|
|
@@ -1,6 +1,7 @@
|
|
|
package correlation
|
|
|
|
|
|
import (
|
|
|
+ "errors"
|
|
|
"fmt"
|
|
|
"github.com/shopspring/decimal"
|
|
|
"hongze/hongze_ETA_mobile_api/models/data_manage"
|
|
|
@@ -210,3 +211,293 @@ func GetChartAndCorrelationInfo(chartInfoId int) (chartInfo *data_manage.ChartIn
|
|
|
}
|
|
|
return
|
|
|
}
|
|
|
+
|
|
|
+// GetChartDataByEdbInfo 相关性图表-根据指标信息获取x轴和y轴
|
|
|
+func GetChartDataByEdbInfo(edbInfoMappingA, edbInfoMappingB *data_manage.ChartEdbInfoMapping, leadValue int, leadUnit, startDate, endDate string) (xEdbIdValue []int, yDataList []data_manage.YData, err error) {
|
|
|
+ xData := make([]int, 0)
|
|
|
+ yData := make([]float64, 0)
|
|
|
+ if leadValue == 0 {
|
|
|
+ xData = append(xData, 0)
|
|
|
+ }
|
|
|
+ if leadValue > 0 {
|
|
|
+ leadMin := 0 - leadValue
|
|
|
+ xLen := 2*leadValue + 1
|
|
|
+ for i := 0; i < xLen; i++ {
|
|
|
+ n := leadMin + i
|
|
|
+ xData = append(xData, n)
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ // 计算窗口，不包含第一天
|
|
|
+ startDateTime, _ := time.ParseInLocation(utils.FormatDate, startDate, time.Local)
|
|
|
+ startDate = startDateTime.AddDate(0, 0, 1).Format(utils.FormatDate)
|
|
|
+
|
|
|
+ //// 2023-03-02 时间序列始终以指标B为基准, 始终是A进行平移
|
|
|
+ //baseEdbInfo := edbInfoMappingB
|
|
|
+ //changeEdbInfo := edbInfoMappingA
|
|
|
+ // 2023-03-17 时间序列始终以指标A为基准, 始终是B进行平移
|
|
|
+ baseEdbInfo := edbInfoMappingA
|
|
|
+ changeEdbInfo := edbInfoMappingB
|
|
|
+
|
|
|
+ // 获取时间基准指标在时间区间内的值
|
|
|
+ aDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ switch baseEdbInfo.EdbInfoCategoryType {
|
|
|
+ case 0:
|
|
|
+ aDataList, err = data_manage.GetEdbDataList(baseEdbInfo.Source, baseEdbInfo.EdbInfoId, startDate, endDate)
|
|
|
+ case 1:
|
|
|
+ _, aDataList, _, _, err, _ = data.GetPredictDataListByPredictEdbInfoId(baseEdbInfo.EdbInfoId, startDate, endDate, false)
|
|
|
+ default:
|
|
|
+ err = errors.New("指标base类型异常")
|
|
|
+ return
|
|
|
+ }
|
|
|
+
|
|
|
+ // 获取变频指标所有日期的值, 插值法完善数据
|
|
|
+ bDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ switch changeEdbInfo.EdbInfoCategoryType {
|
|
|
+ case 0:
|
|
|
+ bDataList, err = data_manage.GetEdbDataList(changeEdbInfo.Source, changeEdbInfo.EdbInfoId, "", "")
|
|
|
+ case 1:
|
|
|
+ _, bDataList, _, _, err, _ = data.GetPredictDataListByPredictEdbInfoId(changeEdbInfo.EdbInfoId, "", "", false)
|
|
|
+ default:
|
|
|
+ err = errors.New("指标change类型异常")
|
|
|
+ return
|
|
|
+ }
|
|
|
+ //changeDataMap := make(map[string]float64)
|
|
|
+ //newChangeDataList, e := HandleDataByLinearRegression(bDataList, changeDataMap)
|
|
|
+ //if e != nil {
|
|
|
+ // err = fmt.Errorf("获取变频指标插值法Map失败, Err: %s", e.Error())
|
|
|
+ // return
|
|
|
+ //}
|
|
|
+
|
|
|
+ // 2023-03-17 时间序列始终以指标A为基准, 始终是B进行平移
|
|
|
+ baseDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ baseDataMap := make(map[string]float64)
