Informatics Publishing Limited

B. Bharathi ¹, G. Kulanthaivel ²

Affiliations
1 Sathyabama University, Chennai-119
2 National Institute of Technical Teacher's Training and Research (NITTTR), Chennai-113

Abstract

The evaluation and performance analysis of software architecture at the design level increases the quality of the software and also reduces the cost of rework during the later stages of the product. The derivation of performance results of a software product, during the early stages of the software life cycle can be achieved by quantitatively evaluating the software performance model. There has been lot of research identifying the methods of evaluating software (Booch, 2001). The evaluation process starts by analyzing the performance model which is derived from the software model annotated with suitable usage profiles. This paper provides a simple approach to convert the software models represented as Unified Modeling Language (UML) diagrams using the profile for Schedulability, Performance and Time specifications (SPT) into Layered Queuing Network (LQN) performance models. The paper mainly illustrates the conversion process from UML to LQN, and also substantiates the method by a simple example.

Keywords

Usage Profiles, Unified Modeling Language, Performance Model, Model Driven Development

Full Text

   

Jyothi Thomas ¹, G. Kulanthaivel ²

Affiliations
1 Department of Computer Science and Engineering, Sathyabama University, Chennai - 600119, Tamil Nadu, IN
2 National Institute of Technical Teachers Training and Research, Taramani, Chennai - 600113, Tamil Nadu, IN

Abstract

Background/Objectives: Classification of preterm birth is a major challenge in the absence of effective tools and lack of domain knowledge with minimized set of rules. Objective of this paper is to classify the preterm birth with optimized rules. Methods/Statistical Analysis: This paper, proposes mutated Particle Swarm Optimization in order to achieve better result. Cognitive component was obtained by applying the mutation technique followed in the evolutionary algorithm. Mutation phase helped in converging to the optimized solution much faster. Findings: Preterm dataset of 1052 records with 5 attributes was applied to the proposed algorithm. Mutation was applied based on the Poisson, Gaussian, Uniform and Exponential distributions. The result shows that Poisson mutation among other distributions applied on the personal best has reduced the number of rules needed for classifying preterm birth datasets for both training and testing. It was observed that the optimized 13 rules out of 28 for training dataset and 9 rules out of 28 rules wereonly needed for classification. Application/Improvements: The proposed algorithm was suitable for preterm birth data set. This can be applied on the dataset with medium initial population. It was observed thatthe efficiency of the algorithm varied depends on the type of data. Therefore a generic method can be developed.

Keywords

Classification, Mutation, Particle Swarm Optimization, Pattern, Preterm Birth

Full Text

   

B. Bharathi ¹, G. Kulanthaivel ²

Affiliations
1 Sathyabama University, Chennai, IN
2 Department of Electronics and Communication, NITTTR, Chennai, IN

Abstract

The evaluation and performance analysis of software architecture at the design level increases the quality of the software and also reduces the cost of rework during the later stages of the product. There has been a lot of research in the area of software architecture and software design evaluations. When the design is to be improved based on the evaluation results, we lack in a feedback mechanism addressing the issues to be concentrated for design improvement. The paper aims at bringing out inter relationships among the various performance attributes, thereby identifying the specific changes that have to be brought about in design to satisfy customer requirements. The paper addresses a simple methodology using ranking and Karl Pearson method to find the critical areas to be updated for design improvements.

Keywords

Karl Pearson Method, Performance Attributes, Software Design, Correlation Coefficients.

Full Text

     

Jyothi Thomas ¹, G. Kulanthaivel ²

Affiliations
1 Department of Computer Science and Engineering Sathyabama University, Chennai-600119, Tamil Nadu, IN
2 National Institute of Technical Teachers Training and Research Chennai-600113, Tamil Nadu, IN

Abstract

Objective: Accurate prediction of pre term birth probability in the deliveries of babies with an effective classifier tool is a big challenging task. This paper talks about a novel competitive co evolution rule prediction classifier for extracting minimum number of rules for identifying the preterm birth. Methods/Analysis: Competitive Co evolution algorithm is applied to the preterm classifier for deriving the pattern governing the classified dataset. In this approach we have used two classes’ namely normal birth dataset and preterm birth dataset as two individuals competing with each other. Fitness of each individual is calculated based on the relative fitness of the other population. Findings: The dataset consist of 1052 records of preterm birth and normal dataset of 1314 each of five attributes. The experimental result shows the total no of rules needed for training and testing is drastically reduced compared to the total rules. The accuracy is also improved to a greater extended by applying the proposed algorithm. Applications/Improvement: The proposed algorithm minimizes the number of training rules to 16 and testing rules to 11 out of total 28 possible rules in the rule set. The accuracy of 0.962938881664499 of correct classification of preterm birth dataset is obtained through method.

Keywords

Co Evolution, Preterm Birth, Reproduction, Rules Extraction

Full Text