MGMT90280 · Managerial Decision AnalyticsMGMT90280 · 管理决策分析（商业数据分析与建模）
Prescriptive + predictive analytics for business decisions — a 12.5-point graduate subject assessed by mid-sem test, group project and a 50% final exam.面向商业决策的 prescriptive（规范式）+ predictive（预测式）分析 —— 12.5 学分的研究生课程，考核含期中测验、小组项目与占 50% 的期末闭卷考试。MGMT90280 equips Melbourne Business School graduate students with the analytical tools used to turn data into effective managerial decisions — built around three pillars (descriptive, predictive and prescriptive analytics) and taught hands-on in Excel (Solver, Analytic Solver, the Analysis ToolPak) against the prescribed text, Camm, Cochran, Fry & Ohlmann, Business Analytics (5th ed.). The 12-week run moves from linear, integer and nonlinear programming, through probability distributions and Monte Carlo simulation, to regression, time-series forecasting and data mining (clustering and classification trees), with applications in international business, marketing, supply chain and strategy. This page summarises the official subject guide and handbook — what the subject covers, how it is taught, and exactly how it is assessed — built from 46 real MGMT90280 course materials in the AskSia Library (past exam papers included).
MGMT90280 帮助墨尔本商学院（MBS）研究生掌握把数据转化为有效管理决策所需的分析工具，课程围绕三大支柱 —— 描述式（descriptive）、预测式（predictive）与规范式（prescriptive）分析 —— 展开，并在 Excel（Solver、Analytic Solver、Analysis ToolPak）中动手实践，配套指定教材 Camm, Cochran, Fry & Ohlmann《Business Analytics》（第 5 版）。12 周课程从线性规划（LP）、整数规划（ILP）、非线性规划（NLP），到概率分布与蒙特卡洛模拟，再到回归、时间序列预测与数据挖掘（聚类、分类树），并应用于国际商务、市场营销、供应链与战略。本页基于墨大官方 subject guide 与 handbook 整理：课程内容、授课方式与真实考核构成 —— 并参考了 AskSia Library 中的 46 份真实 MGMT90280 课程材料（含历年考卷）。
Built from 46 real MGMT90280 course materials in the AskSia Library, including past exam papers and lecture slides.
整理自 AskSia Library 中的 46 份真实 MGMT90280 课程材料，其中包含历年考卷与讲义幻灯片。
What MGMT90280 is aboutMGMT90280 讲什么
MGMT90280 Managerial Decision Analytics is a 12.5-point graduate coursework subject in the Department of Management & Marketing at the University of Melbourne (Parkville / Melbourne Business School), offered in both Semester 1 and Semester 2. It is taught as ten 3-hour seminars plus a revision week and oral presentations, and is built on the INFORMS view of analytics as 'the scientific process of transforming data into insight for making better decisions', organised into three pillars — descriptive, predictive and prescriptive analytics. The prescriptive half develops optimisation: linear programming (LP) and its sensitivity analysis (shadow prices, reduced costs), integer linear programming (ILP) and nonlinear programming (NLP), solved with the Excel Solver Add-in. The descriptive/predictive half covers probability distributions, Monte Carlo simulation, descriptive data mining (clustering, association rules), multiple regression, time-series analysis and forecasting (moving averages, exponential smoothing), and predictive data mining with classification trees. The prescribed text is Camm, Cochran, Fry & Ohlmann, Business Analytics (5th ed., Cengage), and the subject emphasises applications across international business, marketing, supply chain and strategy while teaching students to judge the power and limits of each technique under uncertainty. Assessment combines an individual timed quiz (Assignment 1, 20%), a group assignment (25%) and class presentation (5%), and an individual 2-hour final exam worth 50%.
