternate hypotheses—using the normal distribution and z-scores to assess significance.

Compare one-tailed and two-tailed tests in hypothesis testing, explain rejection regions on the sampling distribution, and illustrate null versus alternate hypotheses with practical examples.

Explore the p value and significance level, and how null and alternative hypotheses guide statistical testing. Illustrate with a coin-toss example how observed results affect rejection decisions.

Explore the main statistical tests—t-test, z-test, ANOVA, chi-square, and correlation—and map them to categorical and numerical data with visualizations like bar charts, box plots, and scatter plots.

Demonstrates performing t tests—one-sample and two-sample—using female and male ages to test hypotheses with p-values and 0.05 significance, via data analysis steps.

Use the z test to compare two sample means and decide at a 0.05 significance level whether they differ, with null vs. alternative hypotheses and excel data analysis toolpak steps.

Apply chi square test to categorical data by comparing observed and expected counts via pivot table, and test the null hypothesis of no association between gender and location with p-value.

Explore the ANOVA (analysis of variance) parametric test to determine if means differ across three dose groups, covering sum of squares, df, F statistic, and p-values.

Explore how correlation and covariance reveal linear relationships between variables, and learn to compute and interpret correlation values using scatter plots and Excel data analysis.

Explore the data analytics and data science process from data collection to model building, highlighting data processing, cleaning, iterative EDA, and visualization that drive insights and decisions.

Explore exploratory data analysis (EDA) as a visual, data-driven approach to understand data characteristics, identify errors and anomalies, and build quick baseline models for business insights.

Learn how data visualization presents data graphically to reveal insights for end customers and management, using bar, pie, line, area, scatter, and other plots to show distributions and trends.

Explore exploratory data analysis fundamentals, from data cleaning and preparation to univariate and bivariate analysis, visualization, and Python-based use cases that reveal insights.

Explore the steps of data sourcing in EDA, differentiating public and private data, and how to obtain access from database administrators for varied company contexts.

Learn how data cleaning turns raw data into high-quality data for analysis and modeling, covering missing values, imputation or deletion, feature scaling, outliers, and invalid data handling.

Master techniques to handle missing values during data analysis and data cleaning, including deleting rows or columns, imputing with mean, median, or mode, and exploring algorithmic and advanced imputation methods.

Explore practical techniques to handle missing values in a churn modeling dataset using pandas and numpy. Impute with mean, median, or mode, apply forward or backward filling, or drop rows.

Learn feature scaling, a data cleaning technique that uses standardization (z-score) for normally distributed data and normalization to 0–1 for data that is not normally distributed to prepare predictive models.

Explains standardization with z scores based on mu and sigma, centering data at zero and typically ranging from -3 to 3, using an income example.

Apply normalization to scale features between 0 and 1 using min-max calculations shown with income values. Learn how feature scaling reduces bias before Python implementations.

Explore feature scaling using normalization and standardization with scikit-learn in a Python workflow, demonstrating min-max scaling and standard scaler on age and tenure after imputing missing values.

Identify outliers with box plots, histograms, scatter plots, and normal distribution cues; decide to remove, keep, or replace them using quantile or interquartile range, and apply robust models.

Identify and treat outliers using the three-sigma rule, assess skewness and normality with transformations, and visualize with box plots, percentiles, and IQR concepts.

Identify and handle invalid values across dates, formats, and logical ranges; convert types, correct encoding, and remove structurally invalid data for robust data cleaning.

Explore qualitative and quantitative data types and how they describe population attributes. Identify nominal and ordinal categories, plus discrete and continuous values, with examples like gender, location, and temperature.

Explore the types of analysis, focusing on univariate analysis that uses a single variable to understand data and its characteristics, mainly for categorical data, with summary statistics for numerical columns.

Explore univariate analysis of numerical and categorical data, revealing distribution and statistics through summaries and visuals like box plots to identify churn patterns.

Explore bivariate analysis to measure relationships between two variables, using scatter plots for numerical pairs and charts for categorical combinations, including correlation and covariance illustrated with churn data.

Explore multivariate analysis, where more than two variables—categorical, numerical, or mixed—drive deeper insights during exploratory data analysis, with complexity increasing as you add features.

Explore numerical analysis of data, from single-variable summaries (mean, median, 25th/75th percentiles, min, max) to multivariate correlations, using scatter plots, histograms, kernel density estimation, and heat maps with pandas.

Explore univariate, bivariate, and numerical analysis of a telco churn dataset using Python, pandas, and seaborn, including age, balance, tenure, geography, gender, and correlation heat maps to derive insights.

Explore how derived metrics create new features from existing data using domain knowledge, binning, and encoding techniques to prepare categorical data for machine learning models.

Learn how feature binning converts continuous variables into categorical bins to reveal patterns, identify outliers, and handle missing values, using equal width and equal frequency methods.

