Ananthi Sheshasayee ¹, P. Sharmila ²

Affiliations
1 Department of Computer Science and Applications, Quaid-e-Milleth government college for Women, Chennai, Tamil Nadu, IN
2 Department of Computer Applications, Vivekanandha Institute of Information and Management Studies, Tiruchengode, Tamil Nadu, IN

Abstract

The Requirement Engineering is the most important phase of the software development life cycle which is used to translate the imprecise, incomplete needs and wishes of the potential users of software into complete, precise and formal specifications. These specifications can be decomposed on application of a data mining techniques, clustering. The process of clustering the requirements allows reducing the cost of software development and maintenance. In this research two most frequently used algorithms in clustering namely k means and fuzzy c means are used. The output generated is then analyzed for evaluating the performance of the two clustering algorithms. The requirements specified by the different stakeholders of the library are used as the input. The data mining tool WEKA was used for clustering. The clustering algorithms were then analyzed for accuracy and performance. On analysis the fuzzy c means algorithm was found to be more suitable for clustering of library requirements. The results proved to be satisfactory.

Keywords

Clustering, Data Mining, Fuzzy C Means, K Means, Library, Requirements

Full Text

   

G. R. Jothilakshmi ¹, P. Sharmila ², Arun Raaza ³

Affiliations
1 Vels University, Pallavaram, Chennai - 43, Tamil Nadu, IN
2 Comm System, Vels University, Pallavaram, Chennai - 43, Tamil Nadu, IN
3 Centre for Advanced Research and Development, Vels University, Pallavaram, Chennai - 43, Tamil Nadu, IN

Abstract

Objectives: Now a days, the major health risk in woman is breast cancer. In order to reduce the death rate, early detection of cancerous region in mammogram images is needed. But finding the lesion part and its spread from the mammogram image is very difficult. To identify the type of cancer, segmentation of lesion region is important. To perform further breast cancer classification, this paper proposes an improved segmentation algorithm using digital mammogram. Method/Analysis: An automated method is used to segment the affected mammogram in a effective manner using split and merging technique based on region based segmentation method by identifying a seed point .The proposed algorithm uses morphological operation to remove the noise digitally and region split and merge technique to remove the background and separate the affected region in the image. Findings: The efficiency of this proposed algorithm is calculated by measuring five different parameters Mean, Variance, standard deviation, entropy, correlation and the output is compared with existing technique and it is observed that proposed method shows better results than previous threshold decomposition method. Improvement: From the segmented output the features are extracted for finding the type of cancer from the mass and Micro calcification region

Keywords

Breast Cancer, Microcalcification, Mammogram Images, Region Based Segmentation, Segmentation, Seed Selection.

Full Text