Informatics Publishing Limited

K. Deeba ¹, K. Thanushkodi ²

Affiliations
1 Department of Computer Science and Engineering, Kalaignar Karunanidhi Institute of Technology, Coimbatore, Tamil Nadu, IN
2 Akshaya College of Engineering and Technology, IN

Abstract

This paper proposes a genetic algorithm optimization approach to solve job scheduling problems in a multiprocessor architecture. Job scheduling problem (JSP) is an extremely difficult hard problem because it requires a large combinatorial search space and also precedence constraints between the processes considered. Efficient assignment and scheduling of tasks is of high importance in the effective utilization of multiprocessor systems. With regard to this kind of JSP problem, a genetic algorithm based approach is utilized. In this approach, the chromosomes are encoded and the population size is devised for the optimized goal of scheduling problem. Then the fitness function is formulated. Next, utilizing the genetic algorithm operators to prevent the premature convergence, a best or satisfactory scheduling path can be determined. The execution results using the devised algorithms in certain multiprocessor architecture are given in the paper.

Keywords

Finishing Time, Genetic Algorithm, Job Scheduling Problem, Multiprocessor Architecture, Waiting Time.

Full Text

     

K. Deeba ¹, B. Amutha ¹

Affiliations
1 Department of Computer Science and Engineering, SRM University, Kattankulathur, Chennai - 603203, Tamil Nadu, IN

Abstract

Objectives: To make a comparative study about different classification techniques of data mining. Methods: In this paper some data mining techniques like Decision tree algorithm, Bayesian network model, Naive Bayes method, Support Vector Machine and K-Nearest neighbour classifier were discussed. Findings: Each algorithm has its own advantages and disadvantages. Decision tree technique do not perform well if the data have smooth boundaries. The Naive Bayesian classifier works with both continuous and discrete attributes and works well for real time problems. This method is very fast and highly scalable. The drawback of this technique is when a data set which has strong dependency among the attribute is considered then this method gives poor performance. KNN can perform well in many situations and it particularly suits well for multi-model classes as well as applications in which an object can have many labels. The drawback of KNN is it involves lot of computation and when the size of training set taken is large then the process will become slow. Support vector machine suites well when the data need to be classified into two groups. Application: This paper is to provide a wide range of idea about different classification algorithms..

Keywords

Bayesian Network, Data Mining, Decision Tree, K-Nearest Neighbour Classifier, Naive Bayes, Support Vector Machine.

Full Text