KMITL

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

Browse

Search Results

Now showing 1 - 10 of 11

Some of the metrics are blocked by your
consent settings
Machine learning approach to predict the strength of concrete confined with sustainable natural FRP composites
(2024-07-01)
Ali Talpur, Shabbir
;
Thansirichaisree, Phromphat
;
Poovarodom, Nakhorn
;
Mohamad, Hisham
;
Zhou, Mingliang
Recent earthquakes have highlighted the need to strengthen existing structures with substandard designs. NFRPs provide a sustainable, cost-effective alternative for strengthening, but accurately predicting their performance remains a challenge. This study investigates the use of machine learning algorithms for predicting the compressive strength concrete specimens confined with various NFRPs. Four algorithms were employed: decision tree, random forest, neural network, and gradient boosting regressor. A diverse dataset encompassing various geometries, material properties, and confinement configurations was used to train and evaluate the models. Gradient boosting regressor (GBR) achieved the highest performance, with an average R-squared value of 0.94 and low mean absolute error (MAE) and root mean squared error (RMSE) during training and k-fold cross-validation. Neural network and random forest also demonstrated satisfactory performance, with average R-squared values of 0.88 and 0.86, respectively, during cross-validation. These results suggest that machine learning holds promise for predicting the compressive strength of concrete confined with NFRPs. GBR offers the most accurate predictions, making it a valuable tool for engineers seeking to optimize the design and performance of strengthened structures using sustainable materials.
Some of the metrics are blocked by your
consent settings
Prediction of CO2 emissions using machine learning
(2024-01-01)
Bussaban, Kanyarat
;
Kularbphettong, Kunyanuth
;
Raksuntorn, Nareenart
;
Boonseng, Chongrag
Carbon dioxide (CO<inf>2</inf>) contributes significantly to climate change as a greenhouse gas. The Earth's atmosphere is naturally kept warm enough to support life by greenhouse gases which trap heat in the atmosphere. However, human activity has significantly increased the amount of CO2 in the atmosphere because of deforestation and the use of fossil fuels. One of the key concerns with human evolution that fuels global climate change is carbon dioxide (CO<inf>2</inf>). It is released as fuels burn and as a result, people worldwide are gradually becoming more conscious of environmental issues. Effective policy formulation requires an investigation of the factors influencing CO<inf>2</inf> emissions, yet tiny datasets and traditional research methodologies have hampered prior investigations. This research uses three prediction models to estimate CO<inf>2</inf> trapping efficiency among CO<inf>2</inf> emissions, energy use and GDP: Multiple Linear Regression (MLR), Support Vector Machine (SVM) and Random Forest (RF). The machine learning (ML) techniques used in this work have demonstrated strong performance with multiple linear regressions, support vector machines and random forest models with mean absolute error (MAE), mean absolute percentage error (MAPE) and root mean square error (RMSE). The investigation has proposed a technique for approximating CO<inf>2</inf> emissions and the results indicate that Support Vector Machine (SVM) can attain the highest degree of precision. The outcome could be a useful model for the decision support system to enhance an appropriate course of action for reducing CO<inf>2</inf> emissions worldwide.
Some of the metrics are blocked by your
consent settings
Indoor Device-free Localization Using Received Signal Strength Indicator and Illuminance Sensor for Random-forest-based Fingerprint Technique
(2021-01-01)
Suroso, Dwi Joko
;
Cherntanomwong, Panarat
;
Sooraksa, Pitikhate
Indoor device-free localization (IDFL) offers more flexibility than conventional indoor localization (device-based) systems, as the targets or objects need not be equipped with any device to be located. In the process of IDFL, the target is passive, enabling applications such as monitoring of elderly people, security systems to detect intruders, and indoor navigation. Despite having more flexibility than device-based systems, IDFL is still inferior in terms of localization performance. The most commonly used technique for IDFL is the fingerprint technique, which uses the uniqueness of spatial information to predict the target's location. The spatial information is a fingerprint database containing information on locations and their corresponding parameters. The most specific parameter for the fingerprint database is the received signal strength indicator (RSSI). RSSI can be obtained directly from many low-cost devices, i.e., Wi-Fi-based devices, without the need to install additional hardware. The fingerprint technique is a two-phase process: The database is constructed in the offline phase, and a matching process to compare the target's current parameter with those in the database is performed in the online phase. We propose fingerprint-technique-based IDFL using RSSI and illumination from an illuminance sensor as the additional parameters of the fingerprint database. Both parameters are recorded by considering two scenarios: An empty room and a person standing in the fingerprint grids. The constructed database is the person-filled room subtracted from the empty room database. We use random forest, one of the machine learning (ML) algorithms, as the pattern-matching algorithm. We evaluate its performance by comparison with two other ML algorithms: K-nearest neighbor (k-NN) and neural networks (NN). The results show that k-NN has better accuracy than the random forest for learning and testing in terms of the root mean square error (RMSE). On the other hand, the random forest has better accuracy than NN and better precision than either k-NN or NN for learning and testing in terms of the standard deviation (STD). The results show the possibility of improving the IDFL performance by adding more parameters to the fingerprint database and using an ML-based pattern-matching algorithm.
