KMITL

Permanent URI for this communityhttps://dspace.kmitl.ac.th/handle/123456789/1

Browse

Search Results

Now showing 1 - 10 of 10

Some of the metrics are blocked by your
consent settings
Artificial Neural Network for Air Pollutant Concentration Predictions Based on Aircraft Trajectories over Suvarnabhumi International Airport
(2025-04-01)
Kamsing, Patcharin
;
Cao, Chunxiang
;
Boonpook, Wuttichai
;
Boonprong, Sornkitja
;
Xu, Min
Air pollutant concentration prediction is essential not only for effective air quality management but also for planning aircraft and ground vehicle route networks in terminal areas. In this work, an artificial neural network (ANN) is used to predict the concentration levels of four types of air pollutants (CO, NO<inf>2</inf>, PM<inf>2.5</inf>, and PM<inf>10</inf>) at Suvarnabhumi International Airport. By leveraging Automatic Dependent Surveillance-Broadcast (ADS-B) historical data, aircraft trajectory pattern clustering is implemented by using K-means and Gaussian mixture model (GMM) clustering algorithms. Then, those trajectory patterns are inputted together with other flight data into ANN computation processes, resulting in an effective air pollutant prediction model for each kind of focus pollutant. The results demonstrate that the mean square errors (MSEs) of the predicted models for CO and PM<inf>2.5</inf> have acceptable values of 51.7622 and 53.9682, respectively, while the predicted model for NO<inf>2</inf> and PM<inf>10</inf> has MSEs of 139.6674 and 124.2517, respectively. This study contributes to the advancement of air pollutant prediction methodologies, facilitating better decision-making processes, proactive air quality management, and route network planning at airports. Although some prediction models for focused air pollutants have slightly high MSEs, further study is needed to enhance the prediction model capacity.
Some of the metrics are blocked by your
consent settings
Anomaly Flying Height Prediction Based on Clustering Techniques in Hard Disk Drive Manufacturing
(2025-01-01)
Kanjanapruthipong, Worawit
;
Konghuayrob, Poom
In this research, we present a method for predicting anomaly flying height (FH) profiles in hard disk drive (HDD) manufacturing by analyzing FH data at the FH1 stage. Anomalies at FH1 can lead to calibration issues at FH2, disrupting the production process. We propose an AI-based approach using unsupervised clustering techniques to group FH profiles of the read/write head. We evaluated four clustering algorithms, KMeans, MiniBatchKMeans, Birch, and BisectingKMeans, along with the Elbow method to determine the optimal number of clusters. By identifying anomalous FH profiles early at FH1, the method enables proactive intervention, reducing calibration process time and improving production efficiency. Our model achieved an accuracy of 0.939 without relying on manual feature selection (e.g., pressure and temperature), which is often difficult to capture using traditional linear or rule-based models owing to the nonlinear nature of FH profiles. These results demonstrate the practical potential of clustering techniques in enhancing HDD manufacturing processes.
Some of the metrics are blocked by your
consent settings
Clustering of Ionospheric Irregularities based on Spatiotemporal ROTI Keogram Images
(2024-01-01)
Mutasov, Gleb
;
Min Myint, Lin Min
;
Supnithi, Pornchai
;
Budtho, Jirapoom
;
Tongkasem, Napat
Ionospheric irregularities associated with Equatorial plasma bubbles (EPB) can significantly impact navigation and communication systems. Therefore, their occurrences need to be studied and predicted. To solve the prediction problem, it is necessary to identify types of spatiotemporal characteristics as reference points for the predictive model. This work employs unsupervised machine learning algorithms to identify types of ionospheric irregularities due to EPB using the rate of total electron content index (ROTI) keograms. Two machine learning methods: two models, the Gaussian mixture model (GMM), and k-means, are considered. Comparative analysis is performed, and the optimal number of clusters is estimated using one classical, k-means and one additional - repeatability score, introduced in this work metric. The optimal GMM model successfully classifies three types of irregularity patterns offering valuable insights for the development of an effective EPB prediction model and enhancing our understanding of ionospheric behavior.
Some of the metrics are blocked by your
consent settings
