International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 08 | Aug 2024
p-ISSN: 2395-0072
www.irjet.net
Requirement risk prediction model(Impact of Fine Tuning) Krishna Kant Bhardwaj Birla Institue of Tecnology and Science Pilani Rajasthan, India ---------------------------------------------------------------------***---------------------------------------------------------------------
Abstract - An This Paper address two major problems that
If these risks are not captured and mitigated timely, the project may suffer a lot. These risks must be eliminated and diluted to control the software cost and schedule. Because in the testing phase beginning, we have already implemented the requirements and the cane requests or any other possible factors causing risk is already known to development team and so it is best time for forecasting risks at this stage. This may improvise software productivity and quality while reducing the probability of project disaster. When risks are appropriately mitigated or their contingency plan is available then, it helps to reduce the possibility of software project failure. There are many solutions available to provide the prediction of software risk at different phases in SDLC. however, as there is lot of variation of data and it depends on the project nature as well. So still there is a scope of improvement in the model implementation. Whereas our timings are apt to predict risky requirements. A Risk prediction model includes classification methods that are projected to predict risks on the Software Specifications of the project.
are faced in the software development projects. Basically, when defects are detected in testing phase, we have to perform Root causal analysis to find out the root cause whether it is injection or removal root cause and then accordingly we perform the preventive and the corrective actions. However, the causal analysis takes significant amount of effort and time of development team. So, using textual analysis we find out the root cause of various defects and accordingly categories defects and take preventive actions. Another problem is that in testing phase if in last phase we found that the defects are more than the expected count and due to which there is a chance of delay in milestone/release of the product so in order to address this issue. We will be using the dataset of defects of previous 5 years products and using ANN model will be trained to predict risky requirements based on various factors, like design complexity, code changes, side case or coverage area of changes, developer’s skill. For example, bigger code changes near the release may risk in quality. Also, one of the factors would be the relevant functionality module. Based on the predicted risky requirements, effort in testing could be increased and also development team will also take action in the relevant module. We will also try using ML algorithm Gradient Boost on same dataset and compare the performance with ANN Key Words: ANN, ML, Gradient Boost, Softmax, SME, SMOTE
1.INTRODUCTION Requirement Engineering is the efficient and systematic approach for gathering user’s requirements to implement software solution. It is a process of describing, understanding and maintaining requirements in the process of engineering design. Mostly all the software saves some basic features and they are basically the requirements which need to be developed by the project team. QCD depends on the requirements of a project directly or indirectly. In the development some of the requirement share straightforward and easy for the development team to understand and implement however in some requirements there some issues or aps or complexity which make these requirements risky for the success of the project. When we say risky that means the either project KPI Quality/ Cost or time is going beyond target.
© 2024, IRJET
|
Impact Factor value: 8.315
Figure 1: Proactive approach with prediction model
1.1 Solution As we have been working for this client for more than 5 years and we have delivered several projects for the same base source code. We have the entire data of the requirements along with their defects. We have dataset of around 250 different requirements developed till date. We have taken inputs from SMEs about how to estimate defects in a particular feature. There are few factors identified by SMEs and based the impact of these factors the defects are estimated. If a feature is having higher than the estimated defects, then that requirement is considered risky requirement and accordingly the testing team takes some corrective action by proactively planning of testing that
|
ISO 9001:2008 Certified Journal
|
Page 519