MGMT90280《管理决策分析》是墨尔本大学（Parkville 校区 / 墨尔本商学院 MBS）管理与市场营销系的一门 12.5 学分研究生（授课型）课程，第一、第二学期均开课。课程以 10 次 3 小时研讨课（seminar）加 1 周复习与口头展示的形式授课，采用 INFORMS 对 analytics 的定义 ——「把数据转化为洞察、从而做出更好决策的科学过程」，并组织为三大支柱：描述式、预测式与规范式分析。规范式部分聚焦优化建模：线性规划（LP）及其敏感性分析（shadow price 影子价格、reduced cost 既约成本）、整数规划（ILP）与非线性规划（NLP），均用 Excel Solver 求解。描述/预测部分涵盖概率分布、蒙特卡洛模拟、描述式数据挖掘（聚类、关联规则）、多元回归、时间序列分析与预测（移动平均、指数平滑），以及基于分类树的预测式数据挖掘。指定教材为 Camm, Cochran, Fry & Ohlmann《Business Analytics》（第 5 版，Cengage）。课程强调在国际商务、市场营销、供应链与战略中的应用，并培养在不确定性下判断各类技术效力与局限的能力。考核由个人限时测验（Assignment 1，20%）、小组作业（25%）与课堂展示（5%）、以及占 50% 的个人 2 小时期末考试组成。
The MGMT90280 syllabus, topic by topicMGMT90280 大纲 · 逐个主题
Intro to analytics & Linear Programming (LP)分析导论与线性规划（LP）
What business analytics is (descriptive / predictive / prescriptive), then formulating a linear program — decision variables, an objective function and constraints — and reading its sensitivity report: shadow prices and reduced costs. Camm et al. (2024) Ch. 14.
先讲清楚什么是商业分析（描述式 / 预测式 / 规范式），再学习线性规划的建模 —— 决策变量、目标函数与约束 —— 并解读其敏感性报告：影子价格（shadow price）与既约成本（reduced cost）。对应 Camm et al. (2024) 第 14 章。
Integer Linear Programming (ILP)整数规划（ILP）
Extending LP to problems where decisions must be whole numbers or yes/no (binary) — capital budgeting, facility location, fixed-charge and set-covering models. Camm et al. (2024) Ch. 14-15.
把线性规划推广到决策必须取整数或 0/1（二元）的问题 —— 资本预算、选址、固定费用与覆盖模型。对应 Camm et al. (2024) 第 14-15 章。
Nonlinear Programming (NLP)非线性规划（NLP）
Optimisation when the objective or constraints are nonlinear (e.g. price-response or portfolio-risk models), including local vs global optima and solving with Excel Solver's GRG engine. Camm et al. (2024) Ch. 16.
当目标函数或约束为非线性时的优化（如价格-响应模型、投资组合风险模型），包括局部最优与全局最优之分，并用 Excel Solver 的 GRG 引擎求解。对应 Camm et al. (2024) 第 16 章。
Probability distributions概率分布
Discrete and continuous distributions (binomial, Poisson, normal, etc.), expected value and variance — the probabilistic language used to describe uncertainty in later models. Assignment 1 due this week. Camm et al. (2024) Ch. 5.
离散与连续分布（二项、泊松、正态等）、期望值与方差 —— 这是后续模型描述不确定性所用的概率语言。Assignment 1 于本周提交。对应 Camm et al. (2024) 第 5 章。
Monte Carlo simulation蒙特卡洛模拟
Drawing many random samples from input distributions to estimate the distribution of an outcome (e.g. project NPV or solar-farm output) and quantify risk when a closed-form answer is impossible. Camm et al. (2024) Ch. 13.
从输入分布中抽取大量随机样本，估计某个结果（如项目 NPV、太阳能电站发电量）的分布并量化风险 —— 当无法求出解析解时尤为有用。对应 Camm et al. (2024) 第 13 章。
Descriptive data mining描述式数据挖掘
Unsupervised methods that find structure without a target variable: clustering (segmenting observations into similar groups) and association rules (e.g. market-basket analysis). Camm et al. (2024) Ch. 6.
无监督方法 —— 在没有目标变量的情况下发现数据结构：聚类（把观测划分为相似的群组）与关联规则（如购物篮分析）。对应 Camm et al. (2024) 第 6 章。
Regression modelling回归建模
Simple and multiple linear regression in Excel — estimating coefficients, reading R-squared, residuals and p-values, and interpreting how drivers (price, promotion) move an outcome (sales). Camm et al. (2024) Ch. 8.
在 Excel 中做简单与多元线性回归 —— 估计系数，解读 R-squared、残差（residual）与 p 值，并解释各驱动因素（价格、促销）如何影响结果（销量）。对应 Camm et al. (2024) 第 8 章。
Time-series analysis & forecasting时间序列分析与预测
Decomposing a series into trend, seasonal and cyclical patterns and forecasting with moving averages and exponential smoothing — e.g. predicting passenger demand or product sales. Camm et al. (2024) Ch. 9.
