http://rdf.ncbi.nlm.nih.gov/pubchem/patent/CN-105589806-B
Outgoing Links
Predicate | Object |
---|---|
classificationCPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentcpc/G06F11-3668 |
classificationIPCInventive | http://rdf.ncbi.nlm.nih.gov/pubchem/patentipc/G06F11-36 |
filingDate | 2015-12-17-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
grantDate | 2018-05-18-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationDate | 2018-05-18-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
publicationNumber | CN-105589806-B |
titleOfInvention | A kind of software defect tendency Forecasting Methodology based on SMOTE+Boosting algorithms |
abstract | The invention discloses a kind of software defects based on SMOTE+Boosting algorithms to be inclined to Forecasting Methodology, for the non-equilibrium sex chromosome mosaicism of software defect data, synthesizes artificial minority class sample using SMOTE algorithms first, is allowed to reach balance with the quantity of most class samples;Then the penalty factor that the density information being distributed according to original sample calculates artificial minority class sample is used for adjusting the weight of artificial sample, makes the study of basic classification device difference original sample and artificial minority class sample and more lays particular stress on the study of original sample and the higher artificial sample of confidence level;Original sample is continued to using original loss function, penalty factor is increased to punish the not high artificial sample of confidence level to the loss function of artificial minority class sample, determines new loss function;Eventually form new Boosting algorithms.The nicety of grading of minority class and most classes is the method increase, in prediction classification field, disaggregated model caused by imbalance of data is low to minority class nicety of grading, solves the problems, such as this to a certain extent. |
priorityDate | 2015-12-17-04:00^^<http://www.w3.org/2001/XMLSchema#date> |
type | http://data.epo.org/linked-data/def/patent/Publication |
Incoming Links
Total number of triples: 14.