Phayung Meesad

King Mongkut's University of Technology North Bangkok, Information Technology Management, Faculty Member

Followers

199

Following

169

Co-authors

Public Views

Interests

Uploads

Papers by Phayung Meesad

1st-degree Atrioventricular (AV-block) and Bundle Branch Block Prediction using Machine Learning

Proceedings of the 11th International Conference on Advances in Information Technology, 2020

Heart block occurs when the flow of electricity interrupted or partially delayed between the top ... more Heart block occurs when the flow of electricity interrupted or partially delayed between the top and bottom chambers of the heart. People are now more often affected by this kind of disease. However, early prediction of heart block can reduce the diagnosis complexity and treatment cost. In this study, a data mining and machine learning model is proposed to predict three types of heart blocks, such as 1st-degree A-V block, Left Bundle Branch Block (LBBB), and Right Bundle Branch Block (RBBB). Experiment data samples are collected from the cardiology department of Chittagong Medical College Hospital (CMCH), Bangladesh. The dataset contains 32 types of numeric and categorical features about the patient's ECG report, daily activities, and food habits. The prediction model has been designed, trained, and tested with some empirical machine learning algorithms namely Decision Tree, Random Forest, K-Nearest Neighbor, and Support Vector Machine. Finally, the experimentation shows that De...

A Model of Knowledge Management and Customer Relationship Management System for Developing Service Quality and Customer Satisfaction in Restaurants

Restaurant investment is playing an important role in Thailand marketplace. In addition, an incre... more Restaurant investment is playing an important role in Thailand marketplace. In addition, an increasing number of restaurants lead to high competition in the market place, therefore; the customer management and knowledge management has become an emerging issue. In this paper, Knowledge Management and Customer Relationship Management restaurant System Model (KM and CRM restaurant System) will be based on European restaurant information; foreigners and Thai customers’ questionnaires in Thailand. The results have shown that KM and CRM restaurant System Model can be applied to employees and customer’s life cycle in restaurants which able to show the measurement of employees’ knowledge, retraining, sharing and customer’s satisfactions. The results are gathered from present customers’ evaluation and developed to a better services quality in restaurants for further enhancing customers. 1 ภาควิชาเทคโนโลยีสารสนเทศ คณะเทคโนโลยีสารสนเทศ มหาวิทยาลัยเทคโนโลยีพระจอมเกลาพระนครเหนือ กรุงเทพฯ 10800 ...

Download

The Application of Thai Natural Language Processing

The purpose of this chapter is to present an example of creating a complete application on Thai N... more The purpose of this chapter is to present an example of creating a complete application on Thai Natural Language Processing to help the readers get some ideas for NLP project development. To achieve this purpose, we create a Natural Language Interface to Database (NLIDB) model that helps users easily retrieve information from a database by natural language.

Thai Word Segmentation

Thai sentences consist of many words that are written consecutively with complicated rules that m... more Thai sentences consist of many words that are written consecutively with complicated rules that make word segmentation to be an arduous task. Thai word segmentation has been researched since the 1980s. Beginning with defining the scope of syllables, then developing techniques and algorithms to word segmentation using various methods that divided into three types which are Rules-based technique, Dictionary-based technique, and Learning-based technique. Details of these algorithms will be presented in the following sections.

An Improved Items Recommendation for Memory-Based Collaborative Filtering Technique

This research aimed to improve Improved Items Recommendation for Memory-based Collaborative Filte... more This research aimed to improve Improved Items Recommendation for Memory-based Collaborative Filtering Technique. In this research: 1) to concern rating item about users and 2) to know new prediction rating based on Memory-based Collaborative Filtering Technique. The experimental results from tested propose equation appear γ is [0, 1) and from the equation, γ is equal to 0.2. Get the minimization then recommend item was improved. Mean Absolute Error (MAE) compared to exiting approach was decrease 99.9 %.

Download

Scanner-Based Optical Mark Recognition

Information Technology Journal, 2009

The objective of this work was to develop Optical Mark Recognition (OMR) software for implementat... more The objective of this work was to develop Optical Mark Recognition (OMR) software for implementation in the simple scanner and for its usage as an OMR machine. It was developed using Java language. The software will help the assessor to capture and score responses of the multiple choices-answer sheets with accuracy and efficiency. Moreover, it was proposed to be a replacement for the costly OMR machine. The developed software was evaluated by simulating an actual implementation. It interpreted 1,000 answer sheers scanned by 5 different scanners in 4 different resolutions. The evaluation’s results were favorable, with an estimated number of mistakes were less than 1 per 1,000 or 0.1 percent. As a result, the quality and accuracy of the software that has been developed is thus within acceptable limits.

