MRMy Research WritingLearning centre
Log inCreate account
Data-driven modelling

Machine learning foundations

Move from a research question to an evaluated model while avoiding leakage, misleading metrics and unsupported conclusions.

View curriculum
X_train, X_test = split(data)

model.fit(X_train, y_train)
prediction = model.predict(X_test)

evaluate(y_test, prediction)
Learning outcomes

Model carefully and interpret honestly.

01

Frame the task

Translate a substantive question into an appropriate prediction or classification problem.

02

Evaluate correctly

Use held-out data, suitable metrics and baselines to assess model performance.

03

Report limitations

Distinguish predictive association from causation and document possible bias.

Curriculum

Four model-building modules

Python basics are recommended before beginning this course.

  1. Supervised and unsupervised learning
  2. Features, labels and baselines
  3. Train, validation and test sets
  1. Linear and logistic regression
  2. Decision trees and ensembles
  3. Clustering and dimensionality reduction
  1. Accuracy, precision, recall and F1
  2. Cross-validation and hyperparameters
  3. Recognise overfitting and data leakage
  1. Inspect subgroup performance
  2. Use feature importance cautiously
  3. Write a transparent model report
Important: Predictive performance alone does not establish causation, fairness or suitability for deployment.