Support Vector Machine Method Research Articles

Comparison of Different Negative-Sample Acquisition Strategies Considering Sample Representation Forms for Debris Flow Susceptibility Mapping

The lack of reliable negative samples is an important factor limiting the quality of machine learning-based debris flow susceptibility mapping (DFSM). The purpose of this paper is to propose multiple negative-sample acquisition strategies for DFSM considering different sample representation forms. The sample representation forms mainly include a single grid, multi-grid, and watershed unit, and the negative-sample acquisition strategies are based on support vector machine (SVM), spy technique, and isolation forest (IF) methods, respectively. These three strategies can assign a value to all the samples based on different assumptions, and reliable, negative samples can be generated from samples with values below a predefined threshold. Combining different sample representation forms with negative sample acquisition strategies, nine datasets were then involved in random forest (RF) modeling. The receiver operating characteristic (ROC) curves and related statistical results were used to evaluate the models. The results show that the strategy based on the spy technique is suitable for multiple datasets, while the IF-based strategy is well-adapted to the watershed unit datasets. This study can provide more options for improving the quality of datasets in DFSM, which can further improve the performance of machine learning models.

Prediction of Customer Switching Using Support Vector Machine Method

Several studies on predicting customer switching focus on the telecommunications industry and online stores. This research aims to predict customer switching to get the best results; customers are the most critical mass; some companies must provide satisfying services so customer flow decreases. The support vector machine (SVM) method uses machine learning to find a hyperplane based on the SRM principle. A hyperplane is a decision boundary that helps classify data points. SVM stands out for its ability to take input data and make predictions based on its characteristics. This study uses data from Kaggle, structured it, cleaned it, identified patterns and inconsistencies (such as skewness, outliers, and missing values), and built and validated hypotheses. From the data processing, the plot shows the imbalance of data classes between churners and non-churners. This research applies several models where the most significant or best performance value is in the SVM model of 0.7996. The Neural Network model can be trained with better patterns to detect data and achieve high accuracy.

Performance Analysis of Neighborhood Component Analysis on Support Vector Machine in Greenhouse Gas Emission Classification

The heatwave phenomenon has hit several countries in various parts of the world, caused by climate change. Climate change leads to greenhouse gas emissions increasing beyond the limits set by the Intergovernmental Panel on Climate Change (IPCC) Fourth Assessment Report Global Warming Potentials. This final project uses a combination of Neighborhood Component Analysis (NCA) and Support Vector Machine (SVM) methods with linear, polynomial, Radial Basis Function (RBF), and sigmoid kernel functions. The purposes of this final project are to evaluate the performance of NCA on SVM and to determine the best kernel function in this combination. Based on the analysis, it was found that classification using a combination of NCA and SVM methods can reduce variables, with the best kernel function being the Polynomial kernel function. This is because the analysis using the Polynomial kernel function achieved the highest accuracy values for training data, testing accuracy, and F1-Score, which are 98,96%, 99,15%, and 98,98% respectively. Additionally, the training analysis time and testing analysis time were the shortest at 0,15 seconds and 0,04 seconds.

Hilbert-Huang Transform and machine learning based electromechanical analysis of induction machine under power quality disturbances

Monitoring and predicting Power Quality (PQ) is crucial for quickly minimizing risk, protecting induction machines (IMs), and increasing productivity. This paper proposes an electromechanical analytical framework for IM working with various PQ disturbances based on machine learning (ML). This work collects the motor characteristics and PQ data, motor vibration signals, and other sensor data to identify the failures. The complicated electromechanical data is broken down into intrinsic mode functions (IMFs) using the Hilbert-Huang Transform (HHT), which reveals the modes present in the signals. This frequency-domain data is retrieved, paired with time-domain characteristics, and used as input features for ML algorithms. Support Vector Machines (SVMs) trained on labeled datasets in which features extracted from PQ disturbances are used to categorize the disturbances into several groups. The PQ disturbances caused by voltage sags, swells, harmonics, and transients can all be efficiently classified using SVM classifiers, allowing for real-time determination of the kind of disturbance influencing the IM. The accuracy for the SVM in the proposed scheme is 97.2 %. The SVM method is trained with PQ data and classification results using MATLAB classifier and Python software.