把一条时间序列分解为趋势（trend）、季节（seasonal）与周期成分，并用移动平均与指数平滑（exponential smoothing）做预测 —— 如预测客流或产品销量。对应 Camm et al. (2024) 第 9 章。
Predictive data mining: classification trees预测式数据挖掘：分类树
Supervised learning that predicts a categorical outcome (e.g. fraud / not-fraud, churn / stay) by recursively splitting the data, with attention to training vs validation and overfitting. Camm et al. (2024) Ch. 11.
有监督学习 —— 通过递归地划分数据来预测类别型结果（如欺诈/非欺诈、流失/留存），并关注训练集与验证集之分以及过拟合（overfitting）。对应 Camm et al. (2024) 第 11 章。
Practice / full integrated session综合练习课
A full practice session pulling the prescriptive and predictive toolkit together on integrated problems, in the style of the final exam.
一节综合练习课，把规范式与预测式工具箱在综合题上串起来，题型贴近期末考试。
Exam review考前复习
Revision of the whole subject and exam-style questions. Group Assignment 2 is due this week.
全课程复习与考试题型讲解。小组 Assignment 2 于本周提交。
Oral presentations口头展示
Groups present their Assignment 2 analysis (10 minutes, 5% of the mark), defending their modelling choices and recommendations.
各小组展示其 Assignment 2 的分析（10 分钟，占 5%），为其建模选择与决策建议做说明与答辩。
How MGMT90280 is assessedMGMT90280 怎么考核
Final exam: Yes期末考试：有| Component考核项 | Weight占比 | Note说明 |
|---|---|---|
| Mid-semester test (60 minutes)期中测验（60 分钟） | 20% | Held in the middle of the teaching period.在教学周期中段进行。 |
| Group assignment (groups of 4-6, 5000 words)小组作业（4-6 人一组，5000 字） | 25% | Due in the second half of the teaching period.在教学周期的后半段提交。 |
| Group presentation (groups of 4-6, 10 minutes)小组展示（4-6 人一组，10 分钟） | 5% | Delivered in the second half of the teaching period.在教学周期的后半段进行。 |
| End of semester exam (2 hours)期末考试（2 小时） | 50% | Sat during the official examination period.在官方考试周内进行。 |
Final exam 50% + mid-semester test 20% + group assignment 25% + group presentation 5%.
期末考试 50% + 期中测验 20% + 小组作业 25% + 小组展示 5%。
When each MGMT90280 task is dueMGMT90280 各项考核时间
Test yourself: MGMT90280 practice questions自测一下：MGMT90280 练习题
- The constraint is non-binding and has slack, so extra capacity is worthless.
- The constraint is binding; relaxing its right-hand side by 1 unit improves the optimal objective by 2.9 (within the allowable range).
- The decision variable must increase by 2.9 to stay feasible.
- The objective coefficient can change by up to 2.9 before the solution changes.
- 该约束非紧、有剩余（slack），扩产能没有价值。
- 该约束是紧约束（binding）；右端项每放松 1 单位，最优目标值改善 2.9（在允许范围内）。
- 决策变量需增加 2.9 才能保持可行。
- 目标系数最多可变动 2.9，解才会改变。
Show answer查看答案
- Trend is not significant; drop the Month term and keep only seasonality.
- Trend is significant: holding the seasonal month fixed, sales rise by about 1.017 (units) per month, and p < 0.05 means the trend coefficient is reliably non-zero.
- R-squared must be below 0.5 because the coefficient is small.
- The 1.017 is the forecast for next month.
- 趋势不显著；应删去 Month 项，只保留季节性。
- 趋势显著：在固定季节月份的情况下，销量每月约上升 1.017（单位），且 p < 0.05 说明趋势系数可靠地不为 0。
- 由于系数很小，R 方一定低于 0.5。
- 1.017 就是下个月的预测值。
Show answer查看答案
- Nothing — 99% accuracy proves the model will generalise well.
- The model is likely overfitting: training-only evaluation gives optimistically biased performance, so a validation/test split (or pruning) is needed to estimate true generalisation.
