K- nearest neighbour (KNN)

The course selection process in higher education has grown more complex as the students face the growing volume of academic options that must align with their personal and career goals. The traditional advising methods often fail to provide the personalized recommendations, resulting in the delayed graduations and the higher dropout rates. Therefore, the automated and intelligent course selection models and systems are becoming more important. These AI-based systems nearly deliver the tailored guidance to the students by analyzing their academic performance. In addition, this guidance is also based on the students' individual characteristics, helping to create the personalized course recommendations. This study presents a predictive model for the personalized course advising, in which it incorporates data from the academic, personal, and the social dimensions. Then, this model uses and employs the advanced machine learning algorithms and the optimization techniques and methods to maximize the accuracy of course recommendations. The feature selection methods like Information Gain and Chi-Square are used, in addition to dedicated classifiers such as the Decision Trees (DT) and the K-Nearest Neighbors (KNN). To balance the dataset, the Synthetic Minority Oversampling Technique (SMOTE) is applied. In the implementation phase, first, the Knapsack Problem algorithm is applied to optimize the students' course loads in order to ensure the recommendations match their abilities and the institutional requirements. Next, the KNN is applied and reaches an accuracy of 91.57% and outperforms the DT, which provides a practical solution for the problem of adaptive course advising. INDEX TERMS Personalized academic advising, educational data mining (EDM), feature selection techniques, knapsack problem algorithm.

This study provides a comprehensive review of machine learning (ML) applications in the fields of business and finance. First, it introduces the most commonly used ML techniques and explores their diverse applications in marketing, stock analysis, demand forecasting, and energy marketing. In particular, this review critically analyzes over 100 articles and reveals a strong inclination toward deep learning techniques, such as deep neural, convolutional neural, and recurrent neural networks, which have garnered immense popularity in financial contexts owing to their remarkable performance. This review shows that ML techniques, particularly deep learning, demonstrate substantial potential for enhancing business decision-making processes and achieving more accurate and efficient predictions of financial outcomes. In particular, ML techniques exhibit promising research prospects in cryptocurrencies, financial crime detection, and marketing, underscoring the extensive opportunities in these areas. However, some limitations regarding ML applications in the business and finance domains remain, including issues related to linguistic information processes, interpretability, data quality, generalization, and the oversights related to social networks and causal relationships. Thus, addressing these challenges is a promising avenue for future research.

Wine quality is traditionally judged by human tasters, which makes the process subjective and inconsistent. This research presents 'Wine IQ: An Exceptional Beverage Selection,' a machine learning-based system that classifies wines into three categories-excellent, good, and poor-using physicochemical properties. The project leverages a Random Forest classifier to achieve accurate quality prediction and introduces a rulebased suggestion engine that provides actionable recommendations for improving lower-quality wines. This dual functionality not only enhances quality prediction but also supports producers in maintaining consistent standards. The system has been validated using accuracy, precision, recall, confusion matrices, and ROC curves, demonstrating its effectiveness in bridging data science and winemaking practices.

Accurate subsurface assessment is critical for ensuring the integrity and longevity of engineered structures, especially in weathered terrains characterized by highly variable soil and rock properties. This challenge is particularly evident in the complex granitic terrain of Ipoh, Perak, Malaysia, where differential weathering and fracturing introduce substantial uncertainty into foundation design. Despite increasing demands for resilient infrastructure, conventional site investigation methods often fall short in capturing the spatial and lithological heterogeneity required for informed decision-making. This study addresses these limitations by integrating electrical resistivity tomography (ERT), induced polarization (IP), standard penetration testing (SPT), and supervised machine learning (ML) algorithms to enhance subsurface characterization and predictive modeling of chargeability. The combined ERT–IP inversion results, supported by borehole logs and SPT data, delineated four distinct lithological units across the study area. Four ML algorithms—SLR, KNN, SVM, and CatBoost—were trained to model relationships between geophysical and geotechnical parameters, with all achieving strong performance (training R² > 0.90; error metrics <10%). Among all the evaluated models, CatBoost exhibited the strongest overall performance, achieving high predictive accuracy and maintaining consistent generalization across validation and test sets, although with mild overfitting observed. Among all the evaluated models, CatBoost exhibited the strongest overall performance, achieving high predictive accuracy and maintaining consistent generalization across validation and test sets, although with mild overfitting observed. From a geotechnical perspective, the topsoil/residual and highly weathered/fractured units were deemed unsuitable for heavy structural loads due to poor consolidation and water retention. In contrast, the relatively weathered bedrock unit, defined by resistivity values of 800 to >1000 Ωm and chargeability >18 msec, was identified as a suitable foundation material after appropriate soil and rock modifications. This research introduces a novel, noninvasive geophysical–geotechnical framework enhanced by ML, offering a cost-effective and scalable approach for subsurface evaluation. The methodology significantly reduces uncertainty in engineering assessments and supports safer, data-driven construction decisions, with potential applicability to other complex weathered terrains, subject to site-specific calibration.

