Temporal Rule Discovery using Genetic Programming and Specialized Hardware

Abstract

Discovering association rules is a well-established problem in the field of data mining, with many existing solutions. In later years, several methods have been proposed for mining rules from sequential and temporal data. This paper presents a novel technique based on genetic programming and specialized pattern matching hardware. The advantages of this method are its flexibility and adaptability, and its ability to produce intelligible rules of considerable complexity.

The Role of Discretization Parameters in Sequence Rule Evolution

“…We used three main techniques to improve the basic evolution presented in [1]: Model selection based on validation performance, majority vote ensembles, and naive Bayesian classifiers. Prior to our empirical study, we expected the ensembles to outperform the simple selection, and the Bayesian classifiers to outperform both of the other methods.…”

“…In this section we first describe the general approach, as developed in [1], as well as the extensions introduced here, that is, combining predictors for several resolutions. Finally, the discretization method is discussed.…”

“…We use a supervised learning scheme, in which we mark the positions where want the rule to make a prediction, and then, for each rule, calculate the correlation between its behaviour and the desired responses. This can be calculated from the confusion matrix of the rule (true and false positives and negatives) [1].…”

“…The process of knowledge discovery and data mining has been the subject of much research interest in later years, and one recent subfield is that of sequence mining. In our previous work [1] we have applied evolutionary algorithms to the problem of mining predictive rules from time series. Our method involves discretizing the time series, in order to be able to evaluate our rules on them (which, in turn, allows us to use genetic programming to find such rules).…”

Evolutionary Rule Mining in Time Series Databases

“…The foundation for this work has been laid in several previous publications (Hetland & Saetrom, 2002. In this paper we have collected our main results, compared them with other machine learning methods (Section 3.1) and extending them by using our method for finding relations between several time series (Section 3.2.4).…”