Elsevier

Expert Systems with Applications

Volume 39, Issue 12, 15 September 2012, Pages 11080-11088
Expert Systems with Applications

A novel chemistry based metaheuristic optimization method for mining of classification rules

https://doi.org/10.1016/j.eswa.2012.03.066Get rights and content

Abstract

When investigated carefully, chemical reactions possess efficient objects, states, process, and events that can be designed as a computational method en bloc. In this study, a novel computational method, which is robust and have less parameters than that of used in the literature, is intended to be developed inspiring from types and occurring of chemical reactions. The proposed method is named as artificial chemical reaction optimization algorithm, ACROA. In this study, one of the first applications of this method has been performed in classification rule discovery field of data mining and efficiency has been demonstrated.

Highlights

► A novel computational method has been developed inspiring from types and occurring of chemical reactions. ► One of the first applications of this method has been performed in classification rule discovery. ► Efficiency has been demonstrated.

Introduction

Classical optimization algorithms are inflexible to adapt the solution procedure to an optimization problem. Generally a given problem is modeled in such a way that a classical algorithm can handle it. This generally requires making several assumptions and/or modifications which might not be easy to validate in many situations. These modifications and/or assumptions on the original problem parameters (rounding variables, softening constraints etc.) certainly affect the solution quality (Baykasoglu, Ozbakir, & Tapkan, 2007). They are insufficient if integer and/or discrete decision variables are required in optimization models (Baykasoglu et al., 2007). Solution strategies of classical optimization algorithms are generally depended on the type of objective and constraint functions (linear, non-linear etc.) and the type of variables used in the problem modeling (integer, real etc.). Their efficiency is also very much dependent on the size of the solution space, number of variables and constraints used in the problem modeling, and the structure of the solution space (convex, non-convex, etc.). Briefly, they do not offer general solution strategies that can be applied to problem formulations where, different type of variables, objective and constraint functions are used. However, most of the optimization problems require different types of variables, objective and constraint functions simultaneously in their formulation (Baykasoglu et al., 2007).

Inefficiency of classical optimization algorithms in solving larger scale and/or highly non-linear problems forced researchers to find more flexible and adaptable general purpose novel algorithms. Problem and model independent heuristic optimization algorithms have been proposed by researchers to overcome the drawbacks of the classical optimization procedures. These algorithms are efficient and flexible and they can be modified and/or adapted to suit specific problem requirements. Researches on these algorithms are still continuing all around the globe.

Sometimes, some effective ideas are inspired by biological, physical, and social processing. In fact, human beings have often simulated natural phenomena to create new technologies resulting in created technologies different from the naturally and ecologically simulated phenomena. There are many computational methods inspired by the human beings and natural processes. Artificial neural networks (Haykin, 1999, Rupérez et al., 2012), evolutionary computation (de Jong, 2006), artificial immunology (de Castro and Timmis, 2002, Laurentys et al., 2010) are biologically based methods. Ant colony optimization (Dorigo, 2004, Xiao and Ping Li, 2011), bee colony optimization (Alatas, 2010, Bozorg et al., 2005), particle swarm optimization (Alatas et al., 2009, Kennedy and Eberhart, 1995), cat swarm optimization (Chu et al., 2006, Pradhan and Panda, 2012), bacterial foraging optimization (Chatzis and Koukas, 2011, Chen et al., 2009), glowworm swarm optimization (Krishnanand and Ghose, 2006, Wu et al., 2012), artificial fish-swarm algorithm (Li, 2003), monkey search optimization (Mucherino & Seref, 2007), firefly algorithm (Horng and Liou, 2011, Yang, 2009) are swarm based optimization methods. Big bang – big crunch optimization (Alatas, 2011, Erol and Eksin, 2006), charged system search (Kaveh and Laknejadi, 2011, Kaveh and Talatahari, 2010), artificial physics optimization algorithm (Xie, Zeng, & Cui, 2009), intelligent water drops algorithm (Shah-Hosseini, 2008), electromagnetism-like algorithm (Birbil and Fang, 2003, Jamili et al., 2011), particle collision algorithm (Sacco, 2005) central force optimization (Formato, 2007, Green et al., 2012), gravitational search algorithm (Rashedi et al., 2009, Yin et al., 2011) are physics based methods. Tabu search (Aladag and Köksoy, 2011, Glover, 1989), imperialist competitive algorithm (Atashpaz-Gargari and Lucas, 2007, Behnamian and Zandieh, 2011), parliamentary optimization algorithm (Borji, 2007), and teaching-learning-based algorithm (Rao, Savsani, & Vakharia, 2012) are social based optimization methods. Some methods can be considered as hybrid.

Similar to these methods, chemistry based a new computational method, entitled as artificial chemical reaction optimization algorithm (ACROA), for searching and optimization problems based on the types and occurring of chemical reactions has been proposed (Alatas, 2011). There is only one paper considering the collision of particles and omitting the conservation of energy as chemistry based heuristics algorithm (Lam & Li, 2010). However, the proposed computational ACROA method is completely different from this paper in terms of reason and principle. Appropriate system can be selected according to the interested optimization problem and modifications may be added to the proposed ACROA. Molecules are encoded in ACROA using appropriate encoding scheme for the optimization problems. It seems robust and has less parameter than that of used in the literature. It also seems a strong algorithm with high in quality initial population generation method and chaotic properties of the operators.