Some of the metrics are blocked by your
consent settings
Automatic detection of pulmonary nodules using three-dimensional chain coding and optimized random forest
(2020-04-01)
Paing, May Phu
;
Hamamoto, Kazuhiko
;
Tungjitkusolmun, Supan
;
Visitsattapongse, Sarinporn
;
Pintavirooj, Chuchart
The detection of pulmonary nodules on computed tomography scans provides a clue for the early diagnosis of lung cancer. Manual detection mandates a heavy radiological workload as it identifies nodules slice-by-slice. This paper presents a fully automated nodule detection with three significant contributions. First, an automated seeded region growing is designed to segment the lung regions from the tomography scans. Second, a three-dimensional chain code algorithm is implemented to refine the border of the segmented lungs. Lastly, nodules inside the lungs are detected using an optimized random forest classifier. The experiments for our proposed detection are conducted using 888 scans from a public dataset, and achieves a favorable result of 93.11% accuracy, 94.86% sensitivity, and 91.37% specificity, with only 0.0863 false positives per exam.
Some of the metrics are blocked by your
consent settings
Computer-assisted screening for cervical cancer using digital image processing of pap smear images
(2020-03-01)
Win, Kyi Pyar
;
Kitjaidure, Yuttana
;
Hamamoto, Kazuhiko
;
Aung, Thet Myo
Cervical cancer can be prevented by having regular screenings to find any precancers and treat them. The Pap test looks for any abnormal or precancerous changes in the cells on the cervix. However, the manual screening of Pap smear in the microscope is subjective with poorly reproducible criteria. Therefore, the aim of this study was to develop a computer-assisted screening system for cervical cancer using digital image processing of Pap smear images. The analysis of Pap smear image is important in the cervical cancer screening system. There were four basic steps in our cervical cancer screening system. In cell segmentation, nuclei were detected using a shape-based iterative method, and the overlapping cytoplasm was separated using a marker-control watershed approach. In the features extraction step, three important features were extracted from the regions of segmented nuclei and cytoplasm. RF (random forest) algorithm was used as a feature selection method. In the classification stage, bagging ensemble classifier, which combined the results of five classifiers-LD (linear discriminant), SVM (support vector machine), KNN (k-nearest neighbor), boosted trees, and bagged trees-was applied. SIPaKMeD and Herlev datasets were used to prove the effectiveness of our proposed system. According to the experimental results, 98.27% accuracy in two-class classification and 94.09% accuracy in five-class classification was achieved using the SIPaKMeD dataset. When the results were compared with five classifiers, our proposed method was significantly better in two-class and five-class problems.
Some of the metrics are blocked by your
consent settings
Effectiveness of Six Text Classifiers for Predicting SET Stock Price Direction
(2020-01-01)
Netisopakul, Ponrudee
;
Saewong, Woranun
Six text classification methods were compared to find the best model for predicting Stock Exchange of Thailand stock prices. News headlines, on individual stocks, were classified as causing “change” and “no-change” based on a preset change threshold, 2.5%. The training dataset was collected by matching stock news in 2018 with stock names and filling in stock price changes. 258 news were associated with a “change” and 636 news with “no-change”. The Thai text news items were preprocessed and converted to TF-IDF vector representation. Six machine learning text classification methods are applied to create six text classifier models and create a confusion matrix, then compared with actual changes to obtain accuracy scores. We found that a deep learning classifier (with 85.6% accuracy) scored better than other classifiers for one day price movement to assist short-term investments.
Some of the metrics are blocked by your
consent settings
Cervical cancer detection and classification from pap smear images
(2019-09-16)
Win, Kyi Pyar
;
Kitjaidure, Yuttana
;
Paing, May Phyu
;
Hamamoto, Kazuhiko
In this paper, we propose a framework for detection and classification of cervical cancer from pap smear images. Early detection and accurate diagnosis of cervical cancer can reduce the death rate of cervical cancer patients. Pap smear or pap test is the most popular technique for early detection of cervical cancer. However, the manual analysis is labor intensive and time consuming process which relies on expert cytologist. Hence, it is needed to develop a computer aided diagnosis system to make the pap smear test more accurate and reliable. The objective of this paper is to present an innovative idea of applying random forest algorithm (RF) as a feature selection method using proposed bagging ensemble classifier for improving the predictive performance. The four basic steps of cervical cancer detection and classification system, image enhancement, segmentation, feature extraction and classification were used. K-means clustering combining with morphology operations obtained good segmentation for cell nuclei and cytoplasm. The most important features, shape, color and texture of nuclei and cytoplasm were applied to detect cervical cancer. To improve the accuracy of prediction results, random forest (RF) algorithm was used as a feature selection method. In classification stage, bagging ensemble classifier was applied which aggregated the results of five classifiers, linear discriminant (LD), support vector machine (SVM), weighted k-nearest neighbor (KNN), boosted trees and bagged trees. Herlev data set was used to prove the effectiveness of our proposed method. According to the experimental results, the high classification accuracy was achieved with top10 features using our proposed combined classifier. The accuracy was 97.83% in two class problem and 81.54% in seven class problem. When the results were compared with five classifiers, our proposed method was significantly better in two class and seven class problems.
