The best way to make the lessons stick is to build something. These five projects take you from predicting house prices with regression, through classifying Titanic passengers, filtering spam with Naive Bayes, and predicting customer churn, to grouping customers with k-means clustering.
Each one is a code-along video, so you can pause, type it out yourself and adapt it to your own datasets. Pick the project that matches the lesson you've just finished and come back for the rest.
Go through the full data science process: load and explore the data, preprocess it, engineer features, then compare a linear regression model with a random forest.
Classify which passengers survived using logistic regression, k-NN, SVM and a random forest, compare their accuracy, and tune each one.
Turn email text into numbers with CountVectorizer, train a Naive Bayes classifier, and wrap the steps in a scikit-learn pipeline.
A complete walkthrough: preprocess the data, select the best features, build a predictive model, and evaluate how well it spots customers who will leave.
Explore customer data, use the elbow method to pick the number of clusters, train k-means, and plot the customer segments.