| Title |
Business Event Forecasting |
| ID_Doc |
75119 |
| Authors |
Appice, A; Malerba, D; Morreale, V; Vella, G |
| Title |
Business Event Forecasting |
| Year |
2015 |
| Published |
|
| DOI |
|
| Abstract |
Contemporary systems record massive amounts of events by making processes visible. Process mining techniques (van del Aalst, 2011) can be used to analyze event logs, in order to extract, modify and extend process models, as well as to check conformance with respect to defined process models. Thus far, process mining techniques have been mainly used in an off-line fashion and rarely for operational decision support. Historical full traces (i.e. instances of the process which have already completed) are rarely processed on-line. Purpose - Recently, van der Aalst et al (2012) demonstrate that process mining techniques are not necessarily limited to the past, but can also be used for the present and the future. Embracing this research direction, we investigate the feasibility of a process mining approach to predict future events of running traces of business processes. Design/methodology/approach - We propose an approach that transforms the task of event forecasting for running traces into a predictive clustering task (Blockeel et al. 1998), where the target variables are the characteristics of future events expected in running traces, while the predictors are characteristics of recent events up to a certain time window. Historical traces can be processed off-line so that a predictive clustering tree (PCT) (Blockeel et al. 1998) can be mined for the predictive aim. A PCT is a tree structured model, which predicts responses of several attributes of an examples at once. In this study, it allows us to foresee the characteristics of future events of business processes based on characteristics of recent time-delayed event elements (Pravilovic et al, 2013). Originality/value - The PCT can be used to predict on-line event elements of any new running trace. Therefore, this approach puts in evidence that process models can be learned for predictive scope providing enterprises with an "intelligent" new monitoring/recommending service. Practical implications - This service can be used to check conformance and recommend appropriate actions of enterprises' users. The proposed approach has been implemented on the OPENNESS platform, the main outcome of the research project VINCENTE (A Virtual collective INtelligenCe ENvironment to develop sustainable Technology Entrepreneurship ecosystems). The platform is a collaborative environment which enables collective intelligence and decision making processes performed by young innovative entrepreneurs and already existing SMEs. The discovered model is used on- line to predict future action(s) of a user of the platform. |
| Author Keywords |
Process mining; Business process modelling; Event prediction |
| Index Keywords |
Index Keywords |
| Document Type |
Other |
| Open Access |
Open Access |
| Source |
Conference Proceedings Citation Index - Social Science & Humanities (CPCI-SSH) |
| EID |
WOS:000357265200111 |
| WoS Category |
Economics; Management; Social Sciences, Interdisciplinary |
| Research Area |
Business & Economics; Social Sciences - Other Topics |
| PDF |
|