Modified K-Means Clustering for Demand-Weighted Locations: A Thailand’s Convenience Store Franchise-Case Study
(2023-03-01)
Leenawong, Chartchai
;
Chaikajonwat, Thanrada
This research applies and modifies K-means clustering analysis from Data Mining to solving the location problem. First, a case study of Thailand’s convenience store franchise in locating distribution centers (DCs) is conducted. Then, the final centroids are served at suggested DC locations. Besides the typical distance, Euclidean, used in K-means, Manhattan, and Chebyshev, is also experimented with. Moreover, due to the stores’ different demands, a modification of the centroid calculation is needed to reflect the center-of-gravity effects. For the proposed centroid calculation, the above three distance metrics incorporating the demands as weights give rise to another three approaches and are thus named Weighted Euclidean, Weighted Manhattan, and Weighted Chebyshev, respectively. Besides the optimal locations, the effectiveness of these six clustering approaches is measured by the expected total distribution cost from DCs to their served stores and the expected Davies– Bouldin index (DBI). Concurrently, the efficiency is measured by the expected number of iterations to the final clusters. All these six clustering approaches are then implemented in the case study of locating eight DCs to distribute to 260 convenience stores in Eastern Thailand. The results show that though all approaches yield locations in close proximity, the Weighted Chebyshev is the most effective one having both the lowest expected distribution cost and lowest expected DBI. In contrast, Euclidean is the most efficient approach, with the lowest expected number of iterations to the final clusters, followed by Weighted Chebyshev. Therefore, the DC locations from Weighted Chebyshev could, ultimately, be chosen for this Thailand’s convenience store franchise.
Some of the metrics are blocked by your
consent settings
Incremental discovery of crowd from evolving trajectory data
(2018-08-13)
Shein, Thi Thi
;
Puntheeranurak, Sutheera
;
Imamura, Makoto
Discovering the useful knowledge from the evolving trajectory data stream contributes to real-world applications such as animal movement behavior analysis, traffic monitoring, and weather forecasting. Among these explorations, moving objects' crowd detection is a challenging task and enabling to find anomalies in the traffic system. In a real application, a large volume of trajectory data stream arrives continuously for immediate data analysis. Due to the changes of these trajectory data, there is a remaining challenge to discover crowd efficiently. In this paper, we propose incremental crowd discovery framework over evolving data stream to reduce computational time complexity. In our framework, firstly we discover the group by proposing micro-group based clustering, and then we incrementally detect the crowd structure form. Experiments of our proposed system will conduct on real taxi trajectory data and synthetic data.
Some of the metrics are blocked by your
consent settings
The fuzzy-based cluster head election algorithm for equal cluster size in wireless sensor networks
(2016-11-18)
Eak-Une, Pichatorn
;
Pornavalai, Chotipat
In Wireless Sensor Network (WSNs), clustering technique is widely used to balance energy usages consumed by the sensors. In each round of operation, a number of sensors are chosen to be candidate cluster head (CCHs) with a fixed and predefined probability value. CCHs then compete among themselves to become Cluster Head (CH) based on some criteria such as its remaining residual energy. CH role is rotated among sensors within the network field to balance their residual energy. However, area near to the corner and edge of the network field usually has less number of sensors to be CHs role than sensors located around center of the field. This will create energy holes problem on the area where sensors that needed to be CHs more often than others. Another problem is almost existing cluster head competition methods cannot precisely control the size of clusters in the networks. Depend on the spatial correlation of sensor nodes, this may impact the quality of data aggregation performed by the CHs. In this paper, we propose fuzzy based cluster head election algorithm (called FuzzCHE) to control and maintain cluster size while balance residual energy of sensors and extend the network lifetime. With FuzzCHE, each sensor can dynamically adjust probability that each sensor becomes CCH in each round by fuzzy logic. Comparing with existing cluster head election algorithm, results from simulations show that FuzzCHE can give more precise control of the average operated cluster size in the network to the deployed size required by the application.
