- author = "Gabriel K. Kronberger",
- title = "Symbolic Regression for Knowledge Discovery - Bloat, Overfitting, and Variable Interaction Networks",
- school = "Johannes Kepler University",
- year = "2010",
- address = "Linz, Austria",
- keywords = "genetic algorithms, genetic programming",
-
URL = "
http://heal.heuristiclab.com/system/files/diss%20gkr2.pdf",
- broken = "http://research.fh-ooe.at/en/publication/2668",
- size = "211 pages",
- abstract = "With the growing amount of data that are collected and recorded in various application areas the need to use these data is also growing. In science, data have always played an important role; in recent years, however, the economic potential of data has also become increasingly important. In combination with methods for data analysis, data can be used to their full potential, whether in the commercial sector to optimise offers, or in the industrial sector to optimize resources and product quality based on process data. This work describes a new approach for the analysis of data which is based on symbolic regression with genetic programming and aims to generate an overall view of the interactions of various variables of a system. By this means, all potentially interesting relationships, which can be detected in a dataset, should be identified and represented as compact and understandable models. In the first part of this work, this approach of comprehensive symbolic regression is described in detail. Important issues that play a role in the process are the prevention of bloat and over-fitting, the simplification of models, and the identification of relevant input variables. In this context, different methods for bloat control and prevention are presented and compared. In particular, the influence of offspring selection on bloat is analysed. In addition, a new way to detect over-fitting is presented. On the basis of this, extensions for the reduction of over-fitting are presented and compared. Pruning of models is featured prominently, on the one hand to prevent over-fitting and on the other hand to simplify complex models. An important aspect is the analysis of the vast amount of different models that results from the proposed approach. In this context, different methods to quantify relevant factors are proposed. These methods can be used to identify interactions of variables of the analysed system. Visualising such interactions provides a general overview of the system in question which would not be possible by analysis of individual models which are concentrated on selected aspects of the problem. Additionally, the prognosis of multivariate time series with genetic programming is described in the first part. The second part of this work shows how the described approach can be applied to the analysis of real-world systems, and how the result of this data analysis process can result in the gain of new knowledge about the investigated system. The analyzed data stem from a blast furnace for the production of steel and an industrial chemical process. In addition the same approach is also applied on a data collection storing economic data in order to identify macro-economic interactions.",
-
notes = "See also
\cite{Kronberger:book}.
http://www.gbv.de/dms/tib-ub-hannover/662622367.pdf",
