多项式曲线拟合:org.apache.commons.math3.fitting.PolynomialCurveFitter类。
用法示例代码:
[java] view plain copy
- // ... 创建并初始化输入数据:
- double[] x = new double[...];
- double[] y = new double[...];
- 将原始的x-y数据序列合成带权重的观察点数据序列:
- WeightedObservedPoints points = new WeightedObservedPoints();
- // 将x-y数据元素调用points.add(x[i], y[i])加入到观察点序列中
- // ...
- PolynomialCurveFitter fitter = PolynomialCurveFitter.create(degree); // degree 指定多项式阶数
- double[] result = fitter.fit(points.toList()); // 曲线拟合,结果保存于双精度数组中,由常数项至最高次幂系数排列
首先要准备好待拟合的曲线数据x和y,这是两个double数组,然后把这两个数组合并到WeightedObservedPoints对象实例中,可以调用WeightedObservedPoints.add(x[i], y[i])将x和y序列中的数据逐个添加到观察点序列对象中。随后创建PolynomialCurveFitter对象,创建时要指定拟合多项式的阶数,注意阶数要选择适当,不是越高越好,否则拟合误差会很大。最后调用PolynomialCurveFitter的fit方法即可完成多项式曲线拟合,fit方法的参数通过WeightedObservedPoints.toList()获得。拟合结果通过一个double数组返回,按元素顺序依次是常数项、一次项、二次项、……。
完整的演示代码如下:
package fitting; import org.apache.commons.math3.fitting.PolynomialCurveFitter; import org.apache.commons.math3.fitting.WeightedObservedPoints; import java.util.ArrayList; import java.util.List; public class TimeCostCalculator { public static void main(String[] args) throws Exception { TimeCostCalculator tcc = new TimeCostCalculator(); double timeCost = tcc.calcTimeCost(new CalcCurveFitting()); System.out.println("--------------------------------------------------------------------------"); System.out.println("time cost is: " + timeCost + "s"); System.out.println("--------------------------------------------------------------------------"); } /** * 计算指定对象的运行时间开销。 * * @param curveFitting 指定被测对象。 * @return 返回sub.run的时间开销,单位为s。 * @throws Exception */ public double calcTimeCost(CurveFitting curveFitting) throws Exception { List<Object> params = curveFitting.getParams(); long startTime = System.nanoTime(); Object result = curveFitting.run(params); long stopTime = System.nanoTime(); curveFitting.printResult(result); System.out.println("start: " + startTime + " / stop: " + stopTime); return 1.0e-9 * (stopTime - startTime); } } interface CurveFitting { public List<Object> getParams(); public Object run(List<Object> params) throws Exception; public void printResult(Object result); } class CalcCurveFitting implements CurveFitting { private WeightedObservedPoints points; private final int degree = 5; // 阶数 public CalcCurveFitting() { int arrayLength = 200000; System.out.println(String.format("本算例用于计算多项式曲线拟合。正在初始化计算数据(%s点,%s阶......", arrayLength, degree)); double[] inputDataX = new double[arrayLength]; // inputDataX = new double[] {1, 2, 3, 4, 5, 6, 7}; double[] inputDataY = new double[inputDataX.length]; double[] factor = new double[degree + 1]; // N阶多项式会有N+1个系数,其中之一为常数项 for (int index = 0; index < factor.length; index++) { factor[index] = index + 1; } for (int index = 0; index < inputDataY.length; index++) { inputDataX[index] = index * 0.00001; inputDataY[index] = calcPoly(inputDataX[index], factor); // y = sum(x[n) * fact[n]) // System.out.print(inputDataY[index] + ", "); } points = new WeightedObservedPoints(); for (int index = 0; index < inputDataX.length; index++) { points.add(inputDataX[index], inputDataY[index]); } System.out.println("init completely"); } @Override public List<Object> getParams() { List<Object> params = new ArrayList<Object>(); params.add(points); return params; } @Override public Object run(List<Object> params) throws Exception { PolynomialCurveFitter fitter = PolynomialCurveFitter.create(degree); WeightedObservedPoints points = (WeightedObservedPoints) params.get(0); double[] result = fitter.fit(points.toList()); return result; } @Override public void printResult(Object result) { for (double data : (double[]) result) { System.out.println(data); } } private double calcPoly(double x, double[] factor) { double y = 0; for (int deg = 0; deg < factor.length; deg++) { y += Math.pow(x, deg) * factor[deg]; } return y; } }
http://blog.csdn.net/kingfox/article/details/44118319
时间: 2024-10-09 16:44:21