IB AI: Quadratic and Polynomial Models

Quadratic and polynomial functions are the workhorses of mathematical modeling in the real world. From predicting the trajectory of a ball to optimizing business revenue or analyzing the growth of a population, these models translate observable patterns into precise, usable equations. In the IB AI course, you move beyond abstract manipulation to focus on selecting, applying, and interpreting these models within authentic, data-driven contexts.

The Foundation: Quadratic Function Properties

A quadratic function is any function that can be written in the standard form $y=a{x}^2+bx+c$, where $a$, $b$, and $c$ are constants and $a\ne 0$. Its graph is a symmetric, U-shaped curve called a parabola. The most important features of a parabola are its vertex and its axis of symmetry.

The vertex is the parabola's turning point—either the maximum or minimum value of the function. The axis of symmetry is the vertical line that passes through the vertex, dividing the parabola into two mirror-image halves. A crucial alternative to standard form is the vertex form: $y=a(x-h{)}^2+k$. Here, the vertex is at the point $(h,k)$, and the axis of symmetry is the line $x=h$. Converting between forms, often by completing the square, is a key skill that allows you to immediately identify these central features.

For example, the profit $P$ (in dollars) from selling $x$ units of a product might be modeled by $P(x)=-2x^2+80x-200$. In vertex form, this becomes $P(x)=-2(x-20{)}^2+600$. You can instantly see that the maximum profit of \$600 occurs when 20 units are sold, and the axis of symmetry is $x=20$.

Solving Quadratic Equations and The Discriminant

Finding where a quadratic function equals zero—solving $a{x}^2+bx+c=0$—tells you its x-intercepts, or roots. These often represent critical points in a model, like break-even points in finance or the time an object hits the ground in physics. You have several tools: factoring, using the quadratic formula, and completing the square.

The quadratic formula, $x=\frac{-b\pm \sqrt{b^2-4ac}}{2a}$, always works. The expression under the radical, $b^2-4ac$, is called the discriminant. Discriminant analysis provides critical information about the roots without calculating them:

• If $b^2-4ac>0$, there are two distinct real roots. The parabola crosses the x-axis twice.
• If $b^2-4ac=0$, there is one real, repeated root (a "double root"). The vertex lies on the x-axis.
• If $b^2-4ac<0$, there are no real roots (only complex ones). The parabola does not intersect the x-axis.

In modeling, the discriminant helps you understand a scenario's feasibility. A projectile's height model with a negative discriminant means it never reaches the ground (perhaps it's launched from a cliff into the sea), while a break-even model with a positive discriminant indicates two production levels where profit is zero.

Extending the Model: Cubic and Higher-Degree Polynomials

While quadratics model relationships with a single turning point, many real-world phenomena are more complex. Cubic models, of the form $y=a{x}^3+b{x}^2+cx+d$ where $a\ne 0$, can have up to two turning points and up to three real roots. Their graphs are S-shaped curves that can model situations with an initial period of rapid change that later slows, or vice-versa.

Consider modeling the growth of a bacterial colony with limited resources. Initially, growth is slow (lag phase), then accelerates rapidly (exponential/log phase), and finally slows and stops as resources are depleted (stationary phase). A cubic function can often approximate this S-shaped growth pattern more effectively than a simple quadratic.

Polynomials extend to quartic (degree 4), quintic (degree 5), and beyond. The general polynomial function of degree $n$ is $y=a_n{x}^n+a_{n-1}{x}^{n-1}+...+a_{1}x+a_0$. The degree dictates the maximum number of turning points ($n-1$) and the maximum number of real roots ($n$).

Selecting an Appropriate Polynomial Model for Data

The core applied skill in IB AI is not just solving equations, but choosing the right model for a given data set. This process involves both technological tool use and critical reasoning.

1. Visualize the Data: Always start by plotting the data points. Look for patterns: a single curve (suggesting a quadratic), an S-shape (suggesting a cubic), or multiple inflections (suggesting a higher-degree polynomial).
2. Use Technology for Regression: Your calculator or software can perform polynomial regression. You input the data and specify a degree (e.g., quadratic regression, cubic regression).
3. Evaluate the Fit: The technology will output a coefficient of determination, $R^2$. This value, between 0 and 1, measures how well the model explains the variation in the data. An $R^2$ closer to 1 indicates a better fit.
4. Apply Contextual Judgment: A higher-degree polynomial will always give you a tighter fit to your specific data points. However, this can lead to overfitting—where the model captures random noise instead of the underlying trend, making it useless for prediction outside the original data. The best model is the simplest one (lowest degree) that provides a sufficiently high $R^2$ and, most importantly, makes sense in the real-world context.

For instance, if you are modeling the distance a car travels over time at constant acceleration, physics dictates a quadratic model ($d=\frac{1}{2}a{t}^2+v_{0}t$). Even if a quartic model gives a slightly higher $R^2$ for your noisy measurements, the quadratic is the contextually appropriate choice.

Common Pitfalls

1. Misapplying the Vertex Formula: The x-coordinate of the vertex in standard form is $x=-\frac{b}{2a}$. A common error is to forget the negative sign or to confuse this with the formula for the axis of symmetry. Remember, they are the same: the axis of symmetry is the vertical line $x=-\frac{b}{2a}$.
2. Misinterpreting the Discriminant: Students often think a positive discriminant means the parabola opens upward, or a negative one means it opens downward. The discriminant only speaks to the number of x-intercepts. The sign of the leading coefficient $a$ determines concavity (upward if $a>0$, downward if $a<0$).
3. Overfitting with High-Degree Polynomials: The temptation is to always use the polynomial with the highest $R^2$. This is a critical mistake in modeling. A complex, wavy polynomial that passes through every data point is often less reliable for prediction than a simpler, smoother model that captures the overall trend. Always ask: "Does this model's shape make sense for the story the data is telling?"
4. Ignoring Domain Restrictions: A model might be mathematically valid but contextually nonsense. If your quadratic model for profit gives roots at $x=-5$ and $x=100$, the negative root represents producing -5 units, which is impossible. The domain of your model must be restricted to non-negative values (e.g., $0\le x\le 100$).

Summary

• Quadratic functions model relationships with a single maximum or minimum. Their key features—vertex and axis of symmetry—are easily identified from the vertex form $y=a(x-h{)}^2+k$.
• Solving $a{x}^2+bx+c=0$ finds the roots, with the discriminant ($b^2-4ac$) revealing the number of real solutions, which has important implications for a model's real-world interpretation.
• Cubic and higher-degree polynomial functions can model more complex behavior with multiple turning points, but selecting the right model requires balancing statistical fit ($R^2$) with simplicity and contextual plausibility to avoid overfitting.
• The complete modeling process involves visualizing data, using technology for regression, and applying critical thinking to choose the simplest appropriate model that accurately reflects the underlying trend.