private void initLearningMetrics() {
if (learningMetricsInitialized) {
return;
}
double defaultWeight = SystemDao.getDefaultWeight();
double lastWeekLift = getDemandUplift(SystemDao.getReviewCycleStartDate());
Sales beginOfRunCycleSales = getSales(SystemDao.getReviewCycleStartDate());
double beginOfRunCycleRcAvgSales = 0;
if (beginOfRunCycleSales != null) {
beginOfRunCycleRcAvgSales = beginOfRunCycleSales.getRcAvgSales();
}
rcAvgSales =
getWeightedWeight1(learningWeekCounter, defaultWeight) * epSales / lastWeekLift
+ (1 - getWeightedWeight1(learningWeekCounter, defaultWeight))
* beginOfRunCycleRcAvgSales;
// Tim's documentation states:
// Its important sales metrics are initialized to AVG_WEEKLY_SALES based on the initialization
// logic.
// This estimates initial values based on different combinations of the PLs product and location
// hierarchies dependent on the specific situation for that PL
rcAvgSalesActual = rcAvgSales;
rcAvgDemand = rcAvgSales;
rcOldAvgDemand = rcAvgSales;
rcAvgDemandActual = rcAvgSales;
epAvgInv = rcAvgSales;
rcWass2 = Math.pow(rcAvgSales, 2);
learningMetricsInitialized = true;
}
private String salesToString() {
StringBuffer epSalesActualBf = new StringBuffer();
epSalesActualBf.append("epSalesActual|");
StringBuffer epSalesBf = new StringBuffer();
epSalesBf.append("epSales|");
StringBuffer rcSalesBf = new StringBuffer();
rcSalesBf.append("rcSales|");
StringBuffer rcAvgSalesBf = new StringBuffer();
rcAvgSalesBf.append("rcAvgSales|");
StringBuffer rcSalesActualBf = new StringBuffer();
rcSalesActualBf.append("rcSalesActual|");
StringBuffer rcAvgSalesActualBf = new StringBuffer();
rcAvgSalesActualBf.append("rcAvgSalesActual|");
Iterator<Years> yrItr = salesMap.keySet().iterator();
Map<LocalDate, Sales> currSalesMap;
LocalDate currDate;
Years currYr;
Sales currSale;
// Populate Sale buffers
while (yrItr.hasNext()) {
currYr = yrItr.next();
currSalesMap = salesMap.get(currYr);
Iterator<LocalDate> itr = currSalesMap.keySet().iterator();
while (itr.hasNext()) {
currDate = itr.next();
currSale = currSalesMap.get(currDate);
// epSaleActual
epSalesActualBf.append(currDate);
epSalesActualBf.append(":");
epSalesActualBf.append(currSale.getEpSalesActual());
epSalesActualBf.append("|");
// epSale
epSalesBf.append(currDate);
epSalesBf.append(":");
epSalesBf.append(currSale.getEpSales());
epSalesBf.append("|");
// rc sales actual
rcSalesActualBf.append(currDate);
rcSalesActualBf.append(":");
rcSalesActualBf.append(currSale.getRcSalesActual());
rcSalesActualBf.append("|");
// rc average sales actual
rcAvgSalesActualBf.append(currDate);
rcAvgSalesActualBf.append(":");
rcAvgSalesActualBf.append(currSale.getRcAvgSalesActual());
rcAvgSalesActualBf.append("|");
// rc sales
rcSalesBf.append(currDate);
rcSalesBf.append(":");
rcSalesBf.append(currSale.getRcSales());
rcSalesBf.append("|");
// rc sales average
rcAvgSalesBf.append(currDate);
rcAvgSalesBf.append(":");
rcAvgSalesBf.append(currSale.getRcAvgSales());
rcAvgSalesBf.append("|");
}
}
epSalesActualBf.append(";\n");
