KMITL

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

Browse

Search Results

Now showing 1 - 8 of 8

Some of the metrics are blocked by your
consent settings
Effect of Class Weights on Imbalanced Classes in Bi-directional LSTM Training for Sleep Apnea Classification
(2024-01-01)
Jansri, Ukkrit
;
Tretriluxana, Suradej
Sleep apnea, which is defined as the repetitive cessations of breathing during sleep, is the common disorder worldwide. The cost and the process of sleep test to obtain the polysomnogram is not optimal for sleep apnea screening in the large population. A deep learning model was developed to classify the normal and apnea events in a single time-series signal exported from the US National Institute of Health (NIH) sponsored database. Our challenge was to train the model with imbalanced dataset between normal and abnormal respiratory events. Three different methods, Synthetic Minority Over-sampling Technique (SMOTE), Random Under-Sampling (RUS), and the Class Weights (CW) were chosen to improve the model performance over the original data on five selected signals from polysomnographic dataset. The binary classification outputs were evaluated by four metrics. Our results showed (1) Matthews Correlation Coefficient was highest (MCC= 0.1385) in the Class Weights method on the nasal airflow signal. (2) Cohen's Kappa score, was highest (k= 0.0819) in SMOTE technique on the abdominal signal, followed by the Class Weights method on the abdominal signal (k= 0.0687) and RUS technique on nasal airflow signal (k= 0.0441). (3) F1-score was highest (F1= 11.89%) in SMOTE technique on the abdominal signal, followed by the Class Weights method on nasal airflow signal (F1 = 11.17%) and RUS technique on nasal airflow signal (F1= 9.16%). The findings suggest that the Class Weights method on nasal airflow and the Class Weights method on abdominal signal were the two combinations to be used in the DL model.
Some of the metrics are blocked by your
consent settings
Enhancing Auto Insurance Fraud Detection Using Convolutional Neural Networks
(2024-01-01)
Wongpanti, Ratchanon
;
Vittayakorn, Sirion
With the increasing number of vehicles in the global fleet, the size of the auto insurance market is projected to reach $1.3 billion USD by 2030. While this growth in the issuance of auto insurance policies brings prosperity to the industry, it also amplifies the risk of fraudulent activities. These fraudulent practices have a significant impact on the industry, resulting in the loss of billions of USD annually. Despite efforts to prevent such activities, the expertise available is often overwhelmed by the sheer volume of cases. In this paper, we propose an auto insurance fraud detection system that leverages a one-dimensional Convolution Neural Network (ID-CNN) model in combination with two data augmentation techniques, Synthetic Minority Over-sampling Technique (SMOTE) and Conditional Tabular Generative Adversarial Networks (CTGAN), to address the class imbalance problem prevalent in fraud detection datasets. Furthermore, we also employ Focal Loss as the loss function in our deep learning model to effectively tackle the difficulty in classifying the minority class. By combining the ID-CNN model with these imbalance manipulation techniques and the Focal Loss function, we aim to enhance the system's ability to accurately identify fraudulent activities, even in the presence of highly imbalanced data. Our proposed approach seeks to mitigate the financial losses incurred by the auto insurance industry due to fraud and provide a more robust and efficient fraud detection system.
Some of the metrics are blocked by your
consent settings
Effect of Resampling Techniques on Deep Learning Model Training in Sleep Apnea Classification
(2022-01-01)
Jansri, Ukkrit
;
Tretriluxana, Suradej
This study is using deep learning model to classify the respiratory events of Sleep Disordered Breathing (SDB) data. Our pilot results showed the missed identification in some classes even the total accuracy is high. This is the result of unbalanced training dataset given to the model. Two different resampling techniques; Synthetic Minority Over-sampling Technique (SMOTE) and Random Under-Sampling (RUS), were introduced to balance the data. One hundred overnight nasal airflow signals were randomly selected from NIH funded polysomnography database. They were used to train and test these two algorithms with Bi-directional Long Short-Term Memory (Bi-LSTM) model. The results showed greater agreement index when compared between with and without data resampling process. However, SMOTE in sum performed better than RUS (93.72% vs 70.01% in overall accuracy and 0.91 vs 0.55 in Cohen's kappa). It demonstrates that the over-sampling technique is more powerful than under-sampling one. Other resampling techniques will be investigated to make the robust conclusion.
Some of the metrics are blocked by your
consent settings
Application of Natural Neighbor-based Algorithm on Oversampling SMOTE Algorithms
(2021-04-01)
Srinilta, Chutimet
;
Kanharattanachai, Sivakorn
