Welcome, champion! 🔥
You've mastered 0 of 134 lessons. Start your first lesson free — no signup needed.
About this course
Most machine learning goes wrong before any model is trained. The target is defined badly, the split leaks, the metric flatters a useless classifier, or the whole problem never needed learning in the first place. This course is built around those decisions rather than around library syntax — you will spend more time on why accuracy misleads under imbalance than on the argument order of a fit call.
What it covers
Eleven topics: foundations and problem framing, data preparation and features, regression, classification, trees and ensembles, distance and margin methods, probabilistic methods, unsupervised learning, model evaluation and selection, and the overfitting, leakage and imbalance traps. A closing practice topic compares methods and asks which one a given situation calls for.
Who this is for
Students and developers preparing for data-science and ML interviews, and anyone who has trained a model with a library but could not defend the choice of algorithm or metric under questioning. It assumes comfort with basic probability, mean and variance, and reading a formula. It does not assume linear algebra beyond matrix intuition, and it is not a deep-learning course.
How MCQ practice works on Abekus
The course is organised as topics, subtopics, and individual lessons. Each lesson pairs a short set of explanation cards with practice questions written for that lesson, so the questions test the idea you just read rather than a general survey. Wrong answers come with a worked explanation, and the AI guide surfaces the mistakes you keep repeating.