Numerical thermal analysis and structural optimization of cascaded PCM heat sinks for thermal management of electronics

In the present study, an innovative cascaded phase change material (PCM) heat sink using organic PCMs was proposed for electronic thermal management. The design arranges multiple PCMs with progressively lower melting points along the heat transfer path, allowing active control of the melt front. A numerical model with 5 % accuracy was developed to evaluate thermal performance. Based on the validated model, the effects of PCM stage numbers, PCM volume fractions, fin types, and fin volume fractions were analyzed. The fin volume fractions were further optimized using a support vector machine (SVM) and genetic algorithm (GA) methods. Results showed that a three-stage PCM heat sink with pin fins and increasing configurations outperformed single-stage PCM heat sinks, enhancing heat transfer by up to 36.7 %. Additionally, the innovative cascaded PCM heat sink with a uniform 20 % fin volume fraction across the three stages achieved a 38.8 % improvement compared to single-stage configurations. When the fin volume fractions were set to 24.1 %, 17.6 %, and 13.0 % for the three stages, thermal performance of the innovative cascaded PCM heat sink was enhanced by up to 39.8 % compared to single-stage configurations. The findings provide valuable insights for the design and application of PCM-based thermal management systems, potentially leading to more efficient solutions for various engineering fields.

Analisis Sentimen Ulasan Wisatawan Terhadap Alun-Alun Kota Batam: Perbandingan Kinerja Metode Naive Bayes dan Support Vector Machine

Batam City, as a rapidly developing tourism destination in Indonesia, continues to strive to enhance the potential of its tourist attractions to attract more visitors. The assessment of reviews from tourists is crucial in identifying necessary development measures to improve the quality of tourist attractions. This research aims to analyze the sentiment of reviews for the Alun-Alun Kota Batam tourist destination by leveraging data from Google Maps. Two classification methods, Naive Bayes and Support Vector Machine, are employed for sentiment analysis, and their performances are compared. From 1140 collected reviews, the data is categorized into three labels: positive, negative, and neutral. The research results indicate that the Support Vector Machine method achieves higher accuracy (94%) compared to Naive Bayes (83%). This study contributes insights into visitor sentiments towards Alun-Alun Kota Batam, with implications for policy development and more effective actions in enhancing local tourism appeal.

A Review on ML and DL Techniques in Detecting Plant Diseases

In Our country, Agriculture is the main occupation of the people. Plant diseases are a major problem today as they affect the quality and production of agricultural production. Most plant-borne diseases are caused by viruses and fungi. Identifying the diseases in early stage by manual power is not possible in large sector of Area. Subsequent maintenance of plants conditions are major challenging task. To overcome this problem in Agriculture, we use Image Processing Technique to identify the disease in beginning stage . To Apply Image Processing Technique, we have to undergo certain Image concepts that deals with image acquisition, image pre-processing, image segmentation, feature extraction and classification of disease. This Methodology reduces the destruction of plants and make the crop production high. The digital Image Processing Technique is the main solution to solve the problems in Plant diseases by identifying the plant diseases in early stage. It is very much useful to the Farmers who are facing plant diseases problem in their Agriculture Area. Many techniques are used to identify the diseases in plants using classifiers such as , K-Nearest Neighbors ,Support Vector Machine methods etc. This paper gives the overview of available methods for plant disease detection

Implementation of Data Mining for Vertigo Disease Classification Using the Support Vector Machine (SVM) Method

