Feature Extraction Techniques Research Articles

Integrating binary classification and clustering for multi-class dysarthria severity level classification: a two-stage approach

Dysarthria, a motor speech disorder, poses challenges in accurate severity assessment. Recent research has excelled in classifying dysarthria based on severity levels, primarily utilizing annotated datasets and achieving high accuracies. However, these classification-based approaches may not readily translate to real-world scenarios without predefined labels. This study follows a different path by proposing a two-stage approach leveraging binary classification and clustering to comprehensively analyze and classify dysarthria severity levels. We begin by employing binary classification to differentiate control from dysarthric cases by experiencing eight different feature extraction techniques and two classifiers in order to support the largest amount of dysarthric cases to the second stage, where k-means clustering uncovers hidden patterns and boundaries within dysarthria severity levels, enabling a more nuanced understanding of the disorder. We applied our methodology to the TORGO dataset, a benchmark in dysarthria research, and evaluated it on the UA Speech dataset. After optimizing the number of clusters, our approach achieved an accuracy of 91% with sentence-based features and 85% with word-based features in clustering. This research extends previous studies by exploring unsupervised clustering to differentiate severity levels in unannotated cases, bridging the gap between controlled datasets and practical applications. Our findings highlight the effectiveness of clustering-driven two-stage analysis in improving dysarthria severity-level classification, with implications for real-world clinical settings.

Utilization of Artificial Intelligence for the automated recognition of fine arts

Fine art recognition, traditionally dependent on human expertise, is undergoing a significant transformation with the integration of Artificial Intelligence (AI) and deep learning. This article introduces a novel AI-based approach for fine art recognition, utilizing Convolutional Neural Networks (CNNs) and advanced feature extraction techniques. Addressing the inherent challenges within this domain, we present a systematic methodology to enhance automated fine art recognition. By leveraging critical dataset characteristics such as objective type, genre, material, technique, and department, our method exhibits exceptional performance in classifying fine art pieces across diverse attributes. Our approach significantly improves accuracy and efficiency by integrating advanced feature extraction techniques with a customized CNN architecture. Experimental validation on a benchmark dataset highlights the efficacy of our method, indicating substantial contributions to the interdisciplinary field of fine art analysis.

Exploring the Impact of Image-Based Audio Representations in Classification Tasks Using Vision Transformers and Explainable AI Techniques

An important hurdle in medical diagnostics is the high-quality and interpretable classification of audio signals. In this study, we present an image-based representation of infant crying audio files to predict abnormal infant cries using a vision transformer and also show significant improvements in the performance and interpretability of this computer-aided tool. The use of advanced feature extraction techniques such as Gammatone Frequency Cepstral Coefficients (GFCCs) resulted in a classification accuracy of 96.33%. For other features (spectrogram and mel-spectrogram), the performance was very similar, with an accuracy of 93.17% for the spectrogram and 94.83% accuracy for the mel-spectrogram. We used our vision transformer (ViT) model, which is less complex but more effective than the proposed audio spectrogram transformer (AST). We incorporated explainable AI (XAI) techniques such as Layer-wise Relevance Propagation (LRP), Local Interpretable Model-agnostic Explanations (LIME), and attention mechanisms to ensure transparency and reliability in decision-making, which helped us understand the why of model predictions. The accuracy of detection was higher than previously reported and the results were easy to interpret, demonstrating that this work can potentially serve as a new benchmark for audio classification tasks, especially in medical diagnostics, and providing better prospects for an imminent future of trustworthy AI-based healthcare solutions.