Learn how to perform feature binning with age using pandas cut, defining bins and labels, handling missing values, and visualizing counts with a bar chart in a practical churn dataset.

Explore feature encoding and convert categorical variables to numerical formats to feed models in predictive analytics. Learn label encoding, one hot encoding, and dummy encoding, with examples and practical guidance.

Explore practical feature encoding techniques on a churn dataset, including missing-value handling, label encoding, one-hot and dummy encoding, target encoding, and hashing encoding.

Explore exploratory data analysis for telecom churn case study, from data cleaning and missing value imputation to univariate, bivariate, and multivariate analysis, leading to insights and predictive modeling.

Explore customer churn through practical data exploration, inspecting the dataset's shape and data types, noting total charges, missing values, and churn imbalance to guide modeling and storytelling.

Back up the data frame, convert total charges to numeric, and address missing values by dropping low-percentage nulls. Bin tenure for clearer insights and remove irrelevant columns like customer ID.

Perform univariate analysis on a cleaned telecom dataset using an automated loop to plot all features, revealing churn patterns by senior citizens, contract type, and payment method.

Explore numerical and bivariate analysis for churn prediction: convert churn to numeric, encode features with dummies, and analyze monthly charges, total charges, tenure, and their correlations.

Explore bivariate analysis for churn data by comparing partner and gender across churners and non churners, reveal insights, and emphasize end to end eda reporting.

Build an end-to-end EDA report for churn analysis, detailing business understanding, data understanding, and findings. Present graphs and insights from the churn study, including final thoughts on churn drivers.

Install MySQL by setting up MySQL Workbench and MySQL Installer, create a root username and password, and connect to the local database; learn about SQL, databases, and DBMS basics.

Compare file server and client server architectures to see how multi-user access, in-memory edits, file locking, and versioning influence data integrity in production databases.

Explore sql, the structured query language for relational databases, covering reading and writing data with commands like select, insert, update, delete, create, and drop, plus data types.

Define constraints in SQL, including unique, not null, and primary key rules. Create and manage tables in MySQL Workbench, applying primary keys and foreign keys.

Explore table basics and data definition language (ddl) essentials, including create, alter, and drop table commands, primary keys and constraints, data types, and practical sql table examples.

Explore data query language (dql) fundamentals with select queries, filtering using where, like, and in, and learn to create tables, insert data, and apply aliases and counts for clear results.

Explore data manipulation language (DML) basics in SQL, learn to insert, update, delete, and select data, create and query tables, and practice with employee datasets.

Learn how SQL joins connect data across tables using a common column, covering inner, left, right, and full (outer) joins, self joins, and cross joins with practical examples.

Master import and export of data in SQL databases using manual methods, Workbench wizards, and command line, with guidance for large datasets and chunked export strategies.

Explore aggregation functions in SQL, including group by, count, minimum, maximum, and average, with practical examples on gender and contract distributions.

Learn how to manipulate SQL string data using concat, trim, substr, mid, upper and lower, and character length to produce readable text and well-structured names.

Explore SQL date and time functions, from date diff and date format to date add and sub date, with practical examples on a transaction details table.

Learn how regular expressions replace the like operator in SQL queries, using regex patterns to match starting characters and ranges with telco churn data examples.

Master nested queries and subqueries in SQL, learning inner and outer queries and their execution order, with practical examples from offices, employees, and telco churn.

Explore how views function as virtual, read-only tables that hide complexity, enable access control, and boost performance; learn about materialized views and practical creation examples.

Learn stored procedures in SQL to create reusable, parameterized code you call to fetch top n players, use in and out parameters, update goals, and count records in MySQL.

Explore window functions in SQL that compute values across related rows using over and partition by, featuring row_number, rank, and first_value in sales and product examples.

Connect SQL databases with Python to extract data from MySQL, load into a Pandas data frame, and perform EDA in notebooks. Connect SQL databases from Power BI.

Explore the fundamentals of machine learning, its scope and use cases, then master core techniques—regression, classification, clustering, association rule learning, time series analysis—and key topics like feature engineering and deployment.

Machine learning is a subset of artificial intelligence that learns from data. Data science blends AI, ML, deep learning, statistics, and data exploration to study data.

Examine the four types of machine learning—supervised, unsupervised, reinforcement, and semi-supervised—along with labeled data, training and testing, and use cases like fraud detection, classification, and regression.

Explore key use cases of machine learning and deep learning across supervised, unsupervised, and reinforcement learning, with examples like recommendation systems, natural language processing, and facial detection.

Explore healthcare use cases powered by machine learning and deep learning, including imaging, disease detection, personalized treatment, fraud prevention, and drug discovery, applying classification, regression, and recommendation systems.

