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

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

Email: xinying@usm.my

Tel: 04-6532668

Text summarization is considered as an operator-based transformation process by which knowledge representation structures, as generated by natural language text understanding system, are mapped into conceptually more abstract, condensed knowledge structures forming a text summary at the representational level. Searching information by reading through tons of articles is very time consuming and not efficient. Text summarization techniques are very useful where it helps to summarize long articles into short paragraphs which contain the essence of the article. This research project helps to review / compare the famous existing text summarization ML techniques and finally propose the best techniques based on different scenarios.

Email: xinying@usm.my

Tel: 04-6532668

Control charting technique is a useful technique in Statistical Process Control (SPC) for improving productivity, reducing defects and providing diagnostic information. It is important for intelligent manufacturing, which is a part of the Industrial Revolution 4.0. SPC techniques have been used to monitoring the quality of various manufacturing processes. The traditional control chart helps industrial practitioners to measure the performance of the production process. However, traditional control chart lies with its insensitivity toward the detection of small shifts in the process mean. This may cause some of the invisible or indirect production cost be spent. Thus, this research project proposed a new adaptive control chart which vary the sampling interval (VSI) of the traditional control chart by integrating machine learning technique (e.g. Support Vector Machine, SVM).   

Requirements:

It will be good if student have mathematics/statistics background.

References:

https://avestia.com/MCM2018_Proceedings/files/paper/ICMIE/ICMIE_129.pdf

https://onlinelibrary.wiley.com/doi/full/10.1002/qre.2536

https://www.sciencedirect.com/science/article/abs/pii/S0169743918304726 

The research is to do investigation of SSD and later applying it into today object detection application. Various object detection domain of your choice such as clothing, leaf disease, fruit recognition, vehicles detection and etc. Few experiments need to be done for comparison performance against various applications.

Compulsory
1) Attend my hands on short course title: Deep learning: Convolutional neural network using Python(Keras & Tensorflow)
2) One hour weekly meeting with me.

2)      Lung cancer

3)      Diabetic retinopathy

4)      Glaucoma detection 

There are plenty algorithms for a above applications. You are required to apply deep learning approach onto one of these applications of your choice. You are required to apply the algorithm with performance measurement.

Compulsory
1) Attend my hands on short course title: Deep learning: Convolutional neural network using Python(Keras & Tensorflow)
2) One hour weekly meeting with me.  

Email: xinying@usm.my

Tel: 04-6532668

 

Text summarization is considered as an operator-based transformation process by which knowledge representation structures, as generated by natural language text understanding system, are mapped into conceptually more abstract, condensed knowledge structures forming a text summary at the representational level. Searching information by reading through tons of articles is very time consuming and not efficient. Text summarization techniques are very useful where it helps to summarize long articles into short paragraphs which contain the essence of the article. This research project helps to review / compare the famous existing text summarization DL techniques and finally propose the best techniques based on different scenarios.

A new predictive model will be proposed to predict readiness to change bad habits among obese patients. Adaptive Neuro-Fuzzy Inference System (ANFIS) needs to be optimized and applied for the new model. A survey will be conducted online to provide dataset. MATLAB needs to be used for implementation.

Path planning is used to solve problems in different fields, from simple spatial route planning to selection of an appropriate action sequence that is required to reach a certain goal. Since the environment is not always known in advance, this type of planning is often limited to the environments designed in advance and environments that we can describe accurately enough before the planning process. Path planning can be used in fully known or partly known environments, as well as in entirely unknown environments where sensed information defines the desired robot motion.

Requirements:
Student must be able to do basic programming either C++/Python/MATLab. The mobile robot can be run through simulation or physical robot.

Reference:
1. Zafar, M. N., & Mohanta, J. C. (2018). Methodology for Path Planning and Optimization of Mobile Robots: A Review. Procedia Computer Science. https://doi.org/10.1016/j.procs.2018.07.018

2. Zhang, H. Y., Lin, W. M., & Chen, A. X. (2018). Path planning for the mobile robot: A review. Symmetry. https://doi.org/10.3390/sym10100450

Input: Abstract or contents of article

Output: Structured data tuples, e.g. [method, knn], [algorithm, heuristics search] etc. 

If you are interested, please drop me an email khgan@usm.my for more information.

Input: Contents of article

Output: Structured Abstract

If you are interested, please arrange a session with me khgan@usm.my for more information. 

This topic is an extended work from my previous student of CCS519 (done with the improved bilateral filter). The image data are all provided. If you want to see the report/dissertation on previous work, or keen to know further, feel free to contact me.

Thank you.

In many machine learning (ML) applications, there is a significant difference between the prior probabilities of different classes, i.e., between the probabilities with which an example belongs to different classes of the classification problem. This situation is known as the unbalanced data or class imbalanced problem (Sun et al., 2009; López et al., 2013). Furthermore, it is worth to point out that the minority class is usually the one that has the highest interest from a learning point of view and it also implies a great cost when it is not well classified. The minority class usually represents the most important concept to be learned, and it is difficult to identify it since it might be associated with exceptional and significant cases (López et al., 2013), or because the data acquisition of these examples is costly (Weiss and Tian, 2008).

Since most of the standard learning algorithms consider a balanced training set, this may generate suboptimal classification models (i.e. a good coverage of the majority examples) whereas the minority ones are misclassified frequently. Therefore, those algorithms, which obtain a good behaviour in the framework of standard classification, do not necessarily achieve the best performance for unbalanced data sets (Fernández et al., 2010).

To this end, the main goal of this study is the development of a novel classification model that addresses unbalanced data which provide practical computational efficiency and high predictive accuracy. The proposed model is also expected to provide a feasible and meaning outcome in its related field.

Objective:

1.     To find the most effective machine learning model for unbalanced class problem 

2.     To optimise the machine learning model to enhance the accuracy.

Note: There is a possibility of using the real industrial dataset