Search

In agriculture, as precision farming increasingly employs robots to monitor crops, the use of weeding and harvesting robots is expanding the need for computer vision. Currently, most researchers and companies address these computer vision tasks with CNN-based deep learning. This technology requires large datasets of plant and weed images labeled by experts, as well as substantial computational resources. However, traditional feature-based approaches to computer vision can extract meaningful parameters and achieve comparably good classification results with only a tenth of the dataset size. This study presents these methods and seeks to determine the minimum number of training images required to achieve reliable classification. We tested the classification results with 5, 10, 20, 40, 80, and 160 images per weed type in a four-class classification system. We extracted shape features, distance transformation features, color histograms, and texture features. Each type of feature was tested individually and in various combinations to determine the best results. Using six types of classifiers, we achieved a 94.56% recall rate with 160 images per weed. Better results were obtained with more training images and a greater variety of features. [ABSTRACT FROM AUTHOR]

Sentiment analysis is a critical component in understanding customer opinions and reactions. This study explores the application of sentiment analysis using Python on the Amazon Fine Food Reviews dataset to classify customer reviews as positive or negative, enabling businesses to gain valuable insight into customer sentiments. This study used and compared the efficiency of Logistic Regression, Support Vector Machines, Random Forest, XGBoost, LSTM, and ALBERT. The comparison results showed that the LSTM and ALBERT classifiers stand out with remarkable accuracy (96%) and substantial support for positive and negative reviews. On the other hand, although the Random Forest classifier had similar accuracy (96%), it exhibited lower support for positive and negative sentiments. [ABSTRACT FROM AUTHOR]

Cloud computing environments, characterized by dynamic scaling, distributed architectures, and complex workloads, are increasingly targeted by malicious actors. These threats encompass unauthorized access, data breaches, denial-of-service attacks, and evolving malware variants. Traditional security solutions often struggle with the dynamic nature of cloud environments, highlighting the need for robust Adaptive Cloud Intrusion Detection Systems (CIDS). Existing adaptive CIDS solutions, while offering improved detection capabilities, often face limitations such as reliance on approximations for change point detection, hindering their precision in identifying anomalies. This can lead to missed attacks or an abundance of false alarms, impacting overall security effectiveness. To address these challenges, we propose ACIDS (Adaptive Cloud Intrusion Detection System)-PELT. This novel Adaptive CIDS framework leverages the Pruned Exact Linear Time (PELT) algorithm and a Support Vector Machine (SVM) for enhanced accuracy and efficiency. ACIDS-PELT comprises four key components: (1) Feature Selection: Utilizing a hybrid harmony search algorithm and the symmetrical uncertainty filter (HSO-SU) to identify the most relevant features that effectively differentiate between normal and anomalous network traffic in the cloud environment. (2) Surveillance: Employing the PELT algorithm to detect change points within the network traffic data, enabling the identification of anomalies and potential security threats with improved precision compared to existing approaches. (3) Training Set: Labeled network traffic data forms the training set used to train the SVM classifier to distinguish between normal and anomalous behaviour patterns. (4) Testing Set: The testing set evaluates ACIDS-PELT's performance by measuring its accuracy, precision, and recall in detecting security threats within the cloud environment. We evaluate the performance of ACIDS-PELT using the NSL-KDD benchmark dataset. The results demonstrate that ACIDS-PELT outperforms existing cloud intrusion detection techniques in terms of accuracy, precision, and recall. This superiority stems from ACIDS-PELT's ability to overcome limitations associated with approximation and imprecision in change point detection while offering a more accurate and precise approach to detecting security threats in dynamic cloud environments. [ABSTRACT FROM AUTHOR]