epSalesBf.append(";\n");
rcSalesActualBf.append(";\n");
rcAvgSalesActualBf.append(";\n");
rcSalesBf.append(";\n");
rcAvgSalesBf.append(";\n");
StringBuffer retBf = new StringBuffer();
retBf.append(epSalesActualBf.toString());
retBf.append(epSalesBf.toString());
retBf.append(rcSalesActualBf.toString());
retBf.append(rcAvgSalesActualBf.toString());
retBf.append(rcSalesBf.toString());
retBf.append(rcAvgSalesBf.toString());
retBf.append("\n");
return retBf.toString();
}
private void processWeeklyMetrics() {
// On the end of the review cycle (Sat night) the rcAvgDemand does not undergo weekly learning
// For this exercise we do not have time granulity so weekly processing is done Sunday for the
// prior week
LocalDate crc = SystemDao.getCrc();
LocalDate prevCRCStartDate = SystemDao.getReviewCycleStartDate(); // getPreviousCRCStartDate();
Sales salesData = getSales(crc);
Sales beginOfPeriodSalesData = getSales(prevCRCStartDate);
double beginOfPeriodRcAvgSales = 0;
double beginOfPeriodRcActualAvgSales = 0;
if (beginOfPeriodSalesData != null) {
beginOfPeriodRcAvgSales = beginOfPeriodSalesData.getRcAvgSales();
beginOfPeriodRcActualAvgSales = beginOfPeriodSalesData.getRcAvgSalesActual();
}
Demand demandData = getDemand(crc);
Demand beginOfPeriodDemandData = getDemand(prevCRCStartDate);
double beginOfPeriodRcAvgDemand = 0;
double beginOfPeriodRcAvgActualDemand = 0;
if (beginOfPeriodDemandData != null) {
beginOfPeriodRcAvgDemand = beginOfPeriodDemandData.getRcAvgDemand();
beginOfPeriodRcAvgActualDemand = beginOfPeriodDemandData.getRcAvgDemandActual();
}
double defaultWeight = SystemDao.getDefaultWeight();
double lastWeekLift = getDemandUplift(prevCRCStartDate);
double weight = 1.0;
if (statusCd == STATUS_CD.LEARNING) {
weight = getWeightedWeight1(learningWeekCounter, defaultWeight);
} else if (statusCd == STATUS_CD.ACTIVE) {
weight = getWeight1(learningWeekCounter);
}
rcAvgSalesActual =
(weight * (salesData.getRcSalesActual() / lastWeekLift))
+ ((1 - weight) * beginOfPeriodRcActualAvgSales);
rcAvgSales =
(weight * (salesData.getRcSales() / lastWeekLift))
+ ((1 - weight) * beginOfPeriodRcAvgSales);
rcAvgDemand =
(weight * (demandData.getRcDemand() / lastWeekLift))
+ ((1 - weight) * beginOfPeriodRcAvgDemand);
rcAvgDemandActual =
(weight * (demandData.getRcDemandActual() / lastWeekLift))
+ ((1 - weight) * beginOfPeriodRcAvgActualDemand);
// error checking
if (rcAvgDemand == 0 && statusCd != STATUS_CD.INACTIVE) {
System.out.println("Error: 0 demand when product status is not inactive");
}
if (rcAvgDemand >= 4 * rcAvgSales) {
rcAvgSales = 4 * rcAvgSalesActual;
System.out.println("Error: RC Actual Sales greater then 4 times RC Average Sales");
}
if (rcAvgDemand >= 3 * rcAvgSalesActual) {
rcAvgSales = 3 * rcAvgSalesActual;
System.out.println("Error: RC Actual Sales greater then 3 times RC Actual Average Sales");
}
storeWeeklyMetrics(crc);
this.learningWeekCounter++;
// this is the end of the review cycle reset hasBeenOffRange
if (hasBeenOffRange) {
hasBeenOffRange = false;
}
resetRcAccumulators();
}