This research aims to implement advanced data mining techniques for the classification of vertigo disorders using the Support Vector Machine (SVM) method. Vertigo, characterized by a spinning sensation, can be triggered by various factors such as nervous system disorders and inner ear infections. With the rising prevalence of vertigo patients, there is a pressing need for more effective and efficient diagnostic tools. This study was conducted at Puskesmas Jangka in Bireuen Regency, involving the collection of vertigo patient data from the years 2023-2024. The collected data underwent a comprehensive preprocessing pipeline, including data cleaning, partitioning into training and testing datasets, and subsequent implementation of the SVM algorithm. The performance of the model was evaluated using the Mean Absolute Percentage Error (MAPE), resulting in a MAPE value of 28.47%.

Assessment of land dynamics transformation into tea plantations using support vector machine

Precisely monitoring land use dynamics and spatial distributions is essential for sustainable development and long-term land management. Tea is one of the leading beverage crops cultivated in Bangladesh, expanding rapidly in northern districts and forcing land use change. This study aims to decipher the expansion of tea cultivation and land dynamics transformation to tea land areas in the northern region of Bangladesh by using Landsat 5 TM and Landsat 8 OLI/TIRS surface reflectance images. The supervised support vector machine (SVM) method was used for classification purposes, resulting in three classified maps for the years 2004, 2013 and 2022, having overall accuracies of 91.43 %, 98.67 %, and 98.48 % and kappa coefficients of 89.51, 98.37, and 98.13, respectively. The images were classified into six land use classes: agriculture, tea cultivation area, settlement, waterbody, bare land, and forest. Land transformation results reveal that overall tea land increased by 41.08 % from 2004 to 2022, experiencing a downward trend during 2004–2013, while from 2013 to 2022, a dramatic rise of 70.01 % (equal to 4683.60 ha) tea cultivation area was found in Panchagarh district. The bare land was among the most highly transformed land classes into tea plantations, followed by the forest area. This study provides evidence of transforming underutilized land class into a profitable land use practice, i.e., tea plantation. Thus, cutting-edge technologies would be imperative in land transformation detection for sustainable land management and policy implications for the sustainable development of small landholding livelihoods and the tea industry.

Comparison of Classification Results of SVM, KNN, Decision Tree, and Ensemble Methods in Diabetes Diagnosis

This study aims to determine which algorithms and test techniques are the most optimal in detecting diabetes mellitus and obtaining the best results based on the value of accuracy, precision, and recall. In this study, approaches were used in early diagnosis of diabetes using KNN, SVM, Decision Tree, and Ensemble Majority Voting methods in Percentage Split and K-Fold Cross Validation methods. Diabetes is a disease characterized by high blood sugar (glucose) levels and can cause a variety of disease complications and damage to the body's organs if not treated immediately. Early diagnosis of diabetes is becoming crucial so that people can take immediate action to the hospital for immediate treatment. The data used is Healthcare-Diabetes from Kaggle. The results of this study have found that the K-Fold Cross Validation method is better because it can provide an average improvement in Ensemble accuracy of 13.42% compared to the Percentage Split method which only gives an average increase in Ensamble accuracy of 9.15%. The best algorithm for classifying diabetes disease is the Ensemble Majority Voting algorithm using the K-Fold Cross Validation method with a 98.81% accuracy rate. These excellent research results may contribute to detecting early symptoms of diabetes before it become too severe.

Comparison of machine learning algorithms for diagnosing diabetes

This study compares various machine learning algorithms for detecting diabetes in patients. The data used included biometric parameters such as blood glucose, blood pressure, body mass index (BMI), and other indicators. Several models were tested, including logistic regression, support vector machine (SVM), k-nearest neighbors (KNN), and ensemble methods such as XGBoost, LightGBM, and CatBoost. Class balancing methods using SMOTE and hyperparameter optimization using GridSearchCV were used to improve the accuracy of the models. The XGBoost model demonstrated the highest prediction accuracy of 90.2%. These results confirm the high efficiency of machine learning algorithms in diagnosing diabetes, which can be used in medical practice to create more accurate prediction tools.