This research presents a comparative analysis of sentiment analysis techniques applied to user comments on YouTube, with a specific focus on TED talks. The proliferation of social media platforms has provided individuals with unprecedented opportunities to express their opinions and emotions. YouTube, as a leading video-sharing platform, has become a significant hub for user-generated content and discussions on a wide range of topics. In light of the exponential growth of unstructured and semi-structured data, sentiment analysis plays a critical role in extracting valuable emotional insights from online interactions. To evaluate sentiments expressed in YouTube comments, a self-created and meticulously labeled dataset comprising user comments was employed. The study compared the performance of five ML techniques: NB, SVM, RF, KNN, and DT. The performance of the classifiers was evaluated using key evaluation metrics such as Precision, Recall, and F1- score. The findings of this research offer valuable insights into the efficacy of various machine learning techniques for sentiment analysis in the context of YouTube comments on TED talks. Among the classifiers, SVM demonstrated the highest Precision, Recall, and F1-score, indicating its effectiveness in accurately identifying sentiment in YouTube comments. Random Forest and Decision tree also displayed competitive performance, while KNN and Naïve Bayes exhibited slightly lower accuracy. These results provide researchers and practitioners with valuable information to make informed decisions regarding the selection of appropriate ML techniques for SA tasks on social media platforms. [ABSTRACT FROM AUTHOR]

Solid propellant, as the energy source for solid rocket engines, it is of great significance to achieve accurate quantitative detection of aging damage of solid propellant. In this paper, a novel approach based on frequency impedance spectroscopy impedance combined with CARS‐SVM algorithm was proposed. First, the temperature, humidity, and pressure of environmental information around the solid rocket motor were sampled, and then the impedance at corresponding frequencies of the propellant was obtained by AD5933 chip. Second, the processed experimental data were subjected to abnormal sample detection before further variables selection using uninformative variables elimination (UVE) competitive adaptive reweighted sampling (CARS), respectively. Finally, support vector machine (SVM), UVE‐SVM and CARS‐SVM quantitative calibration methods were established. The results showed that the determination coefficient (R2), root mean square error (RMSE), and mean absolute percentage error (MAPE) of CARS‐SVM model were 0.9919, 0.7540, and 0.0480, respectively. Therefore, the results prove that impedance of solid propellant combined with CARS‐SVM model can effectively achieve high precision quantitative detection of aging damage of solid propellant, which lays a new method for the application of solid propellants aging damage in the online quantitative detection. [ABSTRACT FROM AUTHOR]

This research introduces an innovative intelligent model developed for predicting and analyzing sentiment responses regarding audio feedback from students with visual impairments in a virtual learning environment. Sentiment is divided into five types: high positive, positive, neutral, negative, and high negative. The model sources data from post-COVID-19 outbreak educational platforms (Microsoft Teams) and offers automated evaluation and visualization of audio feedback, which enhances students’ performances. It also offers better insight into the sentiment scenarios of elearning visually impaired students to educators. The sentiment responses from the assessment to point out deficiencies in computer literacy and forecast performance were pretty successful with the support vector machine (SVM) and artificial neural network (ANN) algorithms. The model performed well in predicting student performance using ANN algorithms on structured and unstructured data, especially by the 9th week against unstructured data only. In general, the research findings provide an inclusive policy implication that ought to be followed to provide education to students with a visual impairment and the role of technology in enhancing the learning experience for these students. [ABSTRACT FROM AUTHOR]

