Adaptive strategy operators based GA for rule discovery
- PDF / 454,698 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 51 Downloads / 176 Views
ORIGINAL RESEARCH
Adaptive strategy operators based GA for rule discovery T. Shobha1 • R. J. Anandhi2
Received: 21 August 2018 / Accepted: 28 March 2019 Ó Bharati Vidyapeeth’s Institute of Computer Applications and Management 2019
Abstract A new variant of genetic algorithm, which provides equal opportunity for all parent solution to produce the offspring solution, has been applied in discovery of classification rules from continuous datasets. The main objective of proposed algorithm is used to discover classification rule with three measures like accuracy, coverage (completeness) and comprehensibility, using which easily understandable, accurate and comprehensible rules can be generated. A new process has been defined to simplify the generated rules by reducing the features dimension, according to their role in the success of discovering rules. Adaptive approach for crossover and mutation operations has been applied to handle the exploration and exploitation in dynamic manner. Algorithm has been tested on UCI benchmark dataset. The results show the better classification accuracy and optimal selection of features. It is also observed that, proposed solution generates rules which are easy to handle and does not require computational machine for applications use. Keywords Adaptive operators Classification Genetic algorithm Rule discovery
& T. Shobha [email protected] 1
Department of CSE, The Oxford College of Engineering, Bommanahalli, 10th Mile Stone, Hosur Main Road, Bangalore, Karnataka 560068, India
2
Department of ISE, New Horizon College of Engineering, Bangalore, India
1 Introduction Data mining is a process of exploring useful information hidden in huge database. Recent development in technology has generated a massive data, which contains useful and interesting information. This hidden information can be used for decision making. The process of extracting the hidden information from databases is a difficult task, for both academia and industry. Hence, data mining has increasing interest in many machine learning areas. One of the widely used data mining technique is classification, which builds model by learning from training datasets that can be used to predict future unknown datasets. Classification rule mining is a technique that aims to find small set of rules from large database. Rules discovered by classification rule mining will be represented in the form If \ A[ Then \C[, where ‘A’ is the combination of predicting conditions, and ‘C’ is the value of predicted class. Representation of extracted knowledge in this form will be advantages to user as the rules will be clearly understandable [1]. There are various machine learning techniques for classification rule mining like decision trees, support vector machines, neural networks, fuzzy logic, evolutionary algorithm (EA) etc. Recent research reveals that evolutionary algorithm provides better classification results. Genetic algorithm (GA), class of EA is widely used method for classification rule mining, that works on Darwin’s princi
Data Loading...
