COMP1961
Essentials of Data Science

This module aims to develop students’ understanding of the theoretical principles, tools, and techniques that underpin contemporary machine learning, while enabling them to apply these concepts effectively in real-world contexts.
It emphasizes critical thinking around model design, evaluation, and ethical considerations such as bias and fairness, alongside practical skills in feature engineering and model selection.
Students will engage with complex scenarios, gaining hands-on experience in deploying robust and ethical machine learning solutions to support predictive analytics and informed decision-making.

1. Critically examine real-world scenarios and translate them into well-defined machine learning problems using appropriate methodological approaches.
2. Design, develop, and implement advanced solutions to domain-specific real-world problems by critically selecting and applying appropriate machine learning techniques.
3. Evaluate, pre-process, and integrate heterogeneous data to formulate and implement effective machine learning solutions to address complex real-world challenges.
4. Analyse and appraise the application-specific machine learning algorithms assessing their usability, reliability and performance using appropriate metrics and evidence-based reasoning.
5. Critically reflect on the ethical and societal implications of machine learning applications.
6. Collaborate effectively with peers to explore, design and critically review machine learning workflows, sharing constructive feedback, effective coordination, and clear communication in the development process.

- Advanced theoretical concepts, mathematical foundations and modern techniques,
- Supervised learning- linear and non-linear models including Ridge, Lasso and polynomial regression; linear discriminant analysis; classification techniques including logistic, SVM, k-NN, decision trees, random forests and boosting models.
- Unsupervised learning – clustering, k-means, dimensionality reduction via PCA, SDV, t-SNE; kernel density estimation
- Neural networks, RNN, GRU, LSTM
- Feature engineering- scaling and normalisation, encoding methods, feature selection methods
- Data quality and preprocessing, model selection and validation, hyperparameter optimisation, overfitting and underfitting, regularisation
- Evaluation metrics for classification and regression, clustering and unsupervised metrics
- Bias, fairness and interpretability – understanding bias; mitigating bias via resampling, synthetic data, threshold adjustment and calibration; local and global interpretability; explainability in high-stakes domains e.g., healthcare, finance, criminal justice; model governance and ethics
- Real-world applications – industry case studies including finance, healthcare, retail & marketing, cybersecurity, engineering & IoT

Presentation: 25% weighting, 50% pass mark.
Learning Outcomes: 1 - 5
Duration: 7 minutes
Outline Details: A pre-recorded presentation of 7 minutes in length max, justifying the project inception, development and evaluation.

Project: 75% weighting, 50% pass mark.
Learning Outcomes: 1 - 6
Word Length: 3000.
Outline Details: Design, implementation and evaluation of ML algorithms to solve real-world problem.