Identify features and their role as x variables in supervised learning, and distinguish y variables as the targets. Use churn examples and Vodafone data to illustrate predictors and model building.

Master train-test split concepts using the 80-20 split, separating training and testing data with X and Y variables, and evaluate models using accuracy for classification and error metrics for regression.

Master feature scaling as a data cleaning step that standardizes or normalizes features to a common scale for predictive modeling and sometimes EDA, using standardization (z-score) or normalization (0–1 range).

Learn standardization by converting data to z-scores using the mean and standard deviation, with Python and scikit-learn, illustrating mu, sigma, and a -3 to 3 z-score range for future normalization.

See how normalization scales diverse features to a 0 to 1 range using (x minus min) over (max minus min), reducing bias from income and unit differences in machine learning.

Explore feature encoding for predictive analytics, converting categorical x variables into numerical y for models using label encoding, one hot encoding, and dummy encoding.

Practicals cover preparing the churn modeling data set, handling missing values, dropping unnecessary columns, and applying feature encoding techniques like label encoding, one-hot and dummy encoding using pandas get_dummies.

Learn regression as a supervised predictive technique that predicts a continuous target from input features, covering linear, multiple linear, lasso, ridge, and polynomial methods with real-world examples.

Explore regression metrics like MAE, MSE, RMSE, R2, and adjusted R2 to evaluate performance while covering data cleaning, train-test split, feature scaling, and model comparison.

Apply practical regression metrics such as MAE, MSE, RMSE, R2 score, and adjusted R2 to evaluate models on a train-test split using sklearn.

Learn simple linear regression to model how area relates to price, fit the best line, and predict prices using y = mx + c.

Extend simple regression to multiple features, forming a plane that predicts price using y equals beta zero plus beta one x1 plus beta two x2, and interpret the beta weights.

Implement linear regression in practice with simple and multiple models on the 50 startups dataset, using one-hot encoding, feature scaling, and train-test splits to predict profit.

Create a 3D two-feature dataset using make regression, split into train and test sets, fit a linear regression model, and evaluate with MAE, MSE, RMSE, and R2 score.

Learn polynomial regression as a form of linear regression, using polynomial features and degrees to capture non-linear relationships, while managing underfitting and overfitting.

Apply polynomial regression practically by building models with polynomial features and comparing degrees from two to ten. Split, train, and evaluate using R2 scores while visualizing results.

Understand the bias-variance trade-off through intuitive examples. Distinguish overfitting and underfitting and identify the sweet spot for model performance, with notes on regularization, bagging, and boosting.

Explore ridge regression as a regularization technique for linear regression, penalizing large slopes with lambda to reduce overfitting and improve generalization, alongside basic cost function updates.

Explore lasso regression and its L1 regularization, which combats overfitting and enables feature selection by adding a lambda times the absolute value of coefficients to the cost function, unlike ridge.

Explore practical ridge and lasso regression by building a linear model with the Boston housing dataset, comparing R2 scores and tuning alpha via manual and grid/randomized search, plus feature selection.

Explore how supervised learning uses labeled training data to predict category membership, with spam filtering and fraud detection as key examples.

Explore binary and multi-class classification in machine learning, with fraud detection and churn examples, and learn how models handle two versus multiple classes.

Explore log loss as a key classification metric, understand prediction probability, thresholds, and model evaluation for binary problems like fraud detection and spam filtering.

Master the confusion matrix, its true/false positives and negatives, and related metrics such as recall, precision, F1 score, and accuracy, with fraud detection examples.

Explore area under the ROC curve (AUC ROC) as a key binary classifier metric, measuring sensitivity and specificity across thresholds.

Learn how to use the classification report in scikit-learn to evaluate precision, recall, F1 score, and support, including handling imbalanced data and minority classes.

Master k nearest neighbors, a supervised classifier using Euclidean distance to vote among k neighbors for churn or active. See binary and multi-class cases with age and salary.

Explore a knn classifier using age and gender to predict cricket or football preference, with dummy encoding and euclidean distance. Neighbor voting determines the predicted class for a new player.

Implement the first KNN classification model by preparing data, handling missing values and numeric conversion, applying one-hot encoding, performing train-test split, and standardizing features.

Learn to implement a kNN classifier with scikit-learn, perform data preparation, encoding, scaling, train-test split, and evaluate accuracy for churn prediction, plus new data and probability predictions.

Understand how a decision tree uses a root node, internal nodes, leaf nodes, and branches with features like age, gender, or smoker status to predict disease.

Explore entropy-based decision trees for classification, identify the root node using information gain, and implement the model in Python with an illustrative profit or loss example.

Explore the gini-based criterion for decision trees, compare it with entropy, and build a root node from age, competition, and type to predict profit.

Learn to implement a decision tree classifier on a churn dataset after data cleaning and dummy encoding. Compare its 72% accuracy to KNN and discuss hyperparameter tuning.