Detection of offshore oil slicks utilizing a novel joint feature extraction technique that integrates limited hyperspectral image data

Detection of offshore oil slicks utilizing a novel joint feature extraction technique that integrates limited hyperspectral image data

Machine Learning Models for the Classification of Histopathological Images of Colorectal Cancer

The aim of this study was to explore the application of computational models for the analysis of histopathological images in the context of colon cancer. A comprehensive dataset of colon cancer images annotated into eight distinct categories based on their representation of cancerous cell portions was used. The primary objective was to employ various image classification algorithms to assess their efficacy in the context of cancer classification. Additionally, this study investigated the use of feature extraction techniques to derive meaningful data from the images, contributing to a more nuanced understanding of cancerous tissues, comparing the performance of different image classification algorithms in the context of colon cancer image analysis. The findings of this research suggested that XGboost provides the highest accuracy (89.79%) and could contribute to the growing body of knowledge in computational pathology. Other algorithms, such as the random forest, SVM, and CNN, also provided satisfactory results, offering insights into the effectiveness of image classification algorithms in distinguishing between different categories of cancerous cells. This work holds implications for the development of more accurate and efficient tools, underscoring the potential of computational models in enhancing the analysis of histopathological images and improving diagnostic capabilities in cancer research.

Detecting Fake Reviews Using BERT and Sublinear_TF Methods on Hotel Reviews in the Lombok Tourism Area

The number of visitors to Lombok, one of the famous tourist destinations in Indonesia, increased from 400,595 in 2020 to 1,376,295 in 2022. Although the government supports the hotel industry, fake reviews are a significant problem that can damage hotel reputations and mislead tourists. This study uses BERT and Sublinear_TF feature extraction techniques to analyze fake reviews from three main areas: Gili Trawangan, Senggigi, and Kuta. BERT detects fake reviews by understanding the context of words, while Sublinear_TF emphasizes more informative words by reducing the weight of irrelevant common words. The results showed that the more extensive and diverse dataset from Gili Trawangan had the best classification results. The combination of BERT and Random Forest achieved the highest accuracy of 0.84. Overall, BERT excels in Gili Trawangan with an accuracy of 0.79 for SVM and 0.84 for Random Forest. In contrast, smaller and more homogeneous datasets such as Senggigi and Kuta have lower accuracy. In addition, Sublinear_TF performed well on Gili Trawangan with an accuracy of 0.82 using SVM and 0.83 using Random Forest; however, its performance declined in Senggigi and Kuta. BERT and Sublinear_TF techniques are more effective on large and diverse datasets such as Gili Trawangan. Sublinear_TF is better for varied data but less effective on more homogeneous datasets, while BERT with Random Forest showed the highest accuracy due to its ability to capture broader language context. This suggests that the size and variety of the dataset highly influence the success of fake review classification techniques.

Hate Comment Detection On Twitter Using Long Short Term Memory (LSTM) With Genetic Algorithm (GA)

In the era of social media like today, one social media that is currently quite popular is Twitter. This study explores the use of the Long Short Term Memory (LSTM) method optimized with the Genetic Algorithm (GA) to detect hate speech in Twitter data in Indonesia. We use TF-IDF and GloVe feature extraction techniques to produce effective word vector representations in natural language processing. This study also introduces feature expansion and similarity corpus construction to improve the performance of the LSTM classification model. Evaluation is carried out through a confusion matrix to measure accuracy, precision, recall, and F1 score. The results show that the LSTM model with TF-IDF and GloVe feature extraction achieves the best performance with an accuracy of up to 92.91%. We also found that the combination of Unigram + Bigram + Trigram, max feature 10000, and Glove corpus with Top 20 similarity gave optimal results. In addition, parameter optimization using genetic algorithms has been shown to improve accuracy and F1-Score. The resulting LSTM model is able to classify test data with high accuracy, which has the potential to help in the detection and handling of hate speech on social media, as well as improving the model's ability to identify and understand text content in the Indonesian language context.

