feat(视图): 数据预测

core/backend/src/main/java/io/dataease/commons/utils/DateUtils.java

@@ -1,33 +1,34 @@
 package io.dataease.commons.utils;
 
+import org.apache.commons.lang3.StringUtils;
+
+import java.text.ParseException;
 import java.text.SimpleDateFormat;
-import java.util.Calendar;
+import java.util.*;
-import java.util.Date;
-import java.util.HashMap;
-import java.util.Map;
 
 
 public class DateUtils {
-    public static final String DATE_PATTERM = "yyyy-MM-dd";
+    public static final String DATE_PATTERN = "yyyy-MM-dd";
     public static final String TIME_PATTERN = "yyyy-MM-dd HH:mm:ss";
 
 
     public static Date getDate(String dateString) throws Exception {
-        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERM);
+        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERN);
         return dateFormat.parse(dateString);
     }
+
     public static Date getTime(String timeString) throws Exception {
         SimpleDateFormat dateFormat = new SimpleDateFormat(TIME_PATTERN);
         return dateFormat.parse(timeString);
     }
 
     public static String getDateString(Date date) throws Exception {
-        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERM);
+        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERN);
         return dateFormat.format(date);
     }
 
     public static String getDateString(long timeStamp) throws Exception {
-        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERM);
+        SimpleDateFormat dateFormat = new SimpleDateFormat(DATE_PATTERN);
         return dateFormat.format(timeStamp);
     }
 
@@ -47,10 +48,10 @@ public class DateUtils {
     }
 
 
-    public static Date dateSum (Date date,int countDays){
+    public static Date dateSum(Date date, int countDays) {
         Calendar calendar = Calendar.getInstance();
         calendar.setTime(date);
-        calendar.add(Calendar.DAY_OF_MONTH,countDays);
+        calendar.add(Calendar.DAY_OF_MONTH, countDays);
 
         return calendar.getTime();
     }
@@ -70,7 +71,7 @@ public class DateUtils {
         try {
             calendar.setTime(date);
             calendar.set(Calendar.DAY_OF_WEEK, calendar.getActualMinimum(Calendar.DAY_OF_WEEK));
-            calendar.add(Calendar.DAY_OF_MONTH,weekDayAdd);
+            calendar.add(Calendar.DAY_OF_MONTH, weekDayAdd);
 
             //第一天的时分秒是 00:00:00 这里直接取日期，默认就是零点零分
             Date thisWeekFirstTime = getDate(getDateString(calendar.getTime()));
@@ -78,12 +79,12 @@ public class DateUtils {
             calendar.clear();
             calendar.setTime(date);
             calendar.set(Calendar.DAY_OF_WEEK, calendar.getActualMaximum(Calendar.DAY_OF_WEEK));
-            calendar.add(Calendar.DAY_OF_MONTH,weekDayAdd);
+            calendar.add(Calendar.DAY_OF_MONTH, weekDayAdd);
 
             //最后一天的时分秒应当是23:59:59。 处理方式是增加一天计算日期再-1
-            calendar.add(Calendar.DAY_OF_MONTH,1);
+            calendar.add(Calendar.DAY_OF_MONTH, 1);
             Date nextWeekFirstDay = getDate(getDateString(calendar.getTime()));
-            Date thisWeekLastTime = getTime(getTimeString(nextWeekFirstDay.getTime()-1));
+            Date thisWeekLastTime = getTime(getTimeString(nextWeekFirstDay.getTime() - 1));
 
             returnMap.put("firstTime", thisWeekFirstTime);
             returnMap.put("lastTime", thisWeekLastTime);
@@ -95,9 +96,9 @@ public class DateUtils {
     }
 