Explore decision tree classifiers, compare Gini and entropy criteria, visualize root, internal, and leaf nodes with plot_tree or Graphviz, and tune max depth for churn data.

Learn how random forest, an ensemble classification method, builds multiple decision trees with random data and feature selections, and uses voting to predict churn more accurately than individual trees.

Implement a random forest classifier using sklearn ensemble and compare its 79% accuracy to decision trees and knn. Explore estimator counts between 100 and 500 and prepare for hyperparameter optimization.

Explore naive Bayes classifiers built on Bayes theorem, assuming independent features, and apply them to a play vs. not play example, with comparisons to other classifiers in later sessions.

Discover how support vector machines classify two classes by maximizing the margin between parallel hyperplanes, identify support vectors, and extend to non-linear data by moving to higher dimensions.

Explore the math behind support vector machines, linking lines, planes, and hyperplanes across dimensions to the generalized equation w^T x + w0 = 0 and interpreting intercepts.

Explore logistic regression as a supervised binary classifier that uses a sigmoid function to estimate p(y=1) from features, contrasts with linear regression, and illustrates with healthcare and fraud examples.

Compare practical classifiers—KNN, decision trees, random forest, Naive Bayes, SVM, and logistic regression—using a churn dataset. Includes end-to-end workflow with EDA, encoding, scaling, and evaluation metrics like accuracy and recall.

Explore underfitting and overfitting in classification, compare training and testing performance, and examine imbalance issues through practical examples to improve model generalization.

Explore imbalanced data challenges in fraud and churn contexts, showing why accuracy is misleading and detailing resampling (upsampling, downsampling), class weights, algorithmic tweaks, and cost-sensitive approaches to improve minority-class performance.

Discover practical strategies for imbalanced data in classification, using oversampling and undersampling, smote variants, and threshold tuning to improve precision, recall, and F1 in real-world churn and fraud problems.

Explore ensemble learning, a group model approach that blends diverse algorithms—bagging, boosting, voting, and stacking—to boost accuracy, with examples like random forest and XGBoost.

Bagging blends bootstrapping and aggregation to turn low-bias, high-variance models into a stronger predictor via majority voting, as seen in random forest and bagged decision trees.

Compare bagging and random forest ensembles, focusing on how max features influence tree splits, visualization of estimators, and the role of data and feature randomness in model performance.

Explore bagging ensembles by implementing bagging classifiers with decision trees on synthetic data, compare accuracy with random forests, and study bootstrap sampling and key parameters.

Apply bagging to regression by averaging predictions from multiple regressors, including decision tree regressor and svr, and evaluate with r2 score, mae, and rmse.

Explore boosting as a sequential ensemble that strengthens weak learners by weighting misclassifications, turning them into a strong predictor, and cover gradient boosting, AdaBoost, CatBoost, and extreme gradient boosting.

Explore AdaBoost and how boosting combines weak learners, especially decision stumps, into a strong learner. Learn how the method updates sample weights to emphasize wrong predictions.

Explore gradient boosting as a sequential boosting method that optimizes the loss function with well-grown trees, differentiating it from AdaBoost.

Learn the difference between loss functions and cost functions with a linear regression example and how they guide model decisions; explore cross entropy, binary cross entropy, and XGBoost topics.

Learn cross entropy as a loss function for classification, contrasted with cost functions, through a weather example and the distinction between binary and multi-class (categorical) cross entropy.

Discover XGBoost, the extreme gradient boosting algorithm, a powerful ensemble method with parallel processing, optimized data structures, cache awareness, and gpu support for fast, accurate predictions.

Apply ensemble learning with random forest, AdaBoost, gradient boosting, and XGBoost to a churn dataset, encode and scale features, compare by accuracy, and plan hyperparameter optimization.

Explore clustering as an unsupervised learning technique that groups similar data into clusters using k-means and hierarchical methods for exploratory data analysis, segmentation, and fraud detection.

Learn k-means clustering, an unsupervised method that groups data by nearest centroids using euclidean distance, updates centers, and uses the elbow method with wcss to choose k.

Learn to perform k-means clustering with the elbow method to identify optimal clusters using spending score and annual income, visualize centroids, and extend to hierarchical clustering.

Explore hierarchical clustering, a bottom-up agglomerative method that builds a dendrogram to reveal a hierarchy of clusters. Compare it with k-means and note pros and cons for large data.

Explore hierarchical clustering in practice by building dendrograms with SciPy, choosing cluster numbers, and comparing agglomerative clustering to k-means using annual income and spending score.

Mean shift clustering is a non-parametric density-based algorithm that shifts points toward the region mean of data. It does not require predefining the number of clusters; bandwidth controls cluster count.