Detection of Thymoma Disease Using mRMR Feature Selection and Transformer Models

Background: Thymoma is a tumor that originates in the thymus gland, a part of the human body located behind the breastbone. It is a malignant disease that is rare in children but more common in adults and usually does not spread outside the thymus. The exact cause of thymic disease is not known, but it is thought to be more common in people infected with the EBV virus at an early age. Various surgical methods are used in clinical settings to treat thymoma. Expert opinion is very important in the diagnosis of the disease. Recently, next-generation technologies have become increasingly important in disease detection. Today’s early detection systems already use transformer models that are open to technological advances. Methods: What makes this study different is the use of transformer models instead of traditional deep learning models. The data used in this study were obtained from patients undergoing treatment at Fırat University, Department of Thoracic Surgery. The dataset consisted of two types of classes: thymoma disease images and non-thymoma disease images. The proposed approach consists of preprocessing, model training, feature extraction, feature set fusion between models, efficient feature selection, and classification. In the preprocessing step, unnecessary regions of the images were cropped, and the region of interest (ROI) technique was applied. Four types of transformer models (Deit3, Maxvit, Swin, and ViT) were used for model training. As a result of the training of the models, the feature sets obtained from the best three models were merged between the models (Deit3 and Swin, Deit3 and ViT, Deit3 and ViT, Swin and ViT, and Deit3 and Swin and ViT). The combined feature set of the model (Deit3 and ViT) that gave the best performance with fewer features was analyzed using the mRMR feature selection method. The SVM method was used in the classification process. Results: With the mRMR feature selection method, 100% overall accuracy was achieved with feature sets containing fewer features. The cross-validation technique was used to verify the overall accuracy of the proposed approach and 99.22% overall accuracy was achieved in the analysis with this technique. Conclusions: These findings emphasize the added value of the proposed approach in the detection of thymoma.

Estimation and modeling of pedestrian dilemma zone at uncontrolled intersections under mixed traffic conditions

Objectives Pedestrian gap acceptance (PGA) theory is the basic concept for pedestrian dilemma zone (PDZ) analysis and modeling and their gap acceptance behavior depends on dilemma behavior also. Uncontrolled intersections are one of the major locations where pedestrians have more dilemma and there is a possibility of an interaction between the pedestrian and vehicle due to the incorrect decision taken by the pedestrian when the vehicle lies within the limits of PDZ. Elimination and modeling of spatial boundaries of pedestrian dilemma stage improve the PGA. Methods The present study is intended to quantify and model the pedestrian dilemma zone (PDZ) boundaries at uncontrolled X-intersections under mixed traffic conditions. Video data were collected from four four-legged uncontrolled intersections in India. Pedestrian and vehicle information was extracted using DataFromSky software and manually from video. The Cumulative Gap Distribution (GCD) and Support Vector Machine (SVM) methods were used to estimate the boundaries of PDZ and developed a binary logistic regression model to estimate the PDZ boundary limits. Results The lower boundary limits of PDZ using GCD and SVM methods are 9.0 m and 6.0, respectively, and the upper boundary limits of PDZ using GCD and SVM methods are 16.5 and 18.5 m, respectively. The GCD method overestimated the lower limit and underestimated the upper limit compared with the SVM method. The binary logistic regression model results confirmed that pedestrian age, gender and crossing speed have a negative correlation and location of pedestrian crossing, vehicle type, and approaching speed have a positive correlation with the boundary limits of PDZ. Conclusions From the present study, it is concluded that SVM better estimated the PDZ boundary limits with the largest margin compared to the GCD method. It is concluded that the boundary limits shift away from the intersection in the case of female and old-age pedestrians compared to male and young-age pedestrians, respectively. The size benefit in the case of 2Ws is the reason for shifting the PDZ boundary limits close to the crosswalk. The lower approaching speeds of the vehicles at uncontrolled intersections are the reason for pedestrians accept the gap at shorter distances.

