University of Melbourne · FACULTY OF STATISTICS

MGMT90280 · Managerial Decision Analytics

- one subject, every graph, every model, every mark

Statistics14 Chapters7-page Bible

Our own words - no uploaded lecturer files

Updated for this semester

Chapter 7 of 10 · MGMT90280

Descriptive Data Mining: Association Rules

Association-rule mining is the market-basket half of descriptive data mining in MGMT90280 Managerial Decision Analytics at the University of Melbourne. It scores an if-antecedent-then-consequent rule with three numbers — support, confidence and the lift ratio — and applies one decision rule (lift greater than 1 means the pattern beats chance). It is the reliable, formula-driven part of the descriptive-data-mining exam question.

In this chapter

What this chapter covers

01Association rules as unsupervised, descriptive mining: no target variable, just transactions
02Rule anatomy: antecedent (if-itemset) → consequent (then-itemset), with A and C disjoint
03Support count of an itemset = number of transactions containing every item in it
04Support of a rule = support count of A∪C divided by N (total transactions)
05Confidence = support(A∪C) / support(A) = P(consequent | antecedent)
06Lift ratio = confidence / (support(C)/N), comparing the rule against the base rate
07Decision rule: lift greater than 1 useful, equal to 1 independent, less than 1 worse than random
08The high-confidence-low-lift trap: a very common consequent inflates confidence but not lift
09Lift is symmetric in A and C; confidence is directional
10Business action: shelf placement, cross-sell bundles, promotion targeting from useful rules

Worked example · free

Score a market-basket rule with support, confidence and lift

Q [6 marks]. A cafe logs N = 20 morning transactions. For the rule {espresso, croissant} then {orange juice} the support counts are: support({espresso, croissant, orange juice}) = 6, support({espresso, croissant}) = 8, and support({orange juice}) = 10. Compute the rule's support, confidence and lift, and give the explicit useful / not-useful decision.

+1Name the parts. Antecedent A = {espresso, croissant}; consequent C = {orange juice}; combined itemset A∪C = {espresso, croissant, orange juice}. Given support(A∪C) = 6, support(A) = 8, support(C) = 10, N = 20.
+1Support of the rule = support(A∪C) / N = 6 / 20 = 0.30 — the whole pattern appears in 30% of baskets.
+1Confidence = support(A∪C) / support(A) = 6 / 8 = 0.75, i.e. P(orange juice | espresso and croissant) = 75%.
+1Expected confidence (base rate) = support(C) / N = 10 / 20 = 0.50: orange juice is in half of all baskets regardless.
+1Lift = confidence / (support(C)/N) = 0.75 / 0.50 = 1.50.
+1Decision: lift = 1.50 is greater than 1, so the rule is useful — an espresso-plus-croissant basket is 50% more likely to include orange juice than a random basket; recommend positioning or bundling them.

Support = 0.30, confidence = 0.75, lift = 1.50. Because lift (1.50) is greater than 1, the rule is a useful (positive) association: espresso-plus-croissant buyers are 1.5 times more likely to buy orange juice than a random customer.

Sia tip — Lay the three numbers out in order — support, then confidence, then lift — each with its formula and arithmetic, so every line banks a mark. Keep confidence and support(C)/N side by side; lift is just one divided by the other. Always compute lift, not just confidence.

Glossary

Key terms

Association rule: A pattern written {antecedent} then {consequent}, meaning baskets containing the antecedent items tend to also contain the consequent items. It is a direction, not an equation.
Antecedent (A): The 'if' itemset on the left of the rule (the body). The consequent must be a disjoint itemset (no shared items).
Consequent (C): The 'then' itemset on the right of the rule (the head), predicted to appear when the antecedent does.
Support count: The number of transactions that contain every item in an itemset. Dividing by N (total transactions) gives the fractional support.
Confidence: support(A∪C) / support(A) = P(consequent | antecedent): of the baskets containing the antecedent, the fraction that also contain the consequent.
Lift ratio: confidence divided by support(consequent)/N. It measures how many times more likely the consequent is given the antecedent, compared with a random basket.
Lift decision rule: Lift greater than 1 = useful rule (beats chance); lift equal to 1 = antecedent and consequent independent; lift less than 1 = worse than random (negative association).
Expected confidence (base rate): support(consequent)/N: the confidence you would expect if the antecedent and consequent were unrelated. It is the denominator of lift.

FAQ

Descriptive Data Mining: Association Rules FAQ

Is association-rule mining on the MGMT90280 final exam?

Yes. In the past papers the final is five compulsory 20-mark questions, and descriptive data mining is one of them — bundling similarity coefficients, hierarchical clustering and association rules. The association-rule part is a formula-driven block: compute support, confidence and lift from given support counts, then state the lift-greater-than-1 verdict. The final is worth 50% and is open-book (Casio FX-82, unrestricted student notes); confirm the exact structure and the ~November 2026 exam date on the LMS.

Why do I need lift when I already have confidence?

Confidence ignores how common the consequent is on its own. If a consequent appears in 90% of baskets, a rule can show 80% confidence yet still be worse than random (lift 0.80/0.90 is about 0.89, below 1). Lift divides out the base rate, so it is the number that tells you whether the association is real. Always check lift, not just confidence.

Can AI help me with association rules in MGMT90280?

Yes, as a study aid. Sia can explain, step by step, how to identify the antecedent and consequent, tally support counts, and apply confidence = support(A∪C)/support(A) and lift = confidence/(support(C)/N), and it can walk through the lift-greater-than-1 decision on practice numbers so you can check your own working. It does not sit your exam, do your assessed assignments, or promise a grade — use it to build the method, then practise unaided within your course's academic-integrity rules.

Study strategy

Exam move

Memorise the three formulae as a chain: support = support(A∪C)/N, confidence = support(A∪C)/support(A), lift = confidence divided by support(C)/N. Practise reading support counts off a small transaction table, because that setup step earns its own mark. In every answer name the antecedent, consequent and combined itemset, then show each ratio with its numbers, then finish with the lift-versus-1 decision in words — markers reward the interpretation line, not just the arithmetic. Rehearse the high-confidence-low-lift trap so your useful / not-useful verdict follows the lift number rather than the confidence number. Pace the paper at roughly 1.2 minutes per mark (120 writing minutes over 100 marks) and confirm the exam length and permitted materials on the LMS.

Working through Descriptive Data Mining: Association Rules in MGMT90280? Sia is AskSia’s AI Statistics tutor — ask any MGMT90280 Descriptive Data Mining: Association Rules question and get a clear, step-by-step explanation grounded in how MGMT90280 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.

Keep going — explore the course

A+Everything unlocked

Unlocks this Bible + all 11 of your University of Melbourne subjects - and 1,000+ Bibles across every Australian university.

Sia - your MGMT90280 tutor, unlimited, worked the way the exam marks it

The full 7-page Bible + practice bank with worked solutions

Chrome extension - sync your LMS so Sia knows your deadlines

Bilingual EN / Chinese on every Bible and every Sia answer

$25/ month

30-day money-back · cancel in one tap · how it works