A novel feature extraction-selection technique for long lead time agricultural drought forecasting

Long-term drought forecasting plays a crucial role in mitigating drought risks by providing early warnings. Researchers have long been interested in achieving accurate long-term drought forecasting, which is challenging since accuracy generally decreases by increasing the forecasting period. The primary aim of this research is to propose a new method for high-accuracy long lead time drought forecasting by combining various Feature Extraction (FE) and selection techniques. In this study, monthly time-series datasets encompassing precipitation, potential evapotranspiration, actual evapotranspiration, runoff, surface and root-zone soil moisture—were utilized to forecast SPEI-6 over various lead times including 1-, 3-, 6-, 9-, 12-, 18-, and 24-months using global gridded products with a 0.5O × 0.5O spatial resolution spanning the years January 1980 to December 2022. The method was evaluated using two different approaches, namely Gaussian Process Regression (GPR) as a simple machine learning technique and Long Short-Term Memory (LSTM) as a deep learning method. The findings provided improved accuracy, particularly for long-term forecasting when employing the proposed methodology. When utilizing LSTM with FE instead of the original datasets as inputs, the error reduced from RMSE = 0.16 to RMSE = 0.07 (a 56 % decrease), while the correlation increased from R = 0.65 to R = 0.90 (a 38 % increase) when forecasting SPEI-6 12 months ahead. The results showed that the GPR with FE and selection model outperformed the LSTM with original datasets model for SPEI-6 (t + 24) with a correlation coefficient (R) of 0.9811 and a Normalized Root Mean Square Error (NRMSE) of 0.1380, compared to R = 0.6517 and NRMSE = 0.4307 for the LSTM with original datasets. These findings can offer valuable insights for early agricultural drought warning in arid areas.

Integrated Anti-Aliasing and Fully Shared Convolution for Small-Ship Detection in Synthetic Aperture Radar (SAR) Images

Synthetic Aperture Radar (SAR) imaging plays a vital role in maritime surveillance, yet the detection of small vessels poses a significant challenge when employing conventional Constant False Alarm Rate (CFAR) techniques, primarily due to the limitations in resolution and the presence of clutter. Deep learning (DL) offers a promising alternative, yet it still struggles with identifying small targets in complex SAR backgrounds because of feature ambiguity and noise. To address these challenges, our team has developed the AFSC network, which combines anti-aliasing techniques with fully shared convolutional layers to improve the detection of small targets in SAR imagery. The network is composed of three key components: the Backbone Feature Extraction Module (BFEM) for initial feature extraction, the Neck Feature Fusion Module (NFFM) for consolidating features, and the Head Detection Module (HDM) for final object detection. The BFEM serves as the principal feature extraction technique, with a primary emphasis on extracting features of small targets, The NFFM integrates an anti-aliasing element and is designed to accentuate the feature details of diminutive objects throughout the fusion procedure, HDM is the detection head module and adopts a new fully shared convolution strategy to make the model more lightweight. Our approach has shown better performance in terms of speed and accuracy for detecting small targets in SAR imagery, surpassing other leading methods on the SSDD dataset. It attained a mean Average Precision (AP) of 69.3% and a specific AP for small targets (APS) of 66.5%. Furthermore, the network’s robustness was confirmed using the HRSID dataset.

SMS Spam Filteration Using Text Features and Supervised Machine Learning Algorithms

Over time, technological advancements have had an immense effect on every aspect of life, including travel, office work, music, healthcare, and communication. In the past, people communicated using telephone lines. With far more functionality than telephone cable technology, wireless technology already prevails. SMS is mostly used by spammers and advertising firms to communicate with the general public and distribute company pamphlets. This explains why over 60% of spam SMS are sent and received every day. Although these spam communications irritate users and occasionally con unsuspecting users, the spammers and ad businesses benefit handsomely from them. This paper suggested a method for categorizing ham and spam SMS using supervised machine learning approaches. Features are extracted from data using feature extraction techniques like bag-of- words and Term Frequency-Inverse Document Frequency (TF-IDF). The imbalance in the SMS dataset we used was addressed by applying both oversampling and under sampling techniques. The support vector classifier, gradient boosting machine, random forest, Gaussian Naive Bayes, and logistics regression are implemented on the using spam SMS and ham SMS data sets, evaluated by F1 score, accuracy, precision and recall are used to assess performance. According to the experiment's findings, the random forest diagnoses spam and ham SMS more precisely-99% of the time.

