Learn Machine Learning & Data Mining in Python

Explore Google Colab for machine learning in Python, coding in your browser with zero configuration, free GPU access, and easy notebook sharing—watch for disconnection and save progress.

Log in to Google Colab with a Gmail account, create a new IPython notebook, and navigate the simple interface to write Python code online.

Learn to create lists in Python with numbers, text, or mixed values, then use len, sort, and append, and access or extract values via indexing.

Learn how to create tuples and dictionaries in Python, explore tuples' immutability, access values by index or key, and build dictionaries with curly braces and key-value pairs.

Learn how to implement loops and functions in Python, including for and while loops, range, conditionals, and simple functions to manipulate lists and data.

Explore the pandas library for data manipulation and analysis in Python, including its data structures for numerical tables and time series, and how to import it in a Python environment.

Learn to use NumPy library for working with arrays and multidimensional arrays like matrices in Python.

Import numpy in a notebook (Colab) as np, create a normal-distribution dataset, and compute its mean. Then import matplotlib.pyplot as plt to plot a histogram inline.

Explore the scikit-learn library, its tools for classification, regression, clustering, and dimensionality reduction. Import the library to start building machine learning models in this course.

Explore supervised learning, where models use known labels to classify or predict, distinguishing classification for categorical outcomes from regression for continuous values with features like height, weight, size, and location.

Learn how r-squared quantifies the portion of the dependent variable's variance explained by predictors, interpret 0–100% values, and note the role of adjusted r-squared in model evaluation.

Learn how to assess predictor significance in regression using p values and a 0.05 threshold, deciding when to keep or remove variables like price, model, year, mileage, age, and gender.

Learn simple linear regression with one independent variable, Y hat equals b0 plus b1 X, and how X predicts Y, for example year of experience to predict salary.

Open a Google Colab notebook, upload a dataset file using the files module, then read it with pandas and print dataset to see our dataset.

Analyze a dataset of 30 students with hours studied and exam scores, and build a simple linear regression model to predict scores from study time.

Learn to prepare data for a simple linear regression in Python by extracting X and Y with iloc, viewing their values, and splitting data into train and test sets.

Build a linear regression model in Python by importing the linear regression class, fitting with inputs and outputs, and predicting with x_test. Compare actual and predicted data to evaluate performance.

Build and evaluate a simple linear regression model in Python by predicting exam scores from study hours and comparing to actual values. Visualize with a regression line and scatter plot.

Discover stepwise regression in multiple linear regression, using forward selection and backward elimination to keep significant variables, guided by p-values and R-squared and adjusted R-squared for model quality.

Explain how multiple linear regression relies on linearity, multivariate normality, and minimal multicollinearity. Assess residuals for autocorrelation and homoscedasticity, and apply nonlinear transformations or collect more data to address deviations.

Learn to predict profit using a multiple linear regression with 71 company data, incorporating staff training expenses, advertising expenses, and city, encoded as dummy variables in Python.

Extract the feature matrix X and target Y from the dataset by selecting all rows and excluding the last column, and convert the first column's categorical data to numeric values.

Build a reusable multiple linear regression model in Python, fit with training data, predict on test data, and evaluate with r-squared to gauge performance.

Explore polynomial regression as an extension of linear regression, modeling curved relationships between inputs and output by using powers of input terms and their coefficients.

Explore the dataset from a reward system to build a polynomial regression model, analyzing hours worked and points earned across twenty observations to predict points in Python.

Import the dataset file in google colab by uploading reward_system.csv and reading it into a pandas dataframe, using the hours and points columns to prepare for a polynomial regression model.

Learn how to build a polynomial regression model in Python by treating hours as input and points as target, compare linear regression and polynomial regression, and visualize results.

In Python, compare linear regression and polynomial regression to predict points from hours. The results show polynomial regression provides predictions closer to the original value, indicating better accuracy.

Learn logistic regression's binary prediction with a sigmoid function and use age, years as a customer, and gender to estimate buying probability above 50 percent.

Learn how the standard scaler standardizes features by subtracting the mean and dividing by the standard deviation. Unify inputs with different units to improve model performance.