Weather prediction plays a vital role in numerous fields, including agriculture, transportation, and emergency response. However, the intricate and nonlinear nature of atmospheric progressions makes accurate forecasting a persistent challenge. This research explores the application of machine learning, particularly Random Forest (RF) algorithm to improve weather prediction using historical meteorological data. The study employs a dataset comprising 13,202 records with 11 weather-related variables. Two distinct models were developed: one using raw, unprocessed data and another incorporating Genetic Algorithm (GA)-based feature selection to identify optimal predictors. Both models were evaluated on a test of 3,960 instances. Results indicate that the GA-enhanced model outperformed the baseline, achieving an accuracy of 92.65% compared to 91.31%. Enhancing Weather Forecasting Accuracy… 284 Additional metrics, including kappa statistics, MAE, RMSE, precision, recall, and F-measure further validated the model's robustness. Notably, both models exhibited strong discriminative ability, with ROC and PRC areas exceeding 0.98, while the optimized model maintained high performance with reduced dimensionality. This study demonstrates that combining Random Forest with Genetic Algorithm-driven feature selection significantly enhances weather prediction accuracy. The proposed approach offers a reliable framework for developing efficient forecasting systems, with potential applications in real-time and longterm meteorological analysis.

Myocarditis is a rare but key adverse event linked to mRNA COVID-19 vaccines, predominantly in young males. Epidemiological data indicate an incidence of approximately 12.6 cases per million doses administered to patients aged 12-39 years, mostly following the second dose of the vaccine. Most patients present with elevated levels of cardiac biomarkers, chest pain, and abnormal ECG findings within a few days of vaccination. Proposed mechanisms for the exact pathophysiology of this include molecular mimicry between the SARS-CoV-2 spike protein and cardiac antigens, activation of immune pathways, and dysregulated cytokine expression. Despite these findings, the overall benefit-risk balance for COVID-19 vaccination remains positive, as the majority of patients recover fully. In contrast, COVID-19-associated myocarditis is more common and more severe, with an estimated incidence of 1,000-1,400 cases per 100,000 infections. Clinical presentation of vaccine-associated myocarditis is usually mild and self-limiting, and most patients do recover without significant long-term effects. Treatment is usually supportive in nature and has an emphasis on ruling out acute coronary syndrome and symptomatic management for heart failure or arrhythmias if present. Given its low incidence and the generally good outcome, vaccination against COVID-19 is recommended from 12 years of age and above, with provision for ongoing surveillance for monitoring and management of rare adverse events like myocarditis.

The pharmaceutical industry, while essential to public health, contributes significantly to environmental degradation through energy-intensive processes, hazardous waste, and high carbon emissions. This review explores the evolving role of quality assurance (QA) as a strategic enabler of sustainability within pharmaceutical manufacturing. Traditionally focused on compliance and product integrity, QA is now expanding to incorporate green chemistry principles, environmental auditing, and sustainability metrics into quality systems. The integration of digital tools such as Digital Quality Management Systems (QMS), Artificial Intelligence (AI), and the Internet of Things (IoT), under the Pharma 4.0 framework, is transforming QA into a real-time, data-driven pillar of sustainable operations. The review emphasizes the implementation of life cycle assessments (LCA), ecodesign strategies, renewable feedstocks, solvent recovery, and green documentation practices across the product lifecycle from R&D to distribution and disposal. It further highlights the importance of cross-functional collaboration, regulatory alignment, and stakeholder engagement in achieving green objectives. Despite promising innovations, the study identifies several barriers, including regulatory inflexibility, a lack of standardized green QA frameworks, and resistance to organizational change. Strategic enablers, including total quality management (TQM), employee empowerment, policy incentives, and environmental governance, are discussed as critical to embedding sustainability in quality systems. Ultimately, QA's alignment with environmental stewardship not only ensures regulatory compliance and operational efficiency but also fosters long-term corporate resilience and societal trust. By embedding sustainability into the pharmaceutical quality framework, this transformation supports the industry's shift toward greener, more ethical, and future-ready manufacturing practices that align with global environmental and public health goals.

The importance of recommendation systems is increasing day by day due to the massive number of data and information-overloaded arising from the internet. This data can be collected in predictive datasets; these datasets can be processed and analysed via data mining methods. In this paper, an efficient hybrid movie recommender system has been designed using the association rules mining technique and K-nearest neighbours (KNN) algorithm as a classification method. The K-nearest neighbours (KNN) algorithm subsystem was used to create the first candidate list through a practical MovieLens dataset, which was retrieved from the source of the NetFlix network. Beside, the Apriori algorithm subsystem is used to analyse the same dataset and create the second list. Finally, the proposed system creates a final recommended list by matching the two lists. The results of the proposed system provide better performance than the existing systems in terms of the important degree. The important degree ...