Automated lung cancer detection using novel genetic TPOT feature optimization with deep learning techniques

Lung cancer remains a leading cause of cancer-related deaths globally. Early and accurate detection is crucial for improving patient outcomes. Traditional methods, relying on manual interpretation of medical images, are time-consuming and prone to errors. Deep learning, particularly convolutional neural networks (CNNs), offers an automated alternative capable of learning intricate patterns from medical images. However, previous deep learning models for lung cancer detection have faced challenges such as limited data, inadequate feature extraction, interpretability issues, and susceptibility to data variability. This paper presents a novel deep learning methodology that addresses these limitations. Our approach leverages expansive datasets, incorporates advanced feature extraction techniques, improves interpretability, and accommodates the diverse nature of lung cancer. Specifically, we develop dedicated models for both chest X-ray and CT images utilizing publicly available datasets from Kaggle. Through the integration of feature selection and model selection techniques—such as employing a genetic algorithm in conjunction with the tree-based pipeline optimization tool (TPOT)—we achieved remarkable accuracy. Our X-ray model attains an overall accuracy of 95.47%, the CT model achieves an accuracy of 98.70%, and the combined model achieves an impressive overall accuracy of 98.93%. Our methodology significantly enhances the performance and efficiency of lung cancer detection and is a valuable tool for early diagnosis and intervention.

Advanced Security Strategies for Software Vulnerability Protection: A Multi-Scale Detection Model With FPGA-RFID

This paper proposes a multi-scale software vulnerability detection model using Radio Frequency Identification (RFID) target recognition technology and Field-Programmable Gate Array (FPGA) hardware design. The systematic analysis of RFID systems and FPGA architectures provides the foundation for constructing robust security solutions against code-based threats. The Scalable Vulnerability Detection Method (SVDM) is developed leveraging multi-scale code metrics and deep feature extraction techniques. The evaluations of vulnerability datasets CWE-119 and CWE-399 demonstrated over 84% accuracy, recall, and F1 score. The precision, recall, and F1 score of the CWE-399 vulnerability dataset in the SVDM were 85.15%, 85.12%, and 84.79%, respectively, while the values of the CWE-199 dataset were 83.12%, 82.16%, and 82.38%, respectively. Compared with existing methods, the functionality of SVDM was further highlighted. This research provides an extensible and adaptable framework for identifying security vulnerabilities with both hardware and software techniques. It can optimize threat detection, risk analysis, and system integrity across software development and operation life cycles, providing technical reference and scientific support for the security protection of software vulnerabilities, while promoting the development of industries such as intelligent monitoring and traceability.

Public opinion mining in social media about Ethiopian broadcasts using deep learning

Now adays people express and share their opinions on various events on the internet thanks to social media. Opinion mining is the process of interpreting user-generated opinion data on social media. Aside from its lack of resources in opinion-mining tasks, Amharic presents numerous difficulties because of its complex structure and variety of dialects. Analyzing every comment written in Amharic is a challenging task. Significant advancements in opinion mining have been achieved using deep learning. An opinion-mining model was used in this study to classify user comments written in Amharic as positive or negative. The domains that we focus on in this study are YouTube and Facebook. From the Ethiopian broadcasts YouTube and Facebook official pages, we gathered 11,872 unstructured data for this study using www.exportcomment.com, and Facebook page tools. Text preprocessing and feature extraction techniques were used, in addition to manual annotation by linguistic specialists. The dataset was prepared for the experiment after annotation, preprocessing, and representation. LSTM, GRU, BiGRU, BiLSTM, and a hybrid of CNN with BiLSTM classifiers from the TensorFlow Keras deep learning library were used to train the model using the dataset, which was split using the 80/20 train-test method, which proved effective for classification problems. Finally, we achieved of 94.27%, 95.20%, 95.49%, 95.62%, and 96.08% using GRU, BiGRU, LSTM, BiLSTM, and CNN with BiLSTM, respectively, in word2vec embedding model.

Combined LinkNet-MBi-LSTM for brain activity recognition with new Stockwell transform features.