Explore the critical role of feature engineering and dimensionality reduction in preparing data, selecting features, and improving model performance using PCA, LDA, and related techniques.

Explore recursive feature elimination and sequential feature selector, comparing how RFE removes least important features and SFS adds or removes features to boost predictive performance.

Explore recursive feature elimination in practice, selecting five features from a 30-feature dataset, compare with a full-model baseline, and learn preprocessing steps like one-hot encoding and handling missing values.

Learn how recursive feature elimination and sequential feature selector optimize logistic regression by progressively pruning and selecting top features, achieving strong accuracy with fewer predictors.

Explore chi-square based feature selection, a statistical test for association between categorical features and the target, enabling significance testing, reduced dimensionality, improved interpretability, handling categorical data, and avoiding overfitting.

Learn chi square test based feature selection to identify the top five features related to the Y variable and compare logistic regression models with and without chi square selection.

Master principal component analysis to reduce high-dimensional data for feature engineering, visualization, and noise filtering; learn how to choose n components and retain 80% variance to combat overfitting.

Explore the practical application of principal component analysis to streamline features, assess explained variance, and upgrade a logistic regression model with PCA-transformed data.

Explore linear discriminant analysis (LDA) as a supervised dimensionality reduction that maximizes class separability and minimizes within-class scatter, unlike PCA.

Explore practical linear discriminant analysis alongside PCA, reuse PCA code for LDA, and compare performance, explained variance, and end-to-end modeling with kernel PCA and QDA.

Explore kernel PCA and quadratic discriminant analysis as nonlinear extensions of PCA and LDA for dimensionality reduction and flexible classification, using hit-and-trial evaluation on real data.

Explore kernel PCA and quadratic discriminant analysis in a hands-on session, compare with PCA and LDA, and discuss when non-linear feature engineering improves or matches linear methods.

Boost model accuracy by tuning hyperparameters across classifiers and algorithms, using manual and automated methods like grid search, random search, and bayesian optimization guided by a validation set.

Explore manual hyperparameter optimization, and demonstrate hit-and-trial tuning for a decision tree, while contrasting manual hpo with grid search, random search, and bayesian approaches, highlighting drawbacks.

Learn hyperparameter optimization with grid search and randomized search cv, comparing exhaustive and flexible exploration, and identify best hyperparameters like criterion, max depth, min sample split, and min samples leaf.

Practice session demonstrates building a breast cancer classification model with a random forest and manual hyperparameter optimization to improve accuracy, then compares grid and randomized search approaches.

Leverage randomized search CV to tune random forest hyperparameters, sampling 100 iterations from a random grid, using 3-fold cross-validation, and compare gains against a base model.

Explore grid search cross-validation and hyperparameter optimization for random forest, tuning bootstrap, max depth, max features, and min samples to achieve about 0.94% accuracy boost.

Explore time series analysis and forecasting, defining time series, identifying trends and seasonality, and predicting future values from past data using ARIMA, SARIMA, Prophet, and LSTM.

Explore the distinction between time series and regression, including when to use each with or without a date column, and learn key algorithms like ARIMA, SARIMA, LSTM, and Prophet.

Explore time series analysis as a specialized forecast on data with a date column at equal intervals, and learn its importance in statistics, plus upcoming anomaly detection (outlier detection).

Explore anomaly detection in time series, identifying data points that deviate from expected patterns as outliers, and discuss handling or removing anomalies with ARIMA, Prophet, or LSTMs.

Explore the four time series components: trend, seasonality, irregularity, and cyclic, and learn how decomposition and autocorrelation reveal them for accurate forecasting.

Explore time series decomposition to separate trend, seasonality, irregularity, and cyclic behavior, using multiplicative and additive methods with Python and statsmodels for practical forecasting.

Explore time series decomposition with statsmodels, comparing additive and multiplicative models, and visualize trend, seasonality, and residuals using the air passengers dataset.

Explore additive and multiplicative decomposition of time series, identifying trend, seasonality, and residuals, and learn when to use each model based on whether seasonal magnitude is proportional to the trend.

Explore why time series must be stationary, with constant mean, variance, and autocorrelation, and learn how differencing and log transformations enable reliable forecasting.

examine time series stationarity using plots, summary statistics, and augmented dickey-fuller test; interpret p-values to decide unit roots, noting ARIMA and exponential smoothing require stationarity while prophet variants may not.

Apply differencing, logarithmic, and smoothing transformations to remove seasonal and trend components and achieve stationarity. Assess stationarity with the ADF test and p-values, then use inverse transformations for forecasts.

Pre-process time series data by cleaning, handling missing values, removing duplicates, and addressing outliers to ensure accurate forecasting. Apply feature scaling, encoding, and feature engineering for robust analysis.

