← Back to Hub ← 返回主页

Week 7: Model Components (Part 1) 第 7 周:模型组件(一)

Learning Objectives 学习目标

  • Classify analytical models in decision support.
  • Identify decision variables, objective functions, and constraints.
  • Formulate linear programming equations.

Step-by-Step Guide 分步指南

5.1 Model and Analytic

  • Decision Model Definition: A simplified mathematical representation of a real-world business situation to evaluate alternative actions.
  • Core Modeling Variables:
    • Decision Variables: Factors under the manager's control (e.g., production quantity, pricing).
    • Uncontrollable Variables: Factors outside manager's control (e.g., inflation rate, competitor pricing, tax rates).
    • Result Variables: Measures of the model's performance (e.g., total profit, customer satisfaction rate).

Real-world Case: Delta Air Lines

  • Delta uses refueling models to control variables:
  • Decision Variables: Fuel quantity to purchase in each city hub.
  • Uncontrollable Variables: Current spot fuel market price per city.
  • Result Variables: Total annual operational fuel costs.

Reference: How Delta Manages Jet Fuel Costs

5.2 Options for Models

  • Optimization Models: Find the best possible solution to a problem under constraints (e.g., using Linear Programming and Excel Solver).
  • Predictive Models: Project future trends based on historical patterns (e.g., regression forecasting).
  • Descriptive Models: Simulate operations to observe behavioral results without identifying a single optimal path (e.g., queue simulation).
  • Heuristics: Rule-of-thumb models for situations where complete math models are unavailable.

Real-world Case: Zara

  • Zara uses Optimization Models (specifically, constrained linear programming) to allocate inventories to stores globally.
  • The model optimizes stock distribution while factoring in shipping limits and display constraints.

Reference: Zara Constrained Global Inventory Optimization

5.3 Problems of Models

  • Simplification Bias: Mathematical models may oversimplify complex market dynamics, leading to inaccurate results.
  • Assumption Risks: Models rely on static assumptions that fail when unexpected structural shifts happen.
  • Data Quality (GIGO): "Garbage In, Garbage Out" - poor input data renders model outcomes useless.

Real-world Case: Long-Term Capital Management (LTCM)

  • The hedge fund LTCM used sophisticated Nobel-prize-winning risk models to guide investments.
  • However, their models assumed static market volatility, causing a complete fund collapse when Russia defaulted in 1998.

Reference: Financial Models Lessons from LTCM Collapse

Decision Model Categories

Model Class Mathematical Focus Solver Goal Typical Use Cases
Optimization Finds the single best solution Maximize profit or minimize cost Linear Programming (LP) product mix
Descriptive Simulates reality, shows what is happening No optimization; lists outcomes Monte Carlo risk simulation, queue models
Predictive Estimates future outcomes Predict value or classify trend Regression analysis, demand forecasting

Linear Programming (LP) Core Elements

  • Decision Variables: The unknowns to solve for (e.g., x1 = quantity of Table, x2 = quantity of Chair).
  • Objective Function: The goal equation to Maximize or Minimize (e.g., Maximize Profit = 50x1 + 30x2).
  • Constraints: Inequalities representing resource limits (e.g., Labor: 2x1 + 1x2 ≤ 100 hours).

5.1 模型与业务分析

  • 决策模型定义: 现实商业情境的简化数学表达,用于评估不同的行动方案。
  • 三大核心模型变量:
    • 决策变量 (Decision): 管理者控制下的因素(例如生产数量、产品定价)。
    • 不可控变量 (Uncontrollable): 管理者无法控制的外部因素(例如通货膨胀率、竞争对手定价、税率)。
    • 结果变量 (Result): 衡量模型绩效的指标(例如总利润、客户满意率)。

真实企业案例:达美航空 (Delta Air Lines)

  • 达美航空使用加油决策模型来控制变量:
  • 决策变量: 在每个城市枢纽购买的燃料数量。
  • 不可控变量: 每个城市目前的燃料市场现货价格。
  • 结果变量: 年度总燃油运营成本。

参考链接: 达美航空如何控制喷气燃料成本

5.2 模型类别与选择

  • 优化模型 (Optimization): 在约束条件下寻找问题的最佳解决方案(例如,使用线性规划与 Excel Solver 求解)。
  • 预测模型 (Predictive): 基于历史模式预测未来趋势(例如回归预测分析)。
  • 描述性模型 (Descriptive): 模拟业务运行以观察行为结果,但不指出单一的最佳路径(例如,排队系统模拟)。
  • 启发式模型 (Heuristics): 适用于无法建立完整数学模型时的经验法则模型。

真实企业案例:飒拉 (Zara)

  • Zara 使用优化模型(具体为约束线性规划)将库存分配到全球的零售门店。
  • 该模型在考虑运输限制和陈列约束的同时,优化了库存分配。

参考链接: Zara 全球约束库存优化

5.3 决策建模中的问题与局限

  • 简化偏差: 数学模型可能会过度简化复杂的市场动态,导致结果不准确。
  • 假设风险: 模型依赖于静态假设,当发生意外的结构性转变时,这些假设会失效。
  • 数据质量风险 (GIGO): “垃圾进,垃圾出” - 劣质的输入数据会使模型生成的结果毫无价值。

真实企业案例:长期资本管理公司 (LTCM)

  • 对冲基金 LTCM 使用获得诺贝尔奖的复杂风险模型来指导投资。
  • 然而,他们的模型假设市场波动率是静态的,导致在 1998 年俄罗斯债务违约时基金发生彻底崩溃。

