📚 Mathematical Models in IB Applications and Interpretation | IB数学应用与解释中的数学模型
In the IB Mathematics: Applications and Interpretation course, mathematical modelling is a central theme that connects abstract mathematics with real-world contexts. A mathematical model is a simplified representation of a real-life situation, expressed using mathematical structures such as functions, equations, and graphs. Students learn to select, construct, and evaluate models based on data, using technology to perform regression and assess the goodness of fit. This article explores the key types of models encountered in the syllabus, the modelling cycle, and how to interpret and critique models effectively.
在IB数学：应用与解释课程中，数学建模是连接抽象数学与现实情境的核心主题。数学模型是用函数、方程、图形等数学结构对现实情况进行的简化表示。学生需要学会根据数据选择、构建和评估模型，并利用技术进行回归分析和拟合优度检验。本文将探讨课程中涉及的主要模型类型、建模循环，以及如何有效地解释和评判模型。
1. What Is a Mathematical Model? | 什么是数学模型？
A mathematical model uses variables and relationships to describe a system or phenomenon. Instead of capturing every detail, a model focuses on the most relevant features to make predictions or inform decisions. For example, a model for the growth of a plant might ignore daily weather variations and only consider average temperature and sunlight hours. Models can be deterministic, giving precise outcomes, or stochastic, incorporating randomness. In the IB AI course, the emphasis is on using functions that fit observed data to create deterministic models.
数学模型使用变量和关系来描述系统或现象。模型不会捕捉每一个细节，而是聚焦于最相关的特征，以便做出预测或为决策提供依据。例如，植物生长模型可能忽略每日的天气波动，只考虑平均温度和日照时长。模型可以是确定性的，给出精确的结果，也可以是随机的，纳入随机性。在IB AI课程中，重点是使用与观测数据拟合的函数来建立确定性模型。
Common mathematical structures used as models include linear, quadratic, exponential, logarithmic, logistic, sinusoidal, and piecewise functions. Each has a characteristic shape and behaviour, making it suitable for different patterns in data. The art of modelling lies in identifying which function type best represents the underlying trend while acknowledging the assumptions and simplifications made.
用作模型的常见数学结构包括线性函数、二次函数、指数函数、对数函数、逻辑斯蒂函数、正弦函数和分段函数。每种函数都有其特有的形状和行为，因此适用于数据中的不同模式。建模的艺术在于识别哪种函数类型最能体现潜在趋势，同时要认识到所做的假设和简化。
2. The Modelling Cycle | 建模循环
The IB AI syllabus promotes a structured approach to modelling, often represented as a cycle: start with a real-world problem, formulate a mathematical model, compute or analyse results, interpret these in the original context, and then refine the model if needed. This iterative process encourages critical thinking and acknowledges that models are rarely perfect on the first attempt.
IB AI 课程倡导一种结构化的建模方法，通常表现为一个循环：从现实世界的问题开始，建立一个数学模型，计算结果或进行分析，在原始情境中解释这些结果，然后根据需要改进模型。这一迭代过程鼓励批判性思维，并承认模型很少在第一次尝试时就完美无缺。
- Step 1 – Identify the problem: Understand the real-world situation and define the variables. 识别问题：理解现实情况并定义变量。
- Step 2 – Set up the model: Choose a suitable function type and state assumptions. 建立模型：选择合适的函数类型并陈述假设。
- Step 3 – Determine parameters: Use regression or algebraic methods to find constants. 确定参数：使用回归或代数方法求出常数。
- Step 4 – Validate and interpret: Check residuals, r², and reasonableness of outputs. 验证与解释：检查残差、r² 和输出的合理性。
- Step 5 – Refine or apply: If the model is inadequate, adjust and repeat; otherwise, use it for prediction. 改进或应用：如果模型不够好，进行调整并重复；否则用于预测。
Technology, particularly spreadsheets and graphing calculators, is essential throughout this cycle for performing regressions and plotting data rapidly. In internal assessments, students are expected to document this cycle and reflect on their model's strengths and weaknesses.
技术，尤其是电子表格和图形计算器，对于在整个循环中快速执行回归和绘制数据至关重要。在内部评估中，学生需要记录这一循环，并反思模型的优点和不足。
3. Linear Models | 线性模型
A linear model assumes a constant rate of change and takes the form y = mx + c, where m is the gradient and c is the y-intercept. Linear models are the simplest to use and interpret. They apply when data points roughly lie on a straight line, indicating that the response variable changes by a fixed amount for each unit increase in the explanatory variable.