Handle missing values in data cleaning for EDA by deleting rows or columns with missing values, imputing with mean, median, or mode in Python, and comparing effects on model accuracy.

Master techniques to handle missing values in real time data. Apply mean, median, and mode imputation, and use forward or backward filling, or drop rows or columns on churn dataset.

identify and treat outliers using box plots, histograms, and scatter plots. decide when to remove or keep outliers in time series forecasting, predictive analytics, using quantile or interquartile range methods.

Identify outliers with the three sigma rule from normal distribution using mu and sigma, reduce skewness with log and double log, and use box plots and percentiles to detect outliers.

Learn how feature scaling standardizes and normalizes numerical features to improve predictive modeling and sometimes data exploration, using standardization and normalization.

Standardization converts data to z-scores using the mean and standard deviation, yielding a zero mean and unit variance within roughly -3 to 3; apply with Python and sklearn's StandardScaler.

Learn how min-max normalization scales features to 0–1, reducing bias from income and other scales, with a hands-on example and a preview of Python tools for standardization.

Explore feature scaling in Python using sklearn, performing normalization with min max scalar and standard scalar on a churn dataset.

Understand feature encoding for predictive analytics. Learn label encoding, one hot encoding, and dummy encoding to convert categorical x variables into numerical features.

Explore practical feature encoding techniques for churn modeling data, including handling missing values, label encoding, and one-hot or dummy encoding using pandas and sklearn.

Explore time series forecasting algorithms like ARIMA, seasonal ARIMA variants, Prophet, LSTMs, Holt-Winters, GARCH, and VAR/ARIMA models, and understand autoregressive integrated moving average concepts.

Learn how Arima, the autoregressive integrated moving average model, analyzes time series with trend, seasonality, and autocorrelation, and how p, d, and q define autoregressive, differencing, and moving average components.

Explore how p, d, and q define AR, differencing, and MA components in ARIMA. Learn to test stationarity, apply transformations and inverse forecasts, and build end-to-end ARIMA models.

Explore the mathematics behind Arima, detailing autoregressive and moving average components, differencing effects, how past values drive forecasts, and how pacf/pcf guide choosing p and q.

Explore the moving average concept and how MA uses previous error terms to forecast, with examples, and note acf and pacf.

Leverage autocorrelation analysis with ACF and PACF plots to identify AR and MA orders for time series forecasting, guiding ARIMA/ARMA models with 95% confidence intervals.

Identify optimal p, d, q values for ARIMA models using grid search, comparing manual ACF/PACF plotting with Auto.arima automation and rmse evaluation in R and Python.

Explore ARIMA fundamentals, ACF and PACF, and practical use with the air passengers time series. Learn to assess stationarity, apply differencing or log differencing, and perform inverse transformations for forecasting.

This lecture covers why stationarity matters for time series forecasting with ARIMA, shows the augmented Dickey-Fuller test, p-value interpretation, and using transformations to achieve stationarity.

Apply transformations such as log, double log, and differencing, plus moving average and exponential weighted moving average, to achieve stationarity for ARIMA modeling, and perform inverse transformations for final forecasts.

Explain multiplicative and additive decomposition of time series using Statsmodels seasonal_decompose. Reveal trend, seasonality, and residuals from log data, show plots, and discuss acf, pacf, and inverse transformations for forecasting.

Learn how to interpret ACF and PACF plots on log-transformed time series to estimate AR and MA orders (p and q), and why grid search matters for production.

Explore end-to-end time series transformations and their inverses, including log, double log, and log differencing, using exponentiation and cumsum to recover original data.

Learn to run a grid search for arima time series, testing p, d, and q values, and use rmse to identify the best pdq for forecasting with practical code workflow.

Grid search evaluates multiple p and q combinations, revealing an optimum pdq of 1, 0, and 1, while highlighting its time and resource costs.

Master final model building for time series with arima, using training and test splits, log transformations, step forecasting, and end-to-end use cases in time series forecasting.

Explore Facebook Prophet for time series forecasting, featuring fast additive regression with yearly, weekly, daily seasonality and holiday effects, plus robust handling of missing data and outliers.

Apply a Facebook Prophet model to air passenger data, fit on transformed series, and forecast future values. Interpret y_hat, y_hat_lower, and y_hat_upper with trend and seasonality insights and offline plots.

Explore holiday effects in Facebook Prophet and how to incorporate holidays via a data frame or built-in methods, enhancing forecasts with seasonality and multivariate considerations.

Explore multivariate time series analysis, compare univariate and multivariate forecasting, and learn how vector autoregressive and Facebook Prophet with regressors improve forecasts by capturing variable interactions.

Explore univariate and multivariate time series forecasting with Facebook Prophet, applying log transformations, inverse transforms, and added regressors like date, humidity, wind speed, and mean pressure to predict mean temperature.