Diabetes, for a long period of time, was misjudged as a trivial concerned disease but has now risen to become one of the fastest-growing chronic diseases, affecting around 463 million people worldwide. In most cases of diabetes, patients are unaware of the disease due to the moderately long asymptomatic stage, and the prevention process becomes complicated with the delay since most of the cases of diabetes remain undiagnosed. Therefore, the initial stage diagnosis of diabetes is an important factor in order to enable clinically meaningful outcomes. To determine the likelihood of having diabetes, our study utilizes a dataset that includes both newly diabetic and would-be diabetic patients and employs five different machine-learning algorithms. Results indicate that Random Forest is the best model with an overall accuracy of around 99%. We also use an interpretable machine learning technique to determine the correlation between the response variable and the predictors.

The competition for postgraduate admissions has intensified due to a rise in the number of applicants. Many students struggle to understand specific admission criteria and rely on costly and biased consultancy services. To address this, a machine learning-based system is proposed to predict admission chances based on individual profiles. The system uses a historical dataset with features like GRE scores, GPA, TOEFL Scores, Statement of purpose, Letter of recommendation, research experience, and professional background. Three models are developed and compared: Linear Regression, Decision Tree Regression, and Logistic Regression. Linear and Decision Tree models explore linear and non-linear patterns, respectively. Logistic Regression, designed for binary classification, predicts admission probabilities effectively. Logistic Regression shows superior accuracy and minimal error, making it the best choice. The system provides a scalable and data-driven alternative to help students apply strategically.

Termín imigrace znamená dobrovolný pohyb z domovského státu do jiného. Každoročně jsou po celém světě registrovány miliony imigrantů (Gregory a kol. 2009). To činí tuto formu migrace za jeden z nejviditelnějších procesů poukazujících na probíhající sociální změny v dnešním světě (Sassen 1996, cit v Gregory a kol. 2009). Imigrace v dnešním slova smyslu souvisí se vznikem národního státu a donedávna byla úzce spojována s kolonizací. V důsledku toho jsou některé státy považovány za imigrantské (Austrálie, USA, Kanada). V případě, že se imigrant usadí bez souhlasu tamní vlády, je tato migrace považována za nelegální (Gregory a kol. 2009). Dle odhadů tvoří nelegální migranti 15-20 % světové migrační populace, tj. 30–40 milionů osob (Drbohlav a kol. 2009, cit. v Drbohlav 2011). Tyto trendy nárůstu nelegální migrace můžeme pozorovat i u nás.

Adopting Machine Learning (ML) in manufacturing quality assurance (QA) has accelerated with Industry 4.0, enabling automated defect detection, predictive maintenance, and real-time process optimization. However, selecting the most effective ML model remains challenging due to performance variability, scalability constraints, and inconsistent evaluation metrics across manufacturing sectors. This systematic review analyzes over 300 peerreviewed studies over the last two decades (mostly analyzing the recent works) to evaluate the effectiveness of widely used ML algorithms-Artificial Neural Networks (ANNs), Support Vector Machines (SVMs), Random Forests (RFs), Decision Trees (DTs), and K-Nearest Neighbors (KNN)-in QA applications. Performance metrics include accuracy, precision, speed, recall, computational efficiency, scalability, and real-time processing capabilities. Findings reveal that ANNs outperform other models in image-based defect detection, while SVMs and RFs excel in predictive maintenance and process parameter optimization. DTs provide better interpretability for process control, and KNN is effective for small-scale QA implementations. In specific case scenarios, RF models showed particular strength in handling high-dimensional sensor data in fault detection in manufacturing quality assurance operations. The study presents a comparative assessment framework, guiding algorithm selection based on industry-specific requirements and operational constraints. This review provides the latest implementation of ML in QA along with quantitative evidence on which algorithm offers the most optimization in specific industrial settings, which would help in algorithm selection in manufacturing quality assurance in future for both researchers and industrial experts. Also, it offers an overview of the major and minor algorithms based on their performance metrics.

This study evaluates several machine learning algorithms—Support Vector Machine (SVM), Random Forest, Logistic Regression, Decision Tree (C4.5), and k-Nearest Neighbors (KNN)—for breast cancer detection using the Breast Cancer Wisconsin Diagnostic dataset. We implemented comprehensive pre-processing and model evaluation with Scikit-learn in Python. Our findings show that SVM achieved the highest accuracy, with 99.9% on the training set and 98.50% on the testing set, indicating superior performance in handling high-dimensional data. Random Forest also performed well, with accuracies of 98.5% and 98.20%, respectively. Logistic Regression and Decision Tree models provided reliable predictions when tuned, while KNN was less effective. SVM and Random Forest are recommended for clinical decision support systems due to their high accuracy and robustness.

The Indian Premier League (IPL) stands as one of the most widely followed T20 cricket leagues, drawing
immense global attention and inspiring fans to anticipate match results. However, predicting match outcomes
remains challenging due to factors such as team dynamics, player performance, venue influence, and toss
decisions. This project introduces a machine learning-based prediction system built using Python to forecast IPL
match winners. Leveraging historical data from IPL seasons spanning 2008 to 2024, the system performs data
preprocessing, feature extraction, and model training using algorithms like Logistic Regression, Random Forest,
and Support Vector Machines (SVM). The system includes a Flask-based web interface that allows users to
enter match details and instantly view the predicted outcome. By focusing on detailed feature engineering and
tuning, the project aims to achieve high prediction accuracy, offering a powerful analytical tool for cricket fans
and sports analysts alike.

