學術出版;;Publications

Permanent URI for this communityhdl:11536/1

收集所有交通大學研究人員與學生個人或團體所發表之各項學術著作，涵蓋期刊論文、教師專書、學位論文等項。;;Collections of Publications of NCTU Researchers and Students Like Journal Papers and Thesis.

Browse

🏛️國立陽明交通大學機構典藏

Search Results

Now showing 1 - 10 of 21

Unsupervised methods for Software Defect Prediction
(2019-01-01)
Ha, Duy-An
;
Chen, Ting-Hsuan
;
Yuan, Shyan-Ming
;
資訊工程學系
;
電機資訊國際碩士學位學程
Software Defect Prediction (SDP) aims to assess software quality by using machine learning techniques. Recently, by proposing the connectivity-based unsupervised learning method, Zhang et al. have been proven that unsupervised classification has great potential to apply to this problem. Inspiring by this idea, in our work we try to replicate the results of Zhang et al.'s experiment and attempt to improve the performance by examining different techniques at each step of the approach using unsupervised learning methods to solve the SDP problem. Specifically, we try to follow the steps of the experiment described in their work strictly and examine three other clustering methods with four other ways for feature selection besides using all. To the best of our knowledge, these methods are first applied in SDP to evaluate their predictive power. For replicating the results, generally results in our experiments are not as good as the previous work. It may be due to we do not know which features are used in their experiment exactly. Fluid clustering and spectral clustering give better results than Newman clustering and CNM clustering in our experiments. Additionally, the experiments also show that using Kernel Principal Component Analysis (KPCA) or Non-Negative Matrix Factorization (NMF) for feature selection step gives better performance than using all features in the case of unlabeled data. Lastly, to make replicating our work easy, a lightweight framework is created and released on Github.
Using Machine Learning Approach to Identify Synonyms for Document Mining
(2019-01-01)
Trappey, Amy J. C.
;
Trappey, Charles V.
;
Wu, Jheng-Long
;
Tsai, Kevin T. -C
;
管理科學系
Technical or knowledge documents, such as research papers, patents, and technical documents, e.g., request for quotations (RFQ), are important knowledge references for multiple purposes. For example, enterprises and R&D institutions often need to conduct literature and patent searches and analyses before, during, and after R&D and commercialization. These knowledge discovery processes help them identify prior arts related to the current R&D efforts to avoid duplicating research efforts or infringing upon existing intellectual property rights (IPRs). It is common to have many synonyms (i.e., words and phrases with near-identical meanings) appeared in documents, which may hinder search results, if queries do not consider these synonyms. For instance, conducting "freedom-to-operate" (FTO) patent search may not find all related patents if synonyms were not taking into consideration. This research develops methodologies of generating domain specific "word" and "phrase" synonym dictionaries using machine learning. The generation and validation of both domain-specific "word" and "phrase" synonym dictionaries are conducted using more than 2000 solar power related patents as testing document set. The testing result shows that, in the solar power domain, both word level and phrase level dictionaries identify synonyms effectively and, thus, significantly improve the patent search results.
Using supervised machine learning on large-scale online forums to classify course-related Facebook messages in predicting learning achievement within the personal learning environment
(2020-01-02)
Wu, Jiun-Yu
;
Hsiao, Yi-Cheng
;
Nian, Mei-Wen
;
教育研究所
;
Institute of Education
This paper demonstrated the use of the supervised Machine Learning (ML) for text classification to predict students' final course grades in a hybrid Advanced Statistics course and exhibited the potential of using ML classified messages to identify students at risk of course failure. We built three classification models with training data of 76,936 posts from two large online forums and applied the models to classify messages into statistics-related and non-statistics-related posts in a private Facebook group. Three ML algorithms were compared in terms of classification effectiveness and congruency with human coding. Students with more messages endorsed by two or more ML algorithms as statistics-related had higher final course grades. Students who failed the course also had significantly fewer messages endorsed by all three ML algorithms than those who passed. Results suggest that ML can be used for identifying students in need of support within the personal learning environment and for quality control of the large-scale educational data.