Communication Protocol Model for Language Game with Multi-Agents and Multi-Languages Using Dynamic Radius of SOM

This paper demonstrates how autonomous systems can learn and communicate with each other in three... more This paper demonstrates how autonomous systems can learn and communicate with each other in three different humane languages. In this work, we proposed a conceptual model based on Language Games and Conceptual Spaces including a special technique: an optimum dynamic radius of Self-Organizing Map (SOM) for the best searching unit between autonomous agents. Our proposed model shows the visualized simulation of the potential ways of a communication protocol during Language Game. Later the performance of the dynamic and the fixed radius of SOM technique is compared and evaluated. The results show that the proposed model enhanced with the optimum of the dynamic radius at the size of 30 can outperform the standard SOM. Finally, we have shown the effects of the number of the players and the number of the topics against the success rate could reduce concept searching process, and helping autonomous learning faster and convergence matching all three languages.

Download

The 1st Joint Seminar on Computational Intelligence (JSCI 2016)

Mean squared error (MSE) loss function is well-suited for regression task. Moreover, it can also ... more Mean squared error (MSE) loss function is well-suited for regression task. Moreover, it can also be applied to classification problem. On the other hand, hinge loss function is one-sided loss function proposed only for classification problem, it allowed data point has value greater than +1 for the positive class and less than-1 for negative class, otherwise it will calculate as error or loss. However, it is not differentiable, to minimize the hinge function uses sub-gradient method or uses quadratic programming (QP) in case of support vector machine (SVM). In this paper, we used simple projection method to minimize hinge loss function for classification case. We modified the least squared method so that it can adopt using hinge loss instead of MSE and we also applied it to ridge regression and least Square-SVM (LS-SVM) for classification problem. The results show the effectiveness of our method in terms of accuracy to the ordinary least squared method and also more robust to the outliers and the distribution of data.

Download

Natural Language Processing with “More Than Words – BERT”

Lecture Notes in Networks and Systems, 2021

Question-Answering (QA) has become one of the most popular natural language processing (NLP) and ... more Question-Answering (QA) has become one of the most popular natural language processing (NLP) and information retrieval applications. To be applied in QA systems, this paper presents a question classification technique based on NLP and Bidirectional Encoder Representation from Transformers (BERT). We performed experimental investigation on BERT for question classification with TREC-6 dataset and a Thai sentence dataset. We propose an improved processing technique called “More Than Words – BERT” (MTW – BERT) that is a special NLP Annotation tags for combining Part-Of-Speech tagging and Named Entities Recognition to be able for learning both pattern of grammatical tag sequence and recognized entities together as input before classifying text on BERT model. Experimental results showed that MTW – BERT outperformed existing classification methods and achieved new state-of-the-art performance on question classification for TREC-6 dataset with 99.20%. In addition, MTW-BERT also applied for ...

A Sequenced Edge Grid Image Technique for Sign Language Recognition

Currently there are many techniques and methods continuously proposed by researchers for Sign lan... more Currently there are many techniques and methods continuously proposed by researchers for Sign language recognition systems based-on machine learning. For data preprocessing for sign language, majority of researchers use single image of hands like static gesture images. Using only static hand images may not be efficient for real-world applications. In this paper, we propose an innovative technique for video processing called Sequenced Edge Grid Images (SEGI) for Sign Language recognition. The proposed SEGI is composed of images that represent the movement of hands within a single image, which can be applied to recognize a word or a sentence. To proof the concept, we have done several experiments with Thai sign language data collected from internet. SEGI was with existing techniques. Data are the Thai sign language learning video clips that are vocabularies to use in daily life. The proposed technique was implemented with convolutional neural network (CNN). For normal CNN, the experim...