Rockfall mapping and susceptibility evaluation based on UAV high-resolution imagery and support vector machine method

Abstract This study aims to develop a comprehensive method for automated rockfall mapping and susceptibility assessment using unmanned aerial vehicle (UAV) tilt photography and the support vector machine (SVM) algorithm. By employing Jinzhong Town in the mountainous forest landscapes of Southwest China as a case study, we leverage photogrammetry principles and computer vision algorithms to generate high-precision, high-resolution digital surface models (DSMs), and digital orthophoto maps through a UAV remote sensing system. The rockfall inventory is accurately and automatically mapped using the object-based classification method and SVM algorithm. The automated rockfall identification method achieves a 93% accuracy with a Kappa coefficient of 0.7967. Statistical analyses of spatial distribution characteristics reveal a significant correlation between rockfall volume and area with a correlation coefficient (R²) of 0.92316 under logarithmic coordinates. In addition, a power function relationship describes the link between rockfall volume and slope height (R² = 0.87), while the relationship with sliding distance is characterized by a weaker linear correlation (R² = 0.65). Rockfall runout distance also shows a significant linear correlation with slope height (R² = 0.79) but exhibits a less-pronounced association with rockfall volume (R² = 0.58). The SVM model employed to assess rockfall susceptibility indicates high accuracy (area under the curve = 0.896), affirming its efficacy in rockfall susceptibility assessment. Our findings underscore the utility of UAV remote sensing for rockfall information extraction and susceptibility evaluation, particularly in challenging mountainous forest environments characterized by intricate topography and geological complexities.

Analysis of Public Sentiment Towards Celebrity Endorsment On Social Media Using Support Vector Machine

Analysis of public sentiment towards celebrity endorsements on social media is very important to understand the public's response to promotional campaigns involving celebrities. In this study, we combine the VADER labeling method with the Support Vector Machine (SVM) method to analyze public sentiment toward celebrity endorsements on social media. Data is taken from various social media sources such as Twitter, Instagram, and Facebook. The data is pre-processed to ensure data accuracy and relevance and then labeled with the VADER method to determine the positive, negative, or neutral sentiment of the text. The labeled data is then extracted for features and used to train the SVM model. The trained SVM model is then validated using test data to measure its accuracy and performance. The results of the analysis provide useful insight into public sentiment towards celebrity endorsements on social media and can provide recommendations for stakeholders regarding this matter. Overall, combining the VADER labeling method with SVM in analyzing public sentiment towards celebrity endorsements on social media shows more accurate results and can provide practical benefits in marketing and promotional strategies. The results shown using the Support Vector Machine method with a ratio of 80:20 can provide average precision results of 77%, recall of 100%, f1-score of 87%, and accuracy of 76.92%. Twitter application user sentiment shows that 77% (338 data) of Twitter user reviews provide positive sentiment and 23% (119 data) provide negative sentiment reviews from a total of 517 data. Suggestions from researchers are that in future research they can add more data to make modeling easier to provide higher accuracy values. Using other classification and performance evaluation methods, such as Naive Bayes, Decision Tree, Fuzzy, or Deep Learning. Use other data processing tools, such as RapidMiner, Jupyter Notebook, RStudio, or others.