Analyze a bank's client data with features like age, accommodation, income, gender, household size, credit lines, and loans. Build a logistic model to predict loan defaults and identify future applicants.

Create a logistic regression model using age, income, household, and credit lines to predict bank client default, then standardize features and split data into 80% train and 20% test.

Import and instantiate a logistic regression model, fit it, predict on x test, and evaluate with confusion metrics to report 75 percent accuracy.

Learn how the k nearest neighbors algorithm classifies a new point by using a selected k and distance measurements, including euclidean distance and collodion methods, with bank client examples.

Learn to build a k-nearest neighbors model in Python with scikit-learn, including data split, scaling, fitting, predicting, and evaluating with confusion matrix and accuracy, exploring neighbor counts and Minkowski distance.

Demonstrate how naive bayes classifies a new bank client as default or not by combining prior defaults with local income and age evidence.

Build a naive Bayes model in python by importing Gaussian and B class, fitting on data, evaluating with a confusion matrix, and reporting 74 percent accuracy.

Explore the random forest classifier, a collection of decision trees. Create nine trees, average their predictions, and experiment with different tree counts in Python to maximize prediction power.

Explore the concept of unsupervised learning, where data lacks predefined targets, and cluster data to discover groupings such as customer segments by location or purchasing behavior.

Explore hierarchical clustering, an unsupervised method that builds nested clusters by merging closest data points using a distance measure. Learn to determine the optimal number of clusters with a diagram.

Learn how a dendrogram guides the optimal number of clusters by mapping pairwise distances with vertical and horizontal lines and cutting at the longest non-crossing vertical line.

Explore a dataset of around 100 movies with budget, genre, and worldwide gross, clustering films to uncover hidden patterns and prepare for unsupervised learning techniques.

Build a supervised clustering model using production budget and worldwide gross income, apply the Hiriko method to choose the number of clusters, and interpret labels with a dendrogram.

Create a three-cluster agglomerated clustering model in python, fit it to data, and visualize production budget versus gross income to interpret high profit, loss, and mid-range clusters.

Explore how the k-means clustering algorithm partitions data into clusters by iteratively updating centers and reassigning points, and how the elbow method helps decide the number of clusters.

Use the elbow method with k-means to determine the optimal number of clusters by locating the elbow on the total within sum of squares curve.

learn to build a k-means clustering model in python by loading the dataset, extracting two inputs, and plotting the elbow method to determine the optimal number of clusters.

Apply k-means clustering in python by setting n_clusters=3, init='k-means++', and random_state=0; predict on X, and display a three-cluster scatter plot with labeled clusters.

Discover association rules and market basket analysis, illustrate left-hand side implies right-hand side, and explain support, confidence, and lift with retail examples like coffee and sugar and beer and diapers.

Investigate association rules on a 100-transaction convenience store dataset, treating each basket as a transaction and using a primary algorithm to uncover associated items.

Install and import the Pury library, load the 100-transaction dataset with five items per basket, and transform each row into a list of item strings for the apriori model.

Create association rules for market basket analysis in Python, identifying base items and add-ons, and evaluate support, confidence, and lift for jam with bread and coffee with donut.

Learn about loss functions, including mean squared error for regression and binary cross entropy or categorical cross entropy for classification, and how they guide weight updates.

Activation functions decide which nodes forward data through hidden layers to the output, shaping decisions across epochs; Relu for hidden layers, sigmoid for output, softmax for multiple outputs.

Explore gradient descent, an optimization algorithm that uses the cost function to guide parameter updates, balancing learning rate to optimize weights through iterations.

Compare gradient descent types—batch, stochastic, and mini-batch—and explain how learning rate, backpropagation, and the cost function guide weight updates to escape local minima toward global minima.

Explore TensorFlow, the open source deep learning library by Google, used in Python to build neural networks, with Keras on top for CPU and GPU speed.

Build a deep learning classifier to predict diabetes from eight input features, reading data with pandas, extracting inputs with iloc, splitting into train and test sets, and scaling with StandardScaler.

See how a Python-based neural network iteratively improves from 38% to 80% accuracy, while using a confusion matrix to validate diabetic vs not diabetic predictions.