Evaluate forecasting performance by training on the training data, predicting on the testing data, and comparing results to select the best model using MAE, MSE, RMSE, MAPE, and R2.

Explore mean absolute error, an evaluation metric that averages the absolute differences between actual values and predictions, illustrated with a simple example.

Explore mean squared error (mse), an evaluation metric that averages the squared differences between actual and predicted values using 1/n sum (y - ŷ)^2, with an example.

Understand root mean squared error, the square root of MSE, as a regression metric using sqrt((1/n) * sum (y - y_hat)^2) to evaluate and compare models.

Explore the mean absolute percentage error (MAPE) as a forecasting and regression evaluation metric, learn its formula, and compare models by the lowest MAPE.

Explore energy demand forecasting with ARIMA, using time series data to predict electricity demand and help utilities optimize power generation and distribution.

Load and visualize energy data with pandas, convert timestamps to datetime, and plot load with solar generation to reveal trend and seasonality; forecast arima and address missing values with ffill.

Assess stationarity using ADF tests and p-values, perform ARIMA modeling with p=2, q=2, d=0, compare RMSE across models, and discuss grid search for hyperparameter tuning in energy demand forecasting.

Predict stock prices using univariate time series with Facebook Prophet, leveraging historical Tesla data and bayesian regression modeling to generate forecasts for trading decisions.

Transfer code to Google Colab, load the Tesla stock data, and build a Prophet forecast with daily, weekly, and yearly seasonality for 365 days, including component plots.

Forecast Tesla stock prices with Prophet, analyze closing price and volume, explore trend, volatility, and seasonality, and apply moving averages.

Explore demand forecasting for e-commerce data by comparing ARIMA, Holt's Winter, and Facebook Prophet, building and evaluating models with training-test splits.

Demonstrate a practical demand forecasting workflow in Google Colab, including data upload, median imputation, weekly aggregation of unit sold, and exploratory analysis of base price and total price.

Analyze the unit sold time series, decompose into trend, seasonality, and residuals to assess stationarity and prepare for autocorrelation, partial autocorrelation, and moving-average based model choices.

Learn to detect seasonality in weekly demand data with autocorrelation plots in pandas, interpreting the sinusoidal pattern and seasonal decomposition to guide model building.

Train Holt-Winters, ARIMA, and Prophet on a cleaned weekly time series, split data into training and test, and evaluate forecasts with RMSE to identify the best algorithm.

Compare arima, Holt-Winters, and Prophet for Facebook profit forecasting on a messy dataset; evaluate plots and rmse, prefer Prophet with potential regressors and interactive Plotly visuals.

Explore how machine learning analyzes patient data and lifestyle factors to classify cancer risk as high, medium, or low. Build algorithms and deploy Streamlit or Flask app to automate prediction.

Examine a generic multi-class classification architecture for predicting three risk levels with an imbalanced dataset. Compare models like tree, random forest, and XGBoost, with feature scaling, PCA/LDA, and hyperparameter optimization.

Understand a Kaggle healthcare dataset to identify independent features and a risk level target, explore cancer types and patient attributes, and plan to train, evaluate, and deploy models.

Perform exploratory data analysis on a cancer risk dataset, examining features from patient id and cancer type to demographics and lifestyle factors, and noting the imbalanced risk level distribution.

Create a seaborn bar plot of gender cancer counts by cancer type and patient count, grouped by gender; breast cancer dominates females, prostate in males.

Transform numerical features, drop patient id, encode risk levels, and one-hot encode cancer type to prepare data for heatmaps, revealing key correlations in cancer risk.

Balance imbalanced data using smot for the model; evaluate with a random forest, achieving 84 percent accuracy and high, low, and medium risk classification, and plan to explore other techniques.

Build, tune, and compare machine learning models using Optuna for hyperparameter optimization of a random forest, aiming to improve macro F1 and recall over SMART, with deployment path considerations.

Apply an imbalanced learn pipeline with SMOTE to a random forest, optimize hyperparameters via threefold stratified cross-validation and macro F1, and compare trials to improve high-risk recall before trying XGBoost.

Build a baseline XGBoost model with Smart inside a leakage-free pipeline, using label encoding, imblearn pipeline, and evaluation with log loss to compare results.

Explore and compare baseline XGBoost, a class-weighted XGBoost, and Optiona-tuned class-weighted XGBoost, with random forest as a baseline, finally identifying the Optiona-tuned model as the winner.

Save the tuned, class-weighted XGBoost model as a pkl file with joblib, then prepare it for deployment in a Streamlit or Flask app.

Convert an Optuna-tuned XGBoost cancer risk prediction model into a Streamlit app, enabling batch CSV predictions and manual input with preprocessing and deployment options.

Deploy a Streamlit cancer risk level detection app to AWS EC2, exploring deployment options across Azure, GCP, and on‑premises, and evaluating databases from Oracle SQL to NoSQL and vector databases.