Previous neuroimaging studies based on electroencephalography (EEG) microstate analysis have identified abnormal neural electric activity in patients with psychiatric diseases. However, the microstate information in individuals with different degrees of fear of heights (FoH) remains unknown so far. The aim of the study was therefore to explore the changes of EEG microstate characteristics in different FoH individuals when exposed to high-altitude stimulated by virtual reality (VR). First, acrophobia questionnaire (AQ) before the experiment and 32-channel EEG signals under the virtual high-altitude exposure were collected from 69 subjects. Second, each subject was divided into one of three levels of FoH including no-FoH, mild or moderate FoH (m-FoH) and severe FoH (s-FoH) groups according to their AQ scores. Third, using microstate analysis, we transformed EEG data into sequences of characteristic topographic maps and computed EEG microstate features including microstate basic parameters, microstate sequences complexity and microstate energy. Finally, the extracted features as inputs were sent to train and test an support vector machine (SVM) for classifying different FoH groups. The results demonstrated that five types of microstates (labeled as A, B, C, D and F) were identified across all subjects, of which microstates A-D resembled the four typical microstate classes and microstate F was a non-canonical microstate. Significantly decreased occurrence, coverage and duration of microstate F and transition probabilities from other microstates to microstate F in m-FoH and s-FoH groups were observed compared to no-FoH group. It was also demonstrated that both m-FoH and s-FoH groups showed a notable reduction in sample entropy and Lempel-Ziv complexity. Moreover, energies of microstate D for m-FoH group and microstate B for s-FoH group in right parietal, parietooccipital and occipital regions exhibited prominent decreases as comparison to people without FoH. But, no significant differences were found between m-FoH and s-FoH groups. Additionally, the results indicated that AQ-anxiety scores were negatively correlated with microstate basic metrics as well as microstate energy. For classification, the performance of SVM reached a relatively high accuracy of 89 % for distinguishing no-FoH from m-FoH. In summary, the findings highlight the alterations of EEG microstates in people with fear of heights induced by virtual high-altitude, reflecting potentially underlying abnormalities in the allocation of neural assemblies. Therefore, the combination of EEG microstate analysis and VR may be a potential valuable approach for the diagnosis of fear of heights.

ObjectivesThe purpose of the present study was to explore the influencing factors of hyperuricemia (HUA) in different populations in Shandong Province based on clinical biochemical indicators. A prediction model for HUA was constructed to aid in the early prevention and screening of HUA.MethodsIn total, 705 cases were collected from five hospitals, and the risk factors were analyzed by Pearson correlation analysis, binary logistic regression, and receiver operating characteristic (ROC) curve in the gender and age groups. All data were divided into a training set and test set (7:3). The training set included age, gender, total protein (TP), low-density lipoprotein cholesterol (LDL-C), and 15 other indicators. The random forest (RF) and support vector machine (SVM) methods were used to build the HUA model, and model performances were evaluated through 10-fold cross-validation to select the optimal method. Finally, features were extracted, and the ROC curve of the test set was generated.ResultsTP, LDL-C, and glucose (GLU) were risk factors for HUA, and the area under the curve (AUC) value of the SVM validation set was 0.875.ConclusionThe SVM model based on clinical biochemical indicators has good predictive ability for HUA, thus providing a reference for the diagnosis of HUA and the development of an HUA prediction model.

Hearing-impaired people use sign language as a means of communication with those with no hearing disability. It is therefore difficult to communicate with hearing impaired people without the expertise of a signer or knowledge of sign language. As a result, technologies that understands sign language are required to bridge the communication gap between those that have hearing impairments and those that dont. Ethiopian Amharic alphabets sign language (EAMASL) is different from other countries sign languages because Amharic Language is spoken in Ethiopia and has a number of complex alphabets. Presently in Ethiopia, just a few studies on AMASL have been conducted. Previous works, on the other hand, only worked on basic and a few derived Amharic alphabet signs. To solve this challenge, in this paper, we propose Machine Learning techniques such as Support Vector Machine (SVM) with Convolutional Neural Network (CNN), Histogram of Oriented Gradients (HOG), and their hybrid features to recognize the remaining derived Amharic alphabet signs. Because CNN is good for rotation and translation of signs, and HOG works well for low quality data under strong illumination variation and a small quantity of training data, the two have been combined for feature extraction. CNN (Softmax) was utilized as a classifier for normalized hybrid features in addition to SVM. SVM model using CNN, HOG, normalized, and non-normalized hybrid feature vectors achieved an accuracy of 89.02%, 95.42%, 97.40%, and 93.61% using 10-fold cross validation, respectively. With the normalized hybrid features, the other classifier, CNN (sofmax), produced a 93.55% accuracy.

