IB数学AI 线性回归 指数模型 正弦拟合 | IB Maths AI Linear Regression Exponential Models Sinusoidal Fitting

IB数学AI 线性回归 指数模型 正弦拟合

1. 课程概览 Course Overview

IB Mathematics Applications and Interpretation (AI) 的 Modelling 单元是课程中最实用的部分之一。本单元要求你掌握如何从现实世界数据出发，选择合适的数学模型，并使用 GDC 进行计算和验证。Modelling 考题通常出现在 Paper 2 中，占分较重，需要你展示完整的建模流程：假设提出、模型选择、参数确定、拟合优度评估以及预测应用。

The Modelling unit in IB Mathematics Applications and Interpretation (AI) is one of the most practical components of the course. This unit requires you to start from real-world data, select appropriate mathematical models, and use your GDC for computation and validation. Modelling questions typically appear in Paper 2 and carry significant weight, demanding a complete modelling workflow: stating assumptions, selecting models, determining parameters, evaluating goodness of fit, and applying predictions.

2. 建模流程五步法 The Five-Step Modelling Framework

IB 建模的核心是一套标准化的五步流程，每一步都有明确的评分标准。第一步：提出假设 (Assumptions) ： 你需要明确说明你对数据行为的假设，例如「数据呈线性增长趋势」或「增长率随时间递减」。第二步：选择模型 (Model Selection) ： 基于散点图的形状判断应该使用线性、二次、指数、正弦还是 Logistic 模型。第三步：确定参数 (Parameter Determination) ： 使用 GDC 的回归功能 (LinReg, QuadReg, ExpReg, SinReg) 求出模型系数。第四步：验证拟合 (Validation) ： 计算 Pearson 相关系数 r 或决定系数 R²，分析残差图 (residual plot)。第五步：预测与应用 (Prediction) ： 使用模型进行插值 (interpolation) 或外推 (extrapolation)，并评估预测的可靠性。

IB modelling revolves around a standardized five-step framework, with each step carrying explicit marking criteria. Step One: Stating Assumptions -- you must clearly articulate your assumptions about data behaviour, such as "the data exhibits a linear growth trend" or "the growth rate decreases over time." Step Two: Model Selection -- based on the scatter plot's shape, determine whether to use linear, quadratic, exponential, sinusoidal, or logistic models. Step Three: Parameter Determination -- use your GDC's regression functions (LinReg, QuadReg, ExpReg, SinReg) to find model coefficients. Step Four: Validation -- calculate the Pearson correlation coefficient r or coefficient of determination R², and analyze the residual plot. Step Five: Prediction and Application -- use the model for interpolation or extrapolation, and assess prediction reliability.

3. 线性模型与最小二乘法 Linear Models and Least Squares

线性模型是最基础的建模工具，形式为 y = ax + b。GDC 使用最小二乘法 (least squares method) 最小化残差平方和来拟合直线。关键指标：Pearson 相关系数 r 衡量线性相关的强度和方向（-1 ≤ r ≤ 1，|r| 越接近 1 线性关系越强）。决定系数 R² 表示模型解释的数据变异比例。需要注意 r 接近 0 不一定意味着没有关系 ： 数据可能呈非线性模式（如二次或正弦）。常见考题：给定散点图，判断线性拟合是否合适；根据 r 值判断相关性强弱；计算预测值并讨论外推的风险。

The linear model is the most fundamental modelling tool, taking the form y = ax + b. The GDC uses the least squares method to minimize the sum of squared residuals when fitting a line. Key indicators: the Pearson correlation coefficient r measures the strength and direction of linear correlation (-1 ≤ r ≤ 1, with |r| closer to 1 indicating stronger linear relationship). The coefficient of determination R² represents the proportion of data variation explained by the model. Note that r close to 0 does not necessarily mean no relationship exists -- the data may follow a nonlinear pattern (such as quadratic or sinusoidal). Common exam questions: given a scatter plot, judge whether linear fitting is appropriate; determine correlation strength from r values; calculate predicted values and discuss the risks of extrapolation.