Human disease prediction systems are becoming increasingly important for early diagnosis and treatment of diseases. The majority of the existence machine learning models of the health care evaluation are focusing on only single disease at the time of analysis. One is like an evaluation of diabetes, one of cancer and another of skin diseases. No such system exists that lets a single evaluation to make multiple disease predictions. More than one analysis can be predicted in our diagnosis system. The system learns through a set of 133 symptoms and 41 diseases. A heatmap describing the relationship between symptoms is calculated, and diseases are predicted by a random forest classifier,svm,naive bayes classifier in respect to symptoms. There is 99 percent accuracy of the system in the training set. These are some of the applications of the system which include assisting doctors, patient education and aid in coming up with diagnostic tools. A system is a useful product in early detection of diseases and it can enhance the standards of healthcare of millions of individuals.

Human disease prediction systems are becoming increasingly important for early diagnosis and treatment of diseases. The majority of the existence machine learning models of the health care evaluation are focusing on only single disease at the time of analysis. One is like an evaluation of diabetes, one of cancer and another of skin diseases. No such system exists that lets a single evaluation to make multiple disease predictions. More than one analysis can be predicted in our diagnosis system. The system learns through a set of 133 symptoms and 41 diseases. A heatmap describing the relationship between symptoms is calculated, and diseases are predicted by a random forest classifier,svm,naive bayes classifier in respect to symptoms. There is 99 percent accuracy of the system in the training set. These are some of the applications of the system which include assisting doctors, patient education and aid in coming up with diagnostic tools. A system is a useful product in early detection of diseases and it can enhance the standards of healthcare of millions of individuals.

This study evaluates the analysis of machine learning (ML) and regression models for maize yield prediction across three districts in India: Raipur, Bilaspur, and Durg, using data from 1997 to 2022. Key climatic parameters, including rainfall (RF), relative humidity (RH), wind speed (WS), maximum temperature (Tmax), and minimum temperature (Tmin), were utilized to train and test four models: Random Forest (RF), Support Vector Machine (SVM), K-Nearest Neighbors (KNN), and Linear Regression (LR). The performance of the models was evaluated using metrics like Mean Absolute Error (MAE), Mean Squared Error (MSE), and Root Mean Squared Error (RMSE)., and R-squared (R²). Results reveal that Random Forest consistently outperformed other models across all districts, achieving the highest R² values (0.87 in Raipur, 0.90 in Bilaspur, and 0.95 in Durg) and the lowest error rates. In contrast, Linear Regression exhibited the poorest performance, highlighting its limitations in handling complex agricultural datasets. The study demonstrates the superiority of ML techniques, particularly Random Forest, in accurately predicting maize yields, offering valuable insights for farmers and policymakers to optimize crop planning and resource allocation under varying climatic conditions. These findings underscore the potential of advanced ML models in enhancing agricultural productivity and decision-making in the face of climate variability. The analysis was conducted using R-software, ensuring robust and reproducible results.

Depression is a mental disorder that is characterized by a general mood or feeling of low self-esteem, loss of interest towards daily activities and low energy within a particular person. It is a very serious mental condition and its automatic detection through online social media platforms like Twitter could help identifying depressed individuals remotely. This paper suggests a novel method to extract tweets indicating depression using word lists. Various classification algorithms like SVM, KNN, Naive Bayes and Random Forests have been used to classify the individual tweets as to whether they indicate depression in the subject or not. Metrics like F1-Score has been used the verify and compare the results of the models using an unseen test dataset.

The evaluation of creditworthiness and loan eligibility remains a fundamental yet complex task within financial institutions. Traditional loan assessment processes often involve manual analysis of applicant profiles, which can be time-consuming, prone to inconsistencies, and influenced by subjective judgment. To address these limitations, this study presents the development of an intelligent, automated loan approval prediction system that leverages machine learning techniques to enhance decision-making efficiency and reliability. The proposed model employs a K-Nearest Neighbors (KNN) classification algorithm, trained on synthetically generated datasets designed to simulate realistic applicant scenarios. This system is integrated into a web-based platform built using Flask, HTML, CSS, and JavaScript, enabling users to interactively submit personal and financial information. Upon submission, the model evaluates the application and determines the likelihood of loan approval. For approved applications, the system further computes an estimated loan amount, tenure, and monthly instalment using dynamic interest rate inputs. A distinguishing feature of the system is its ability to provide interpretable feedback. Users receive either a detailed breakdown of loan terms in the case of approval or a clear explanation supported by visual analytics when the application is declined. This transparency fosters user understanding and builds trust in the automated process. The proposed solution demonstrates the potential to improve the accuracy, consistency, and speed of loan processing in financial institutions. Moreover, its explainable nature aligns with the principles of responsible AI in financial technology applications.