 
-
     /**
      * 获取当天的起始时间Date
+     *
      * @param time 指定日期  例： 2020-12-13 06:12:42
      * @return 当天起始时间 例： 2020-12-13 00:00:00
      * @throws Exception
@@ -105,4 +106,81 @@ public class DateUtils {
     public static Date getDayStartTime(Date time) throws Exception {
         return getDate(getDateString(time));
     }
+
+    public static List<String> getForecastPeriod(String baseTime, int period, String dateStyle, String pattern) throws ParseException {
+        String split = "-";
+        if (StringUtils.equalsIgnoreCase(pattern, "date_split")) {
+            split = "/";
+        }
+
+        List<String> result = new ArrayList<>(period);
+        switch (dateStyle) {
+            case "y":
+                int baseYear = Integer.parseInt(baseTime);
+                for (int i = 1; i <= period; i++) {
+                    result.add(baseYear + i + "");
+                }
+                break;
+            case "y_Q":
+                String[] yQ = baseTime.split(split);
+                int year = Integer.parseInt(yQ[0]);
+                int quarter = Integer.parseInt(yQ[1].split("Q")[1]);
+                for (int i = 0; i < period; i++) {
+                    quarter = quarter % 4 + 1;
+                    if (quarter == 1) {
+                        year += 1;
+                    }
+                    result.add(year + split + "Q" + quarter);
+                }
+                break;
+            case "y_M":
+                String[] yM = baseTime.split(split);
+                int y = Integer.parseInt(yM[0]);
+                int month = Integer.parseInt(yM[1]);
+                for (int i = 0; i < period; i++) {
+                    month = month % 12 + 1;
+                    if (month == 1) {
+                        y += 1;
+                    }
+                    String padMonth = month < 10 ? "0" + month : "" + month;
+                    result.add(y + split + padMonth);
+                }
+                break;
+            case "y_W":
+                String[] yW = baseTime.split(split);
+                int yy = Integer.parseInt(yW[0]);
+                int w = Integer.parseInt(yW[1].split("W")[1]);
+                for (int i = 0; i < period; i++) {
+                    Calendar calendar = Calendar.getInstance();
+                    calendar.setMinimalDaysInFirstWeek(7);
+                    calendar.setFirstDayOfWeek(Calendar.MONDAY);
+                    calendar.set(Calendar.YEAR, yy);
+                    calendar.set(Calendar.MONTH, Calendar.DECEMBER);
+                    calendar.set(Calendar.DAY_OF_MONTH, 31);
+                    int lastWeek = calendar.get(Calendar.WEEK_OF_YEAR);
+                    w += 1;
+                    if (w > lastWeek) {
+                        yy += 1;
+                        w = 1;
+                    }
+                    result.add(yy + split + "W" + w);
+                }
+                break;
+            case "y_M_d":
+                SimpleDateFormat sdf = new SimpleDateFormat("yyyy" + split + "MM" + split + "dd");
+                Calendar calendar = Calendar.getInstance();
+                Date baseDate = sdf.parse(baseTime);
+                calendar.setTime(baseDate);
+                for (int i = 0; i < period; i++) {
+                    calendar.add(Calendar.DAY_OF_MONTH, 1);
+                    Date curDate = calendar.getTime();
+                    String date = sdf.format(curDate);
+                    result.add(date);
+                }
+                break;
+            default:
+                break;
+        }
+        return result;
+    }
 }

core/backend/src/main/java/io/dataease/commons/utils/MathUtils.java

@@ -1,8 +1,17 @@
 package io.dataease.commons.utils;
 
+import groovy.lang.Tuple2;
+import org.apache.commons.math4.legacy.stat.regression.SimpleRegression;
+import org.apache.commons.statistics.distribution.NormalDistribution;
+import org.apache.commons.statistics.distribution.TDistribution;
+
 import java.math.BigDecimal;
+import java.math.RoundingMode;
+import java.util.ArrayList;
+import java.util.List;
 
 public class MathUtils {
+    private static final NormalDistribution NORMAL_DISTRIBUTION = NormalDistribution.of(0.0, 1.0);
 
