Keywords machine learning algorithms, Bayesian statistics, neural networks, inductive algorithms, inductive logic, genetic algorithms
Start Date: 02-OCT-90 / Duration: 30 months
[ contact / participants ]
STATLOG has completed an evaluation of the performance of machine learning, neural and statistical algorithms on large-scale, complex commercial and industrial problems. The overall aim has been to give an objective assessment of the potential for classification algorithms in solving significant commercial and industrial problems, and to widen the foundation for commercial exploitation of these and related algorithms both old and new.
Historically, the three main approaches to decision problems have been (i) machine learning algorithms using decision trees; (ii) Bayesian methods of classical statistics and (iii) discrimination or regression methods generally. Partly due to the limited field of application of these methods, more recent methods have emerged in response to new problems: relational learning algorithms deal with complex data in the form of rules; neural net algorithms are linked to the fascination of mankind with understanding and emulating the human brain; while genetic algorithms solve problems by following an evolutionary path. The fact that the various methods may sometimes be applied to the same dataset with contradictory results is partly due to their treatment of the data but is more to do with the different emphasis put on the classification/prediction/optimisation aspects of the problem.
By testing around 23 algorithms from this list on about 22 large-scale and commercially important problems, this project has determined to what extent the various algorithms meet the needs of industry and has provided improved software designed to extend the commercial exploitation of advanced data analysis, including machine learning type algorithms.
The objectives of the project have been to:
The major results are:
Other results include:
Daimler-Benz has gained a lot of experience from STATLOG to help realise two application projects in cooperation with two of its departments. The first project was about the prediction of registered cars and trucks in different countries, of interest to the marketing department of Mercedes-Benz. The second project was about the fault diagnostic in automatic transmissions, again of interest to Mercedes-Benz.
Due to the interest of many East German companies in the development area of machine learning, and in order to enable them to reach a competitive level, Brainware plans to exploit the STATLOG results in a project entitled "Online Multi-Transputer Systems for Quality Inspection of Composite Materials (COMPO-Q)". COMPO-Q aims to provide statistical and neural methods based on techniques developed by Brainware within STATLOG for the high-level processing of (pre-processed) sensor data in order to produce a system output which can be a warning, advice, an instruction, or the automatic initiation of an action.
Isoft is strongly involved in the knowledge acquisition market (design of knowledge bases, modelling of data, analysis of symbolic data, etc.). Results from STATLOG are being added to Isoft's commercial offererings. Isoft has carried out an application for Bussel Uclaf (a chemical group) mixing data analysis and machine learning techniques.
Fraunhofer Gesellschaft has included the results of STATLOG in the project WISCON (Knowledge Based Process Control) supported by the German Ministry of Science and Technology, and reported on them at several workshops of that project.
Dr Gholamreza Nakhaeisadeh
D - 89013 ULM
tel: + 49/ 731-5052860
fax: + 49/ 731-5054210
DAIMLER-BENZ AG - D - C
BRAINWARE GMBH - D - P
ISOFT SA - F - P
TURING INSTITUTE - UK - P (till July 1992)
UNIVERSITY OF STRATHCLYDE - UK - P
IBK (UNIVERSITÄT LUBECK) - D - A
UNIVERSIDAD DE GRANADA - E - A
UNIVERSIDADE DE PORTO - P - A
MBB GMBH - D - A (till July 1992)
STATLOG - 5170, December 1993
please address enquiries to the ESPRIT Information Desk
html version of synopsis by Nick Cook