Stres merupakan keadaan dimana seseorang merasakan adanya tekanan yang berlebih pada dirinya. Pemantauan tingkat stres menjadi hal yang penting bagi manusia. Tingkat stres yang tinggi dapat menimbulkan dampak negatif terhadap kesehatan manusia. Deteksi dini stres menjadi sesuatu yang sangat penting untuk dilakukan. Salah satu cara mengetahui tingkat stres seseorang adalah melalui analisis teks. Penelitian ini dilakukan untuk melakukan klasifikasi tingkat stres berdasarkan data berupa teks menggunakan algoritma Support Vector Machine (SVM) dan Random Forest. Pada penelitian ini melakukan perbandingan beberapa metode transformasi. Transformasi yang dilakukan pada penelitian ini menggunakan TF-IDF, CountVectorizer, NRCLex, dan Word Affect Intensities. Data yang digunakan dalam penelitian ini berupa sebuat teks berbahasa Inggris yang diambil dari media sosial Twitter. Total data yang digunakan yaitu 8439 data. Pelatihan model baik untuk Support Vector Machine dan Random Forest menggunakan 6751 data. Sedangkan untuk pengujian menggunakan 1688 data. Hasil penelitian menunjukkan bahwa algoritma SVM dengan pembobotan menggunakan TF-IDF memiliki performa yang paling baik dibandingkan dengan algoritma Random Forest dan metode transformasi lainnya yang digunakan dalam penelitian. Model algoritma SVM dengan transformasi TF-IDF yang dibangun berhasil mendapatkan akurasi sebesar 84%. Model ini mendapatkan akurasi yang lebih tinggi dibanding model Random Forest yang memperoleh akurasi tinggi sebesar 80% dengan menggunakan transformasi CountVectorizer. Abstract Stress is a condition where a person feels excessive pressure on himself. Monitoring stress levels is important for humans. High levels of stress can have a negative impact on human health. Early detection of stress is something that is very important to do. One way to find out someone's stress level is through text analysis.This research was conducted to classify stress levels based on text data using the Support Vector Machine (SVM) and Random Forest algorithms. This research compares several transformation methods. The transformation performed in this study uses TF-IDF, CountVectorizer, NRCLex, and Word Affect Intensities. The data used in this research is an English text taken from Twitter social media. The total data used is 8439 data. Model training for both Support Vector Machine and Random Forest uses 6751 data. While for testing using 1688 data. The results showed that the SVM algorithm with weighting using TF-IDF had the best performance compared to the Random Forest algorithm and other transformation methods used in the study. The SVM algorithm model with TF-IDF transformation that was built managed to get an accuracy of 84%. This model obtained a higher accuracy than the Random Forest model which obtained a high accuracy of 80% using the CountVectorizer transformation.

Coral reefs, despite covering less than 0.2 % of the ocean floor, harbor approximately 35 % of all known marine species, making their conservation critical. However, coral bleaching, exacerbated by climate change and phenomena such as El Niño, poses a significant threat to these ecosystems. This study focuses on the Red Sea, proposing a generalized machine learning approach to detect and monitor changes in coral reef cover over an 18-year period (2000–2018). Using Landsat 7 and 8 data, a Support Vector Machine (SVM) classifier was trained on depth-invariant indices (DII) derived from the Gulf of Aqaba and validated against ground truth data from Umluj. The classifier was then applied to Al Wajh, demonstrating its robustness across different sites and times. Results indicated a significant decline in coral cover: 11.4 % in the Gulf of Aqaba, 3.4 % in Umluj, and 13.6 % in Al Wajh. This study highlights the importance of continuous monitoring using generalized classifiers to mitigate the impacts of environmental changes on coral reefs.