     /**
      * 获取百分比
@@ -12,8 +21,57 @@ public class MathUtils {
      * @return
      */
     public static double getPercentWithDecimal(double value) {
-        return new BigDecimal(value * 100)
+        return new BigDecimal(value * 100).setScale(1, RoundingMode.HALF_UP).doubleValue();
-                .setScale(1, BigDecimal.ROUND_HALF_UP)
-                .doubleValue();
     }
+
+    /**
+     * 获取预测数据的置信区间，这边计算的是预测值的置信区间，还有一个是预测值的预测区间，公式不一样，注意区分.
+     * 参考资料 <a href="https://zhuanlan.zhihu.com/p/366307027">知乎</a>,
+     * <a href="https://real-statistics.com/regression/confidence-and-prediction-intervals/">real-statistics</a>
+     * @param data 原始数据
+     * @param forecastValue 预测得到的数据
+     * @param forecastData 将原数据 x 代入回归方程得到的拟合数据
+     * @param alpha 置信水平
+     * @param degreeOfFreedom 自由度，t分布使用
+     * @return 预测值的置信区间数组
+     */
+    public static double[][] getConfidenceInterval(double[][] data, double[] forecastValue, double[][] forecastData, double alpha, int degreeOfFreedom) {
+        // y 平均方差
+        double totalPow = 0;
+        double xTotal = 0;
+        for (int i = 0; i < data.length; i++) {
+            double xVal = data[i][0];
+            xTotal += xVal;
+            double realVal = data[i][1];
+            double predictVal = forecastValue[i];
+            totalPow += Math.pow((realVal - predictVal), 2);
+        }
+        double xAvg = xTotal / data.length;
+        double yMseSqrt = Math.sqrt(totalPow / (forecastValue.length - 2));
+        // x 均值方差
+        double xSubPow = 0;
+        for (int i = 0; i < data.length; i++) {
+            double xVal = data[i][0];
+            xSubPow += Math.pow(xVal - xAvg, 2);
+        }
+        // t/z 值, 样本数 < 30 选 t 分布， > 30 选 z 分布,
+        double tzFactor;
+        if (data.length <= 30) {
+            tzFactor = TDistribution.of(degreeOfFreedom).inverseCumulativeProbability(1 - (1 - alpha) / 2);
+        } else {
+            tzFactor = NORMAL_DISTRIBUTION.inverseCumulativeProbability(1 - (1 - alpha) / 2);
+        }
+        double[][] result = new double[forecastData.length][2];
+        for (int i = 0; i < forecastData.length; i++) {
+            double xVal = forecastData[i][0];
+            double curSubPow = Math.pow(xVal - xAvg, 2);
+            double sqrt = Math.sqrt(1.0 / data.length + curSubPow / xSubPow);
+            double lower = forecastData[i][1] - tzFactor * yMseSqrt * sqrt;
+            double upper = forecastData[i][1] + tzFactor * yMseSqrt * sqrt;
+            result[i][0] = lower;
+            result[i][1] = upper;
+        }
+        return result;
+    }
+
 }

core/backend/src/main/java/io/dataease/dto/chart/ChartSeniorForecastDTO.java

@@ -0,0 +1,35 @@
+package io.dataease.dto.chart;
+
+import lombok.Data;
+
+@Data
+public class ChartSeniorForecastDTO {
+    /**
+     * 是否开启预测
+     */
+    private boolean enable;
+    /**
+     * 预测周期
+     */
+    private int period;
+    /**
+     * 是否使用所有数据进行预测
+     */
+    private boolean allPeriod;
+    /**
+     * 用于预测的数据量
+     */
+    private int trainingPeriod;
+    /**
+     * 置信区间
+     */
+    private float confidenceInterval;
+    /**
+     * 预测用的算法/模型
+     */
+    private String algorithm;
+    /**
+     * 多项式阶数
+     */
+    private int degree;
+}

core/backend/src/main/java/io/dataease/dto/chart/ForecastDataDTO.java

@@ -0,0 +1,14 @@
+package io.dataease.dto.chart;
+
+import com.fasterxml.jackson.annotation.JsonIgnore;
+import lombok.Data;
+
+@Data
+public class ForecastDataDTO {
+    @JsonIgnore
+    private double xVal;
+    @JsonIgnore
+    private double yVal;
+    private double lower;
+    private double upper;
+}