Explore statistics as the science of collecting, analyzing, interpreting, and presenting empirical data and methods for driving decisions, with examples from medicine, exams, and pricing.

Explore when to use a sample versus a population, and how a random sample from the population ensures each member has an equal chance of being selected.

Explore descriptive statistics to describe and summarize data with mean, median, and variance, visualize frequency and time-series patterns, and learn inferential statistics to generalize from samples to populations.

Visualize data in Excel using charts and descriptive statistics to reveal patterns and trends, identifying Texas as the state with the most Walmart stores and Nevada with the fewest.

Explore how to group continuous data into bins using histograms, set bin ranges, and interpret frequency for age data in Excel with the data analysis toolpak.

Explore how standard deviation measures data dispersion around the mean, and how variance equals the square of the standard deviation, with sigma and s for population and sample.

Discover how to compute sum, mean, variance, and standard deviation for population and sample data using Excel, including stdev.p and stdev.s functions.

Learn how symmetry and skewness shape data distributions, compare mean, median, and mode across left-skew, right-skew, and symmetric cases, and recognize outliers at distribution tails.

learn inferential statistics by using probability distributions to draw inferences about a population from a random sample, and differentiate discrete and continuous distributions.

Explore discrete probability distributions, including the uniform distribution with a fair coin and a fair die, and introduce the binomial distribution before the normal distribution.

Standardize a variable by subtracting its mean and dividing by the standard deviation to obtain a z-score, enabling use of the standard normal distribution for probability and hypothesis testing.

See in this standardization demo how to transform a normal distribution by shifting the mean to zero, and setting the standard deviation to one, producing a standard normal distribution.

Learn how the central limit theorem makes the sampling distribution of the mean approach a normal distribution as sample size grows, with standard error and population mean clarified.

Compare point estimates and confidence intervals to understand accuracy. Higher confidence widens intervals—from 68–72 at 90% to 65–75 at 95% and 60–80 at 99%, reflecting the alpha error.

Calculate a confidence interval for a single sample with known population variance using z-values and a z-table, covering 90%, 95%, and 99% levels and rule of thumb for clinical studies.

Compute the confidence interval for clinic income in a city where the standard deviation is known, using the sample mean from 30 clinics and the z-margin of error, with alpha/2.

Explore the t distribution, or student distribution, which handles small samples with unknown population standard deviation, using the t table and degrees of freedom to form confidence intervals.

Increase the sample size to reduce the margin of error and bring the sample mean, x-bar, closer to the population mean, as shown by a pizza pricing example.

Explore confidence intervals for two dependent samples using a before-and-after medical study, such as testing a blood pressure medicine, with a 99 confidence level and Excel methods.

Learn to compute confidence intervals for two independent samples with known variances, using sample means of 85% and 91% from two schools and their population standard deviations, at 95% confidence.

Compute a 95% confidence interval for the difference between two independent means with known variances using a z-test, from two school samples (n=60 and n=40).

derive a confidence interval for the difference between two independent means with unknown but equal variances, using pooled standard deviation, for salaries of engineers in Dallas and New York City.

Define null and alternative hypotheses, testable with data observations, to verify claims about population means; decide to accept or reject the null based on sample mean comparisons.

Explore one- and two-tailed hypothesis tests using Z and T distributions, and understand alpha as the significance level, type I and type II errors, with examples from medicine and finance.

Learn how p-values decide whether to reject the null hypothesis at 10%, 5%, and 1% significance with 90%, 95%, and 99% confidence, using regression of price on sales volume.

Learn to test a one-sample hypothesis with known variance, evaluating whether the city’s average clinic income equals 160,000, and define the null and alternative hypotheses.

Use the one-sample t-test with unknown variance for a small n=10 to test if city clinic income exceeds $170,000; the 95% test does not reject the null.

Test the hypothesis that a medicine lowers blood pressure using two dependent samples before and after treatment, with a 1% significance level and 14 degrees of freedom, yielding no evidence.

Apply a z-test for the difference of means between two independent samples with known variances to assess whether school B outperforms school A by 8%, at the 95% confidence level.