|
|
|
+ changeDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ changeDataMap := make(map[string]float64)
|
|
|
+
|
|
|
+ // 先把低频指标升频为高频
|
|
|
+ {
|
|
|
+ frequencyIntMap := map[string]int{
|
|
|
+ "日度": 1,
|
|
|
+ "周度": 2,
|
|
|
+ "旬度": 3,
|
|
|
+ "月度": 4,
|
|
|
+ "季度": 5,
|
|
|
+ "年度": 6,
|
|
|
+ }
|
|
|
+
|
|
|
+ // 如果A指标是高频，那么就需要对B指标进行升频
|
|
|
+ if frequencyIntMap[edbInfoMappingA.Frequency] < frequencyIntMap[edbInfoMappingB.Frequency] {
|
|
|
+ tmpNewChangeDataList, e := HandleDataByLinearRegression(aDataList, baseDataMap)
|
|
|
+ if e != nil {
|
|
|
+ err = fmt.Errorf("获取变频指标插值法Map失败, Err: %s", e.Error())
|
|
|
+ return
|
|
|
+ }
|
|
|
+ baseDataList = tmpNewChangeDataList
|
|
|
+ } else {
|
|
|
+ baseDataList = aDataList
|
|
|
+ for _, v := range baseDataList {
|
|
|
+ baseDataMap[v.DataTime] = v.Value
|
|
|
+ }
|
|
|
+ }
|
|
|
+ // 如果B指标是高频，那么就需要对A指标进行升频
|
|
|
+ if frequencyIntMap[edbInfoMappingA.Frequency] > frequencyIntMap[edbInfoMappingB.Frequency] {
|
|
|
+ tmpNewChangeDataList, e := HandleDataByLinearRegression(bDataList, changeDataMap)
|
|
|
+ if e != nil {
|
|
|
+ err = fmt.Errorf("获取变频指标插值法Map失败, Err: %s", e.Error())
|
|
|
+ return
|
|
|
+ }
|
|
|
+ changeDataList = tmpNewChangeDataList
|
|
|
+ } else {
|
|
|
+ changeDataList = bDataList
|
|
|
+ for _, v := range changeDataList {
|
|
|
+ changeDataMap[v.DataTime] = v.Value
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ }
|
|
|
+
|
|
|
+ // 计算不领先也不滞后时的相关系数
|
|
|
+ baseCalculateData := make([]float64, 0)
|
|
|
+ baseDataTimeArr := make([]string, 0)
|
|
|
+ for i := range baseDataList {
|
|
|
+ baseDataTimeArr = append(baseDataTimeArr, baseDataList[i].DataTime)
|
|
|
+ baseCalculateData = append(baseCalculateData, baseDataList[i].Value)
|
|
|
+ }
|
|
|
+
|
|
|
+ zeroBaseData := make([]float64, 0)
|
|
|
+ zeroCalculateData := make([]float64, 0)
|
|
|
+ for i := range baseDataTimeArr {
|
|
|
+ tmpBaseVal, ok1 := baseDataMap[baseDataTimeArr[i]]
|
|
|
+ tmpCalculateVal, ok2 := changeDataMap[baseDataTimeArr[i]]
|
|
|
+ if ok1 && ok2 {
|
|
|
+ zeroBaseData = append(zeroBaseData, tmpBaseVal)
|
|
|
+ zeroCalculateData = append(zeroCalculateData, tmpCalculateVal)
|
|
|
+ }
|
|
|
+ }
|
|
|
+ if len(zeroBaseData) != len(zeroCalculateData) {
|
|
|
+ err = fmt.Errorf("相关系数两组序列元素数不一致, %d-%d", len(baseCalculateData), len(zeroCalculateData))
|
|
|
+ return
|
|
|
+ }
|
|
|
+ zeroRatio := utils.CalculateCorrelationByIntArr(zeroBaseData, zeroCalculateData)
|
|
|
+ if leadValue == 0 {
|
|
|
+ yData = append(yData, zeroRatio)
|
|
|
+ }
|
|
|
+
|
|
|
+ // 计算领先/滞后N期
|
|
|
+ if leadValue > 0 {
|
|
|
+ // 平移变频指标领先/滞后的日期(单位天)
|
|
|
+ moveUnitDays := utils.FrequencyDaysMap[leadUnit]
|
|
|
+
|
|
|
+ for i := range xData {
|
|
|
+ if xData[i] == 0 {
|
|
|