core/backend/src/main/java/io/dataease/dto/chart/ForecastDataVO.java

@@ -0,0 +1,11 @@
+package io.dataease.dto.chart;
+
+import lombok.Data;
+import lombok.EqualsAndHashCode;
+
+@Data
+@EqualsAndHashCode(callSuper = true)
+public class ForecastDataVO<D, Q> extends ForecastDataDTO {
+    private D dimension;
+    private Q quota;
+}

core/backend/src/main/java/io/dataease/service/chart/ChartViewService.java

@@ -11,6 +11,7 @@ import io.dataease.commons.constants.JdbcConstants;
 import io.dataease.commons.model.PluginViewSetImpl;
 import io.dataease.commons.utils.AuthUtils;
 import io.dataease.commons.utils.BeanUtils;
+import io.dataease.commons.utils.DateUtils;
 import io.dataease.commons.utils.LogUtil;
 import io.dataease.controller.request.chart.*;
 import io.dataease.controller.response.ChartDetail;
@@ -52,6 +53,8 @@ import io.dataease.plugins.view.service.ViewPluginService;
 import io.dataease.plugins.xpack.auth.dto.request.ColumnPermissionItem;
 import io.dataease.provider.query.SQLUtils;
 import io.dataease.service.chart.util.ChartDataBuild;
+import io.dataease.service.chart.util.dataForecast.ForecastAlgo;
+import io.dataease.service.chart.util.dataForecast.ForecastAlgoManager;
 import io.dataease.service.dataset.*;
 import io.dataease.service.datasource.DatasourceService;
 import io.dataease.service.engine.EngineService;
@@ -71,6 +74,7 @@ import javax.annotation.Resource;
 import java.lang.reflect.Type;
 import java.math.BigDecimal;
 import java.math.RoundingMode;
+import java.text.ParseException;
 import java.text.SimpleDateFormat;
 import java.util.*;
 import java.util.concurrent.atomic.AtomicReference;
@@ -1090,7 +1094,7 @@ public class ChartViewService {
             logger.info("plugin_sql:" + sql);
             Map<String, Object> mapTableNormal = ChartDataBuild.transTableNormal(fieldMap, view, data, desensitizationList);
 
-            return uniteViewResult(datasourceRequest.getQuery(), mapChart, mapTableNormal, view, isDrill, drillFilters, dynamicAssistFields, assistData);
+            return uniteViewResult(datasourceRequest.getQuery(), mapChart, mapTableNormal, view, isDrill, drillFilters, dynamicAssistFields, assistData, Collections.emptyList());
             // 如果是插件到此结束
         }
 
@@ -1361,6 +1365,16 @@ public class ChartViewService {
         }
         tempYAxis.addAll(yAxis);
 
+        // forecast
+        List<? extends ForecastDataVO<?, ?>> forecastData = Collections.emptyList();
+        JSONObject senior = JSONObject.parseObject(view.getSenior());
+        JSONObject forecastObj = senior.getJSONObject("forecast");
+        if (forecastObj != null) {
+            ChartSeniorForecastDTO forecastCfg = forecastObj.toJavaObject(ChartSeniorForecastDTO.class);
+            if (forecastCfg.isEnable()) {
+                forecastData = forecastData(forecastCfg, data, xAxis, yAxis, view);
+            }
+        }
         for (int i = 0; i < tempYAxis.size(); i++) {
             ChartViewFieldDTO chartViewFieldDTO = tempYAxis.get(i);
             ChartFieldCompareDTO compareCalc = chartViewFieldDTO.getCompareCalc();
@@ -1622,12 +1636,43 @@ public class ChartViewService {
 
             mapTableNormal = ChartDataBuild.transTableNormal(xAxis, yAxis, view, data, extStack, desensitizationList);
         }
-        chartViewDTO = uniteViewResult(datasourceRequest.getQuery(), mapChart, mapTableNormal, view, isDrill, drillFilters, dynamicAssistFields, assistData);
+        chartViewDTO = uniteViewResult(datasourceRequest.getQuery(), mapChart, mapTableNormal, view, isDrill, drillFilters, dynamicAssistFields, assistData, forecastData);
         chartViewDTO.setTotalPage(totalPage);
         chartViewDTO.setTotalItems(totalItems);
         return chartViewDTO;
     }
 