Heart as one of the essential organ of the human body and with its related disease such as cardiovascular diseases accounts for the death of many in our society over the last decades, and also regarded as one of the most life-threatening diseases in the world. Hence we seek to predict a system for Heart disease using a supervised Machine Learning (ML) trained model in MATLAB2018 workflow in a real-time environment. To develop the system, 299 heart sounds from patients were obtained and labeled as normal and abnormal heart sound. Features were extracted and labeled as dataset; K Nearest Neighbour (KNN), Support Vector Machine (SVM) and Decision Tree (DT) algorithm were used as the training platform. From the classification analysis developed using the supervised ML trained model in MATLAB2018 in conjunction with system software features for the prediction of the heartbeat for both current and predefined of a heart condition algorithms used in training the dataset for the prediction w...

Handling missing data is a critical challenge in data processing, as it can significantly impact the performance of machine learning models. This paper proposes a Correlation Assisted Support Vector Machine (CA-SVM) based imputation method that integrates correlation analysis with the predictive power of Support Vector Machines to enhance missing data prediction accuracy. The CA-SVM approach identifies highly correlated attributes to guide the SVM in estimating missing values more effectively, preserving the underlying structure of the dataset. A comparative study is conducted between the proposed CA-SVM, standard SVM, and K-Nearest Neighbors (KNN) imputation techniques using benchmark datasets. Experimental results demonstrate that the CA-SVM method achieves a prediction accuracy of approximately 97 per cent, outperforming both traditional SVM and KNN methods in terms of accuracy, robustness, and consistency. The findings confirm that correlation-based feature guidance significantly improves the imputation performance of SVM models. This research highlights the potential of hybrid imputation techniques in building more reliable and accurate data-driven systems.

With its increasing occurrence, diabetes has become a major global health concern that presents serious difficulties for healthcare systems everywhere. Diabetes must be identified early and managed proactively to improve patient outcomes and lessen the disease's toll. This initiative offers an original method for predicting diabetes using cutting-edge machine learning algorithms.

This study used Artificial Neural Network (ANN) for the prediction of power required to inflate different tyre sizes and inflation times. ANN is a widely accepted machine learning method that uses past data to predict future trend. An existing database obtained experimentally from a tyre pressure control test rig was optimized using genetic algorithm(GA) which is an optimization tool that can find better subsets of input variables for importing into ANN. The ANN results were compared with the results obtained experimentally. The results show that the model can be implemented in modern day tyre pressure control designs and be used to predict inflation times and power required to inflate different tyre sizes.

Rest is essential for a high-level physiological and psychological performance. It is also necessary for the muscles to repair, rebuild, and strengthen. There is a significant correlation between the quality of rest and the resting posture. Therefore, identification of the resting position is of paramount importance to maintain a healthy life. Resting postures can be classified into four basic categories: Lying on the back (supine), facing of the left / right sides and free-fall position. The later position is already considered to be an unhealthy posture by researchers equivocally and hence can be eliminated. In this paper, we analyzed the other three states of resting position based on the data collected from the physiological parameters: Electrogastrogram (EGG), Electrocardiogram (ECG), Respiration Rate, Heart Rate, and Oxygen Saturation (SpO2). Based on these parameters, the resting position is classified using a hybrid stacked ensemble machine learning model designed using the Decision tree, Random Forest, and Xgboost algorithms. Our study demonstrates a 100% accurate prediction of the resting position using the hybrid model. The proposed method of identifying the resting position based on physiological parameters has the potential to be integrated into wearable devices. This is a low cost, highly accurate and autonomous technique to monitor the body posture while maintaining the user's privacy by eliminating the use of RGB camera conventionally used to conduct the polysomnography (sleep Monitoring) or resting position studies.

Early detection of life-threatening diseases such as cancer and thyroid disorders plays a critical role in reducing mortality and improving treatment outcomes. With traditional diagnostic methods often being time-consuming, resourceintensive, and dependent on specialist expertise, the integration of intelligent systems has become essential. This study explores the application of various machine learning (ML) and deep learning (DL) algorithms to predict diseases at an early stage using structured clinical datasets. Algorithms like Random Forest, Decision Tree, K-Nearest Neighbors (KNN), Naïve Bayes, and Artificial Neural Networks (ANN) were implemented and evaluated based on accuracy, precision, recall, and F1-score. The results show that all models achieved above 85% accuracy, with ANN and Random Forest models performing exceptionally well. A Streamlit-based web application was developed for real-time prediction, enabling easy clinical integration. The research underscores the effectiveness of ML/DL in enhancing diagnostic efficiency and recommends further development involving real clinical data and advanced model interpretability.

In this work, we propose a novel boosting-based machine learning algorithm called EvoBoost, invented by Sudip Barua. Gradient boosting has emerged as a cornerstone technique in machine learning, achieving state-of-the-art performance in both classification and regression tasks. While existing models such as XGBoost, LightGBM, and CatBoost are widely adopted, they present challenges including excessive hyperparameter tuning, high memory consumption, and suboptimal handling of imbalanced data. EvoBoost addresses these limitations through a streamlined boosting framework that is both effective and easy to implement. It introduces probabilistic residuals for classification and a clean, interpretable residual computation for regression. Extensive empirical evaluations across six benchmark datasets demonstrate that EvoBoost consistently outperforms or matches the performance of established models in terms of accuracy, R² score, and log loss, while maintaining superior interpretability and implementation simplicity.