线性模型假定变化率恒定，形式为 y = mx + c，其中 m 是斜率，c 是 y 轴截距。线性模型是最易于使用和解释的。它适用于数据点大致落在一条直线上的情况，表明响应变量随着解释变量每增加一个单位而改变一个固定量。
For example, a spring's extension (E) is directly proportional to the applied force (F) within the elastic limit: E = kF + d, where k is the spring constant and d accounts for any initial reading error. Linear regression can be used to find m and c that minimise the sum of squared residuals. The strength of the fit is quantified by the coefficient of determination r², where a value close to 1 suggests a strong linear relationship.
例如，在弹性限度内，弹簧的伸长量 (E) 与施加的力 (F) 成正比：E = kF + d，其中 k 是弹簧常数，d 用于修正初始读数误差。线性回归可用于求出使残差平方和最小的 m 和 c。拟合的强度由决定系数 r² 量化，r² 接近 1 表明存在很强的线性关系。
It is crucial to remember that a high r² does not prove causation. Many real-life relationships appear linear only over a limited domain, and extrapolation beyond the observed range may lead to meaningless predictions.
必须牢记，高 r² 并不能证明因果关系。许多现实生活中的关系仅在有限的定义域内呈现线性，在观测范围之外进行外推可能导致无意义的预测。
4. Quadratic Models | 二次模型
Quadratic models involve a parabolic curve described by y = ax² + bx + c, with a ≠ 0. The sign of a determines whether the parabola opens upwards (a > 0) or downwards (a < 0). These models suit phenomena that have a single maximum or minimum, such as the trajectory of a projectile, revenue optimisation, or the shape of a suspension cable.
二次模型涉及由 y = ax² + bx + c 描述的抛物线，其中 a ≠ 0。a 的符号决定了抛物线的开口方向：a > 0 向上，a < 0 向下。这类模型适用于具有单一最大值或最小值的现象，例如抛射体的轨迹、收益最优化或悬索的形状。
h(t) = –4.9t² + v₀t + h₀
In this projectile height model, h₀ is the initial height, v₀ is the initial vertical velocity, and –4.9 represents half the gravitational acceleration in m/s². The vertex of the parabola gives the maximum height reached and the positive root represents the time of flight. Quadratic regression on a set of points returns the best-fitting coefficients a, b, c.
在这个抛射体高度模型中，h₀ 是初始高度，v₀ 是初始垂直速度，–4.9 表示重力加速度的一半（单位 m/s²）。抛物线的顶点给出了达到的最高点，正根代表飞行时间。对一组数据点进行二次回归可以得到最佳拟合的系数 a、b、c。
When using a quadratic model, examine the vertex to ensure it makes sense in context (e.g. predicting a negative maximum height is obviously unrealistic). Also, be aware that the parabola is symmetric, which may not be true for the real data; thus, a quadratic model is best used over a restricted domain where symmetry is reasonable.
在使用二次模型时，应检查顶点是否符合实际意义（例如预测出负的最大高度显然不切实际）。此外，要注意抛物线是对称的，但实际数据可能并不对称；因此，二次模型最适合在对称性合理的有限定义域内使用。
5. Exponential Models | 指数模型
Exponential models describe growth or decay where the rate of change is proportional to the current value. The general form is y = abˣ or y = aeᵏˣ, where a is the initial amount, b is the growth/decay factor, and k is the continuous growth rate. If b > 1 (or k > 0), the model represents growth; if 0 < b < 1 (or k < 0), it represents decay.
指数模型描述变化率与当前值成正比的增长或衰减现象。一般形式为 y = abˣ 或 y = aeᵏˣ，其中 a 是初始量，b 是增长/衰减因子，k 是连续增长率。若 b > 1（或 k > 0），模型表示增长；若 0 < b < 1（或 k < 0），则表示衰减。
Typical applications include population growth with unlimited resources, radioactive decay, compound interest, and the cooling of an object. A classic example is the model for the number of bacteria after t hours: N(t) = N₀ × 2t/d king logarithms and applying linear regression to the transformed data (ln y = ln a + kx).
典型应用包括资源无限条件下的人口增长、放射性衰变、复利计算以及物体的冷却。一个经典例子是 t 小时后的细菌数量模型：N(t) = N₀ × 2t/d，其中 d 是倍增时间。指数回归可以通过取对数并对变换后的数据（ln y = ln a + kx）进行线性回归来实现。
Exponential models eventually predict values that grow or decay extremely fast. In reality, unbounded growth is unsustainable, so an exponential model is often a good fit for a limited time frame but requires careful justification when used for long-term predictions.
指数模型最终会预测出极快增长或衰减的值。在现实中，无限制的增长是不可持续的，因此指数模型通常在有限的时间范围内拟合良好，但用于长期预测时需要谨慎说明理由。
6. Logarithmic Models | 对数模型
Logarithmic models have the form y = a + b ln x or y = a + b log₁₀ x. These functions increase rapidly at first and then level off, making them suitable for phenomena where the rate of change decreases as x becomes large. Common examples include the perceived loudness of sound (decibels), the pH scale in chemistry, and the measurement of earthquake magnitudes (Richter scale).