Recognizing brain activity from EEG waves is an important field of study in biomedical engineering and neuroscience. Conventional approaches usually begin with signal processing techniques to extract features from the EEG data, and then machine learning algorithms are applied to classify the data. However, the spatial resolution of these EEG signals is low, which makes it difficult to pinpoint the exact location of the neural activity source in the brain. There are ongoing initiatives to use DL-based brain activity recognition algorithms to overcome these constraints. Therefore, this work presents a novel hybrid framework for brain activity detection using the enhanced Stockwell transform and an EEG signal that is called LinkNet and modified bidirectional-long short-term memory (LN-MBi-LSTM) model. This framework follows a methodical approach that includes stages for feature extraction, brain activity recognition and preprocessing. Firstly, the improved Weiner filtering (IWF) approach is used to preprocess the EEG input signal. The relevant features are then extracted using a feature extraction technique from the preprocessed EEG signal. To identify the brain activity, these recovered feature sets are subsequently processed separately using LinkNet and modified bidirectional-long short-term memory (MBi-LSTM). A thorough analysis that takes into account both simulation and experimental calculations is part of the validation process for the LN-MBi-LSTM model. Finally, this study demonstrates the therapeutic potential of the LN-MBi-LSTM framework by presenting a strong and verified model for brain activity recognition. With the highest precision of 0.997, the LinkNet-MBi-LSTM model distinguishes itself from other models and confirms its exceptional capacity to produce accurate positive predictions.

Data-driven AI for the automated classification of the isothermal heat-treated thermal barrier coatings using pulsed infrared thermography

Abstract Development of reliable age prediction models are crucial in monitoring the formation of oxide layer and degradation of TBC at regular intervals. This study proposes an automated classification of isothermal heat-treated TBC samples using temperature data, which helps in predicting the TBC life and monitoring the TBC degradation. TBC-coated samples are isothermal heat-treated at 1000 °C, and the initial growth of thermally grown oxide is monitored using a non-destructive thermal imaging technique. The proposed study integrates data-driven AI (DAI) models and feature extraction techniques to interpret complex thermal patterns measured from the TBC coating surface. The performance of the proposed classification framework is tested using deep learning and classical machine learning models with different types and window sizes of input data. Input data used for validation are raw experiment data, logarithmic of experiment data, polynomial fit data, and thermal signal reconstruction fit coefficients. The maximum classification performance is obtained with gated recurrent unit with accuracy and F1-score of 89.2% and 89.0%, respectively with raw temperature data as input of window 300. The study demonstrates that the proposed DAI approach effectively predicts the age of thermal barrier coatings under isothermal heat-treatment conditions by correlating the thermal response with coating degradation.

SOCIAL MEDIA FAKE ACCOUNT IDENTIFICATION USING ML

The proliferation of social media platforms has led to an increase in the creation of fake accounts. These accounts are used for various malicious activities, such as spreading false information, phishing, and identity theft. As a result, there is a growing need for effective methods to identify and eliminate fake accounts. This paper proposes a machine learning-based approach for social media fake account identification. This paper proposes a machine learning-based approach to identify fake accounts on social media platforms. Our method leverages a combination of feature extraction techniques and supervised learning algorithms to classify accounts as genuine or fake. We collect a large dataset of labeled social media accounts, including both genuine and fake profiles, and extract features from various sources such as profile information, network behavior, and content analysis. We experiment with multiple machine learning models, including Support vector machines(SVM),K-Nearest Neighbors Algorithm(KNN),Random forest, Logistic Regression & Artificial Neural Network(ANN) to evaluate their performance in identifying fake accounts. Our proposed method has significant implications for social media platform operators, policymakers, and researchers seeking to combat fake accounts and maintain online trust. The approach can be integrated with existing social media moderation tools to enhance the accuracy and efficiency of fake account identification. Key Words: Support vector machines(SVM),K-Nearest Neighbors Algorithm(KNN),Random forest, Logistic Regression & Artificial Neural Network(ANN),Python.

