كرار عبد الأمير حسون الجمالي

Abstract— Software requirement is something which is required or we can say that it is something wanted or needed from the software. In Webster’s dictionary requirements are basically, that is difference between needed and wanted, are called requirements? Software requirement defects are those which an inspector found in any aspect of the software during the inspection. There are many different types of defect detection techniques some are based on reading, some on the language used for requirement specifications and some are experimental replication, etc. When we talk about readingbased techniques then we found that Perspective Based Reading (PBR) is one of the best techniques for software inspection and to find defects and errors in it. PBR has been anticipated to advance the efficiency of software requirement inspection. This paper starts discussion from, the very beginning concept, what are defects to the concept of inspection and its importance. It also elaborates software defects, types, and reasons with its main focus on the PBR technique and how it improves the effectiveness of the software. In the end, I try to add a test case that shows the practical implementation of PBR and ad hoc techniques and try to prove my argument that PBR is one of the best.

This main focus for this thesis is to develop different prediction models for skin cancer survivability. In data mining, there is no an optimal model for the prediction process which is different based on datasets. After developing different kind of models, we will evaluate them using different performance metrics such as accuracy, F-measure, precision, etc. The datasets which will be used in this thesis will be collected from different resources such as UCI machine learning website. In the prediction process, some problems may happen like over-fitting when there is a small number of instances and a large number of parameters. To avoid these problems, K-fold cross validation will be used in which we will divide the dataset to different parts and for each part, we will register the performance metrics. This research of the dissertation will be achieved using open source software tools .The toolset will be used for different tasks like preprocessing of datasets which will be the first task before trying different algorithms as the datasets may have missing values, categorical attributes ,etc. Finally, we will compare the machine algorithms to find the “best ones” for the selected datasets considering the selected parameters for performance and quality measurements.