4. 二次模型与多项式回归 Quadratic Models and Polynomial Regression

当散点图呈现出明显的弯曲趋势（如先增后减或先减后增），线性模型就不够了，需要二次模型 y = ax² + bx + c。GDC 的 QuadReg 可以直接拟合抛物线。二次模型的关键特征：顶点 (vertex) 代表最大值或最小值点；对称轴 (axis of symmetry) x = -b/(2a)；开口方向由 a 的正负决定。实际应用包括抛体运动轨迹、利润最大化问题、桥梁拱形设计。考试中常要求：从数据判断应该使用二次而非线性模型；解释二次模型参数的实际意义（如重力加速度）；使用顶点坐标解决优化问题。

When the scatter plot shows a clear curved trend (such as increasing then decreasing, or vice versa), a linear model is insufficient, and a quadratic model y = ax² + bx + c is needed. The GDC's QuadReg function can directly fit a parabola. Key features of quadratic models: the vertex represents the maximum or minimum point; the axis of symmetry is x = -b/(2a); the direction of opening is determined by the sign of a. Real-world applications include projectile motion trajectories, profit maximization problems, and bridge arch design. Exams frequently require: judging from data that a quadratic model should be used rather than linear; interpreting the real-world meaning of quadratic parameters (such as gravitational acceleration); using vertex coordinates to solve optimization problems.

5. 指数模型与对数变换 Exponential Models and Logarithmic Transformation

指数模型适用于增长率（或衰减率）近似恒定的数据，形式为 y = a · bˣ 或 y = a · eᵏˣ。典型应用包括：人口增长、放射性衰变、复利计算、细菌繁殖。GDC 的 ExpReg 提供两种拟合模式：y = a·bˣ 和 y = a·eᵏˣ，两者等价（b = eᵏ）。判断是否使用指数模型的关键：计算连续数据点之间的比率，如果近似恒定则适合指数拟合。对数变换技巧：对 y 取自然对数，如果 ln(y) 与 x 呈线性关系，则原始数据适合指数模型 ： 这是 IB 考试中的常见分析题。半衰期 (half-life) 和倍增时间 (doubling time) 是高频考点。

Exponential models are suitable for data with an approximately constant growth rate (or decay rate), taking the form y = a · bˣ or y = a · eᵏˣ. Typical applications include: population growth, radioactive decay, compound interest calculations, and bacterial reproduction. The GDC's ExpReg offers two fitting modes: y = a·bˣ and y = a·eᵏˣ, which are equivalent (b = eᵏ). The key to deciding whether to use an exponential model: calculate the ratio between consecutive data points -- if it is approximately constant, exponential fitting is appropriate. Logarithmic transformation technique: take the natural logarithm of y; if ln(y) shows a linear relationship with x, the original data is suitable for an exponential model -- this is a common analysis question in IB exams. Half-life and doubling time are high-frequency exam topics.

6. 正弦模型与周期性数据 Sinusoidal Models and Periodic Data

正弦模型用于拟合具有周期性波动的数据，形式为 y = a·sin(bx + c) + d 或 y = a·cos(bx + c) + d。四个参数各有明确的几何意义：a 是振幅 (amplitude)，控制波峰波谷的高度；b 与周期 (period) 相关（period = 2π/b 或 360°/b）；c 是相位偏移 (phase shift)；d 是垂直平移 (vertical shift)，即平衡位置。典型应用：温度的年周期变化、潮汐高度、日照时长、声波振动。GDC 的 SinReg 可以自动拟合，但考试中常要求从图像手动提取参数。关键技巧：振幅 = (最大值 - 最小值) / 2；垂直平移 = (最大值 + 最小值) / 2；周期从相邻波峰或波谷的距离读取。

Sinusoidal models are used to fit data with periodic fluctuations, taking the form y = a·sin(bx + c) + d or y = a·cos(bx + c) + d. Each of the four parameters has a clear geometric meaning: a is the amplitude, controlling the height of peaks and troughs; b relates to the period (period = 2π/b or 360°/b); c is the phase shift; d is the vertical shift, representing the equilibrium position. Typical applications: annual temperature cycles, tide heights, daylight duration, and sound wave vibrations. The GDC's SinReg can fit automatically, but exams often require manually extracting parameters from graphs. Key technique: amplitude = (maximum - minimum) / 2; vertical shift = (maximum + minimum) / 2; period is read from the distance between adjacent peaks or troughs.