Some of the metrics are blocked by your
consent settings
Predicting sheet and rill erosion of Shihmen reservoir watershed in Taiwan using machine learning
(2019-07-01)
Nguyen, Kieu Anh
;
Chen, Walter
;
Lin, Bor Shiun
;
Seeboonruang, Uma
;
Thomas, Kent
Shihmen Reservoir watershed is vital to the water supply in Northern Taiwan but the reservoir has been heavily impacted by sedimentation and soil erosion since 1964. The purpose of this study was to explore the capability of machine learning algorithms, such as decision tree and random forest, to predict soil erosion (sheet and rill erosion) depths in the Shihmen reservoir watershed. The accuracy of the models was evaluated using the RMSE (Root Mean Squared Error), MAE (Mean Absolute Error), and R<sup>2</sup>. Moreover, the models were verified against the multiple regression analysis, which is commonly used in statistical analysis. The predictors of these models were 14 environmental factors which influence soil erosion, whereas the target was 550 erosion pins installed at 55 locations (on 55 slopes) and monitored over a period of approximately three years. The data sets for the models were separated into 70% for the training data and 30% for the testing data, using the simple random sampling and stratified random sampling methods. The results show that the random forest algorithm performed the best of the three methods. Moreover, the stratified random sampling method had better results among the two sampling methods, as anticipated. The average error (RMSE relative to 1:1 line) of the stratified random sampling method of the random forest algorithm is 0.93 mm/yr in the training data and 1.75 mm/yr in the testing data, respectively. Finally, the random forest algorithm predicted that type of slope, slope direction, and sub-watershed are the three most important factors of the 14 environmental factors collected and used in this study for splits in the trees and thus they are the three most important factors affecting the depth of sheet and rill erosion in the Shihmen Reservoir watershed. The results of this study can be employed by decision-makers to improve soil conservation planning and watershed remediation.
Some of the metrics are blocked by your
consent settings
Comparison of Sampling Methods for Imbalanced Data Classification in Random Forest
(2019-01-10)
Paing, May Phu
;
Pintavirooj, C.
;
Tungjitkusolmun, Supan
;
Choomchuay, Somsak
;
Hamamoto, Kazuhiko
Imbalanced data classification is a serious and challenging task for most of the medical image diagnosis applications. They usually produce a larger number of false samples compared to the actual ones. That is the number of samples for the class of interest (minority) is significantly fewer than other types of class (majority). The classification performed using such data is called imbalanced data classification. As a consequence, the learning model bias towards the majority class and fails the classification of the minority class. Data sampling and ensemble methods are common ways to compensate for this issue. Random forest (RF), an ensemble of multiple decision trees, is very famous in both of the classification and regression problems because of its robust and accurate predictions. However, it also suffers class bias in the imbalanced data classification problems. This paper proposes and compares different sampling methods to solve the imbalanced data classification in RF.
Some of the metrics are blocked by your
consent settings
Detection and classification of overlapping cell nuclei in cytology effusion images using a double-strategy random forest
(2018-09-11)
Win, Khin Yadanar
;
Choomchuay, Somsak
;
Hamamoto, Kazuhiko
;
Raveesunthornkiat, Manasanan
Due to the close resemblance between overlapping and cancerous nuclei, the misinterpretation of overlapping nuclei can affect the final decision of cancer cell detection. Thus, it is essential to detect overlapping nuclei and distinguish them from single ones for subsequent quantitative analyses. This paper presents a method for the automated detection and classification of overlapping nuclei from single nuclei appearing in cytology pleural effusion (CPE) images. The proposed system is comprised of three steps: nuclei candidate extraction, dominant feature extraction, and classification of single and overlapping nuclei. A maximum entropy thresholding method complemented by image enhancement and post-processing was employed for nuclei candidate extraction. For feature extraction, a new combination of 16 geometrical and 10 textural features was extracted from each nucleus region. A double-strategy random forest was performed as an ensemble feature selector to select the most relevant features, and an ensemble classifier to differentiate between overlapping nuclei and single ones using selected features. The proposed method was evaluated on 4000 nuclei from CPE images using various performance metrics. The results were 96.6% sensitivity, 98.7% specificity, 92.7% precision, 94.6% F1 score, 98.4% accuracy, 97.6% G-mean, and 99% area under curve. The computation time required to run the entire algorithm was just 5.17 s. The experiment results demonstrate that the proposed algorithm yields a superior performance to previous studies and other classifiers. The proposed algorithm can serve as a new supportive tool in the automated diagnosis of cancer cells from cytology images.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results