Applied machine learning

Objectives

• Prepare a dataset for learning (cleaning, encoding, feature creation) and document every choice.
• Train and compare at least three supervised regression or classification models, using cross-validation and metrics suited to the problem.
• Apply an unsupervised method (segmentation) and interpret the resulting groups for a business use.
• Diagnose and correct overfitting, data leakage or class imbalance.
• Explain a model's predictions (feature importance, SHAP values) and write a one-page model card for a non-specialist.

Audience

Engineers, experienced data analysts, actuaries, technically minded management controllers, and developers who want to move into data science.

Prerequisites

Able to program in Python with pandas (reading, filtering, aggregating) and comfortable with descriptive statistics (mean, standard deviation, correlation); DATA-101 covers these prerequisites. Checked by a placement test. One laptop per participant, 8 GB of memory minimum; an online computing environment (notebooks, with scikit-learn, XGBoost and pandas preinstalled) provided by Yolea for the length of the session, with no local installation; for remote delivery, a stable connection of at least 5 Mbit/s, microphone and camera. Training datasets provided; the participant's own dataset is accepted if it is anonymised.

Programme

• Data preparation
• Supervised and unsupervised modelling
• Rigorous evaluation
• Diagnosing classic errors
• Model explainability and documentation

Format

In-company or open enrolment. 21 hours, 3 consecutive days of 7 hours, or 2 + 1 days one week apart, with the project moving forward in between. On site, live remote or hybrid; 8 to 12 participants. 30% input, 70% notebook workshops; one running project per participant, from the raw dataset to the model card; the principle is a simple baseline first, the complex model second.

Assessment

Beforehand, a placement test. During the course, check quizzes and a practical challenge (model an imposed dataset; the result is assessed on the metric and on the quality of the approach). At the end, a presentation of the running project's model card, a same-day feedback survey and a named certificate of course completion. Three weeks later, a follow-up check-in.