• Communications for Statistical Applications and Methods
• /
• v.31 no.3
• /
• pp.309-322
• /
• 2024

Identifying algae in water is important for managing algal blooms which have great impact on drinking water supply systems. There have been various microscopic approaches developed for algae classification. Many of them are based on the morphological features of algae. However, there have seldom been mathematical frameworks for comparing the shape of algae, represented as a planar continuous curve obtained from an image. In this work, we describe a recent framework for computing shape distance between two different algae based on the elastic metric and a novel functional representation called the square root velocity function (SRVF). We further introduce statistical procedures for multiple shapes of algae including computing the sample mean, the sample covariance, and performing the principal component analysis (PCA). Based on the shape distance, we classify six algal species in watersheds experiencing algal blooms, including three cyanobacteria (Microcystis, Oscillatoria, and Anabaena), two diatoms (Fragilaria and Synedra), and one green algae (Pediastrum). We provide and compare the classification performance of various distance-based and model-based methods. We additionally compare elastic shape distance to non-elastic distance using the nearest neighbor classifiers.

• The Journal of the Acoustical Society of Korea
• /
• v.29 no.2
• /
• pp.133-140
• /
• 2010

We improve the performance of cardiac disorder classification by adding new temporal features extracted from continuous heart sound signals. We add three kinds of novel temporal features to a conventional feature based on mel-frequency cepstral coefficients (MFCC): Heart sound envelope, murmur probabilities, and murmur amplitude variation. In cardiac disorder classification and detection experiments, we evaluate the contribution of the proposed features to classification accuracy and select proper temporal features using the sequential feature selection method. The selected features are shown to improve classification accuracy significantly and consistently for neural network-based pattern classifiers such as multi-layer perceptron (MLP), support vector machine (SVM), and extreme learning machine (ELM).

• Asia pacific journal of information systems
• /
• v.30 no.1
• /
• pp.31-52
• /
• 2020

Stock markets are popular investment avenues to people who plan to receive premium returns compared to other financial instruments, but they are highly volatile and risky due to the complex financial dynamics and poor understanding of the market forces involved in the price determination. A system that can forecast, predict the stock prices and automatically create a portfolio of top performing stocks is of great value to individual investors who do not have sufficient knowledge to understand the complex dynamics involved in evaluating and predicting stock prices. In this paper the authors propose a Stock prediction, Portfolio Generation and Selection model based on Machine learning algorithms, Artificial neural networks (ANNs) are used for stock price prediction, Mathematical and Statistical techniques are used for Portfolio generation and Un-Supervised Machine learning based on K-Means Clustering algorithms are used for Portfolio Evaluation and Selection which take in to account the Portfolio Return and Risk in to consideration. The model presented here is limited to predicting stock prices on a long term basis as the inputs to the model are based on fundamental attributes and intrinsic value of the stock. The results of this study are quite encouraging as the stock prediction models are able predict stock prices at least a financial quarter in advance with an accuracy of around 90 percent and the portfolio selection classifiers are giving returns in excess of average market returns.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2007.11a
• /
• pp.34-37
• /
• 2007

Face recognition is still a challenging problem in pattern recognition field which is affected by different factors such as facial expression, illumination, pose etc. The facial feature such as eyes, nose, and mouth constitute a complete face. Mouth feature of face is under the undesirable effect of facial expression as many factors contribute the low performance. We proposed a new approach for face recognition under facial expression applying two cascaded classifiers to improve recognition rate. All facial expression images are treated by general purpose classifier at first stage. All rejected images (applying threshold) are used for adaptation using GA for improvement in recognition rate. We apply Gabor Wavelet as a general classifier and Gabor wavelet with Genetic Algorithm for adaptation under expression variance to solve this issue. We have designed, implemented and demonstrated our proposed approach addressing this issue. FERET face image dataset have been chosen for training and testing and we have achieved a very good success.

• Journal of information and communication convergence engineering
• /
• v.22 no.2
• /
• pp.165-171
• /
• 2024

In this study, we present a novel approach for enhancing chest X-ray image classification (normal, Covid-19, edema, mass nodules, and pneumothorax) by combining contrastive learning and machine learning algorithms. A vast amount of unlabeled data was leveraged to learn representations so that data efficiency is improved as a means of addressing the limited availability of labeled data in X-ray images. Our approach involves training classification algorithms using the extracted features from a linear fine-tuned Momentum Contrast (MoCo) model. The MoCo architecture with a Resnet34, Resnet50, or Resnet101 backbone is trained to learn features from unlabeled data. Instead of only fine-tuning the linear classifier layer on the MoCopretrained model, we propose training nonlinear classifiers as substitutes for softmax in deep networks. The empirical results show that while the linear fine-tuned ImageNet-pretrained models achieved the highest accuracy of only 82.9% and the linear fine-tuned MoCo-pretrained models an increased highest accuracy of 84.8%, our proposed method offered a significant improvement and achieved the highest accuracy of 87.9%.