+    private List<? extends ForecastDataVO<?,?>> forecastData(ChartSeniorForecastDTO forecastCfg, List<String[]> data, List<ChartViewFieldDTO> xAxis, List<ChartViewFieldDTO> yAxis, ChartViewDTO view) throws ParseException {
+        List<String[]> trainingData = data;
+        if (!forecastCfg.isAllPeriod() && data.size() > forecastCfg.getTrainingPeriod()) {
+            trainingData = data.subList(data.size() - forecastCfg.getTrainingPeriod(), data.size() - 1);
+        }
+        if (xAxis.size() == 1 && xAxis.get(0).getDeType() == 1) {
+            // 先处理时间类型, 默认数据是有序递增的
+            String lastTime = data.get(data.size() - 1)[0];
+            ChartViewFieldDTO timeAxis = xAxis.get(0);
+            List<String> forecastPeriod = DateUtils.getForecastPeriod(lastTime, forecastCfg.getPeriod(), timeAxis.getDateStyle(), timeAxis.getDatePattern());
+            if(!forecastPeriod.isEmpty()){
+                ForecastAlgo algo = ForecastAlgoManager.getAlgo(forecastCfg.getAlgorithm());
+                List<ForecastDataDTO> forecastData = algo.forecast(forecastCfg, trainingData, view);
+                if (forecastPeriod.size() == forecastData.size()) {
+                    List<ForecastDataVO<String, Double>> result = new ArrayList<>();
+                    for (int i = 0; i < forecastPeriod.size(); i++) {
+                        String period = forecastPeriod.get(i);
+                        ForecastDataDTO forecastDataItem = forecastData.get(i);
+                        ForecastDataVO<String, Double> tmp = new ForecastDataVO<>();
+                        BeanUtils.copyBean(tmp, forecastDataItem);
+                        tmp.setDimension(period);
+                        tmp.setQuota(forecastDataItem.getYVal());
+                        result.add(tmp);
+                    }
+                    return result;
+                }
+            }
+        }
+        return List.of();
+    }
+
     // 对结果排序
     public List<String[]> resultCustomSort(List<ChartViewFieldDTO> xAxis, List<String[]> data) {
         List<String[]> res = new ArrayList<>(data);
@@ -1684,11 +1729,12 @@ public class ChartViewService {
         return "SELECT " + stringBuilder + " FROM (" + sql + ") tmp";
     }
 
-    public ChartViewDTO uniteViewResult(String sql, Map<String, Object> chartData, Map<String, Object> tableData, ChartViewDTO view, Boolean isDrill, List<ChartExtFilterRequest> drillFilters, List<ChartSeniorAssistDTO> dynamicAssistFields, List<String[]> assistData) {
+    public ChartViewDTO uniteViewResult(String sql, Map<String, Object> chartData, Map<String, Object> tableData, ChartViewDTO view, Boolean isDrill, List<ChartExtFilterRequest> drillFilters, List<ChartSeniorAssistDTO> dynamicAssistFields, List<String[]> assistData, List<? extends ForecastDataVO<?, ?>> forecastData) {
 
         Map<String, Object> map = new HashMap<>();
         map.putAll(chartData);
         map.putAll(tableData);
+        map.put("forecastData", forecastData);
 
         List<DatasetTableField> sourceFields = dataSetTableFieldsService.getFieldsByTableId(view.getTableId());
         map.put("sourceFields", sourceFields);

core/backend/src/main/java/io/dataease/service/chart/util/dataForecast/ForecastAlgo.java

