A Knowledge Discovery Method to Predict the Economical Sustainability of a Company

Abstract

In this paper we are building a prototype of a machine-learning system using an inductive supervised approach to predict the logistical performance of a company. Focus lies on the learning phase, the handling of different types of data, the creation of new concepts in order to provide better measurable information. In this system numeric financial data are combined with categorical data creating symbolic data, distinguishing the phase of model generation from examples, and the phase of model classification & interpretation. The system has been implemented in vector spaces. Our data are benchmarking surveys on concurrent engineering measuring the usage of in total 302 best practices in Belgian manufacturing companies. The general purpose for implementing a best practice is the statement that the company will improve his product processing and that this way the company will establish his economical existence on the market. Our model processes a limited number of predefined steps generating value factors for the 302 best practices. The best practices are grouped into 30 subjects, the value factors combined in linear combinations. These value factors and their linear combinations are then subject to pattern interpretation relating concurrent engineering performance to past financial state of the company but also to an economical well doing of the company on a longer term i.e. we also refer to the sustainability of the company on the market.