org.apache.commons.math3.stat.regression.SimpleRegression X 和 Y 轴

Question

数据集 --> [[2019-05-01,4],[2019-05-02,9],[2019-05-03,13],[2019-05-04,17],[2019 -05-05,21],[2019-05-06,25],[2019-05-07,29],[2019-05-08,33],[2019-05-09,37],[2019 -05-10,41],[2019-05-11,45],[2019-05-12,49],[2019-05-13,53],[2019-05-14,57],[2019 -05-15,61],[2019-05-16,65],[2019-05-17,69],[2019-05-18,73],[2019-05-19,77],[2019 -05-20,81],[2019-05-21,85],[2019-05-22,89],[2019-05-23,93],[2019-05-24,97],[2019 -05-25,101],[2019-05-26,105]];

我有一个像上面这样的数据集，其中日期是 x 轴和一些在 y 轴上每天递增的随机数。

如何使用 SimpleRegression java 类 predict() 方法预测未来日期的值。

我被困在这里了。任何想法都会有所帮助。

谢谢。

Answer 1

We can add the date in milliseconds in the X-axis to get the more accurate predicted data. 

Posting the whole code here.


============================================================
import java.time.Instant;
import java.time.temporal.ChronoUnit;

import org.apache.commons.math3.stat.regression.SimpleRegression;

public class TestRegression {
    public static void main(String args[])
    {
        SimpleRegression s = new SimpleRegression();
        Instant now = Instant.now();

        //s.addData(new Date(),3);
        //s.addData(3,4);
        s.addData(now.toEpochMilli(),5);

        now = Instant.now();
        Instant yesterday = now.minus(1, ChronoUnit.DAYS);

        s.addData(yesterday.toEpochMilli(),3);

        now = Instant.now();
        yesterday = now.minus(2, ChronoUnit.DAYS);

        s.addData(yesterday.toEpochMilli(),1);

        now = Instant.now();
        Instant 2daysFromNow = now.plus(2, ChronoUnit.DAYS);
        System.out.println(s.predict(2daysFromNow.toEpochMilli()));

        now = Instant.now();
        Instant nextDay = now.plus(1, ChronoUnit.DAYS);
        System.out.println(s.predict(nextDay.toEpochMilli()));
    }
}


============================================================

Answer 2

如果您已将数据点加载到某个可迭代集合中，则可以使用类似类型的代码进行预测：

public Iterable<DataPoint> getNextDataPoints(long returnTime, Iterator<DataPoint> dataPointRange)
        {
            long start = -1L;
            long stop = -1L;
            DataPoint first = null;
            DataPoint second = null;
            int count = 0;
            SimpleRegression simpleRegression = new SimpleRegression(true);

            while (dataPointRange.hasNext())
            {
                count ++;
                DataPoint dp = dataPointRange.next();
                if (second == null)
                {
                    if (first == null)
                        first = dp;
                    else
                        second = dp;
                }

                stop = dp.getTimestamp();
                if (start == -1L)
                    start = dp.getTimestamp();

                simpleRegression.addData(dp.getTimestamp(), dp.getDoubleValue());
            }

            List<DataPoint> ret = new ArrayList<DataPoint>();

            if (count == 1)
            {
                ret.add(first);
            }
            else if (count == 2)
            {
                ret.add(first);
                ret.add(second);
            }
            else if (count != 0)
            {
                ret.add(m_dataPointFactory.createDataPoint(start, simpleRegression.predict(start)));
                ret.add(m_dataPointFactory.createDataPoint(stop, simpleRegression.predict(stop)));
            }

            return (ret);
        }

感谢kairosdb

另一个例子here