[Edit]

JavaScript - Mean absolute error (MAE)

1 contributors

2 contributions

0 discussions

4 points

Created by:

mkrieger1

726

In this short article, we would like to show how to calculate Mean absolute error (MAE) using JavaScript.

MAE is kind of error formula that is useful in many cases.

Mean absolute error formula is:

Where:

n is the number of the samples,
y is expected sample values,
x is actual sample values.

Mean absolute error can be used in the cases where we are looking for approximated values - used for solved problems where it is not problem if once x(i) < y(i) and other time x(i) > y(i). That formula can be used in any modeling technique, e.g. Artificial Inteligence, Kriging, any approximation technique, etc.

Practical example

Edit

Let's suppose we have parabola function values: 0, 1, 4, 9 that were approximated by some technique giving: 0.1, 0.9, 3.5, 10. The below source code shows how to calulate MAE for them.

xxxxxxxxxx
 
const calculateMeanAbsoluteError = (y, x) => {
    if (y.length === 0 || y.length !== x.length) {
        return NaN;
    }
    let sum = 0;
    for (let i = 0; i < y.length; ++i) {
        sum += Math.abs(y[i] - x[i]);
    }
    return sum / y.length;
};
​
​
// Usage example:
​
const y = [0,   1,   4,    9]; // expected values
const x = [0.1, 0.9, 3.5, 10]; // actual values
​
const error = calculateMeanAbsoluteError(y, x); // MAE
​
console.log(error);  // 0.425

Auto running

Alternative solution

Edit

This solution is useful when we collect results during computation, and we don't want to create additional temporary array.

xxxxxxxxxx
 
const createMeanAbsoluteErrorCalculator = () => {
    let sum = 0;
    let count = 0;
    return {
        update: (expected, actual) => {
            sum += Math.abs(expected - actual);
            count += 1;
        },
        reset: () => {
            sum = 0;
            count = 0;
        },
        calculate: () => sum / count
    };
};
​
​
// Usage example:
​
const calculator = createMeanAbsoluteErrorCalculator();
​
//         expected  actual
//                |  |
//                v  v
calculator.update(0, 0.1);
calculator.update(1, 0.9);
calculator.update(4, 3.5);
calculator.update(9, 10);
​
const error = calculator.calculate(); // MAE
​
console.log(error);  // 0.425