7. Logistic 模型与增长极限 Logistic Models and Growth Limits

Logistic 模型描述许多真实系统中观察到的 S 形增长曲线：初始阶段缓慢增长，中间阶段快速增长，最终趋于一个上界（承载能力 carrying capacity）。标准形式为 y = L / (1 + C·e⁻ᵏˣ)，其中 L 是最大承载量。典型应用包括：流行病传播（感染人数最终饱和）、技术采用率（从早期使用者到市场饱和）、种群在有限资源下的增长。与指数模型的区别：指数模型假设无限增长，Logistic 模型假设存在上限。GDC 的 Logistic 回归可以直接拟合。IB 考试中的分析题常给出 Logistic 模型方程，要求解释 L 和 k 的实际含义，并判断某时刻的增长率。

Logistic models describe the S-shaped growth curve observed in many real systems: slow growth in the initial phase, rapid growth in the middle phase, and eventual approach to an upper bound (carrying capacity). The standard form is y = L / (1 + C·e⁻ᵏˣ), where L is the maximum capacity. Typical applications include: epidemic spread (number of infections eventually saturating), technology adoption rates (from early adopters to market saturation), and population growth under limited resources. The key difference from exponential models: exponential models assume unlimited growth, while logistic models assume an upper bound exists. The GDC's Logistic regression can fit directly. IB exam analysis questions often provide a logistic model equation and ask you to explain the real-world meaning of L and k, and determine the growth rate at a given time.

8. 模型比较与选择策略 Model Comparison and Selection Strategy

IB 考试中最具挑战性的部分是模型比较 ： 你需要对同一组数据尝试多种模型，并论证哪个最佳。比较工具：首先看 R² 值，越接近 1 拟合越好；其次分析残差图 (residual plot)，残差应随机分布在零线两侧，不应呈现任何模式（如 U 形、喇叭形）。常见陷阱：高阶多项式总能得到更高的 R²，但会导致过拟合 (overfitting) ： 模型复杂度过高而泛化能力差。IB 偏好原则：在拟合优度相近时，选择更简单的模型（奥卡姆剃刀原则）。此外，检查模型在数据范围之外的合理性 ： 例如指数模型预测的人口在 100 年后可能达到天文数字，缺乏现实意义。

The most challenging part of IB exams is model comparison -- you need to try multiple models on the same dataset and justify which is best. Comparison tools: first check R² values, with closer to 1 indicating better fit; second, analyze the residual plot, where residuals should be randomly distributed around the zero line without any discernible pattern (such as U-shaped or fan-shaped). Common pitfall: higher-degree polynomials always yield higher R² but lead to overfitting -- the model has excessive complexity with poor generalization. IB's preferred principle: when goodness of fit is similar, choose the simpler model (Occam's Razor). Additionally, check the model's reasonableness beyond the data range -- for example, an exponential model predicting population might reach astronomical numbers after 100 years, lacking real-world plausibility.

9. GDC 操作技巧 GDC Calculator Techniques

熟练掌握 GDC 是 Modelling 拿高分的前提。关键操作：输入数据到 List (Stat -> Edit)；画散点图 (2nd -> Stat Plot, 选择 scatter, Xlist 和 Ylist)；各种回归：LinReg(ax+b), QuadReg, CubicReg, QuartReg, ExpReg, SinReg, Logistic（需确认你的 GDC 型号是否支持）；存储回归方程到 Y= 以便后续画图和预测。TI-Nspire 用户：使用 Lists & Spreadsheet 输入数据，Data & Statistics 画散点图，Analyze -> Regression 选择模型。特别注意：考试要求给出完整精度（至少 3 位有效数字），不要在中间步骤四舍五入。

Proficiency with your GDC is a prerequisite for scoring well in Modelling. Key operations: input data into Lists (Stat -> Edit); plot scatter diagrams (2nd -> Stat Plot, select scatter, Xlist and Ylist); various regressions: LinReg(ax+b), QuadReg, CubicReg, QuartReg, ExpReg, SinReg, Logistic (check whether your GDC model supports it); store the regression equation to Y= for subsequent graphing and prediction. TI-Nspire users: use Lists & Spreadsheet for data entry, Data & Statistics for scatter plots, Analyze -> Regression to select models. Special note: exams require full precision (at least 3 significant figures) -- do not round in intermediate steps.

