Vandana Bhattacherjee

Applicability of Weyuker Property 9 to Object-Oriented Inheritance Tree Metric-A Discussion

In the metric suite for object-oriented design put forward by Chidamber and Kemerer(C&K) (1994), it is observed that Weyuker property 9 (1998) is not satisfied by any of the structural inheritance complexity metrics. The same is also observed for candidate structural inheritance complexity metric by Brito and Carapuca (1994) and on the applicability of Weyuker property 9 to object-oriented structural inheritance complexity metrics by Gursaran and Roy (2001). In this paper we present a new inheritance tree metric (ITM) which satisfies the Weyuker property 9 (interaction increases complexity) i.e. when two classes are combined, the interaction between classes can increase the complexity of ITM value. Examples supporting the applicability of the property are also presented.This paper presents a novel approach of generating test cases from UML design diagrams. We consider use case and sequence diagram in our test case generation scheme. Our approach consists of transforming a UML use case diagram into a graph called use case diagram graph (UDG) and sequence diagram into a graph called the sequence diagram graph (SDG) and then integrating UDG and SDG to form the system testing graph (STG). The STG is then traversed to generate test cases. The test cases thus generated are suitable for system testing and to detect operational, use case dependency, interaction and scenario faults.

Modeling using K-means clustering algorithm

Modeling is an abstract representation of real world process. Predicting the likely behavior from observed behavior would be entirely legitimate if the relationship were found in the data. Two common data mining techniques for finding hidden patterns in data are clustering and classification analyses. Classification is supposed to be supervised learning and clustering is an unsupervised classification with no predefined classes. Clustering tries to group a set of objects and find whether there is some relationship between those objects. In this paper we have used the numerical results generated through the Probability Density Function algorithm as the basis of recommendations in favor of the K-means clustering for weather-related predictions. We propose a model for predicting the probability of the outcome of the Play class as YES or NO through K-means clustering on weather data. The main reason for our choice in favor of K-means clustering algorithm is that it is robust.

A survey in the area of machine learning and its application for software quality prediction

This paper explores software quality improvement through early prediction of error patterns. It summarizes a variety of techniques for software quality prediction in the domain of software engineering. The objective of this research is to apply the various machine learning approaches, such as Case-Based Reasoning and Fuzzy logic, to predict software quality. The system predicts the error after accepting the values of certain parameters of the software. This paper advocates the use of case-based reasoning (i.e., CBR) to build a software quality prediction system with the help of human experts. The prediction is based on analogy. We have used different similarity measures to find the best method that increases reliability. This software is compiled using Turbo C++ 3.0 and hence it is very compact and standalone. It can be readily deployed on any configuration without affecting its performance.

The Soft Computing Approach to Program Development Time Estimation

Software effort modeling is one of the fields in which machine learning techniques have proved effective. In this paper, attempt has been made to establish a relation between program development time with respect to its dependence on various program and personnel attributes. A neural network model has been developed to predict the development time of various software programs. The model is a three layer feed forward network with 17 neurons in the hidden layer. The model has 5 inputs and 1 output.

An analysis of dependency of coupling on software defects

Functional independence is a key to good software design and a good design results in high quality software. Functional independence is the refined form of the design concept of modularity, abstraction and information hiding. Coupling is a measure of relative interconnection among modules. Coupling in software has been linked with maintainability and existing metrics are used as predictors of external software quality (e.g., fault -proneness, impact analysis, ripple effect of changes, changeability).In this paper we demonstrate the defects of software due to coupling by studying five different attributes of coupling and measured its impact on software defects.

Application of K-Medoids with Kd-Tree for Software Fault Prediction

Software fault prediction area is subject to problems like non availability of fault data which makes the application of supervised techniques difficult. In such cases unsupervised approaches like clustering are helpful. In this paper, K-Medoids clustering approach has been applied for software fault prediction. To overcome the inherent computational complexity of KMedoids algorithm a data structure called Kd-Tree has been used to identify data agents in the datasets. Partitioning Around Medoids is applied on these data agents and this results in a set of medoids. All the remaining data points are assigned to the nearest medoids thus obtained to get the final clusters. Software fault prediction error analysis results show that our approach outperforms all unsupervised approaches in the case of one given real dataset and gives best values for the evaluation parameters. For other real datasets, our results are comparable to other techniques. Performance evaluation of our technique with other techniques has been done. Results show that our technique reduces the total number of distance calculations drastically since the number of data agents is much less than the number of data points.

Evaluation and Application of Package Level Metrics in Assessing Software Quality

Today almost all the software industries are overloaded with the maintenance work of already developed software. When any new demand of software arrives, company matches the new problem with the existing product, so that the new product can be easily developed with some new modification in existing products. The reuse of the existing product is only possible when it is measured accurately and efficiently for a longer period. In this paper, first we will calculate the class level metrics viz. CBO, RFC, WMC etc for the entire package then we will calculate the average value for each class level metric (selected) by dividing the value for each metric from the total number of classes for each package. The new resultant metrics are named as CBOavg, RFCavg, WMCavg and so on. The importance of these metrics is to accurately measure the complexity at package level. We then map these package level class metrics with the quality attributes and finally validate these metrics upon three open source projects i.e. Jedit, FreeCS and Llamma chart. Data extraction has been done through an automated tool JHawk and analysis of data has been done using SPSS 10.0 as statistical tool. General Terms Software Engineering

Software Quality Estimation using Machine Learning: Case-based Reasoning Technique

Software quality estimation is one of the most interesting research areas in the domain of software engineering for last few decades. Large numbers of techniques and models have already been worked out in the area of error estimation. The aim of software quality estimation is to identify error prone tasks as the cost can be minimized with advance knowledge about the errors and this early treatment of error will enhance the software quality. In this paper we have explored a set of data in university setting. This paper advocates the use of case-based reasoning (i.e., CBR) to make a software quality estimation system by the help of human experts. CBR relies on historical information from similar past projects, whereby similarities are determined by comparing the projects, and key attributes. We have used different similarity measures to find the best method which increases estimation accuracy & reliability. This paper presents a work in which we have expanded our previous work [24]. The software is a console based application and thus does not use the GUI functions of the Operating System, which makes it very fast in execution. In order to obtain results we have used an indigenous tool for software quality estimation, run in c++ compiler. General Terms Software Engineering: Quality

Are Our Students Prepared for Testing Based Software Development

The pervasive impact of software in systems design as well as its changing character presents immense challenges for the education of software engineers. In the twenty first century, software engineers face the challenges of rapid change and uncertainty alongwith dependability and diversity. This paper presents the results of a study conducted to assess the pair programmers’ as well as individual programmers’ ability and eagerness to begin with early testing while writing programs or developing semester projects.

A New Complete Class Complexity Metric

Software complexity metric is essential for minimizing the cost of software maintenance. Package level and system level complexity cannot be measured without class level complexity. This research addresses the class complexity metrics. This paper studies the existing class complexity metrics and proposes a new class complexity metric CCC (Complete class complexity metric), CCC metric is then analytically evaluated by Weyukers property. An automated CCCMETRIC tool was developed for empirical sample of these metrics.