Exercise is essential for a healthy lifestyle, thus it is important to consider how to keep proper posture when performing arm exercises at home. This work uses wrist-worn bands with the MPU6050 sensor to address these issues, which collects motion data using acceleration measurements. The individuals in the dataset are completing a variety of activities at varying ranges of motion. Machine learning-based classification methods are then applied after the pre-processing and feature extraction of the gathered data. An App prototype integrated with a WiFi module and Cloud infrastructure is created to enable real-time data collecting and storage. The Arduino IDE is used to send the collected data to the ThingSpeak platform, where it is subsequently sent to MATLAB for additional analysis. The studied data is then returned to ThingSpeak, where the program displays the findings. This approach reduces the risk of injuries caused by bad posture by enabling people to continue regular workouts at home without requiring a personal trainer or a particular environment. The findings of this work shed important light on the performance of Boosted Trees, Quadratic SVM, Subspace KNN, and Fine KNN algorithms for arm exercises employing a wrist-worn band with an MPU6050 sensor. The Fine KNN has the highest accuracy of 91.3% among all implemented algorithms.

Multiple Input and Output-Orthogonal Time–Frequency Selective (MIMO-OTFS) is considered one of the leading candidates for the beyond fifth generation (B5G) radio framework. The signal detection process is complex due to the large number of antennas, which also increases the framework’s latency. Signal detection algorithms such as Recurrent Neural Networks (RNNs), Neural Networks (NNs), Support Vector Machines (SVMs), Minimum Mean Square Error (MMSE), Maximum Likelihood Detection (MLD), Expectation-Maximization (EM), and Zero-Forcing Equalization (ZFE) are analyzed for Rayleigh and Rician channels. Currently available methods involve intricate identification and receivers with lower spectral efficiency. Experimental results indicate that RNNs, NNs, and SVM detectors, which have lower complexity, are recommended to improve the bit error rate (BER) and power spectral density (PSD) of the MIMO-OTFS system. It is also noted that RNNs offer diversity in received data, achieving a significant gain of 5 dB to 7 dB compared to existing OTFS systems across different MIMO frameworks. Furthermore, the utilization of machine learning algorithms significantly obtained a gain of −305 and −330 (RNNs) for the Rayleigh and Rician channels, respectively. These findings underscore the benefits of integrating sophisticated detection methods in B5G communication channels, indicating a valuable direction for future research and advancements in this area.

A photovoltaic panel without cooling capabilities shows low efficiency and this efficiency deteriorates further during peak periods. In contrast, a photovoltaic/thermal (PV/T) system achieves higher electrical efficiency through a cooling mechanism that limits efficiency drops even during peak periods. This research centers on exploring the influence of temperature on the current, voltage, power, and efficiency of PV/T modules and the associated thermal system. Additionally, experimental investigations are conducted to examine the enhancement of PV/T power. Furthermore, the study involves the implementation of an Artificial Neural Network (ANN) for modeling the system, aiming to predict and analyze the performance of PV/T systems under the weather conditions prevalent in Oman and the Gulf region.Neural network models, such as multilayer perceptron (MLP), Second Order Volterra Model (SOVM), and Support Vector Machine (SVM) have been constructed to simulate and forecast the output power of the PV/T system. The SVM model demonstrated the lowest mean absolute percentage error (MAPE), registering at 0.0358, surpassing both Self-Organizing Feature Map (SOFM) and MLP with MAPE values of 0.0524 and 0.0601, respectively. In terms of root mean squared error (RMSE), SOFM exceeded MLP and SVM, yielding a lower error of 0.2235. However, when considering R2, MLP exhibited a notably high value of 0.9554, signifying a robust relationship. It is worth noting that R2 values approaching one indicate a strong correlation. The MLP model highlighted superior performance, achieving outstanding metrics with values of 99.67 %, 0.0631, and 0.2512 for R2, MAPE, and RMSE, respectively. This collective evidence solidifies the effectiveness and reliability of the proposed neural network models in predicting the PV/T system's output power.