Adaptive service node placements in P2P-like architectures

Content delivery networks (CDNs) are so far efficient tools to distribute a fast growing amount o... more Content delivery networks (CDNs) are so far efficient tools to distribute a fast growing amount of information to the users. To ensure the required Quality of Service (QoS), more and more distributed and even hybrid P2P system architectures are used to cope with the not sufficient bandwidth and capacities of a single server. The proposed algorithm solves the problem to find an optimal position for the server or service nodes in a hybrid P2P system depending on network parameters and especially user activities. Moreover, simulations show that the customer peer assignment to the service nodes results in an almost equal partition of the network. Furthermore, an automatically movement of the service node can be induced in case of significant load changes.

Download

Thai News Clustering Based on Machine Learning Approach

There are many online news available on the web; however, not every news articles is telling a tr... more There are many online news available on the web; however, not every news articles is telling a true story, some news websites are not trusted. Identification of trusted online news is a challenged task. Currently, there is no tool for checking the reliability of online news. In this research, we propose a methodology for online news trusted identification based on a text clustering technique. To find trusted news, collections of news articles from websites is needed as data mining sources. The collected news articles are then grouped according to the similarity of the words in the articles with the same meaning. The performance of the developed methodology is measured using sum squared error (SSE). The best performance is based on Hierarchical clustering algorithm with SSE lowest to 0.00. Keywords—machine learning, text mining, clustering, online news

Download

การหาค่าที่เหมาะสมที่สุดด้วยขั้นตอนวิธีเคออสมีมีติก แบบหลายวัตถุประสงค์ Optimization by Multi-Objective Chaos-Memetic Algorithm

In this research, we discuss the efficacy improvement for Multi-Objective Memetic Algorithms: MOM... more In this research, we discuss the efficacy improvement for Multi-Objective Memetic Algorithms: MOMA by using Chaos theory. Chaos theory is used to improve the population ภรัณยา อําามฤครัตน์ (Paranya Ammaruekarat)* และพยุง มีสัจ (Phayung Meesad)* to achieve a better result. The technique can be referred to in abbreviation as MOCMA: Multi-Objective Chaos Memetic Algorithms. In this research, such technique is applied to solve Multi-Objective equa-tion:2 objective DTLZ1-4. The resulted outcomes will then be measured for their capabilities to find out the best outcome group in 2 aspects namely; the capability to converge to the true outcome and the capability to spread the outcome groups found. The capabilities of the technique will be compared with those of NSGAII(Non-dominated Sorting Genetic Algorithm II). The research shows that the Convergence Measurement of MOCMA process has a better rate of convergence than that of NSGAIIand Measurement of MOCMA is scattered more evenly than that ...

Download

Cell-DROS : A Fast Outlier Detection Method for Big Datasets

Outlier detection is one of the obstacles of big dataset analysis because of its time consumption... more Outlier detection is one of the obstacles of big dataset analysis because of its time consumption issues. This paper proposes a fast outlier detection method for big datasets, which is a combination of cell-based algorithms and a ranking-based algorithm with various depths. A cell-based algorithm is proposed to transform a very large dataset to a fairly small set of weighted cells based on predefined lower and upper bounds. A ranking-based algorithm with various depths is modified and applied on weighted cells to calculate outlier scores and sort cells based on their outlier scores. Finally, an outlier obtaining algorithm is proposed to identify ids of outliers from the ranked cells and eliminate outliers from the given datasets. Experiment results show that the proposed method can produce the same results when compared to the previous rank-difference outlier detection algorithm but it can reduce up to 99% of executing time.

Download

Thai Fake News Detection Based on Information Retrieval, Natural Language Processing and Machine Learning

SN Computer Science

Fake news is a big problem in every society. Fake news must be detected and its sharing should be... more Fake news is a big problem in every society. Fake news must be detected and its sharing should be stopped before it causes further damage to the country. Spotting fake news is challenging because of its dynamics. In this research, we propose a framework for robust Thai fake news detection. The framework comprises three main modules, including information retrieval, natural language processing, and machine learning. This research has two phases: the data collection phase and the machine learning model building phase. In the data collection phase, we obtained data from Thai online news websites using web-crawler information retrieval, and we analyzed the data using natural language processing techniques to extract good features from web data. For comparison, we selected some well-known classification Machine Learning models, including Naïve Bayesian, Logistic Regression, K-Nearest Neighbor, Multilayer Perceptron, Support Vector Machine, Decision Tree, Random Forest, Rule-Based Classifier, and Long Short-Term Memory. The comparison study on the test set showed that Long Short-Term Memory was the best model, and we deployed an automatic online fake news detection web application.

