Müügiprotsessi optimeerimine läbi ennustava protsessiseire
Date
2016
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Äriprotsesside toetamiseks on üha laiemalt kasutusele võetud ettevõtte ressursside planeerimise (ERP) tööriistad, sealhulgas CRM süsteemid müügiprotsessi jaoks. ERP süsteemid salvestavad oma töö käigus protsesside logisid, mille oskuslik käsitlemine võimaldab efektiivistada äriprotsesse. Protsessilogide analüüsimiseks on välja töötatud protsessikaeve meetodid, mis oskavad logidest pöördprojekteerida tegelikult käivitatud protsesside mudeleid. Neid meetodeid on rakendatud koos ennustava seire meetoditega protsesside tulemuste soovitud ja soovimatute tulemuste varajaseks tuvastamiseks.\n\rKuigi ennustav seire on hiljuti rohkelt tähelepanu saanud ja leidnud rakendamist soovitusmootorites, mis pakuvad välja soovitusi äriprotsesside parendamiseks, ei ole seni palju uuritud kontekstiandmete, nt müügisüsteemi kirjetes klientide finantsandmed, mõju ennustava seire tulemustele soovituste kontekstis.\n\rKäesolevas magistritöös uuritakse kontekstiandmete mõju ennustava seire mudelite kvaliteedile müügiprotsessi optimeerimise kontekstis. Eksperimendid näitavad, et välistel kontekstiandmetel on pigem negatiivne mõju, samas kui sisemistel, protsessi käigus kogutud kontekstiandmetel on positiivne mõju mudelite kvaliteedile. Muuhulgas selgub eksperimentidest, et juba kolme esimese sündmuse baasil saab müügiprotsessis ennustada müügi õnnestumist.
Business processes today are supported by enterprise systems such as Enter-\n\rprise Resource Planning systems. These systems store large amounts of process execution\n\rlog data that can be used to improve business processes across the organization. The\n\rprocess mining methods have been developed to analyze such logs, which are capable of\n\rextracting process models. These methods, in turn, have been applied in conjunctions\n\rwith predictive monitoring methods for early differentiation of desired and undesired\n\routcomes. Although predictive monitoring approach has recently caught attention and\n\rfound application in recommendation engines, which suggest cases to improve business\n\rprocess outcomes, there is no much research on how contextual data, such as clients fi-\n\rnancial indicators and other external data, may improve the quality of recommendations.\n\rThis thesis examines whether including the external data with the event data affects the\n\raccuracy of predictive monitoring for early predictions positively. More specifically, this\n\rthesis reveals usage of context data had the adverse effect on the performance of learned\n\rmodels. Furthermore, the study indicated that the usage of first three events from the\n\revent logs with internal data is sufficient to predict the label of an opportunity in the\n\rsales funnel.
Business processes today are supported by enterprise systems such as Enter-\n\rprise Resource Planning systems. These systems store large amounts of process execution\n\rlog data that can be used to improve business processes across the organization. The\n\rprocess mining methods have been developed to analyze such logs, which are capable of\n\rextracting process models. These methods, in turn, have been applied in conjunctions\n\rwith predictive monitoring methods for early differentiation of desired and undesired\n\routcomes. Although predictive monitoring approach has recently caught attention and\n\rfound application in recommendation engines, which suggest cases to improve business\n\rprocess outcomes, there is no much research on how contextual data, such as clients fi-\n\rnancial indicators and other external data, may improve the quality of recommendations.\n\rThis thesis examines whether including the external data with the event data affects the\n\raccuracy of predictive monitoring for early predictions positively. More specifically, this\n\rthesis reveals usage of context data had the adverse effect on the performance of learned\n\rmodels. Furthermore, the study indicated that the usage of first three events from the\n\revent logs with internal data is sufficient to predict the label of an opportunity in the\n\rsales funnel.