对数模型的形式为 y = a + b ln x 或 y = a + b log₁₀ x。这类函数起初增长迅速，随后趋于平缓，因此适用于变化率随着 x 增大而减小的现象。常见例子包括声音的感知响度（分贝）、化学中的 pH 值标度以及地震震级的测量（里氏震级）。
For instance, the relationship between the magnitude M of an earthquake and the energy E released can be modelled as M = c + d log₁₀ E. A logarithmic model can be fitted to data by taking the logarithm of the independent variable and then applying linear regression. Students should note that the domain of a logarithmic model is x > 0, so any prediction for non-positive x is meaningless.
例如，地震震级 M 与释放的能量 E 之间的关系可以建模为 M = c + d log₁₀ E。对数模型可以通过对自变量取对数，然后应用线性回归来拟合。学生应注意对数模型的定义域是 x > 0，因此对非正数 x 的任何预测都是无意义的。
Logarithmic models are often confused with exponential ones. One way to distinguish is that exponentials model rapid upward or downward trends that accelerate, whereas logarithmic models show a sharp initial change that gradually slows. Examining the data graph helps in choosing the appropriate form.
对数模型常与指数模型混淆。区分的一种方法是，指数模型模拟的是加速上升或下降的趋势，而对数模型则表现出初始剧烈、随后逐渐放缓的变化。观察数据图有助于选择合适的形式。
7. Logistic Models | 逻辑斯蒂模型
The logistic model describes growth that is initially exponential but slows down as it approaches a carrying capacity L. Its general equation is:
y = L / (1 + Ce⁻ᵏˣ)
or equivalently y = L / (1 + e⁻ᵏ⁽ˣ⁻ˣ⁰⁾). Here L is the upper horizontal asymptote, k > 0 controls the steepness, and C relates to the initial value. Logistic curves are S‑shaped and are used extensively in population ecology, the spread of diseases, and the adoption of new technologies.
逻辑斯蒂模型描述最初呈指数增长，但随着接近环境承载力 L 而逐渐减缓的增长过程。其一般方程为：
y = L / (1 + Ce⁻ᵏˣ)
或等价地 y = L / (1 + e⁻ᵏ⁽ˣ⁻ˣ⁰⁾)。其中 L 是上水平渐近线，k > 0 控制曲线的陡峭程度，C 与初始值有关。逻辑斯蒂曲线呈 S 形，广泛应用于种群生态学、疾病传播以及新技术的采用率。
Fitting a logistic model is more complex than linear or exponential regression because it is non‑linear in its parameters. Usually, software is used to estimate L, k, and C simultaneously. The model is valuable because it respects real-world limits, unlike an exponential model which forecasts unbounded growth. When using a logistic model, the predicted carrying capacity L should be justified by reasoning or external constraints.
拟合逻辑斯蒂模型比线性或指数回归更复杂，因为它在参数上是非线性的。通常需要借助软件同时估计 L、k 和 C。该模型的价值在于它尊重现实世界的限制，而指数模型却预测无限制的增长。使用逻辑斯蒂模型时，应通过推理或外部约束来证明预测的承载力 L 的合理性。
8. Sinusoidal Models | 正弦模型
Many natural phenomena repeat in regular cycles, such as temperature over a year, tides, and hours of daylight. These can be modelled using sinusoidal functions of the form:
y = A sin(B(x – C)) + D
rizontal shift, and D is the vertical shift (the midline). Cosine functions can be used interchangeably by adjusting the phase shift.
许多自然现象呈现出规律的周期循环，例如一年中的气温变化、潮汐和日照时长。这些现象可以用以下形式的正弦函数来建模：
y = A sin(B(x – C)) + D
其中 A 是振幅，B 与周期有关（周期 = 2π/B），C 是水平位移，D 是垂直位移（中线）。余弦函数可以通过调整相位差来互换使用。
refining with regression. For example, in modelling average monthly temperatures: A = (max – f the first peak. Sinusoidal regression on a calculator yields values that minimise error.
在建立正弦模型时，学生通常先从数据中估计参数，然后用回归进行优化。例如，在模拟月平均气温时：A = (最大值 – 最小值)/2，D = (最大值 + 最小值)/2，周期为 12 个月，因此 B = 2π/12。水平位移 C 确定第一次高峰出现的时间。计算器上的正弦回归能给出使误差最小化的参数值。
Sinusoidal models assume perfect symmetry between the rise and fall, which may not hold exactly. They work best for data that show a clear, regular wave pattern without long-term upward or downward trend. If a trend exists, a combination of a linear and sinusoidal model may be required.