+ yData = append(yData, zeroRatio)
|
|
|
+ continue
|
|
|
+ }
|
|
|
+ xCalculateData := make([]float64, 0)
|
|
|
+ yCalculateData := make([]float64, 0)
|
|
|
+
|
|
|
+ // 平移指定天数
|
|
|
+ mDays := int(moveUnitDays) * xData[i]
|
|
|
+ _, dMap := MoveDataDaysToNewDataList(changeDataList, mDays)
|
|
|
+
|
|
|
+ // 取出对应的基准日期的值
|
|
|
+ for i2 := range baseDataTimeArr {
|
|
|
+ if yVal, ok := dMap[baseDataTimeArr[i2]]; ok {
|
|
|
+ xCalculateData = append(xCalculateData, baseCalculateData[i2])
|
|
|
+ yCalculateData = append(yCalculateData, yVal)
|
|
|
+ }
|
|
|
+ }
|
|
|
+ if len(yCalculateData) <= 0 {
|
|
|
+ //err = fmt.Errorf("领先滞后相关系数两组序列元素数不一致, %d-%d", len(baseCalculateData), len(yCalculateData))
|
|
|
+ //return
|
|
|
+ // 领先滞后后，没有可以计算的数据了
|
|
|
+ continue
|
|
|
+ }
|
|
|
+
|
|
|
+ // 公式计算出领先/滞后频度对应点的相关性系数
|
|
|
+ ratio := utils.CalculateCorrelationByIntArr(xCalculateData, yCalculateData)
|
|
|
+ yData = append(yData, ratio)
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ xEdbIdValue = xData
|
|
|
+ yDataList = make([]data_manage.YData, 0)
|
|
|
+ yDate := "0000-00-00"
|
|
|
+ yDataList = append(yDataList, data_manage.YData{
|
|
|
+ Date: yDate,
|
|
|
+ Value: yData,
|
|
|
+ })
|
|
|
+ return
|
|
|
+}
|
|
|
+
|
|
|
+// GetRollingCorrelationChartDataByEdbInfo 滚动相关性计算
|
|
|
+func GetRollingCorrelationChartDataByEdbInfo(edbInfoMappingA, edbInfoMappingB *data_manage.ChartEdbInfoMapping, leadValue int, leadUnit string, calculateValue int, calculateUnit string, startDate, endDate string) (xDateTimeValue []string, yDataList []data_manage.YData, err error) {
|
|
|
+ xDateTimeValue = make([]string, 0)
|
|
|
+ yData := make([]float64, 0)
|
|
|
+
|
|
|
+ // 计算窗口，不包含第一天
|
|
|
+ startDateTime, _ := time.ParseInLocation(utils.FormatDate, startDate, time.Local)
|
|
|
+ startDate = startDateTime.AddDate(0, 0, 1).Format(utils.FormatDate)
|
|
|
+
|
|
|
+ baseEdbInfo := edbInfoMappingA
|
|
|
+ changeEdbInfo := edbInfoMappingB
|
|
|
+
|
|
|
+ // 获取时间基准指标在时间区间内的值
|
|
|
+ aDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ switch baseEdbInfo.EdbInfoCategoryType {
|
|
|
+ case 0:
|
|
|
+ aDataList, err = data_manage.GetEdbDataList(baseEdbInfo.Source, baseEdbInfo.EdbInfoId, startDate, endDate)
|
|
|
+ case 1:
|
|
|
+ _, aDataList, _, _, err, _ = data.GetPredictDataListByPredictEdbInfoId(baseEdbInfo.EdbInfoId, startDate, endDate, false)
|
|
|
+ default:
|
|
|
+ err = errors.New("指标base类型异常")
|
|
|
+ return
|
|
|
+ }
|
|
|
+
|
|
|
+ // 获取变频指标所有日期的值, 插值法完善数据
|
|
|
+ bDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ switch changeEdbInfo.EdbInfoCategoryType {
|
|
|
+ case 0:
|
|
|
+ bDataList, err = data_manage.GetEdbDataList(changeEdbInfo.Source, changeEdbInfo.EdbInfoId, "", "")
|
|
|
+ case 1:
|
|
|
+ _, bDataList, _, _, err, _ = data.GetPredictDataListByPredictEdbInfoId(changeEdbInfo.EdbInfoId, "", "", false)