Test the Streamlit app locally by running app.py with Anaconda, then prepare for aws deployment using the final xgb class weighted pickle and a sample prediction.

Deploy a Streamlit app on AWS EC2 using an Ubuntu instance, open port 8501, configure security groups, install dependencies, and access it via the public IP.

Master the data analytics lifecycle from business understanding through data understanding, collection, preparation, and exploratory data analysis, guiding BI or AI deployment paths.

Analyze telco churn using four customer segments, key loyalty drivers and churn triggers, and apply a data science led bronze-to-gold pipeline to predict churn and drive proactive retention campaigns.

Examine data types for telecom churn analysis, including customer demographics, usage, reload data, and call center records. Identify must-have (yellow) versus good-to-have data to build a churn model.

Explore a telecom churn dataset with features like tenure, monthly charges, payment method, and contract type; analyze churn distribution, identify high churners, and outline eda and predictive modeling directions.

Explore data through exploratory data analysis (eda) within telecom churn analysis to derive insights for dashboards or churn prediction models, outlining business understanding, data understanding, and data collection and preparation.

Explore customer churn through an end-to-end eda: load the dataset, inspect shape and data types, identify total charges as a numeric issue, and visualize churn distribution to surface insights.

Clean the data by converting invalid types to numeric, handling nulls, backing up data, dropping few missing values, and binning tenure; remove noninformative columns to prepare for analysis.

Explore univariate analysis to identify churn drivers from gender, senior citizen status, contract type, and payment method, revealing that senior citizens and monthly contracts are linked to higher churn.

Investigate numerical analysis of churn by comparing active and churned users, applying dummy encoding, correlation, and kernel density plots to link monthly charges, total charges, and tenure.

Perform bivariate analysis to compare churners and non-churners by partner status, gender, and payment method, and summarize insights in an end-to-end EDA report for data storytelling.

Learn to build an end-to-end EDA report for a churn analysis use case, covering business and data understanding, missing data handling, variable conversion, and graph-driven insights.

Build a telecom churn prediction model by data prep, stratified 80-20 split, scaling, feature engineering, and multiple model trials with hyperparameter tuning, then deploy via Flask or Streamlit on AWS.

Build a base model for customer churn using a train-test split, a decision tree classifier, and one-hot encoding, then address imbalanced data and data type issues like total charges.

Clean telco data, convert total charges to numeric, scale features, bin tenure, and encode variables to train and evaluate model, while balancing data with upsampling and smart ENN or Adacene.

Explore class-imbalance handling for XGBoost, comparing smart upsampling, Adacene, and Exibust, and apply hyperparameter optimization to boost recall and F1 toward a 0.7 target.

Explains hyperparameter optimization for models like AdaBoost and XGBoost, comparing randomized search and grid search, highlighting time costs, and the role of parameter grids and tuning methods.

Save the best model to a PKL file using Joblib or pickle to back up before Google Colab session timeouts, and prepare for deployment with Flask and Streamlit.

Tune hyperparameters for churn prediction using Adaboost, LightGBM, and Cadboost, guided by Optuna Bayesian optimization. Save, compare, and deploy the best models with Streamlit or Flask.

Identify Adaboost as the top performer on F1 and recall, and plan hyperparameter optimization and saving as a PKL file for the next Streamlit app.

Turn a best AdaBoost model into an interactive Streamlit application, exploring deployment with Streamlit versus Flask, API endpoints, and churn prediction with a 44% probability.

Deploy a Streamlit application on AWS using EC2, configure port 8501 and security groups, and publish with a public IP for front-end access.

Describe post-churn workflow after deployment on AWS or Azure, prioritizing hot, amber, and lower-priority customers by probability scores and using sentiment analysis of call data to guide retention offers.

Learn how to turn standalone Python notebooks into deployable apps with Flask, build a front end, expose models via REST APIs, and map end-to-end deployment workflows.

Learn to build visually appealing flask web apps that connect front-end interfaces to ML models using get and post methods, wsgi, and URL binding.

Create a basic Flask web app by installing Flask, creating app.py, defining routes, returning text responses, and running with optional debug mode and port configuration.

Build a breast cancer prediction model using a random forest classifier with a CSV health dataset, apply label encoding, train-test split, and expose predictions via a Flask app.

Transform a notebook-based breast cancer model into a Flask web app with a simple html front end, handling post submissions, rendering templates, and frontend-backend integration.

Deploy your flask application on AWS EC2 to expose it publicly instead of localhost. Launch an EC2 instance, select an AMI, configure security groups, upload and deploy your code, verify.

Deploy a flask ml app on aws ec2 by creating an account, launching an instance, securing access via security groups, installing dependencies, and running with nohup.