@@ -0,0 +1,21 @@
+package io.dataease.service.chart.util.dataForecast;
+
+import io.dataease.dto.chart.ChartSeniorForecastDTO;
+import io.dataease.dto.chart.ChartViewDTO;
+import io.dataease.dto.chart.ForecastDataDTO;
+import lombok.Getter;
+
+import java.util.List;
+
+@Getter
+public abstract class ForecastAlgo {
+    private final String algoType;
+
+    public ForecastAlgo() {
+        this.algoType = this.getAlgoType();
+        ForecastAlgoManager.register(this);
+    }
+
+    public abstract List<ForecastDataDTO> forecast(ChartSeniorForecastDTO forecastCfg, List<String[]> data, ChartViewDTO view);
+
+}

core/backend/src/main/java/io/dataease/service/chart/util/dataForecast/ForecastAlgoManager.java

@@ -0,0 +1,17 @@
+package io.dataease.service.chart.util.dataForecast;
+
+import io.dataease.service.chart.util.dataForecast.impl.LinearRegressionAlgo;
+
+import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
+
+public class ForecastAlgoManager {
+    private static final Map<String, ForecastAlgo> FORECAST_ALGO_MAP = new ConcurrentHashMap<>();
+    public static ForecastAlgo getAlgo(String algoType) {
+        return FORECAST_ALGO_MAP.get(algoType);
+    }
+
+    public static void register(ForecastAlgo forecastAlgo) {
+        FORECAST_ALGO_MAP.put(forecastAlgo.getAlgoType(), forecastAlgo);
+    }
+}

core/backend/src/main/java/io/dataease/service/chart/util/dataForecast/impl/LinearRegressionAlgo.java

@@ -0,0 +1,59 @@
+package io.dataease.service.chart.util.dataForecast.impl;
+
+import com.alibaba.fastjson.JSONObject;
+import io.dataease.commons.utils.MathUtils;
+import io.dataease.dto.chart.ChartSeniorForecastDTO;
+import io.dataease.dto.chart.ChartViewDTO;
+import io.dataease.dto.chart.ForecastDataDTO;
+import io.dataease.plugins.common.dto.chart.ChartViewFieldDTO;
+import io.dataease.service.chart.util.dataForecast.ForecastAlgo;
+import lombok.Getter;
+import org.apache.commons.math4.legacy.stat.regression.SimpleRegression;
+import org.springframework.stereotype.Component;
+
+import java.util.ArrayList;
+import java.util.List;
+
+@Component
+@Getter
+public class LinearRegressionAlgo extends ForecastAlgo {
+
+    private final String algoType = "linear-regression";
+
+    @Override
+    public List<ForecastDataDTO> forecast(ChartSeniorForecastDTO forecastCfg, List<String[]> data, ChartViewDTO view) {
+        List<ChartViewFieldDTO> xAxis = JSONObject.parseArray(view.getXAxis(), ChartViewFieldDTO.class);
+        final List<double[]> forecastData = new ArrayList<>(data.size());
+        // 先按连续的数据处理
+        for (int i = 0; i < data.size(); i++) {
+            String val = data.get(i)[xAxis.size()];
+            double value = Double.parseDouble(val);
+            forecastData.add(new double[]{i, value});
+        }
+        double[][] matrix = forecastData.toArray(double[][]::new);
+        SimpleRegression regression = new SimpleRegression();
+        regression.addData(matrix);
+        double[][] forecastMatrix = new double[forecastCfg.getPeriod()][2];
+        for (int i = 0; i < forecastCfg.getPeriod(); i++) {
+            double xVal = data.size() + i;
+            double predictVal = regression.predict(xVal);
+            forecastMatrix[i] =  new double[]{xVal, predictVal};
+        }
+        final double[] forecastValue = new double[forecastData.size()];
+        for (int i = 0; i < forecastData.size(); i++) {
+            double xVal = forecastData.get(i)[0];
+            forecastValue[i] = regression.predict(xVal);
+        }
+        double[][] confidenceInterval = MathUtils.getConfidenceInterval(forecastData.toArray(new double[0][0]), forecastValue, forecastMatrix, forecastCfg.getConfidenceInterval(), forecastData.size() - 2);
+        final List<ForecastDataDTO> result = new ArrayList<>(forecastCfg.getPeriod());
+        for (int i = 0; i < forecastMatrix.length; i++) {
+            ForecastDataDTO tmp = new ForecastDataDTO();
+            tmp.setXVal(forecastMatrix[i][0]);
+            tmp.setYVal(forecastMatrix[i][1]);
+            tmp.setLower(confidenceInterval[i][0]);
+            tmp.setUpper(confidenceInterval[i][1]);
+            result.add(tmp);
+        }
+        return result;
+    }
+}

core/backend/src/main/java/io/dataease/service/chart/util/dataForecast/impl/PolynomialRegressionAlgo.java