Spectrometric and remote sensing investigations of some granitic rocks in the Egyptian north Eastern Desert: Insights on environmental radiogenic heat production

Granitic rocks dominate the Neoproterozoic outcrops in the northern Egyptian Eastern Desert, prominently featuring two main categories: Arc Granitoids (AG) and late-to post-collision granites (LPCG). The AG range from granodiorite and tonalite to quartz-diorite. In contrast, LPCG comprise syenogranite, monzogranite, and alkali-feldspar granite. This study leverages Landsat-8 remote sensing data to effectively discriminate between these rock types using several advanced image processing techniques. False color composite and decorrelation stretch methods highlighted geological and structural features, revealing distinct spectral signatures for each rock type. Principal Component Analysis and band rationing further refined distinguishing various varieties within the LPCG and detailed mapping. Supervised classification using the Support Vector Machine method yielded precise delineation of rock units. The investigated granitic rocks exhibited estimated radiogenic heat production values ranging from 6.02 to 1.41 μW/m3, surpassing the average values observed in the Earth's crust. The reason behind these noteworthy surpassing values of radiogenic heat production can be directly attributed to the relatively high Gamma-ray measurements in the LPCG outcrops. Gamma-ray spectrometric analysis indicated varying distributions of radioelements, particularly between AG and LPCG. The equivalent uranium (eU) concentrations range from 2.8 to 7 ppm in AG, while LPCG exhibited broader variability from 5.1 to 34 ppm. The equivalent thorium (eTh) values range from 16.1 to 34.1 ppm, with an overall average of 23 ppm. Conversely, within the ferruginated-silicified domains, the LPCG display slightly elevated levels of eU, reaching 31.4, 35.5, and 27.8 ppm for the monzogranites, syenogranites, and alkali-feldspar granites, respectively. These elevated levels suggest the potential for iron oxy-hydroxide minerals to adsorb uranium within alteration zones. Additionally, radioactive minerals such as zircon, columbite, uranothorite, allanite, euxenite, and samarskite contribute to the observed spot anomalies.

Pemodelan Text Mining untuk Analisis Sentimen Terhadap Program Makan Siang Gratis di Media Sosial X Menggunakan Algoritma Support Vector Machine (SVM)

The free lunch program is an initiative that provides free lunches and milk in schools and Islamic boarding schools, as well as providing nutritional support for children under five and pregnant women. Even though efforts to overcome stunting with this program are considered too late, providing free milk to pregnant women is considered more relevant for dealing with the stunting problem. Based on experience from various countries, the free lunch program shows various positive impacts, including increasing children's nutritional intake, increasing school participation, developing student character, and supporting the family and national economy. Discussions about this program received various responses from netizens, including platform X users. Online platforms have an important role in reaching and disseminating information about various social programs. In this context, the support vector machine method can be used to classify netizen responses to the free lunch program into positive and negative categories. It is hoped that this analysis will provide a deeper understanding of how this program is received by the general public, as well as provide suggestions for policy makers in designing social programs in the future.

Advanced prediction of rice yield gaps under climate uncertainty using machine learning techniques in Eastern India

The current study focuses on applying machine learning approaches to forecast future Kharif rice yield gaps in eastern India while accounting for climate change implications. To achieve the United Nations Sustainable Development Goals (SDGs), food security must be prioritized. Rice yield has been predicted using Cubist, GBM, MARS, RF, SVM, and XGB machine learning methods, considering six factors: elevation, soil moisture, precipitation, temperature, soil temperature, and actual evapotranspiration. Climatic change scenarios were generated using the latest climatic Coupled Model Intercomparison Project Phase 6 (CMIP6 MIROC6) Shared Socioeconomic Pathways (SSP) 2–4.5 and SSP5-8.5 datasets between 1990 and 2030. Finally, machine learning algorithms were used to identify rice yield gaps to achieve sustainable agricultural intensification. The rice yield validation was completed with 1889 field-based farmer observation records. The results suggest that Murshidabad and Purba Bardhaman districts had very high rice yields (5.60–3.45 t/ha) when using the Cubist model compared to another model. The findings also reveal a poor rice-yielding zone (1.44–0.39 t/ha) in the western region (Purulia) and a northwestern region (half of the west of Birbhum). The Cubist and RF models' maximum and minimum R2 values were 0.73 and 0.72, respectively. The R2 values were also negligible for the XGB, GBM, SVM, and MARS, machine learning models. Projections for rice production in 2030 indicate that the northern and north-eastern regions (Murshidabad and Purba Bardhaman) as well as the southeastern areas (Jhargram and Paschim Medinipur) have the highest yields, categorized as extremely very high (5.56–3.49 t/ha) and high (3.49–2.49 t/ha). A significant rice yield gap (50-40 %) was found in the center and south-east areas (Bankura, Jhargram, and Paschim Medinipur), the northern region (Murshidabad and Birbhum), and the western region (Purulia). In 2030, the north-western region (Birbhum), as well as the middle and south-eastern regions (Bankura, Jhargram, and Paschim Medinipur districts), had the highest proportion of high (50%–40 %) and very high (60%–50 %) rice yield gap. Our findings can contribute to a new viewpoint on agricultural planning and management for sustainable growth in the face of changing climate circumstances.