This paper explores the potential of Retrieval Augmented Generation (RAG) to enhance search and retrieval capabilities within academic libraries. RAG integrates large language models (LLMs) with structured knowledge retrieval, creating a hybrid system that can deliver accurate, context-aware, and conversational search results. The study evaluates how this approach can modernize digital library services and improve user engagement with scholarly content. Design/methodology/approach We examine the technical architecture and requirements of RAG systems-including vector databases, embedding pipelines, and API middleware integration with existing Integrated Library Systems (ILS). The paper also evaluates ethical considerations, copyright issues, and data privacy risks in implementing RAG in academic contexts. Use cases focus on enhanced semantic search, multimodal research interfaces, and personalized AI research assistance. Findings RAG demonstrates strong potential to transform academic library systems by enabling semantic indexing, natural language query processing, and interactive AI-driven interfaces. Benefits include improved precision in search results, better alignment with user intent, and enhanced accessibility to interdisciplinary content. However, implementation requires addressing key challenges in scalability, transparency, cost-efficiency, and regulatory compliance. Originality/value This is one of the first comprehensive studies to investigate RAG applications in academic library systems. The paper provides a roadmap for library technologists, digital infrastructure teams, and AI researchers to leverage RAG models for next-generation search services. It emphasizes the transformative impact of LLM-powered retrieval in academic knowledge discovery.

This study aims to equip students with a foundational understanding of the business data analysis process. By utilizing Python and widely used libraries such as pandas, scikit-learn, matplotlib, and seaborn, it demonstrates key steps including data preprocessing, exploration, feature transformation, and the development and evaluation of machine learning models. Th e dataset, obtained from Kaggle, comprises 200,000 records related to marketing campaigns with variables such as campaign type, customer segment, cost, and impressions. Through the implementation of regression models-Linear Regression, Decision Tree, and Random Forest-the study helps learners grasp practical data analysis workflows and strengthens their ability to apply Python to real-world business challenges.

Lightning-related outages are great concern for power utility companies especially in the tropical regions. Tenaga Nasional Berhad as the system operator in Malaysia realized the importance of this issue. In this paper, statistical information on transmission line outages together with lightning stroke data are presented. Lightning data was obtained from TNB Lightning Detection Network which generates lightning density map and other information such as peak current, polarity and location of the detected strokes with 500 m location accuracy. A case study on 500 kV A to B line represent statistical information of collected strokes where 18 kA and 14 kA were reported as median values of current peak for negative and positive strokes, respectively. Mean values are 22 kA for negative strokes and 17 kA for positive strokes. About 93% of 658,619 collected strokes from 2004 to 2015 are negative whereas the remaining are positive. A total of 582 (90%) negative strokes and 65 (10%) positive strokes were recorded along selected case line where 83% of outages were reported to occur on top phase. Lightning stroke density, terrain characteristics, tower footing resistance and monsoon activities were highlighted as contributing factors on line performance due to lightning in this study.

This paper introduces a technique for reducing the sampling size necessary to explore a design space via building performance simulation. The strategy is based on a novel algorithmic procedure for creating a metamodel from an incomplete data series of a multivariate design problem. The algorithm interpolates the near-neighbours of the unknown vectors in a design space by the means of the near-neighbouring gradients, and progressively creates a global network of local relations involving vectors and gradients. The procedure produces low deviations with respect to the fully simulated series and performs at its best with sparse clusters of adjacent vectors (of the kind produced by star samplings or block-coordinate searches), unsmooth performance landscapes, and small sampling densities. The technique can also be utilized advantageously in optimization: post-metamodelling on the basis of the samples taken in an optimization search can indeed increase both the breadth of the information produced and the optimization efficiency.

One of the main challenges in bioinformatics is predicting the structures of macromolecules, particularly nucleic acids and proteins. In this study, we propose a hybrid approach integrating K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Neural Network (NN) algorithms. We perform an in-depth analysis using various metrics, including accuracy, Q3 score, ROC, and precision-recall curves. Based on the RS126 dataset, we compared our hybrid model with individual approaches, revealing that our model achieves an accuracy of 80% and a Q3 score of 86%, outperforming each of the algorithms separately. These results validate the effectiveness of combining models for protein secondary structure prediction (PSSP). We show that the hybrid model outperforms the other models for this task. We also discuss the implications of these results and propose future work to further improve the accuracy and robustness of the model. This approach could have important implications for protein structure modeling, in particular for understanding their three-dimensional structure and function.

Cardiac and Circulatory Disorders are among the primary causes of death globally, making early detection crucial for successful treatment and prevention, This research presents an advanced predictive cardiac and circulatory Disorders risk assessment model using machine learning and deep learning techniques. Three different datasets were analyzed, and the most accurate dataset was selected to train the models. Various algorithms, including Decision Tree, XGBoost(Extreme Gradient Boosting), Ada-Boost, Multi-Layer Perceptron (MLP), and Tabnet were implemented and compared. The MLP model attained the highest accuracy of 98% among all evaluated approaches, surpassing traditional machine learning models. The proposed system is designed to support healthcare professionals in identifying at-risk patients early, enabling timely intervention and improved healthcare results.