10. 常见错误与避坑指南 Common Mistakes and How to Avoid Them

错误一：忘记陈述假设 (Assumptions)。建模题的评分标准中明确包含假设陈述，遗漏直接失分。每次开始建模前，至少写出两条合理的假设。错误二：混淆 r 和 R²。r 衡量线性相关的强度和方向（有正负），R² 衡量模型整体的解释力（始终为正）。报告结果时必须明确区分。错误三：外推 (extrapolation) 不加限制。在已知数据范围之外使用模型进行预测时，必须注明「假设当前趋势持续」并讨论不确定性。错误四：残差分析忽略。只看 R² 不看残差图是严重的建模缺陷 ： 即使 R² 很高，残差的非随机模式也意味着模型选择不当。错误五：参数解释脱离实际。二次模型的顶点坐标必须与实际情境联系（如最高点代表最大利润），不能只给出数学答案。

Mistake One: forgetting to state assumptions. The modelling marking criteria explicitly include assumption statements -- omitting them directly loses marks. Before starting any modelling, write down at least two reasonable assumptions. Mistake Two: confusing r and R². r measures the strength and direction of linear correlation (with sign), while R² measures the overall explanatory power of the model (always positive). You must clearly distinguish between them when reporting results. Mistake Three: extrapolation without caveats. When using a model to predict beyond the known data range, you must note "assuming current trends continue" and discuss uncertainty. Mistake Four: neglecting residual analysis. Checking only R² without looking at the residual plot is a serious modelling flaw -- even with a high R², non-random patterns in residuals indicate inappropriate model choice. Mistake Five: parameter interpretation detached from context. The vertex coordinates of a quadratic model must be connected to the real situation (e.g., the peak represents maximum profit) -- don't just give a mathematical answer.

11. 分阶段学习计划 Phase-Based Study Plan

Phase 1 第1-2周

Linear and Quadratic models + GDC basics. 线性与二次模型，GDC基础操作。完成教材中的基础练习题，确保能独立完成 LinReg 和 QuadReg 操作。

Phase 2 第3-4周

Exponential and Sinusoidal models. 指数与正弦模型。重点练习对数变换法和手动提取正弦参数，这是 Paper 2 的高频考点。

Phase 3 第5-6周

Logistic models + Model Comparison. Logistic模型与模型比较。集中练习评分标准中的比较与论证环节，提升书面表达能力。

Phase 4 第7-8周

Past Paper Practice + Exam Technique. 真题训练与考试技巧。计时完成近5年的 Paper 2 Modelling 题目，标注常见陷阱。

12. 考试答题策略 Exam Answer Strategy

Paper 2 中的 Modelling 题通常是一道大题 (extended response)，分值为 15-20 分，分配约 25-30 分钟作答。答题结构：先写假设 (2-3条)，然后展示散点图判断模型类型，接着给出回归方程及 r/R² 值，随后进行残差分析验证模型，最后使用模型预测并讨论可靠性。时间分配建议：前 5 分钟读题 + 画散点图；中间 15 分钟进行回归计算和验证；最后 5-10 分钟写结论和讨论。记住：过程分比结果分更多 ： 即使最终预测值有偏差，只要建模流程完整且论证充分，仍能获得大部分分数。

Modelling questions in Paper 2 are typically an extended response worth 15-20 marks, with approximately 25-30 minutes allocated. Answer structure: first state assumptions (2-3 items), then present the scatter plot and justify model type, followed by the regression equation with r/R² values, then residual analysis for validation, and finally model-based predictions with a reliability discussion. Suggested time allocation: first 5 minutes for reading the question and plotting the scatter diagram; middle 15 minutes for regression calculations and validation; final 5-10 minutes for conclusions and discussion. Remember: process marks outweigh result marks -- even if your final prediction has some deviation, a complete modelling workflow with thorough justification will still earn most of the marks.

🛍️ 微信小店好物推荐

fx-991CN X 中文版计算器-老王推荐

¥168.00 🛒 微信小店购买

🎓 Need one-on-one tutoring? 需要一对一辅导?

📱 16621398022 同微信

Follow tutorhao on WeChat for more learning resources 关注公众号获取更多学习资源

更多咨询请联系16621398022（同微信）