Encoding of the reactants for ACROA depends on the interested problem. It can be binary, real, string, and etc. These encoding schemes plays role in the formation of reaction rules. Reaction rules define the interaction among one or two reactants which may lead to production of a new reactant. ACROA begins with set of uniform initial reactants in a solution. Then reactants are consumed and produced via chemical reactions. Algorithm is terminated when the termination criterion is met similar to the state when no more reactions can take place (inert solution).

In this study, one of the first applications of ACROA has been performed in classification rule discovery field of data mining. Databases have been conceived as search space and ACROA has been designed as search method for mining accurate and comprehensible classification rules.

The organization of this paper is as follows. In the Section 2, ACROA and the computational operators inspired by chemical reactions are presented. The application of this novel computational method in data mining and experimental results are presented in Section 3. Finally, Section 4 concludes this paper.

Section snippets

Artificial chemical reaction optimization algorithm (ACROA)

Atoms and molecules move and collide in a continuous manner in a viscous fluid filling a 2D cellular space. Atoms are elementary particles possessing a type, mass, radius, charge, orientation, position, and velocity. A molecule is a set of atoms connected by bonds. Chemical reactions are mappings of discrete cellular configurations to parameterized actions on atoms. Actions allow atom creation and destruction, bonding and un-bonding to make and break molecules, orientation, type change, and

Application of ACROA

Heuristics search algorithms are popular due to the strength and good ability to search in huge and complex search space in data mining applications (Alatas and Akin, 2005, Alatas and Akin, 2006, Alatas and Akin, 2008). An application of ACROA that searches the comprehensible classification rules within data sets has been performed. This application discovers comprehensible IF-THEN rules within two public domain real-world data sets, Zoo and Nursery (they have more than one potential class

Conclusions

A novel, chemistry inspired computational method has been developed by inspecting the efficient objects, states, process, and events en bloc in the different types of chemical reactions. This method includes both global and local search ability and does not need a local search method to refine the search. This method does not use an extra function or relation such as fitness function for determination of quality of reactants. Only objective function is needed. Furthermore, this method does not

Acknowledgements

This study is supported by Turkish Scientific and Technical Research Council – TUBITAK (Project Number: MAG-110M309).

References (55)

  • A. Jamili et al.

    A hybridization of simulated annealing and eectromagnetism-like mechanism for a periodic job shop scheduling problem

    Expert Systems with Applications

    (2011)
  • A. Kaveh et al.

    A novel hybrid charge system search and particle swarm optimization method for multi-objective optimization

    Expert Systems with Applications

    (2011)
  • C.A. Laurentys et al.

    Design of an artificial immune system based on Danger Model for fault detection

    Expert Systems with Applications

    (2010)
  • P.M. Pradhan et al.

    Solving multiobjective problems using cat swarm optimization

    Expert Systems with Applications

    (2012)
  • R.V. Rao et al.

    Teaching-learning-based optimization: An optimization method for continuous non-linear large scale problems

    Inform Sciences

    (2012)
  • E. Rashedi et al.

    GSA: A gravitational search algorithm

    Inform Sciences

    (2009)
  • M.J. Rupérez et al.

    Artificial neural networks for predicting dorsal pressures on the foot surface while walking

    Expert Systems with Applications

    (2012)
  • B. Wu et al.

    The improvement of glowworm swarm for continuous optimization problems

    Expert Systems with Applications

    (2012)
  • J. Xiao et al.

    A hybrid ant colony optimization for continuous domains

    Expert Systems with Applications

    (2011)
  • M. Yin et al.

    A novel hybrid K-harmonic means and gravitational search algorithm approach for clustering

    Expert Systems with Applications

    (2011)
  • B. Alatas et al.

    Mining fuzzy classification rules using an artificial immune system with boosting

    Lecture Notes in Computer Science Springer-Verlag

    (2005)
  • B. Alatas et al.

    An efficient genetic algorithm for automated mining of both positive and negative quantitative association rules

    Soft Computing, Springer Verlag

    (2006)
  • B. Alatas et al.

    Rough particle swarm optimization algorithm and its applications in data mining

    Soft Computing, Springer

    (2008)
  • E. Atashpaz-Gargari et al.

    Imperialist competitive algorithm: An algorithm for optimization inspired by imperialistic competition

    IEEE Congress on Evolutionary Computation

    (2007)
  • A. Baykasoglu et al.

    Artificial bee colony algorithm and its application to generalized assignment problem

  • S.I. Birbil et al.

    An electromagnetism-like mechanism for global optimization

    Journal of Global Optimization

    (2003)
  • Blake, C., Merz, C. J. (1998). UCI Repository of machine learning databases...
  • Cited by (0)

    View full text