Handwritten character recognition remains a challenging task in various domains, including handwriting analysis and document digitization. This task poses a significant challenge because of the variability and complexity of handwriting. To address this, it is crucial to develop different methodologies that enhance accuracy and efficiency in recognition. This research evaluates the discrimination capacity of a classifier that combines together the Hu Moments features extraction technique and the k-nearest neighbors (KNN) algorithm for the purpose of handwritten character recognition. The authors worked with handwritten characters from the English alphabet where each character is represented in two handwritten forms. The average accuracy of recognition was 82.69%, that conforms the efficiency of the proposed classifier. This research contributes to the field of automated handwritten recognition and highlights the necessity for improved techniques in character analysis, which has implications for handwriting recognition in different contexts, linguistic studies, document digitization, and other applications where character recognition is crucial. This is the first study that aims to apply the combination of Hu Moments features extraction technique in combination with the KNN algorithm for purpose of handwritten character recognition.

In the rapidly evolving landscape of technology-driven commerce, laptops have become indispensable for both personal and professional applications, with a vast array of models presenting varied specifications and features. The intricate interplay of hardware configurations and pricing frameworks underscores the necessity for robust predictive models that empower consumers and manufacturers to make well-informed choices. This study delves into the critical challenge of accurately forecasting laptop prices by evaluating three machine learning methodologies: Linear Regression (LR), Histogram Gradient Boosting Regression (HGBR), and XGBoost Regression (XGBR). The research's importance is rooted in its capacity to refine pricing strategies, bolster market efficiency, and provide consumers with deeper insights into the value dynamics associated with different laptop specifications. The study leveraged an extensive dataset comprising 1,303 laptop entries, each characterized by 11 pivotal attributes encompassing processor type, RAM, storage capacity, screen dimensions, and graphical performance. Analytical techniques encompassed correlation assessment, feature significance determination, and comparative evaluation of model efficacy, employing key performance indicators such as the R² coefficient, Mean Squared Error (MSE), Root Mean Squared Error (RMSE), and Mean Absolute Error (MAE). XGBoost demonstrated a clear dominance over other predictive models, securing an R² value of 0.93559 on training data and 0.77524 on testing data. This superiority was further underscored by its markedly lower error margins, with an RMSE of 9,334.9 for training data, starkly contrasting the significantly higher 23,506.3 observed in Linear Regression. A thorough correlation analysis pinpointed RAM and processor specifications as the most decisive variables influencing price determination. The study asserts that ensemble learning methodologies, particularly XGBoost, represent the most dependable strategy for forecasting laptop prices. Nonetheless, the research highlights key areas for refinement, especially in narrowing the discrepancy between training and testing performance. These insights hold substantial implications for stakeholders in the laptop industry, paving the way for the advancement of more sophisticated predictive frameworks. Furthermore, the study enriches the broader discourse on consumer electronics pricing, emphasizing the transformative role of machine learning in optimizing market dynamics and strategic decision-making.

Esai ini disusun untuk memenuhi tugas akhir mata kuliah Filsafat Ilmu. Judul ini diangkat dari perkembangan teknologi wireless charging untuk kendaraan listrik (EV) di Indonesia yang kerap dipandang sebagai kemajuan teknis semata, tanpa mempertimbangkan nilai-nilai sosial dan etis yang menyertainya. Melalui pendekatan filsafat ilmu, tulisan ini mengkaji apakah ilmu pengetahuan bersifat bebas nilai atau terikat nilai, khususnya dalam konteks inovasi teknologi. Dengan menelaah penerapan wireless charging, dikaji bagaimana faktor kenyamanan, efisiensi, dan modernitas bersinggungan dengan tantangan keberlanjutan, ketimpangan infrastruktur, dan dampak lingkungan. Esai ini menunjukkan bahwa keputusan ilmiah dan teknologis tidaklah netral, melainkan dibentuk oleh konteks sosial, ekonomi, dan politik tertentu. Di akhir, ditekankan pentingnya pendekatan responsible innovation serta perlunya edukasi publik dan transparansi kebijakan demi pembangunan teknologi yang berkeadilan dan berkelanjutan di Indonesia.

