Flockhart, I. W. (1995). GA-MINER: Parallel Data Mining with Hierarchical Genetic Al- gorithms (Final Report). Technical Report EPCCAIKMS -GA-MINER-REPORT 1.0, University of Edinburgh.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....The idea of using genetic algorithms (GA s) or genetic programming for classification tasks in KDD is not new. In GABIL [2] a Genetic Algorithm is used so that the bit string represents a set of rules. Other approaches to data mining using evolutionary computing include REGAL [4] and GA MINER [3]. The main difference between existing evolutionary approaches to data mining and the new algorithm presented in this paper is that we use a direct representation of solutions compared to an indirect representation. The decision trees are considered the building blocks of the final outcome of ....

Flockhart, I.W., Radcliffe, N.J. GA-MINER: parallel data mining with hierarchical genetic algorithms - final report. From EPCC-AIKMS-GA-MINER-Report 1.0. University of Edinburgh. 1995.

.... processing can be regarded as a natural solution to the problem of scalability in data mining [Freitas Lavington 1998] Since genetic algorithms (GAs) tend to be slow, in comparison with most rule induction methods, the design of parallel GAs for data mining is an important research area [Flockhart Radcliffe 1995], Giordana Neri 1995] Anglano et al. 1997] Anglano et al. 1998] Araujo et al. 1999] This paper presents GA PVMINER, a parallel GA for rule discovery in data mining. In this paper we focus only on the parallelization aspects of GA PVMINER. In particular, we report only results concerning ....

Flockhart, I.W. and Radcliffe, N.J. (1995) GA-MINER: parallel data mining with hierarchical genetic algorithms -- final report. EPCC-AIKMS-GA-MINERReport 1.0. University of Edinburgh, UK.

.... and Has a job, in this order, the two above individuals would be encoded as follows: Age 25) Marital Status = married ) empty conditon ) Age 21) empty condition ) Has a job = yes ) Now each attribute occupies the same position in the two individuals, i.e. attributes are aligned [21]. Hence, crossover will produce only valid individuals. This example raises the question of how to determine, for each gene, whether it represents a normally expressed condition or an empty condition. A simple technique for solving this problem is as follows. Suppose the data being mined contains ....

....proportional to the number of conditions in the rule antecedent i.e. the shorter the rule, the simpler it is. Several fitness functions that take into account both the predictive accuracy and the comprehensibility of a rule are described in the literature see e.g. 37] 28] 51] [21]. Noda and his colleagues [49] have proposed a fitness function which takes into account not only the predictive accuracy but also a measure of the degree of interestingness of a rule. Their GA follows the Michigan approach and was developed for the task of dependence modeling. Their fitness ....

Flockhart IW and Radcliffe NJ. GA-MINER: parallel data mining with hierarchical genetic algorithms - final report. EPCC-AIKMS-GA-MINERReport 1.0. University of Edinburgh, UK, 1995.

....individual . data subset p proc. p The data being mined is distributed across the processors Individuals are evaluated one at a time Each individual is evaluated by accessing the data being mined in parallel Each processor access only its local data and partial results are combined GA MINER [Flockhart Radcliffe 95] It discover three kinds of pattern: prediction rules, distribution comparison patterns, and correlation patterns It is a fine grain, distributed population GA Two parallel implementations: On a shared memory, 8 processor SGI Challenge XL, mining 51,140 tuples, with a population of 400 ....

I.W. Flockhart and N.J. Radcliffe. GA-MINER: parallel data mining with hierarchical genetic algorithms - final report. EPCC-AIKMS-GAMINER -Report 1.0. University of Edinburgh, UK, 1995.

....then it is said that the EA has completed a generation. At first sight, a simple selection based on just the fitness of the chromosomes would give desirable enough parents. But, what if we were working in a structured population The idea behind structure and its very implementation is locality (Flockhart and Radcliffe, 1995). Locality states that the population is divided into groups of chromosomes, called demes and chromosomes are selected based not only on their fitness but on how close they are too. The main advantage of locality is that it improves the effectiveness of the algorithm, because it reduces the number ....

....form (see sub section II in the next section) THE SOLUTION The problem has many facets. It would be better to deal with one facet at a time to preserve scrutability. We begin with the configuration of the EA. The Evolutionary Algorithm The EA is mostly based on that of the GA Miner project (Flockhart and Radcliffe, 1995), with a few differences. First of all, it uses a two dimensional structured population. It contains haploid organisms with a single chromosome and that, of course, is their genome also. Structured means that each chromosome mates with a chromosome that belongs to its surrounding deme. This ....

[Article contains additional citation context not shown here]

I. W. Flockhart, and N.J. Radcliffe, 1995. GA-MINER: Parallel Data Mining with Hierarchical Genetic Algorithms. Final Report, The University of Edinburgh.

No context found.

Flockhart, I. W. (1995). GA-MINER: Parallel Data Mining with Hierarchical Genetic Al- gorithms (Final Report). Technical Report EPCCAIKMS -GA-MINER-REPORT 1.0, University of Edinburgh.

No context found.

Ian W. Flockhart. GA-MINER: Parallel Data Mining with Hierarchical Genetic Algorithms (Final Report). Technical Report EPCC-AIKMS-GAMINER -REPORT 1.0, University of Edinburgh, 1995.

No context found.

I.W. Flockhart and N.J. Radcliffe. GA-MINER: parallel data mining with hierarchical genetic algorithms - final report. EPCC-AIKMS-GA-MINERReport 1.0. University of Edinburgh, 1995.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC