Beyond Aggregation - Real-Time Analytics - page 459

amount of added noise. In the case of Figure 11.1 , a window of 30, perhaps
35 would best recover the underlying signal.

Figure 11.1

Weighted Moving Average

The weighted moving average is a generalization of the standard moving
average that uses different weights for each of the elements in the window.
This collection of weights is known as the kernel ; various kernels are
standard in different industries. The normal moving average is simply a
weighted moving average with a kernel that assigns 1/k to all values.
Implementing a weighted moving average is a little more complex:

public class WeightedMovingAverage {
double [] kernel;
double [] values;
double

kernelSum = 0.0;

int k = 0;

long N = 0;