Potential project for CCS590 -  

VISUALIZING REQUIREMENTS TRACEABILITY (Data & Knowledge engineering focus Area)
Requirements traceability is a part of requirements management which is one of the main process in  Requirement Engineering. 

Requirements traceability in particular, is defined as "the ability to describe and follow the life of a requirement in both a forwards and backwards direction (i.e., from its origins, through its development and specification, to its subsequent deployment and use, and through periods of ongoing refinement and iteration in any of these phases)" 

One goal of traceability is to visualize the relationship between artifacts. As the number and complexity of trace links increases, techniques for traceability visualization are necessary. A visualization can include information about the artifacts (e.g. artefact type, metadata, attributes) and links (e.g. link typ, metadata, link strength). 

So this research is aiming at studying how to visualize and follow the life of a requirement, in both forward and backward directions. 

 

STUDENT ARE ALSO WELCOME TO PROPOSE THEIR OWN PROJECT. 

My Current research Interest: 

1.       Visual Computing 

·         Data Visualization 

·         Visual Data Mining 

·         Computer Graphics & Animations 

 

2.       Software Engineering 

·         Requirement Engineering 

·         Software Reuse 

 

Dr. Wan Mohd Nazmee Wan Zainon
 nazmee@usm.my
 Room 713, Level 7, School of Computer Sciences 

 

 Descriptions: 

 

In the past years, the research works aim to designate online learning environment based on learning style increased substantially (Feldman et al., 2015). This is because, by determining the students learning style, an adaptive learning environment can be obtained. Based on the previous research, it is stated that, the first step to achieve an adaptive learning environment is by identifying students’ learning style (Chang et al., 2009; Da˘g and Geçer, 2009; Feldman et al., 2015). Different classification algorithms have been used in order to have automated learning style identification. One of the most popular methods is rule-based in which researchers "translated" different learning styles based on the learning style model (Graf et al., 2009). 

 

In machine learning, hyper-parameter optimization is the problem of choosing a set of optimal hyper-parameters for a learning algorithm. The critical step in hyper-parameter optimization is to choose the set of trials of the parameter. The most commonly used technique in hyper-parameter optimization is by doing a grid search technique. Grid search technique is simple to implement and parallelization is trivial. Other than that, it is also reliable in low dimensional spaces. It is required to choose a set of values for each parameter. Then, the set of trials is formed by assembling every possible combination of these values which leads in determining the most possible optimal value of the parameters.

 

Therefore, to improve the percentage of accuracy in the learning style detection, this research focuses on using the artificial neural network with hyper-parameter. This is because, from the previous research, the percentage of accuracy obtained in the learning style detection is still low which are between the range of 75%-85% (Bernard et al., 2017; Maaliw III, 2016; Özpolat and Akar, 2009). The increasing of accuracy in learning style detection will help improving adaptive learning system which will increase the performance of the user in fulfilling their course.

 

Dataset:

·   Dataset is available. However, we suggest the student to get more dataset to support the work. 

 

The outcome of the study includes, but not limited to

1)  To develop a learning style detection using artificial neural network with hyper parameter optimization

2)   To obtain higher percentage of accuracy value in the learning style detection.

 

 Dr Umi Kalsom Yusof

Room 631

Descriptions: 

Diabetic Retinopathy (DR) is one of the serious global health issues nowadays. People with DR has the potential to loss the sight permanently. The number prevalence of DR is increasing year on year. This increasing trend raises concern among all the people around the world. As a precaution, people with diabetes are advised to get a comprehensive dilated eye screening regularly or at least once in a year. This precaution step is good for the patients, but it is become a challenging task for the ophthalmologist as they need to deal with a large number of cases to be diagnosed every day. In addition, current screening program requires additional equipment which quite cost-prohibitive or sometimes unavailable especially in rural area.

 

This situation requires a solution that can ease the burden of ophthalmologist and with a considerable reduction in health care cost. Therefore, a new method has been proposed which is diagnosis of diabetic retinopathy through blood test. The researchers take benefit of diabetic patients’ blood test result to proposed algorithm that can automatically classify patients into stages of DR using the machine learning technique.

 

However, machine learning algorithms that have been proposed previously have low accuracy result. Reducing the data dimensions to the minimal set of features is one solution to the problem of low accuracy. Therefore, it is good to discover which clinical feature play the key role in determining the diabetic retinopathy. Even this subject has been studied previously, but it remains the subject of on-going research. 

 

Therefore, this study aims to obtain optimal or near-optimal accuracy value in the study of diabetic retinopathy classification based on feature selection with association rule mining. The idea is to select the features that are closely related using association rule mining and perform the DR classification using the selected features. 

 

Classification of DR with machine learning can help health care organization to anticipate trends in the patient’s medical record and produce meaningful knowledge that could significantly enhance understanding of disease progression and management. Thus, easier for doctors to advise the patients about the disease care, drug dosage and time of next check-up.

 

 Datasets:  

·   Few datasets are available. However, we suggest the student to get more dataset to support the work. 

 

 The outcome of the study includes, but not limited to

1)  To perform feature selection using association rule mining method

2)  To obtain optimal or near-optimal accuracy value of the diabetic retinopathy classification using the selected features

 

Dr Umi Kalsom Yusof

Room 631

1. Framework for an autonomous wheelchair with an evacuation module.

2. A prototype of autonomous wheelchair with the designed framework from output (1).

Track: Data & Knowledge Engineering

Keywords: Graphics Processing, Image Processing, Font Reconstruction

Title: An Objective Performance Evaluation Method for Digital Font Reconstruction

Objective: To propose a more refined digital font reconstruction objective performance measure/evaluation method

Problem Background: We attempt to establish a more refine objective performance evaluation method for digital font reconstruction so that more appealing and refine reconstructed fonts can be produced. The method will involves identifying the curves that make up a particular character/font. The performance measure will then involves comparing the set of curves that makes up the reconstructed font/character against the set of curves that makes up the original font/character.

Status & Detail of the Project: A student is required to work on the final phase of this research project since substantial work has been done in the earlier phases. This final phase will propose the overall metrics of performance evaluation of character reconstruction by including the full set of alphabet (a to z) and thus establishing an objective method/measure for performance evaluation of digital font reconstruction. In this phase, you will be continuing the earlier work and expected to do programming using Python programming language.

 

Period of Research: I am accepting student who is going to start the dissertation in Semester 2. The student is expected to start working on the topic in the CCS591 class in Semester 1. 

Prerequisites: It would be a good advantage if you have taken or planned to take courses related to image processing and analysis or multimedia information retrieval  although it is not really necessary.