Transportation Type Identification by using Machine Learning Algorithms with Cellular Information
(2019-01-01)
Lin, Yi-Hao
;
Chen, Jyh-Cheng
;
Lin, Chih-Yu
;
Su, Bo-Yue
;
Lee, Pei-Yu
It is crucial for future 5G networks to intelligently understand how users move so that the networks can allocate different resources efficiently. In this paper, we try to find practical features to identify four common types of motorized transportations, including High-Speed Rail (HSR), subway, railway, and highway. We propose a system architecture that can provide accurate, real-time, and adaptive solution by using cellular information only. Because we do not use GPS as that in most of the prior studies, we can reduce energy consumption, size of log data, and computational time. Around 500-hour data are collected for performance evaluation. Experimental results confirm the effectiveness of the proposed algorithm, which can improve well-known machine learning algorithms to approximately 98% classification accuracy. The results also show that battery consumption can be reduced about 37%.
Applying Big Data, Machine Learning, and SDN/NFV to 5G Traffic Clustering, Forecasting, and Management
(2018-01-01)
Luong-Vy Le
;
Sinh, Do
;
Lin, Bao-Shuh Paul
;
Tung, Li-Ping
;
資訊工程學系
Traffic clustering, forecasting, and management play a crucial role in improving network efficiency, network quality, load balancing (LB), and energy saving of mobile networks. Especially, in 5G networks, a dense heterogeneous architecture of various types of cells (macro cells and small cells) make traffic management become more complicated. Moreover, investigating and understanding traffic patterns of a huge number of cells are challenging issues, but valuable for network operators. On the other hand, big data, machine learning (ML), software defined network (SDN), and network functions virtualization (NFV) have recently been proposed as emerging technologies and the necessary tools for empowering the SON of 5G to address the intensive computation and optimization issues. In this study, the authors applied those technologies to build a practical and powerful framework for clustering, forecasting, and managing traffic behaviors for a huge number of base stations with different statistical traffic characteristics of different types of cells (GSM, 3G, 4G). Besides, several applications based on traffic forecasting were also introduced. Finally, the performance of the proposed models was evaluated by applying them to a real dataset that collected traffic KPIs (key performance indicators) of more than 6000 cells of a real network during the years, 2016 and 2017.
Nonvolatile Crossbar 2D2R TCAM with Cell Size of 16.3 F-2 and K-means Clustering for Power Reduction
(2018-01-01)
Zhou, Keji
;
Xue, Xiaoyong
;
Yang, Jianguo
;
Xu, Xiaoxin
;
Lv, Hangbing
A 28nm 16Kb nonvolatile ternary content addressable memory (nvTCAM) test chip based on logic-compatible resistive memory (ReRAM) is demonstrated with high density and low power for packet routing in network applications. The crossbar array exhibits a smallest cell size of 16.3F(2) by using 2-diode-2-ReRAM (2D2R) nvTCAM cell, useful for further 3-dimension (3D) stacking. K-means clustering is employed to allocate the storage of routing table entry for given bank count. Then the search of destination IP address can be classified and confined to a specific bank, reducing active banks for power saving. Evaluations show >3X improvement in cell density and >70% reduction in search energy with limited overhead in silicon area for bank count of four.
SDN/NFV, Machine Learning, and Big Data Driven Network Slicing for 5G
(2018-01-01)
Le, Luong-Vy
;
Lin, Bao-Shuh Paul
;
Tung, Li-Ping
;
Sinh, Do
;
資訊工程學系
5G networks are expected to be able to satisfy a variety of vertical services for mobile users, business demands, and automotive industry. Network slicing is a promising technology for 5G to provide a network as a service (NaaS) for a wide range of services that run on different virtual networks deployed on a shared network infrastructure. Moreover, the SON (self-organizing network) in 5G is expected as a significant evolution to guarantee for full intelligence, automatic, and faster management and optimization. To deal with those requirements, recently, software-defined networking (SDN), network functions virtualization (NFV), big data, and machine learning have been proposed as emerging technologies and the necessary tools for 5G, especially, for network slicing. This study aims to integrate various machine learning (ML) algorithms, big data, SDN, and NFV to build a comprehensive architecture and an experimental framework for the future SONs and network slicing. Finally, based on this framework, we successfully implemented an early state traffic classification and network slicing for mobile broadband traffic applications implemented at Broadband Mobile Lab (BML), National Chiao Tung University (NCTU).
