COMP4318 · Machine Learning and Data Mining
Introduction to Machine Learning & Data Mining
Week 1 sets up the whole unit: what machine learning and data mining are, the end-to-end pipeline (data → preprocessing → model → evaluation → deployment), and the three learning paradigms — supervised, unsupervised and reinforcement. It fixes the vocabulary (features, labels, generalisation, train/validation/test) that every later chapter reuses. In the exam this shows up as short definition and true/false parts, and as the framing for 'which method would you use here' questions.
What this chapter covers
- 01What machine learning is (building models that generalise from data) and how data mining relates to it
- 02The ML pipeline: data → preprocessing/normalisation → model → evaluation → deployment
- 03Supervised (labelled → predict) vs unsupervised (unlabelled → group) vs reinforcement learning
- 04Attribute types: categorical/nominal vs numeric/continuous
- 05Training vs test sets; accuracy = proportion of test examples correctly classified
- 06Generalisation, features and labels — the vocabulary the unit reuses
- 07Choosing a method for a task (e.g. real-time spam filtering: naïve Bayes for fast per-example prediction)
Matching a task to its learning paradigm and method
- +1(a) A labelled archive plus the goal of predicting a class for each new email = supervised learning (classification).
- +1(a) A good choice is naïve Bayes: classification cost is roughly O(number of attributes), so it is fast per email and is a standard, robust text classifier — well suited to real-time filtering.
- +1(b) No labels, grouping by similarity = unsupervised learning (clustering, e.g. k-means).
- +1(c) Learning from a reward signal through interaction, with no labelled correct outputs = reinforcement learning.
Key terms
- Machine learning
- Automatically building mathematical models that explain and generalise a dataset; it blends statistics with algorithm development.
- Data mining
- The exploration and analysis of large data by (semi-)automatic means to discover useful patterns; machine learning supplies many of its methods.
- Supervised learning
- Learning from labelled examples to predict the label of new examples (classification for classes, regression for numeric values).
- Unsupervised learning
- Learning from unlabelled data — grouping similar examples (clustering) or reducing dimensions — with no target to predict.
- Generalisation
- How well a model performs on new, unseen data rather than the data it was trained on; the whole goal of learning.
- Training/test split
- Building the model on a training set and estimating accuracy on a held-out, labelled-but-unseen test set; accuracy = proportion of test examples classified correctly.
Introduction to Machine Learning & Data Mining FAQ
What's the difference between machine learning and data mining?
They overlap heavily. Machine learning is about algorithms that build models which generalise from data; data mining is the broader activity of exploring and analysing large datasets to find useful patterns, and it uses machine-learning methods to do so. In this unit they are taught together — you learn the algorithms (machine learning) and how to apply them to discover patterns, classify, and predict (data mining).
How is Week 1 material examined in COMP4318?
As short definition and true/false parts, and as the setup for 'which method would you use' questions across the paper. You should be able to define the paradigms, name the pipeline stages, and justify a method for a given task — the marks are for correct vocabulary used precisely.
Supervised or unsupervised — how do I tell which a task is?
Ask whether the training examples carry labels. If you are given labelled examples and must predict a label for new ones, it is supervised (classification or regression). If there are no labels and you are grouping by similarity or reducing dimensions, it is unsupervised. Learning from a reward through interaction is the third paradigm, reinforcement learning.
Can AI help me with the COMP4318 basics?
Yes. Sia can quiz you on the pipeline and the paradigms, explain why a task is supervised or unsupervised, and check your reasoning on 'which method' questions — step by step, as a tutor. It explains the method and never completes graded work for you; confirm assessment details on Canvas.
Exam move
Do not skim Week 1 — its vocabulary is worth easy marks and underpins every later chapter. Make a single page that names the pipeline stages, defines supervised/unsupervised/reinforcement with one example each, and lists the method-to-task matches taught in lectures (fast real-time → naïve Bayes; images → CNN; unlabelled grouping → clustering). Because this material appears as short-answer and framing parts across the paper, rehearse stating definitions crisply. When a distinction blurs, ask Sia to contrast the two with a fresh example and check your wording.
Working through Introduction to Machine Learning & Data Mining in COMP4318? Sia is AskSia’s AI Machine Learning tutor — ask any COMP4318 Introduction to Machine Learning & Data Mining question and get a clear, step-by-step explanation grounded in how COMP4318 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