正弦模型假定上升和下降完全对称，但实际数据可能并非如此。它们最适用于显示出清晰、规律的波形且没有长期上升或下降趋势的数据。如果存在趋势，则可能需要将线性模型与正弦模型结合使用。
9. Piecewise Models | 分段模型
Not all real-world situations can be described by a single equation. Piecewise models use different functions for different intervals of the independent variable. This allows modelling of phenomena that have distinct phases, such as a graduated tax rate, a shipping cost that changes with weight, or the cooling of an object that is re-heated at some point.
并非所有现实情况都能用单一方程描述。分段模型为自变量的不同区间使用不同的函数。这使得我们可以对具有明显不同阶段的现象进行建模，例如分级税率、随重量变化的运费，或在某个时刻被重新加热的物体冷却过程。
An example of a piecewise model for a mobile phone plan:
Cost = { 30 for 0 ≤ data ≤ 5 GB; 30 + 2(x – 5) for x > 5 }
Here the cost is constant for the first 5 GB, then increases linearly with additional usage. When fitting a piecewise model, the boundaries (called knots) must be chosen based on either the context (e.g. tax brackets) or by detecting change points in the data.
以下是手机套餐的分段模型示例：
费用 = { 30，0 ≤ 数据 ≤ 5 GB；30 + 2(x – 5)，x > 5 }
前 5 GB 费用固定，之后随额外用量线性增加。在拟合分段模型时，边界（称为节点）必须根据情境（如税率区间）或通过检测数据中的变化点来选择。
Piecewise models can combine different function types, such as a linear part and a constant part, or two quadratic pieces. While they offer great flexibility, they also introduce more parameters and require careful justification for each segment. Overly complicated piecewise models can overfit the data and lose predictive power.
分段模型可以组合不同类型的函数，例如线性段和常数段，或两个二次段。虽然它们提供了极大的灵活性，但也引入了更多参数，并且需要为每一段提供细致的说明。过于复杂的分段模型可能会过度拟合数据，从而丧失预测能力。
10. Assessing Model Fit and Recognising Limitations | 评估模型拟合与认识局限性
No model is perfect; every model involves assumptions and simplifications that limit its applicability. The IB AI course expects students to evaluate models critically. Two key tools are the coefficient of determination r² and residual plots. A high r² value (close to 1) indicates that a large proportion of the variation in the response variable is accounted for by the model. However, a high r² does not guarantee that the model is appropriate; a curved residual pattern suggests that a different function might fit better, even if r² is high.
没有任何模型是完美的；每个模型都包含假设和简化，这限制了其适用性。IB AI 课程要求学生批判性地评价模型。两个关键工具是决定系数 r² 和残差图。高 r² 值（接近 1）表明响应变量中的大部分变异已被模型解释。但高 r² 并不保证模型是合适的；若残差图呈现弯曲模式，即使 r² 较高，也可能提示另一个函数能更好地拟合数据。
Residuals (observed value – predicted value) should be randomly scattered around zero with no apparent pattern. If they show a trend, the model is missing some underlying structure. For instance, using a linear model for parabolic data produces residuals that form a U‑shape. Analysing residuals thus guides model refinement.
残差（观测值 – 预测值）应随机分布在零附近且无明显模式。如果残差显示出某种趋势，说明模型遗漏了某些潜在结构。例如，对抛物线数据使用线性模型会产生 U 形残差。因此，分析残差可以指导模型的改进。
Limitations of mathematical models include:
- Assumptions may not hold: e.g. a population growth model assuming unlimited resources will fail when resources become scarce. 假设可能不成立：例如，假设资源无限的人口增长模型在资源稀缺时会失效。
- Extrapolation danger: Using a model far beyond the range of observed data can produce absurd predictions. 外推的风险：在远离观测数据范围的区域使用模型可能得出荒谬的预测。
- Oversimplification: Real systems are often affected by numerous variables not included in the model. 过度简化：真实系统往往受到许多模型未包含的变量影响。
- Measurement errors: Inaccurate data lead to unreliable models. 测量误差：不准确的数据会导致模型不可靠。
- Causation vs. correlation: A good fit does not mean the independent variable causes changes in the dependent variable. 因果与相关：良好的拟合并不意味着自变量导致了因变量的变化。
In the internal assessment, students are rewarded for thoughtful reflection on these limitations and for suggesting concrete improvements, such as collecting more data, trying alternative models, or considering confounding variables. By understanding both the power and the pitfalls of mathematical models, IB learners develop the sophistication needed to apply mathematics responsibly in real-world scenarios.
在内部评估中，学生若能深入反思这些局限性，并提出具体的改进建议——例如收集更多数据、尝试其他模型或考虑混杂变量——将会获得好评。通过学习数学模型的威力与缺陷，IB 学生能够培养在现实情境中负责任地应用数学所需的成熟思维。
Published by TutorHao | Mathematics Revision Series | aleveler.com
更多咨询请联系16621398022（同微信）
或截图后用「扫一扫 → 相册」选择本图