• Asia pacific journal of information systems
• /
• v.33 no.3
• /
• pp.737-767
• /
• 2023

With the availability of big customer data and advances in machine learning techniques, the prediction of customer behavior at the session-level has attracted considerable attention from marketing practitioners and scholars. This study aims to predict customer purchase conversion at the session-level by employing customer profile, transaction, and clickstream data. For this purpose, we develop a multimodal deep learning fusion model with dynamic and static features (i.e., DS-fusion). Specifically, we base page views within focal visist and recency, frequency, monetary value, and clumpiness (RFMC) for dynamic and static features, respectively, to comprehensively capture customer characteristics for buying behaviors. Our model with deep learning architectures combines these features for conversion prediction. We validate the proposed model using real-world e-commerce data. The experimental results reveal that our model outperforms unimodal classifiers with each feature and the classical machine learning models with dynamic and static features, including random forest and logistic regression. In this regard, this study sheds light on the promise of the machine learning approach with the complementary method for different modalities in predicting customer behaviors.

• Asia pacific journal of information systems
• /
• v.29 no.4
• /
• pp.789-816
• /
• 2019

Text classification is a challenging task, especially when dealing with a huge amount of text data. The performance of a classification model can be varied depending on what type of words contained in the document corpus and what type of features generated for classification. Aside from proposing a new modified version of the existing algorithm or creating a new algorithm, we attempt to modify the use of data. The classifier performance is usually affected by the quality of learning data as the classifier is built based on these training data. We assume that the data from different domains might have different characteristics of noise, which can be utilized in the process of learning the classifier. Therefore, we attempt to enhance the robustness of the classifier by injecting the heterogeneous data artificially into the learning process in order to improve the classification accuracy. Semi-supervised approach was applied for utilizing the heterogeneous data in the process of learning the document classifier. However, the performance of document classifier might be degraded by the unlabeled data. Therefore, we further proposed an algorithm to extract only the documents that contribute to the accuracy improvement of the classifier.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.1
• /
• pp.617-623
• /
• 2024

Identifying emotions through speech poses a significant challenge due to the complex relationship between language and emotions. Our paper aims to take on this challenge by employing feature engineering to identify emotions in speech through a multimodal classification task involving both speech and text data. We evaluated two classifiers-Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM)-both integrated with a BERT-based pre-trained model. Our assessment covers various performance metrics (accuracy, F-score, precision, and recall) across different experimental setups). The findings highlight the impressive proficiency of two models in accurately discerning emotions from both text and speech data.

• The Journal of the Convergence on Culture Technology
• /
• v.10 no.3
• /
• pp.145-154
• /
• 2024

In order to operate predictive models utilizing administrative big data, it is crucial to consider policy changes and the characteristics of highly volatile data. Considering this scenario, this study proposes the Cut-off Voting Classifier (CVC) algorithm. This proposed algorithm prevents a sharp decline in accuracy by utilizing multiple weak classifiers. The study validates the proposed algorithm's performance through experiments. The performance evaluation demonstrates the ability to maintain stable prediction rates even in situations with a sharp decline in predictive model accuracy.

• Journal of the Microelectronics and Packaging Society
• /
• v.31 no.2
• /
• pp.45-53
• /
• 2024

The maintenance of semiconductor equipment is crucial for the continuous growth of the semiconductor market. System management is imperative given the anticipated increase in the capacity and complexity of industrial equipment. Ensuring optimal operation of manufacturing processes is essential to maintaining a steady supply of numerous parts. Particularly, monitoring the status of substrate transfer robots, which play a central role in these processes, is crucial. Diagnosing failures of their major components is vital for preventive maintenance. Fault diagnosis methods can be broadly categorized into physics-based and data-driven approaches. This study focuses on data-driven fault diagnosis methods due to the limitations of physics-based approaches. We propose a methodology for data acquisition and preprocessing for robot fault diagnosis. Data is gathered from vibration sensors, and the data preprocessing method is applied to the vibration signals. Subsequently, the dataset is trained using Gradient Tree-based XGBoost machine learning classification algorithms. The effectiveness of the proposed model is validated through performance evaluation metrics, including accuracy, F1 score, and confusion matrix. The XGBoost classifiers achieve an accuracy of approximately 92.76% and an equivalent F1 score. ROC curves indicate exceptional performance in class discrimination, with 100% discrimination for the normal class and 98% discrimination for abnormal classes.