Analyzing Students' Attention in Class Using Wearable Devices
(2017-01-01)
Zhang, Xin
;
Wu, Cheng-Wei
;
Fournier-Viger, Philippe
;
Van, Lan-Da
;
Tseng, Yu-Chee
Detecting students' attention in class provides key information to teachers to capture and retain students' attention. Traditionally, such information is collected manually by human observers. Wearable devices, which have received a lot of attention recently, are rarely discussed in this field. In view of this, we propose a multimodal system which integrates a head-motion module, a pen-motion module, and a visual-focus module to accurately analyze students' attention levels in class. These modules collect information via cameras, accelerometers, and gyroscopes integrated in wearable devices to recognize students' behaviors. From these behaviors, attention levels are inferred for various time periods using a rule-based approach and a data-driven approach. The former infers a student's attention states using user-defined rules, while the latter relies on hidden relationships in the data. Extensive experimental results show that the proposed system has excellent performance and high accuracy. To the best of our knowledge, this is the first study on attention level inference in class using wearable devices. The outcome of this research has the potential of greatly increasing teaching and learning efficiency in class.
Optimisation and Evaluation of Hyperspectral Imaging System using Machine Learning Algorithm
(2017-01-01)
Suthar, Gajendra
;
Huang, Jung Y.
;
Chidangil, Santhosh
;
光電工程學系
;
Department of Photonics
Hyperspectral imaging (HSI), also called imaging spectrometer, originated from remote sensing. Hyperspectral imaging is an emerging imaging modality for medical applications, especially in disease diagnosis and image-guided surgery. HSI acquires a three-dimensional dataset called hypercube, with two spatial dimensions and one spectral dimension. Spatially resolved spectral imaging obtained by HSI provides diagnostic information about the objects physiology, morphology, and composition. The present work involves testing and evaluating the performance of the hyperspectral imaging system. The methodology involved manually taking reflectance of the object in many images or scan of the object. The object used for the evaluation of the system was cabbage and tomato. The data is further converted to the required format and the analysis is done using machine learning algorithm. The machine learning algorithms applied were able to distinguish between the object present in the hypercube obtain by the scan. It was concluded from the results that system was working as expected. This was observed by the different spectra obtained by using the machine-learning algorithm. Software tool used: -Spectral Python, SciKit-image, IPython, MATLAB, ENVI 5.0, LabVIEW.
Big Data and Machine Learning Driven Handover Management and Forecasting
(2017-01-01)
Vy, Le Luong
;
Tung, Li-Ping
;
Lin, Bao-Shuh Paul
;
材料科學與工程學系
;
資訊工程學系
Handover (HO), as a key aspect of mobility management, plays an important role in improving network quality and mobility performance in mobile networks. Especially, in 5G networks, heterogeneous networks (HetNets) deployment of macro cells and small cells, and the deployment of ultra-dense networks (UDNs) make HO management become more challenging. Besides, the understanding of HO behavior in a cell is quite limited in existing studies, thus the forecasting HO for an individual cell is complicated, even impossible. This challenge led the authors to propose a practical process for managing and forecasting HO for a huge number of cells, based on machine-learning (ML) algorithms and big data. Moreover, based on HO forecasting, the authors also propose an approach to detect any abnormal HO in cells. The performance of the proposed approaches was evaluated by applying it to a real dataset that collected HO KPI of more than 6000 cells of a real network during the years, 2016 and 2017. The results show that the study was successful in identifying, separating HO behavior, forecasting the future number of HO attempts, and detecting abnormal HO behaviors of cells.

學術出版;;Publications

Browse

Filters

Settings

Sort By

Results per page

🏛️國立陽明交通大學機構典藏

Search Results