@@ -0,0 +1,71 @@
+package io.dataease.service.chart.util.dataForecast.impl;
+
+import com.alibaba.fastjson.JSONObject;
+import io.dataease.commons.utils.MathUtils;
+import io.dataease.dto.chart.ChartSeniorForecastDTO;
+import io.dataease.dto.chart.ChartViewDTO;
+import io.dataease.dto.chart.ForecastDataDTO;
+import io.dataease.plugins.common.dto.chart.ChartViewFieldDTO;
+import io.dataease.service.chart.util.dataForecast.ForecastAlgo;
+import lombok.Getter;
+import org.apache.commons.math4.legacy.analysis.function.Logistic;
+import org.apache.commons.math4.legacy.fitting.PolynomialCurveFitter;
+import org.apache.commons.math4.legacy.fitting.WeightedObservedPoints;
+import org.springframework.stereotype.Component;
+
+import java.util.ArrayList;
+import java.util.List;
+
+@Component
+@Getter
+public class PolynomialRegressionAlgo extends ForecastAlgo {
+    private final String algoType = "polynomial-regression";
+
+    @Override
+    public List<ForecastDataDTO> forecast(ChartSeniorForecastDTO forecastCfg, List<String[]> data, ChartViewDTO view) {
+        List<ChartViewFieldDTO> xAxis = JSONObject.parseArray(view.getXAxis(), ChartViewFieldDTO.class);
+        WeightedObservedPoints points = new WeightedObservedPoints();
+        double[][] originData = new double[data.size()][2];
+        // 先按连续的数据处理
+        for (int i = 0; i < data.size(); i++) {
+            String val = data.get(i)[xAxis.size()];
+            double value = Double.parseDouble(val);
+            points.add(i, value);
+            originData[i] = new double[]{i, value};
+        }
+        PolynomialCurveFitter filter = PolynomialCurveFitter.create(forecastCfg.getDegree());
+        // 返回的是多次项系数, y = 3 + 2x + x*2 则为 [3,2,1]
+        double[] coefficients = filter.fit(points.toList());
+        double[][] forecastMatrix = new double[forecastCfg.getPeriod()][2];
+        for (int i = 0; i < forecastCfg.getPeriod(); i++) {
+            double xVal = data.size() + i;
+            double predictVal = getPolynomialValue(xVal, coefficients);
+            forecastMatrix[i] = new double[]{xVal, predictVal};
+        }
+        final double[] forecastValue = new double[data.size()];
+        for (int i = 0; i < data.size(); i++) {
+            double xVal = originData[i][0];
+            forecastValue[i] = getPolynomialValue(xVal, coefficients);
+        }
+        int df = data.size() - forecastCfg.getDegree() - 1;
+        double[][] confidenceInterval = MathUtils.getConfidenceInterval(originData, forecastValue, forecastMatrix, forecastCfg.getConfidenceInterval(), df);
+        final List<ForecastDataDTO> result = new ArrayList<>(forecastCfg.getPeriod());
+        for (int i = 0; i < forecastMatrix.length; i++) {
+            ForecastDataDTO tmp = new ForecastDataDTO();
+            tmp.setXVal(forecastMatrix[i][0]);
+            tmp.setYVal(forecastMatrix[i][1]);
+            tmp.setLower(confidenceInterval[i][0]);
+            tmp.setUpper(confidenceInterval[i][1]);
+            result.add(tmp);
+        }
+        return result;
+    }
+
+    private double getPolynomialValue(double x, double[] coefficients) {
+        double result = 0.0;
+        for (int i = 0; i < coefficients.length; i++) {
+            result += coefficients[i] * Math.pow(x, i);
+        }
+        return result;
+    }
+}