Raffaella Settimi

Goal-centric traceability for managing non-functional requirements

This paper describes a goal centric approach for effectively maintaining critical system qualities such as security, performance, and usability throughout the lifetime of a software system. In goal centric traceability (GCT) non-functional requirements and their interdependencies are modeled as softgoals in a softgoal interdependency graph (SIG). A probabilistic network model is then used to dynamically retrieve links between classes affected by a functional change and elements within the SIG. These links enable developers to identify potentially impacted goals; to analyze the level of impact on those goals; to make informed decisions concerning the implementation of the proposed change; and finally to develop appropriate risk mitigating strategies. This paper also reports experimental results for the link retrieval and illustrates the GCT process through an example of a change applied to a road management system.

Utilizing supporting evidence to improve dynamic requirements traceability

Requirements traceability provides critical support throughout all phases of a software development project. However practice has repeatedly shown the difficulties involved in long term maintenance of traditional traceability matrices. Dynamic retrieval methods minimize the need for creating and maintaining explicit links and can significantly reduce the effort required to perform a manual trace. Unfortunately they suffer from recall and precision problems. This paper introduces three strategies for incorporating supporting information into a probabilistic retrieval algorithm in order to improve the performance of dynamic requirements traceability. The strategies include hierarchical modeling, logical clustering of artifacts, and semi-automated pruning of the probabilistic network. Experimental results indicate that enhancement strategies can be used effectively to improve trace retrieval results thereby increasing the practicality of utilizing dynamic trace retrieval methods.

Best Practices for Automated Traceability

Automated traceability applies information-retrieval techniques to generate candidate links, sharply reducing the effort of manual approaches to build and maintain a requirements trace matrix as well as providing after-the-fact traceability in legacy documents.The authors describe nine best practices for implementing effective automated traceability.

The Detection and Classification of Non-Functional Requirements with Application to Early Aspects

This paper introduces an information retrieval based approach for automating the detection and classification of non-functional requirements (NFRs). Early detection of NFRs is useful because it enables system level constraints to be considered and incorporated into early architectural designs as opposed to being refactored in at a later time. Candidate NFRs can be detected in both structured and unstructured documents, including requirements specifications that contain scattered and non-categorized NFRs, and freeform documents such as meeting minutes, interview notes, and memos containing stakeholder comments documenting their NFR related needs. This paper describes the classification algorithm and then evaluates its effectiveness in an experiment based on fifteen requirements specifications developed as term projects by MS students at DePaul University. An additional case study is also described in which the approach is used to classifying NFRs from a large free form requirements document obtained from Siemens Logistics and Automotive Organization

Automated classification of non-functional requirements

This paper describes a technique for automating the detection and classification of non-functional requirements related to properties such as security, performance, and usability. Early detection of non-functional requirements enables them to be incorporated into the initial architectural design instead of being refactored in at a later date. The approach is used to detect and classify stakeholders’ quality concerns across requirements specifications containing scattered and non-categorized requirements, and also across freeform documents such as meeting minutes, interview notes, and memos. This paper first describes the classification algorithm and then evaluates its effectiveness through reporting a series of experiments based on 30 requirements specifications developed as term projects by MS students at DePaul University. A new and iterative approach is then introduced for training or retraining a classifier to detect and classify non-functional requirements (NFR) in datasets dissimilar to the initial training sets. This approach is evaluated against a large free-form requirements document obtained from Siemens Logistics and Automotive Organization. Although to the NFR classifier is unable to detect all of the NFRs, it is useful for supporting an analyst in the error-prone task of manually discovering NFRs, and furthermore can be used to quickly analyse large and complex documents in order to search for NFRs.

Improving automated requirements trace retrieval: a study of term-based enhancement methods

Automated requirements traceability methods that utilize Information Retrieval (IR) methods to generate and maintain traceability links are often more efficient than traditional manual approaches, however the traces they generate are imprecise and significant human effort is needed to evaluate and filter the results. This paper investigates and compares three term-based enhancement methods that are designed to improve the performance of a probabilistic automated tracing tool. Empirical studies show that the enhancement methods can be effective in increasing the accuracy of the retrieved traces; however the effectiveness of each method varies according to specific project characteristics. The analysis of such characteristics has lead to the development of two new project-level metrics which can be used to predict the effectiveness of each enhancement method for a given data set. A procedure to automatically extract critical keywords and phrases from a set of traceable artifacts is also presented to enhance the automated trace retrieval algorithm. The procedure is tested on two new datasets.

Poirot: A Distributed Tool Supporting Enterprise-Wide Automated Traceability

Poirot is a Web-based tool supporting traceability of distributed heterogeneous software artifacts. A probabilistic network model is used to generate traces between requirements, design elements, code and other artifacts stored in distributed 3rd party case tools such as DOORS, rational rose, and source code repositories. The tool is designed with extensibility in mind, so that additional artifact types and 3rd party case tools can be easily added. Trace results are displayed in both a textual and visual format. This paper briefly describes the underlying probabilistic model, and the user interface of the tool, and then discusses Poirots deployment and use in an industrial setting

Oligonucleotide microarray identification of Bacillus anthracis strains using support vector machines

The capability of a custom microarray to discriminate between closely related DNA samples is demonstrated using a set of Bacillus anthracis strains. The microarray was developed as a universal fingerprint device consisting of 390 genome-independent 9mer probes. The genomes of B. anthracis strains are monomorphic and therefore, typically difficult to distinguish using conventional molecular biology tools or microarray data clustering techniques. Using support vector machines (SVMs) as a supervised learning technique, we show that a low-density fingerprint microarray contains enough information to discriminate between B. anthracis strains with 90% sensitivity using a reference library constructed from six replicate arrays and three replicates for new isolates.

Online Active Learning with Imbalanced Classes

This paper proposes an online algorithm for active learning that switches between different candidate instance selection strategies (ISS) for classification in imbalanced data sets. This is important for two reasons: 1) many real-world problems have imbalanced class distributions and 2) there is no ISS that always outperforms all the other techniques. We first empirically compare the performance of existing techniques on imbalanced data sets and show that different strategies work better on different data sets and some techniques even hurt compared to random selection. We then propose an unsupervised score to track and predict the performance of individual instance selection techniques, allowing us to select an effective technique without using a holdout set and wasting valuable labeled data. This score is used in a simple online learning approach that switches between different ISS at each iteration. The proposed approach performs better than the best individual strategy available to the online algorithm over data sets in this paper and provides a way to build practical and effective active learning system for imbalanced data sets.