الفهرس 
Table of ContentsOnly 14 pages are availabe for public view
 |  | from | 83 |  |  |
| | | from | 83 | | |
Abstract
According to the growing evolution in complex systems and their integrations, Internet of things, communication, massive information flows and big data, a new type of systems has been raised to software engineers, known as Ultra Large Scale (ULS) Systems. Hence, it’s requires dramatic change in all aspects of “Software Engineering” practices and their artifacts. In this respect, ULS system that’s considered as a system of systems, have gained considerable reflection on system development activities, as the scale is incomparable to the traditional systems since there are thousands of stakeholders from different national and organizational cultures and time zones are involved in developing software, each of them have different, interests, complex and changing needs beside there are already new services are being integrated simultaneously to the current running ULS systems. The scale of ULS systems makes a lot of challenges for Requirements Engineers. As a result, they are working on some automatic tools to support requirement engineering activities to overcome many challenges. Requirements Management and Change Management specifically needs new strategies. As consider one change or more in ULS requirements may result in a lot of side effects in other running or planned requirements.This thesis points to the limitations of the current Requirements Engineering (RE) practices for the challenges forced by ULS nature and proposed new model that help ULS systems stakeholders for requirements change evolution to be able to measure the impact of several changes requests on ULS requirements in addition to requirements amendments changes that occur on the running systems represented in Natural Language utilizing Similarity models, and Clustering techniques. In order to achieve that goal, three modules were conducted through the thesis. The first module was developed to represent ULS requirements in a structured format using Documents Representation methods. The second module was built an automatically measure the impact of change request submitted in a free natural language against a large set of represented requirements. The key characteristics of the module that is exploits the synonyms of text in which the user is representing the change against the list of defined ULS requirements through Natural Language Processing (NLP) similarity models. The third module was to build a clustering model to overcome the problem of changing the current in place requirements without being exposed by any change request, using k clustering technique in order to allocate the similar requirements in different cluster with automatic calculating the optimum number of clusters according to requirement similarity Experiments for the models have been conducted using ULS ERP for around 4000 textual requirements, and the results were measured using the NLP documents processing activities, standard similarity models, and k-means clustering technique.The first experiment was directed in order to be able to extract the list of textual requirements in a single requirements repository for one of ecosystems, then to measure the impact of different raised changes issued from different crowdsourcing parties using similarity models. Then the ULS development communities can extract the list of impacted requirements according to the similarity score ranging from 0 to 1 according to the strength of impact. The second study was conducted to upgrade the change impact analysis model to consider some other scenarios in which the requirements could be amended or deleted without any raised change requests, so to tackle this problem clustering techniques have been utilized using k means in chapter 4, By using two technique, the first one is to find the optimum number of k clusters automatically and the second is to allocate all ULS requirements to clusters based on their similarity in order to connect or the associate the requirements to k cluster, the experiments showed that using this upgrade has increased the accuracy of the proposed model drastically. Finally, it’s preferred to test the performance and the accuracy of this model in different business domains to have a valid assessment of the benefits and cost of such as a decision support system for requirements engineering change management.