int ZLStatistics::correlation(const ZLStatistics& candidate, const ZLStatistics& pattern) {
if (&candidate == &pattern) {
return 1000000;
}
const unsigned long long candidateSum = candidate.getVolume();
const unsigned long long patternSum = pattern.getVolume();
const unsigned long long candidateSum2 = candidate.getSquaresVolume();
const unsigned long long patternSum2 = pattern.getSquaresVolume();
fb::shared_ptr<ZLStatisticsItem> ptrA = candidate.begin();
fb::shared_ptr<ZLStatisticsItem> ptrB = pattern.begin();
const fb::shared_ptr<ZLStatisticsItem> endA = candidate.end();
const fb::shared_ptr<ZLStatisticsItem> endB = pattern.end();
size_t count = 0;
long long correlationSum = 0;
while ((*ptrA != *endA) && (*ptrB != *endB)) {
++count;
const int comparison = ptrA->sequence().compareTo(ptrB->sequence());
if (comparison < 0) {
ptrA->next();
} else if (comparison > 0) {
ptrB->next();
} else {
correlationSum += ptrA->frequency() * ptrB->frequency();
ptrA->next();
ptrB->next();
}
}
while (*ptrA != *endA) {
++count;
ptrA->next();
}
while (*ptrB != *endB) {
++count;
ptrB->next();
}
const long long patternDispersion = patternSum2 * count - patternSum * patternSum;
const long long candidateDispersion = candidateSum2 * count - candidateSum * candidateSum;
const long long numerator = correlationSum * count - candidateSum * patternSum ;
if ((patternDispersion == 0) || (candidateDispersion == 0)) {
return 0;
}
const long long quotient1 = (1000 * numerator / patternDispersion);
const long long quotient2 = (1000 * numerator / candidateDispersion);
const int sign = (numerator >= 0) ? 1 : -1;
return sign * quotient1 * quotient2;
}
int ZLStatistics::correlation(const ZLStatistics& candidate, const ZLStatistics& pattern) {
if (&candidate == &pattern) {
return 1000000;
}
const unsigned long long candidateSum = candidate.getVolume();
const unsigned long long patternSum = pattern.getVolume();
const unsigned long long candidateSum2 = candidate.getSquaresVolume();
const unsigned long long patternSum2 = pattern.getSquaresVolume();
shared_ptr<ZLStatisticsItem> ptrA = candidate.begin();
shared_ptr<ZLStatisticsItem> ptrB = pattern.begin();
const shared_ptr<ZLStatisticsItem> endA = candidate.end();
const shared_ptr<ZLStatisticsItem> endB = pattern.end();
size_t count = 0;
long long correlationSum = 0;
while ((*ptrA != *endA) && (*ptrB != *endB)) {
++count;
const int comparison = ptrA->sequence().compareTo(ptrB->sequence());
if (comparison < 0) {
ptrA->next();
} else if (comparison > 0) {
ptrB->next();
} else {
correlationSum += ptrA->frequency() * ptrB->frequency();
ptrA->next();
ptrB->next();
}
}
while (*ptrA != *endA) {
++count;
ptrA->next();
}
while (*ptrB != *endB) {
++count;
ptrB->next();
}
const long long patternDispersion = patternSum2 * count - patternSum * patternSum;
const long long candidateDispersion = candidateSum2 * count - candidateSum * candidateSum;
const long long numerator = correlationSum * count - candidateSum * patternSum ;
if ((patternDispersion == 0) || (candidateDispersion == 0)) {
return 0;
}
int orderDiff = ::log10(patternDispersion) - ::log10(candidateDispersion);
int patternMult = 1000;
if (orderDiff >= 5) {
patternMult = ::power10(6);
} else if (orderDiff >= 3) {
patternMult = ::power10(5);
} else if (orderDiff >= 1) {
patternMult = ::power10(4);
} else if (orderDiff <= -1) {
patternMult = ::power10(2);
} else if (orderDiff <= -3) {
patternMult = ::power10(1);
} else if (orderDiff <= -5) {
patternMult = ::power10(0);
}
int candidateMult = 1000000 / patternMult;
const long long quotient1 = (patternMult * numerator / patternDispersion);
const long long quotient2 = (candidateMult * numerator / candidateDispersion);
const int sign = (numerator >= 0) ? 1 : -1;
return sign * quotient1 * quotient2;
}