KMITL

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

Browse

Search Results

Now showing 1 - 7 of 7

Some of the metrics are blocked by your
consent settings
Correlations Between Subjective Icon Characteristics and Entropy From A User Perspective
(2026-03-20)
Satcharoen, Kleddao
;
Tangtisanon, Pikulkaew
Some of the metrics are blocked by your
consent settings
A Portable Electronic Nose for Real-time Monitoring of Food Spoilage Using Multiple Machine Learning Models
(2025-01-01)
Tangtisanon, Pikulkaew
;
Grodniyomchai, Boonyawee
In this study, we present the design and development of a portable electronic nose (E-nose) system for detecting and classifying spoiled household food through the application of machine learning (ML) techniques. The targeted odors include fungi from bread, spoiled rice, spoiled milk, yoghurt, rotten egg, rotten boiled egg, rotten pork, and rotten beef, totaling eight odor classes. A total of 1800 samples were collected using three gas sensors and one temperature sensor. After outlier removal with Isolation Forest, 1000 samples remained. Multiple ML models were trained and evaluated over ten iterations, comparing classification accuracy and processing time. Among all the models, the k-nearest neighbor (KNN) achieved the highest performance, with an average accuracy of 99.889% and an average processing time of 0.167477 s. The decision tree (DT) model followed closely with an accuracy of 99.843% and required a significantly less processing time of 0.012080 s. Although DT has a slightly lower accuracy than KNN, its processing time is 13.86 times faster. For our scenario that requires real-time results, DT is a better choice than KNN. The proposed portable E-nose demonstrates strong potential for realworld applications such as food spoilage detection, environmental monitoring, and health diagnostics.
Some of the metrics are blocked by your
consent settings
Enhancing Odor Classification of Essential Oils with Electronic Nose Data
(2024-01-01)
Grodniyomchai, Boonyawee
;
Satcharoen, Kleddao
;
Tangtisanon, Pikulkaew
In the current business landscape, the fragrance industry has gained substantial prominence. In this context, there is a requirement to create the most compact and sufficiently accurate model possible, suitable for deployment on a portable device. The aim is to develop a model capable of effectively classifying various fragrance types based on data pertaining to air properties and fragrance component attributes. This paper presents the feature extraction from the dataset electronic node to classify odor types using a machine learning model compared before and after the feature extraction of the dataset. In our investigation, we employed datasets of varying sizes, including small datasets (composed of 1000 samples), large datasets (composed of 10000 samples), and raw datasets (composed of 21000 samples). This methodology was employed to discern disparities in model performance, average accuracy, and computational runtime across these different dataset sizes. We observed that the decision tree model, post-training with principal component analysis, showed a performance improvement when compared to the basic machine learning model. Specifically, the decision tree model achieved accuracy rates of 100.00%, 99.97%, and 97.00% respectively.
Some of the metrics are blocked by your
consent settings
A Generational Cohort Comparison of Icon Selection Accuracy under Varying Conditions of Icon Entropy and Concreteness
(2023-01-01)
Satcharoen, Kleddao
;
Tangtisanon, Pikulkaew
—The objective of this research was to compare icon selection accuracy under varying icon entropy and concreteness conditions between different generational cohorts (Millennial, Generation X, and Baby Boomers). These generational cohorts have different levels of experience with technology, with younger generations often being framed as “digital natives” and holding stronger technological experience and competence in comparison to older groups. Generational groups also have variations in physiological factors including visual acuity and reaction time. Despite these differences between user groups, many user interaction systems and processes are designed for a single user, rather than considering differences in user processing between different groups. Therefore, this study compares generational cohorts in their icon selection accuracy under varying icon conditions, to help identify what generational differences can be observed in this task. The study selected a sample of 150 participants (n = 50 for each generational cohort). The experiment was a 2☓2☓3 design (entropy (high/low) ☓ abstractness (abstract/concrete) ☓ time (9/6/3 seconds), with each participant completing 60 trials (five questions per entropy/abstractness pair over three timed runs). Results showed that there were significant differences in mean accuracy per trial under all of the time conditions and icon entropy and concreteness conditions. Mean differences showed that under most conditions, Millennial and Generation X participants did not have a significant mean difference, but Baby Boomers were significantly slower under almost all conditions. The implication of this finding is that Baby Boomers are more sensitive to icon abstractness and entropy conditions than other age groups tested.