Mamba-VNPS: A Visual Navigation and Positioning System with State-Selection Space

This study was designed to address the challenges of autonomous navigation facing UAVs in urban air mobility environments without GPS. Unlike traditional localization methods that rely heavily on GPS and pre-mapped routes, Mamba-VNPS leverages a self-supervised learning framework and advanced feature extraction techniques to achieve robust real-time localization without external signal dependence. The results show that Mamba-VNPS significantly outperforms traditional methods across multiple aspects, including localization error. These innovations provide a scalable and effective solution for UAV navigation, enhancing operational efficiency in complex spaces. This study highlights the urgent need for adaptive positioning systems in urban air mobility (UAM) and provides a methodology for future research on autonomous navigation technologies in both aerial and ground applications.

Employing convolutional neural networks and explainable artificial intelligence for the detection of seizures from electroencephalogram signal

A seizure is a rapid, uncontrolled burst of electrical activity in the brain. Epilepsy is a neurological disorder caused by repeated seizures. Effective management requires early detection. This research aims to create a convolutional neural network (CNN) and explainable artificial intelligence (XAI) integrated system for epileptic seizure detection, using techniques such as Shapley additive explanation (SHAP) for better interpretability. The relevant data is acquired from a variety of electroencephalogram (EEG) dataset to facilitate the development of the model. To identify the EEG data, the proposed system combines deep learning models with feature extraction technique. Model visualization and XAI approaches like feature importance analysis from SHAP values offer clear insights into the model's decision-making process. Evaluation criteria like specificity and accuracy are used to assess the models' performance. This framework's objective is to create simple seizure detection systems that assist early epilepsy patient identification and individualized treatment plans. This research aims in opening the door to improved patient care and treatment by developing the field of epilepsy seizure detection.

Fault Diagnosis in a Four-Arm Delta Robot Based on Wavelet Scattering Networks and Artificial Intelligence Techniques

This paper presents a comprehensive fault diagnosis approach for a delta robot utilizing advanced feature extraction and classification techniques. A four-arm delta robot prototype is designed in SolidWorks for realistic fault analysis. Two case studies investigate faults through control effort and vibration signals, with control effort detecting motor and encoder faults, while vibration signals identify bearing faults. This study compares time-domain signal features and wavelet scattering networks, applied by classification algorithms including wide neural networks (WNNs), efficient linear support vector machine (ELSVM), efficient logistic regression (ELR), and kernel naive Bayes (KNB). Results indicate that a WNN, using wavelet scattering features ranked by one-way anova, is optimal due to its consistency and reliability, while these features enhance computational efficiency by reducing classifier size. Sensitivity analysis demonstrates the classifier’s capacity to detect untrained faults, highlighting the importance of effective feature extraction and classification methods for fault diagnosis in complex robotic systems. This research significantly contributes to fault diagnosis in delta robots and lays the groundwork for future studies on fault tolerance control and predictive maintenance planning. Future work will focus on the physical implementation of the delta robot in laboratory settings, aiming to improve operational efficiency and reliability in industrial applications.

Analysis of K-NN with the Integration of Bag of Words, TF-IDF, and N-Grams for Hate Speech Classification on Twitter

Social media has emerged as one of the primary communication channels in the modern world, but it has simultaneously become a platform where hate speech can spread easily. This study attempts to evaluate the performance of a hate speech classification model using the K-Nearest Neighbors (K-NN) algorithm along with various feature extraction techniques, specifically Bag of Words (BoW), Term Frequency-Inverse Document Frequency (TF-IDF), and N-Grams. The dataset used in this study consists of 13169 entries, which represent a diverse range of hate speech examples commonly encountered on social media platforms. In this experimental investigation, we assess the efficacy of the model using each feature extraction technique. The findings reveal that the K-NN model exhibits optimal performance when the k parameter is set to 3 (k=3). Under this configuration, the model achieves an accuracy of 86.88%, with a precision of 88.27%, a recall of 86.88%, and an F1-Score of 86.50%. These results show that the integration of TF-IDF feature extraction technique with K-NN algorithm produces superior performance in hate speech classification.