- AUC near 1 means the classes are imbalanced and the tree is invalid.
- Classification trees cannot be evaluated with a ROC curve.
- 没问题——99% 的准确率证明模型泛化能力很强。
- 模型很可能过拟合：仅用训练集评估会给出乐观偏高的表现，需要用验证集/测试集划分（或剪枝）来估计真实泛化能力。
- AUC 接近 1 说明类别不平衡，这棵树无效。
- 分类树不能用 ROC 曲线评估。
Show answer查看答案
High-value exam questions in MGMT90280MGMT90280 高频考点 · 考试风格题
A worked MGMT90280 problemMGMT90280 例题
Signature problem: the Par Inc. product-mix LP招牌例题：Par Inc. 产品组合线性规划
Par Inc. makes two golf bags, Standard (S) and Deluxe (D). Each bag passes through four departments — cutting & dyeing, sewing, finishing, and inspection & packaging — and each department has a fixed number of hours available over the next three months. Each Standard bag contributes $10 profit and each Deluxe bag $9. Given the per-bag hours each department needs and the hours available, how many of each bag should Par produce to maximise total profit contribution?
Par Inc. 生产两款高尔夫球包：标准款（S）与豪华款（D）。每个球包都要经过四个部门 —— 裁剪与染色、缝制、整理、检验与包装 —— 每个部门在未来三个月内可用的工时是固定的。每个标准款贡献 $10 利润、每个豪华款贡献 $9。已知每款球包在各部门所需工时以及各部门的可用工时，Par 应各生产多少个球包才能使总利润贡献最大化？
Formulate it as a linear program. (1) Decision variables: S = number of Standard bags, D = number of Deluxe bags. (2) Objective: Max 10S + 9D. (3) Constraints: one 'hours used <= hours available' inequality per department (cutting & dyeing, sewing, finishing, inspection & packaging), plus non-negativity S, D >= 0. Because the objective and every constraint are linear in S and D, this is a valid LP and can be solved with the Excel Solver Add-in (which runs Dantzig's simplex algorithm to walk the corner points of the feasible region). The real payoff is in reading Solver's sensitivity report: the shadow price of each department tells you how much extra profit one more hour of that resource would buy, so you know which department is the true bottleneck — and a binding constraint with a positive shadow price is where extra capacity is worth paying for, while a non-binding one has slack and a zero shadow price.
把它建模成一个线性规划。（1）决策变量：S = 标准款数量，D = 豪华款数量。（2）目标函数：Max 10S + 9D。（3）约束：每个部门一条「已用工时 <= 可用工时」的不等式（裁剪染色、缝制、整理、检验包装），再加非负约束 S, D >= 0。由于目标函数与每条约束对 S、D 都是线性的，这是一个合法的 LP，可用 Excel Solver 求解（它运行 Dantzig 的单纯形算法，在可行域的角点之间移动寻优）。真正的价值在于解读 Solver 的敏感性报告：每个部门的影子价格（shadow price）告诉你「该资源再多 1 小时能多赚多少利润」，由此判断哪个部门才是真正的瓶颈 —— 紧约束（binding）且影子价格为正之处，才是值得花钱扩产能的地方；而非紧约束有剩余（slack）、影子价格为 0。
MGMT90280 glossaryMGMT90280 术语表
- Prescriptive analytics规范式分析
- Analytics that recommends the best action to take, typically via optimisation models under constraints.
- 推荐"应该采取什么行动"的分析，通常通过带约束的优化模型实现。
- Predictive analytics预测式分析
- Analytics that uses historical data to forecast future outcomes or behaviours.
- 利用历史数据预测未来结果或行为的分析方法。
- Optimisation优化（最优化）
- Finding the decision variables that maximise or minimise an objective subject to constraints.
- 在约束条件下求使目标函数最大化或最小化的决策变量。
- Objective function目标函数
- The quantity (e.g. profit or cost) an optimisation model maximises or minimises.
- 优化模型要最大化或最小化的量（如利润或成本）。
- Constraint约束条件
- A limit on the decision variables (e.g. budget, capacity) that any feasible solution must satisfy.
- 对决策变量的限制（如预算、产能），任何可行解都必须满足。
- Decision variable决策变量
- A quantity the decision-maker controls and the optimisation model solves for.