|
|
|
+ default:
|
|
|
+ err = errors.New("指标change类型异常")
|
|
|
+ return
|
|
|
+ }
|
|
|
+
|
|
|
+ // 数据平移变频指标领先/滞后的日期(单位天)
|
|
|
+ // 2023-03-17 时间序列始终以指标A为基准, 始终是B进行平移
|
|
|
+ //baseDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ baseDataMap := make(map[string]float64)
|
|
|
+ changeDataList := make([]*data_manage.EdbDataList, 0)
|
|
|
+ changeDataMap := make(map[string]float64)
|
|
|
+
|
|
|
+ // A指标不管三七二十一，先变个频再说
|
|
|
+ {
|
|
|
+ _, e := HandleDataByLinearRegression(aDataList, baseDataMap)
|
|
|
+ if e != nil {
|
|
|
+ err = fmt.Errorf("获取变频指标插值法Map失败, Err: %s", e.Error())
|
|
|
+ return
|
|
|
+ }
|
|
|
+ //baseDataList = tmpNewChangeDataList
|
|
|
+ }
|
|
|
+ // B指标不管三七二十一，先变个频再说
|
|
|
+ {
|
|
|
+ tmpNewChangeDataList, e := HandleDataByLinearRegression(bDataList, changeDataMap)
|
|
|
+ if e != nil {
|
|
|
+ err = fmt.Errorf("获取变频指标插值法Map失败, Err: %s", e.Error())
|
|
|
+ return
|
|
|
+ }
|
|
|
+ changeDataList = tmpNewChangeDataList
|
|
|
+
|
|
|
+ // 平移下日期
|
|
|
+ moveUnitDays := utils.FrequencyDaysMap[leadUnit]
|
|
|
+ _, changeDataMap = MoveDataDaysToNewDataList(changeDataList, leadValue*moveUnitDays)
|
|
|
+ }
|
|
|
+
|
|
|
+ // 计算计算时，需要多少个日期内数据
|
|
|
+ calculateDay := utils.FrequencyDaysMap[calculateUnit] * calculateValue
|
|
|
+
|
|
|
+ // 计算 每个日期的相关性值
|
|
|
+ {
|
|
|
+ startDateTime, _ := time.ParseInLocation(utils.FormatDate, startDate, time.Local)
|
|
|
+ endDateTime, _ := time.ParseInLocation(utils.FormatDate, endDate, time.Local)
|
|
|
+
|
|
|
+ for currDay := startDateTime; !currDay.After(endDateTime); currDay = currDay.AddDate(0, 0, 1) {
|
|
|
+ yCalculateData := make([]float64, 0)
|
|
|
+ baseCalculateData := make([]float64, 0)
|
|
|
+
|
|
|
+ // 取出对应的基准日期的值
|
|
|
+ for i := 0; i < calculateDay; i++ {
|
|
|
+ iDay := currDay.AddDate(0, 0, i).Format(utils.FormatDate)
|
|
|
+
|
|
|
+ tmpBaseValue, ok1 := baseDataMap[iDay]
|
|
|
+ tmpChangeValue, ok2 := changeDataMap[iDay]
|
|
|
+ if ok1 && ok2 {
|
|
|
+ baseCalculateData = append(baseCalculateData, tmpBaseValue)
|
|
|
+ yCalculateData = append(yCalculateData, tmpChangeValue)
|
|
|
+ } else {
|
|
|
+ continue
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ // 公式计算出领先/滞后频度对应点的相关性系数
|
|
|
+ var ratio float64
|
|
|
+ if len(baseCalculateData) > 0 {
|
|
|
+ ratio = utils.CalculateCorrelationByIntArr(baseCalculateData, yCalculateData)
|
|
|
+ }
|
|
|
+ yData = append(yData, ratio)
|
|
|
+ xDateTimeValue = append(xDateTimeValue, currDay.AddDate(0, 0, calculateDay-1).Format(utils.FormatDate))
|
|
|
+ }
|
|
|
+ }
|
|
|
+
|
|
|
+ yDataList = make([]data_manage.YData, 0)
|
|
|
+ yDate := "0000-00-00"
|
|
|
+ yDataList = append(yDataList, data_manage.YData{
|
|
|
+ Date: yDate,
|
|
|
+ Value: yData,
|
|
|
+ })
|
|
|
+ return
|
|
|
+}
|
ziwen