Résumé:
Given the massive amount of data that business processes bring, which in turn are an inevitable part of today's information systems and enterprises (manufacturing, administration...), Business Process Management (BPM) has become a crucial technology that provides a set of techniques and tools for process management. It aims to to deal with several factors/problems, such as:
The nature of the data and their heterogeneity.
The time needed to analyze and process this immense data.
As well as the technological advances in the field of ICT and the democratization of the use of the Internet have upset the modes of operation of organizations and the modes of consumption of people. An immediate consequence of this intensive use is the explosion of the mass of data generated, known as Big-Data. From a business intelligence perspective, the rational exploitation of this mass of data requires their integration in adequate formats and supports for their analysis and to facilitate decision making.
Indeed, the traditional Extract, Transform and Load (ETL) process aims to respond to this concern by offering models and tools to extract data from different sources and to integrate them into homogeneous formats for their exploitation. Nevertheless, given the diversity of data, their speed of evolution as well as their volume which is more and more consistent, the traditional ETL approaches have shown their limits and have become inadequate, as they can no longer meet the new requirements.
In this work we have proposed an improvement of the ETL architecture in order to take care of the three properties volume, speed and variety of massive data. We expose a solution which will have to allow to recover heterogeneous data from different sources, to analyze their structure and to formalize the process of their integration. The distributive aspect of the data will have to be taken into account in order to allow the storage and exploitation of large volumes of data stored in traditional structured databases (RDB) or semi-structured databases (XML, CSV) as well as data in (EXCEL).
The proposed approach has been implemented under the PyCharm environment and we have modeled the business process of management of a commercial company.