Some of the metrics are blocked by your
consent settings
The Factors that Affect the Selecting of Sticker Image on Social Media Application: A Case Study of Retired People Purchasing Behavior in Thailand
(2022-01-01)
Satcharoen, Kleddao
;
Tangtisanon, Pikulkaew
This research aimed to investigate factors that influence retirees in Thailand on selection of Line stickers image to purchase. The study uses an integrated model of attitude-behavior decision-making, incorporating the theory of planned behavior (TPB) and technology acceptance model (TAM) along with personal and demographic factors. The data were collected via online questionnaire survey, from 400 retirees who live in Thailand. The results showed that an integrated model that incorporated TPB and TAM was generally successful at predicting consumer attitudes, behavioral intentions, and behaviors toward the purchase of Line stickers image. However, there was a notable exception, which was that subjective norms did not have a significant effect on attitudes except for perceived ease of use.
Some of the metrics are blocked by your
consent settings
COVID-19 Pandemic Prevention Mobile Application for on Campus Classroom
(2021-04-23)
Tangtisanon, Pikulkaew
COVID-19 pandemic is a novel coronavirus that has not been found in humans before. This virus can be transmitted to other humans primarily through respiratory secretions when an infected person coughs or talks. To avoid human to human transmission of this pandemic, the government extend the state of emergency policies which cause vital damage for many business sections worldwide including educational institution. Many schools decide to let the students learn at home. However, in practical courses such as a chemical workshop, students must come to the laboratory room to perform experiments which may increase the risk of infection. In order to prevent the spread of COVID-19 between students and staff, anybody entering the school must conduct a risk assessment, measure a body temperature and wear a face mask at all times. Many COVID-19 contact tracing platforms allow users to assess infection risk and notify if they have been exposed to infected persons. Unfortunately, they cannot be used effectively with the on-campus education system. The proposed mobile application was developed to handle the needs of the onsite education system during the ongoing COVID-19 situation in schools. The application contains three main functions which are a COVID-19 self-assessment, a roll-call, and a social distancing function. This paper focused on the roll-call function using face recognition and Global Positioning System (GPS). In a normal situation, the student just opens an application, shows his or her face to a smartphone camera then the application will detect a face part and easily recognize the student's identification. However, in the new normal situation where everyone must wear a mask, it will be a very difficult task to perform face recognition since almost half of the face is hidden. The convolutional neural network (CNN) was applied to train a CNN model using a dataset of 18 peoples with non face mask wearing and face mask wearing. The face mask wearing consisted of three different face mask types: Disposable surgical mask (DS), N95 face respirators (N95) and general 3D mask (3D). After that, the model was exported to the proposed mobile application. Experimental results on a realworld dataset show that the proposed model can be used with a high accuracy rate in non face mask samples. In face mask samples, the 3D mask has the highest accuracy rate.
Some of the metrics are blocked by your
consent settings
Holy Basil Curl Leaf Disease Classification using Edge Detection and Machine Learning
(2020-02-14)
Tangtisanon, Pikulkaew
;
Kornrapat, Suttipong
Holy basil (Ocimum basilicum L.) is one of the most vital economic crops that has a significant impact on export earnings. However, the holy basil prices could be dropped due to a curl leaf disease caused by pests. Several previous studies focused on plant leaf disease detection based on the leaf color. Unfortunately, the leaf curl disease sometimes changes a shape of the leaf not the color so it cannot be detected with those schemes. We proposed a novel approach aims to automatically detect a curling leaf on holy basil. This paper presents a Neural Network (NN) model and Logistic Regression (LR) model to automatically detect a curling leaf on holy basil. To be able to detect the infected one not by colors but by its shape, we have applied edge detection algorithms which are Canny and Sobel model. To speed up processing time, images were resized and converted to grayscale before passing them to machine learning models. Moreover, NN and LR were modified with mini-batch technique in order to increase the speed of the processing time. The dataset contains 600 images of holy basil leaves with 300 images of healthy leaves and 300 images of infected leaves. The experimental results indicate that the proposed method effectively detects the curling leaves on holy basil.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results