- 决策者可控制、由优化模型求解的量。
- Sensitivity analysis敏感性分析
- Examining how a model's recommended solution changes as inputs or assumptions vary.
- 考察当输入或假设变化时，模型推荐解如何变化。
- Decision under uncertainty不确定性下的决策
- Choosing among actions when outcomes are probabilistic rather than known with certainty.
- 在结果具有概率性、而非确定已知时，于多个行动方案中作出选择。
- Shadow price影子价格
- In an LP sensitivity report, the change in the optimal objective value per one-unit increase in a constraint's right-hand side — the marginal value of one more unit of that resource.
- 在线性规划敏感性报告中，约束右端项每增加一单位所带来的最优目标值变化 —— 即该资源每多一单位的边际价值。
- Reduced cost既约成本
- For a decision variable that is zero in the optimal solution, how much its objective coefficient must improve before it becomes worth using.
- 对于在最优解中取值为 0 的决策变量，其目标系数需要改善多少，该变量才值得被采用。
- Binding constraint紧约束
- A constraint satisfied with equality at the optimum (no slack); it actively limits the objective and typically has a non-zero shadow price.
- 在最优解处取等号、没有剩余（slack）的约束；它实际制约着目标，且通常有非零的影子价格。
- Integer / binary programming整数 / 0-1 规划
- Optimisation in which some decision variables must take whole-number or yes/no (0-1) values, used for project selection, location and fixed-charge problems.
- 部分决策变量必须取整数或 0/1（是/否）值的优化，用于项目选择、选址与固定费用问题。
- Simplex algorithm单纯形算法
- Dantzig's method (used inside Excel Solver) that finds an LP's optimum by moving between corner points (extreme points) of the feasible region.
- Dantzig 提出的算法（Excel Solver 内部使用），通过在可行域的角点（极点）之间移动来求得线性规划的最优解。
- Monte Carlo simulation蒙特卡洛模拟
- Estimating the distribution of an outcome by drawing many random samples from the input probability distributions and aggregating the results.
- 通过从输入概率分布中抽取大量随机样本并汇总结果，来估计某个结果的分布。
- R-squared (coefficient of determination)R 方（决定系数）
- The proportion of variation in the dependent variable explained by a regression model; closer to 1 means a better fit.
- 回归模型解释的因变量变异占总变异的比例；越接近 1，拟合越好。
- Exponential smoothing指数平滑
- A time-series forecasting method that weights recent observations more heavily than older ones via a smoothing constant.
- 一种时间序列预测方法，通过平滑常数对较近的观测赋予更大权重、对较旧的观测赋予较小权重。
- Clustering聚类
- An unsupervised data-mining technique that segments observations into similar groups based on their observed variables, with no target label.
- 一种无监督数据挖掘技术，依据观测变量把样本划分为相似的群组，无需目标标签。
- Classification tree分类树
- A supervised predictive model that classifies observations by recursively splitting on input variables; managed against overfitting using a validation set.
- 一种有监督的预测模型，通过对输入变量递归划分来对样本分类；用验证集来控制过拟合。
- Excel SolverExcel Solver（规划求解）
- The spreadsheet add-in used throughout MGMT90280 to build and solve LP, ILP and NLP optimisation models and produce sensitivity reports.
- MGMT90280 全程使用的 Excel 加载项，用于构建并求解 LP、ILP 与 NLP 优化模型，并生成敏感性报告。
MGMT90280 — common questionsMGMT90280 常见问题
How is MGMT90280 assessed?MGMT90280 怎么考核？
Does MGMT90280 have a final exam?MGMT90280 有期末考试吗？
What does MGMT90280 actually cover?MGMT90280 到底讲什么？
Are there prerequisites for MGMT90280?MGMT90280 有先修课要求吗？
Is using AskSia allowed under University of Melbourne academic integrity policy?墨尔本大学的学术诚信政策允许使用 AskSia 吗？
Other UniMelb course guides墨大 其他课程指南
AskSia is an independent study aid and is not affiliated with, endorsed by, or sponsored by The University of Melbourne. Course details may change — always confirm against the official handbook. Read about how this guide is built. AskSia 是独立的学习辅助工具，与墨尔本大学没有任何隶属、背书或赞助关系。课程信息可能变动，请始终以官方 handbook 为准。了解本指南的编写方法。