Download

Improving the Evolutionary Computation for General Keyboard Arrangement Problem

Applied Mechanics and Materials, 2015

Thai traditional computer keyboard layouts are derived from the Thai typewriter keyboard's la... more Thai traditional computer keyboard layouts are derived from the Thai typewriter keyboard's layout, which supports typing with ten fingers. Hence, this layout is not optimized for typing on touch screen device with one finger. This article presents a new method in arrangement the position of Thai characters on the virtual keyboard to support typing by a single finger. The process of the new method consist of analysing Thai Language to create difference digraph table from Benchmark for Enhancing the Standard of Thai language processing corpus (BEST corpus), designed keyboard layout using evolutionary computation technique and estimated time speed of typing by model human hand movement with Fitts' Law. The experiment shows that new keyboard designed by the proposed method has achieved a typing speed faster than the Thai traditional keyboard layout.

Recent Advances in Information and Communication Technology

Advances in Intelligent Systems and Computing, 2014

This book presents recent research work and results in the area of communicationand information t... more This book presents recent research work and results in the area of communicationand information technologies. The book includes the main results of the 11th InternationalConference on Computing and Information Technology (IC2IT) held during July2nd-3rd, 2015 in Bangkok, Thailand. The book is divided into the two main parts DataMining and Machine Learning as well as Data Network and Communications.New algorithms and methods of data mining asr discussed as well as innovativeapplications and state-of-the-art technologies on data mining, machine learning anddata networking.

Leader-Following Discrete-Time Consensus Protocol in a Buyer-Seller Network

A discrete-time consensus protocol with the leader-following control is proposed, which is for ag... more A discrete-time consensus protocol with the leader-following control is proposed, which is for agents to follow either static or time-varying reference state of the leader. An agent updates its state based on only the current information available from its neighbors. This proposed protocol is applied to the problem of the buyer-seller network, and simulations will illustrate the effectiveness of the proposed consensus protocol.

Download

An optimal design for type–2 fuzzy logic system using hybrid of chaos firefly algorithm and genetic algorithm and its application to sea level prediction

Journal of Intelligent & Fuzzy Systems

ABSTRACT This paper proposes an optimal design for interval type-2 Takagi-Sugeno-Kang (TSK) fuzzy... more ABSTRACT This paper proposes an optimal design for interval type-2 Takagi-Sugeno-Kang (TSK) fuzzy logic system. In this method, the fuzzy c-means clustering algorithm is used to determine structure of fuzzy rule as well as number of rules. A hybrid between chaos firefly algorithm and genetic algorithms (CFGA) is developed, which is used to find the desirable parameters of membership functions and consequents parameters of the fuzzy logic system. The obtained optimal fuzzy logic system is used to predict sea water level in short-term and long-term horizontal. To demonstrate the superiority of the hybrid algorithm in design the fuzzy logic system, comparison between CFGA with genetic algorithms and firefly algorithm applied to optimize the fuzzy logic system for sea water level prediction is investigated. Results illustrate CFGA approach to design fuzzy logic system to be highly comparative, outperforming both genetic algorithms and firefly algorithm.

The Evaluated Measurement of a Combined Genetic Algorithm and Artificial Immune System

International Journal of Electrical and Computer Engineering (IJECE)

This paper demonstrates a hybrid between two optimization methods which are the Artificial Immune... more This paper demonstrates a hybrid between two optimization methods which are the Artificial Immune System (AIS) and Genetic Algorithm (GA). The novel algorithm called the immune genetic algorithm (IGA), provides improvement to the results that enable GA and AIS to work separately which is the main objective of this hybrid. Negative selection which is one of the techniques in the AIS, was employed to determine the input variables (populations) of the system. In order to illustrate the effectiveness of the IGA, the comparison with a steady-state GA, AIS, and PSO were also investigated. The testing of the performance was conducted by mathematical testing, problems were divided into single and multiple objectives. The five single objectives were then used to test the modified algorithm, the results showed that IGA performed better than all of the other methods. The DTLZ multiobjective testing functions were then used. The result also illustrated that the modified approach still had the b...

Download

Phayung Meesad

Uploads

Papers by Phayung Meesad

Log In