Learning path
Machine Learning in Depth
Regression, trees and boosting, k-NN, Naive Bayes, SVM, clustering, PCA, metrics, tuning, imbalance and explainability, with animated films and runnable Python. Part of the Data Scientist premium track.
-
1
Linear and Logistic Regression, Step by Step
Predict numbers with linear regression and yes/no outcomes with logistic regression. Understand the line, the sigmoid, coefficients and probabilities, with scikit-learn.
-
2
Decision Trees, Random Forests and Gradient Boosting
How a decision tree asks yes/no questions, why Gini impurity picks the best split, and how random forests (bagging) and gradient boosting combine many trees into strong models.
-
3
k-Nearest Neighbours, Naive Bayes and SVM
Three classic classifiers with three different ideas: ask your neighbours (k-NN), count evidence with probability (Naive Bayes), and draw the widest boundary (SVM). With scikit-learn.
-
4
Clustering with k-Means and DBSCAN
Group customers without labels using k-means, choose k with the elbow method and silhouette score, and find odd shapes and outliers with DBSCAN and hierarchical clustering.
-
5
PCA and Dimensionality Reduction
The curse of dimensionality, how PCA finds the directions of greatest variance using eigenvectors, how to read explained variance and a scree plot, and when to use PCA.
-
6
Metrics, Cross-Validation and Hyper-Parameter Tuning
Confusion matrix, precision, recall, F1, ROC-AUC, RMSE; why accuracy lies; k-fold cross-validation; and tuning models with GridSearchCV without fooling yourself.
-
7
Imbalanced Data and Model Explainability
Handle rare classes with class weights, resampling and threshold tuning, then explain predictions with feature importance, permutation importance and partial dependence.