Some of the metrics are blocked by your
consent settings
Unequal Initial Energy Assignment in Wireless Sensor Networks
(2015-08-24)
Eak-Une, Pichatorn
;
Pornavalai, Chotipat
In multi-hop clustering Wireless Sensor Networks (WSNs), Cluster Head (CH) role is rotated among sensor nodes in order to balance their residual energy usages. However, to forward data from CHs to sink or Base Station (BS), CH nodes that are placed near to the sink consume more energy than CH nodes that are further away. This is known as 'Energy Hole' problem in WSNs. This paper proposes an algorithm called 'Unequal Initial Energy Assignment' or UIEA to determine initial energy for sensors based on their distances to the sink. This unequal energy assignment is to assign different values of initial energy to sensors that belong to different regions in the network field. Since CHs perform data aggregation, this approach allows sensors to operate using same cluster sizes throughout deployed network field, which is one of the important requirements for some specific applications. Performance from simulation shows that UIEA can get more detail and precise data after aggregation by deploying smaller cluster size than average EC cluster size while the Stable Operation Period (SOP) is approximately the same as EC solution.
Some of the metrics are blocked by your
consent settings
Clustering Top-10 malware/bots based on download behavior
(2013-01-01)
Yukonhiatou, Chaxiong
;
Kittitornkun, Surin
;
Kikuchi, Hiroaki
;
Sisaat, Khamphao
;
Terada, Masato
Malware can be spread over the Internet via especially download mechanism to the victim computers. This work tries to cluster malware/bots download behavior of Top-10 malware based on 2010 and 2011 CCC (Cyber Clean Center) datasets. The datasets contain more than one million download logs collected from several independent honeypots in Japan to observe malware/bot traffic and activities. Although the daily and hourly patterns are quite similar in 2010, those of 2011 are quite different. As a result, the proposed Integral Correlation Coefficient can cluster 3 and 4 groups of Top-10 malware/bots in 2010 and 2011, respectively. © 2013 IEEE.
Some of the metrics are blocked by your
consent settings
Iterative PCA for population structure analysis
(2011-08-18)
Limpiti, T.
;
Intarapanich, A.
;
Assawamakin, A.
;
Wangkumhang, P.
;
Tongsima, S.
An extension of principal component analysis called ipPCA has been proposed earlier for analyzing structure in genetic data. This non-parametric framework iteratively classifies individuals into subpopulations. However, it is prone to false positives when dealing with large datasets and mixed-type genetic markers. We address these shortcomings by introducing a unified encoding scheme and suggesting a new terminating criterion for ipPCA. To validate the improvements, simulated datasets as well as real bovine and large human genetic datasets are analyzed. It is observed that the estimation of the number of subpopulations and the individual assignment accuracy have been improved. Furthermore, the structure resolved by this approach can be used to identify subset of individuals for further parametric population structure analysis. © 2011 IEEE.
Some of the metrics are blocked by your
consent settings
A combination of decision tree learning and clustering for data classification
(2011-07-21)
Kaewchinpom, Chinnapat
;
Vongsuchoto, Nattakan
;
Srisawat, Anantapom
In this paper, we present a new classification algorithm which is a combination of decision tree learning and clustering called Tree Bagging and Weighted Clustering (TBWC). The TBWC algorithm was developed to enhance a classification performance of a clustering algorithm. In the experiments, five datasets were used to evaluate the predictive performance. The experimental results show that the TBWC algorithm yields the highest accuracies when compared with decision tree learning and clustering for all datasets. In addition, this algorithm can improve the predictive performance especially for multi-class datasets which can increase the accuracy up to 36.67%. Finally, it can reduce attributes up to 59.82%. © 2011 IEEE.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results