参考链接: 对冲基金与金融模型:LTCM 的教训

决策模型大分类

模型类别 数学侧重点 求解目标 典型应用场景
优化模型 寻找唯一的最佳数学解 利润最大化或成本最小化 线性规划 (LP) 产品组合决策
描述性模型 模拟现实,展示系统运行状态 不自动寻找最优解,仅列出结果 蒙特卡洛风险模拟、排队论模型
预测模型 评估未来发展趋势和结果 预测数值或对趋势分类 回归分析、未来产品需求预测

线性规划 (LP) 三大核心要素

  • 决策变量 (Decision Variables): 待求解的未知变量(如 x1 = 桌子的生产数量,x2 = 椅子生产数量)。
  • 目标函数 (Objective Function): 最大化或最小化的目标公式(如 最大化利润 = 50x1 + 30x2)。
  • 约束条件 (Constraints): 表示资源限制的不等式方程(如 人工工时限制:2x1 + 1x2 ≤ 100 小时)。

Concept Visualization 概念可视化

1. Decision Var (X)2. Uncontrollable VarMathematical Model3. Result Var (Y)

Educational Infographic 教学信息图

Visual concept map explaining the key module topic: 解释该模块核心主题的视觉概念图:

Variables (Inputs) e.g. Units Produced Objective Function Maximize Total Profit Constraints Resource limits <= Max

📊 Practical Decision Support Suite (Week 7) 📊 实用决策支持套件(第 7 周)

📗 Excel Solver Add-In Setup for Linear Programming (LP)

Learn how to formulate and solve a product-mix optimization decision in Excel:

  1. Enable Excel Solver Add-in: File → Options → Add-Ins → Manage Excel Add-ins → Check "Solver Add-in".
  2. Set up Objective Cell: =SUMPRODUCT(Unit_Profits, Decision_Variables).
  3. Open Solver (Data Tab → Solver):
    • Set Objective Cell: $G$7 To: Max.
    • By Changing Variable Cells: $F$4:$F$6.
    • Subject to Constraints: $C$4:$C$6 <= $E$4:$E$6 (Dev Hours & Server Limits).
    • Select Solving Method: Simplex LP. Click Solve.

学习如何在 Excel 中构建并求解产品组合优化决策:

  1. 启用 Excel Solver 插件:文件 → 选项 → 加载项 → 管理 Excel 加载项 → 勾选“规划求解加载项 (Solver Add-in)”。
  2. 设置目标单元格:=SUMPRODUCT(单位利润, 决策变量)。
  3. 打开规划求解(数据选项卡 → 规划求解):
    • 设置目标单元格:$G$7 目标为:最大值 (Max)。
    • 通过更改变量单元格:$F$4:$F$6。
    • 遵守约束条件:$C$4:$C$6 <= $E$4:$E$6(工时与服务器上限)。
    • 选择求解方法:单纯形 LP (Simplex LP)。点击求解。

🐍 Python SciPy Optimization Decision Engine

Run this Python script to optimize DSS product licensing allocation under resource constraints:

运行此 Python 脚本,以在资源约束下优化 DSS 产品许可分配:

Tips 提示

Differentiate Decision variables (pricing, quantity under your control) from Uncontrollable variables (inflation, competitors outside your control). This is a highly frequent test topic!
区分决策变量(控制下的定价、生产量)与不可控变量(超出控制的通货膨胀率、竞争对手价格)。这是非常高频的考点!

Knowledge Check 知识自测

1. Inflation rate and tax rate are examples of which variables? 通货膨胀率和税率是哪种变量的示例?

Show Explanation 显示解析
Correct Answer: Option 2 正确选项:选项 2

2. What modeling category finds the best choice under mathematical constraints? 哪种模型类别在数学约束条件下寻找最佳选择?

Show Explanation 显示解析
Correct Answer: Option 1 正确选项:选项 1

3. The term GIGO in decision support modeling stands for: 决策支持建模中的术语 GIGO 代表什么?

Show Explanation 显示解析
Correct Answer: Option 2 正确选项:选项 2

Practical Exercise 实践练习

Task 1: Objective cell formula calculating total profit if units produced are in B1 and profit per unit is 10. 任务 1: Excel 目标单元格公式:如果产量在 B1,每单位利润为 10,计算总利润。

Task 2: SQL statement to select all rows from 'Models' where Type is 'Linear'. 任务 2: SQL 语句:从 'Models' 表中选择 Type 为 'Linear' 的所有行。

Task 3: DAX query to calculate the maximum value in column 'Sales[Amount]'. 任务 3: DAX 查询:计算 'Sales[Amount]' 列中的最大值。

Revision Guide 复习指南

Revision Checklist:

  • I can distinguish between Optimization, Descriptive, and Predictive models.
  • I can define Decision Variables, Constraints, and Objective Functions.
  • I can formulate a complete linear programming mathematical model.

Key Concepts to Remember:

  • Model and Analytic:
    • Decision Model Definition: A simplified mathematical representati...
    • Options for Models:
      • Optimization Models: Find the best possible solution to a problem...
      • Problems of Models:
        • Simplification Bias: Mathematical models may oversimplify complex...

Matching Game 配对游戏

Match the term on the left with its definition on the right. 将左侧的术语与右侧的定义进行配对。

GIGO
Inputs outside management control (e.g. inflation)管理层无法控制的输入(如通货膨胀)
Uncontrollable Variables
Output measures of model performance (e.g. profit)模型性能的输出度量(如利润)
Decision Variables
Poor input data leads to incorrect model results劣质输入数据导致错误的模型结果
Result Variables
Inputs under direct management control (e.g. price)管理层直接控制的输入(如价格)