A universal method for composing business transaction models using logs

In this paper, a novel procedural framework for increasing the efficiency of the model composition process for the end-to-end transactional behavior of various classes of business-level computer applications is presented. The objectives of the framework are reducing the composition time and the level of engagement of domain experts in deploying business management solutions. Starting with application footprints (i.e., raw log data of various types) and without the need for a priori understanding of their syntax and semantics, the framework permits the analysis of the footprints based on an agnostic tokenization of log data. The framework then produces candidate states and relationships that form the basis for composing the transactional models via the framework's state and relationship manipulation utilities. The main architectural components of the framework are presented and their underlying principles discussed. A case study highlighting the use of the framework, based on an implementation of it, is given and a discussion regarding additional features of the framework and its relation to other research activities is also provided.