Predictive cheminformatics modeling of reorganization energy (RE) for p-type organic semiconductors: Integration of quantitative read-across structure-property relationship (q-RASPR) and stacking regression analysis

Organic semiconductors (OSCs), being light in weight, decomposable, cheap, and flexible, can be an excellent replacement for inorganic semiconductors. Reorganization energy (RE) is an essential parameter that can help determine charge carriers' mobility. Understanding and controlling the reorganization energy (RE) of OSCs is necessary for the design and optimization of organic electronic devices such as optoelectronic devices, organic photovoltaics, organic light-emitting diodes (OLEDs), photodetectors, and organic field-effect transistors (OFETs). The quantitative read-across structure-property relationship (q-RASPR) is an emerging computational methodology that can efficiently predict the material's property using structural and physiochemical features, and similarity measures. This method has also shown an enhancement in the predictivity of the models so developed. The stacking regression methodology uses the aggregates/predictions from the individual models and uses them according to their weights to produce an output in the new model. This study elaborates on the sequential steps for the development of a stacking q-RASPR model for the prediction of reorganization energy (RE) of the p-type organic semiconductors (OSCs) comprising 171 diverse organic structures of acenes, thiophenes, thienoacenes, and some pantalenes moiety as well. The predictions from the individual q-RASPR models developed using different similarity measures were used to perform the final stacking regression. Different machine learning (ML) algorithms were also used to enhance the model quality and predictivity. The model so developed through the stacking regression using support vector machine (SVM) method was of statistically good quality (R2 = 0.735), robustness (Q2LOO = 0.668), and effective predictive ability (Q2F1 = 0.801). The error in the model's predictions is also low compared to the other models developed earlier. This q-RASPR model was also used to predict an external set of 10888 compounds, and it shows an error (MAE) of only 87.95 meV, much lower than that reported earlier. The model developed in this study may be used to predict RE during high throughput screening of OSCs.

Implementation of Natural Language Processing for Bullying Complaints with Voice to Text Conversion

Bullying in high school is frequent and negatively affects students' psychological well-being. The lack of effective reporting mechanisms makes students hesitant to report bullying cases for fear of their identity being exposed, which can lead to stigma or retaliation. The lack of data on bullying incidents also hampers prevention and intervention measures. The study designed and implemented a mobile application that uses natural language processing (NLP) for speech-to-text conversion, enabling anonymous and convenient reporting of bullying cases. The app ensures the anonymity of whistleblowers and facilitates the collection of accurate data on bullying incidents, helping schools respond with appropriate preventive measures. Software engineering methodologies are used with a focus on requirements analysis, system design, implementation, and application testing. NLP technology is used to interpret verbal instructions into text, with the Support Vector Machine (SVM) method for text classification, ensuring high accuracy in detecting bullying incidents. The trial application in several high schools showed the relevance and effectiveness of the application. Ethical and security considerations are top priorities, with an emphasis on whistleblower identity protection and data security. The test results showed that the application achieved 92% accuracy, 90% precision, and 88% recall, demonstrating its effectiveness in collecting bullying reports anonymously and accurately.