Classification performance depends highly on data distribution. In real life, data often come imbalanced where one class is found more often than others. SMOTE-based algorithms are usually used to handle the class imbalance problem. One key parameter that algorithms in SMOTE family require is k-the number of nearest neighbors with respect to a certain data point. K that fits the dataset the most gives the optimum performance. This paper proposes an approach to suggest a value of the parameter k using Natural Neighbor algorithm. Datasets are made balanced by four SMOTE-based algorithms-standard SMOTE, Safe-Level-SMOTE, ModifiedSMOTE and Weighted-SMOTE. The F-measure and Recall matrices are used to evaluate classification performance of a Support Vector Machine classifier running against six datasets with different imbalance ratios. The results show that, the average classification performance achieved by the proposed k's is closer to the optimum when compared with the performance given by the default value of k.
Some of the metrics are blocked by your
consent settings
A New Hybrid Sampling Approach for Classification of Imbalanced Datasets
(2018-09-11)
Hanskunatai, Anantaporn
Nowadays it is an era of data driven. Many organizations around the world including bank, industry, commercial, and medical intend to extract knowledge from a huge of data. But in the real-word datasets, most of them occur class imbalance problems. This paper presents a new algorithm to handle an imbalanced classification. The proposed technique is a hybrid sampling approach which is the combination of a well know oversampling algorithm called SMOTE and the undersampling technique by removing the ambiguous instances from the majority class instances. The experimental results show that the new hybrid sampling method yields the better predictive performance in term of F-measure when compare with other sampling techniques. In addition, it can improve f-measure up to 59.73% and 412.26% when compare with the original dataset based on decision tree learning and naïve bayes classifiers respectively.
Some of the metrics are blocked by your
consent settings
Auto-Tuning of parameters in hybrid sampling method for class imbalance problem
(2017-02-21)
Sanguanmak, Yotsathon
;
Hanskunatai, Anantaporn
The class imbalance is a major problem in machine learning. This problem affects the performance of a model prediction. The DBSM algorithm, a hybrid-sampling technique, was developed to deal with the class imbalance for two-class classification problem. Although the DBSM algorithm is the effective solution, there are too many parameters for tuning in the algorithm. Thus, this paper proposes an automatic parameter tuning for the DBSM algorithm by using a genetic algorithm (GA), called GADBSM. The experimental results of GADBSM are compared with the DBSM algorithm. The results show that the GADBSM can enhance the classification performance of the DBSM algorithm. Moreover, the GADBSM provides the best in F-measure and AUC in all datasets.
Some of the metrics are blocked by your
consent settings
DBSM: The combination of DBSCAN and SMOTE for imbalanced data classification
(2016-11-18)
Sanguanmak, Yotsathon
;
Hanskunatai, Anantaporn
Many applications in the real world encounter the class imbalance problem. This problem affects the performance of the model prediction. Nowadays, resampling technique is a popular technique to handle the class imbalance problem such as oversampling, undersampling, and hybridsampling. Thus, this paper proposes a new hybrid resampling technique to deal with the class imbalance problem, called DBSM. The concept of DBSM is to use DBSCAN algorithm for undersampling and apply SMOTE technique for oversampling. The experimental results of the DBSM algorithm are compared with an original datasets and other sampling techniques, which are SMOTE, Tomek Links, SMOTE+Tomek Links and DBSCAN. The results show that the DBSM can improve the predictive performance of the classifiers. In addition, it yields the best in the average of AUC, F-measure, and accuracy.
Some of the metrics are blocked by your
consent settings
Bank direct marketing analysis of asymmetric information based on machine learning
(2015-08-24)
Ruangthong, Pumitara
;
Jaiyen, Saichon
The bank direct marketing campaign for offering products that meet the customers' needs is the challenge problems. The bank direct marketing data analysis is important work that helps the banks predict whether customers will sign a long term deposits with the banks. The method that can predict such customers' needs can be profitable to the banks for improving their marketing campaign strategies. Unfortunately, it is very hard to predict the customers' needs because the available information is asymmetric. In this paper, we propose the method to analyze asymmetric information using SMOTE algorithm and Rotation Forest (PCA)-J48. The SMOTE method is used to modify the data and improve the accuracy of the prediction. The performance of the proposed method is evaluated and compared to Decision Tree, Rotation Forest, Navie Bayes, BayesNet, Multilayer Perceptron Neural Network, RBF Neural Network. The experimental results show the predicting accuracies of all predictors. The experiments show that Rotation Forest (PCA)-J48 can achieve the highest value of accuracy and specificity. However, the sensitivity of Rotation Forest (PCA)-J48 is higher than all methods except BayesNet and Rotation Forest (PCA) RandomTree.

KMITL

Browse

Filters

Settings

Sort By

Results per page

Search Results