The purpose of this study was to develop methods to localize forest windstorm damages, assess their severity and estimate the total damaged area using space-borne SAR data. The development of the methods is the first step towards an operational system for near-real-time windstorm damage monitoring, with a latency of only a few days after the storm event in the best case. Windstorm detection using SAR data is not trivial, particularly at C-band. It can be expected that a large-area and severe windstorm damage may affect backscatter similar to clear-cutting operation, that is, decrease the backscatter intensity, while a small area damage may increase the backscatter of the neighboring area, due to various scattering mechanisms. The remaining debris and temporal variation in the weather conditions and possible freeze-thaw transitions also affect observed backscatter changes. Three candidate windstorm detection methods were suggested, based on the improved k-nn method, multinomial logistic regression, and support vector machine classification. The approaches use multitemporal ESA Sentinel-1 C-band SAR data and were evaluated in Southern Finland using wind damage data from the summer of 2017, together with 27 Sentinel-1 scenes acquired in 2017 and other geo-referenced data. The stands correctly predicted severity category corresponded to 79% of the number of the stands in the validation data, and already 75% when only one Sentinel-1 scene after the damage was used. Thus, the damaged forests can potentially be localized with proposed tools within less than one week after the storm damage. In this study, the achieved latency was only two days. Our preliminary results also indicate that the damages can be localized even without separate training data.

In this study, the capability of Landsat-8 (L8), Sentinel-2 (S2), Sentinel-1 (S1), and their combination was investigated for estimating aboveground biomass (AGB). A pure stand of Fagus Orientalis located in the Hyrcanian forest of Iran was selected as the study area. The performance of a parametric approach, i.e., Multiple Linear Regression (MLR) model, and non-parametric approaches, i.e., k-Nearest Neighbor (k-NN), Random Forest (RF), and Support Vector Regression (SVR), was also evaluated for AGB estimations. Our results indicated that among S2 metrics, the FAPAR canopy biophysical index and NDVI index based on the red-edge band (NIR-b8a) have the highest correlation coefficient (r) of 0.420 and 0.417, respectively. The results of AGB estimation showed that a combination of S2 and S1 datasets using the k-NN algorithm had the best accuracy (R2 of 0.57 and rRMSE of 14.68%). The best rRMSE using L8, S2, and S1 datasets was 18.95, 16.99, and 19.17% using k-NN, k-NN, and MLR algorithms, respectively. The combination of L8 with the S1 dataset also improved the rRMSE relative to L8 and S1 separately by 0.96 and 1.18%, respectively. We concluded that the combination of optical data (L8 or S2) with SAR data (S1) improves the broadleaved Hyrcanian AGB estimation.

Forest height (FH) is an important driver for aboveground biomass (AGB) that can be obtained using interferometric synthetic aperture radar (InSAR). However, the limited access to the quad-polarimetric data or high-accuracy terrain model makes FH retrieval a challenging task. This study aimed to retrieve FH and further predict AGB by combining TanDEM-X InSAR coherence, Sentinel-1 (S-1), and Sentinel-2 (S-2) data. A total of 125 sample plots with a size of 900 m 2 were established in a broadleaved forest of Kheyroud, Iran. The linear and sinc models obtained by simplification of the random volume over ground model were used for deriving FH Lin and FH Sinc. Further investigation was conducted when S-1 and S-2 features, including backscatters and multispectral information, were added to FH predictions. Using the above-mentioned datasets and FH as an additional predictor, AGB was also predicted. K-nearest neighbor (k-NN), random forest (RF), and support vector regression (SVR) were employed for prediction. Lorey's mean height and AGB at sample plots were used in the accuracy assessment. Using the SVR method and synergy of FH Sinc, S-1, and S-2 features, the FH prediction was improved (FH imp) with RMSE of 3.18 m and R 2 = 0.59. The AGB prediction with RF and the combination of S-1 and S-2 features resulted in RMSE = 62.88 Mg•ha-1 (19.77%), which was improved to RMSE = 51.27 Mg•ha-1 (16.12%) when FH imp included. This study highlighted the capability of TanDEM-X InSAR coherence with certain geometry for FH prediction.

yang umum terjadi dan berdampak signifikan terhadap penurunan kualitas citra, seperti berkurangnya kontras, perubahan warna, dan hilangnya detail visual. Kabut disebabkan oleh penyebaran dan penyerapan cahaya oleh partikel halus seperti uap air dan debu di udara, menghasilkan efek yang dikenal sebagai airlight. Degradasi visual ini menjadi tantangan dalam berbagai aplikasi visi komputer, seperti pengenalan objek, navigasi otonom, dan penginderaan jauh. Untuk mengatasi permasalahan tersebut, penelitian ini mengkaji metode Dark Channel Prior (DCP), sebuah teknik penghapusan kabut pada citra tunggal yang telah terbukti efektif. DCP didasarkan pada pengamatan statistik bahwa dalam sebagian besar patch lokal gambar luar ruangan yang bebas kabut, terdapat setidaknya satu saluran warna dengan intensitas piksel sangat rendah. Dengan memanfaatkan karakteristik ini, DCP mampu memperkirakan peta transmisi dan cahaya atmosfer untuk menghilangkan kabut secara akurat. Studi ini juga memaparkan implementasi algoritma DCP dalam lingkungan mobile menggunakan bahasa pemrograman Kotlin untuk aplikasi Android, memungkinkan penghapusan kabut secara real-time pada perangkat portabel. Hasil pengujian menunjukkan peningkatan kualitas visual citra secara signifikan, termasuk kontras, ketajaman, dan detail [1], sehingga metode ini berpotensi diterapkan secara luas dalam teknologi citra mobile sehari-hari.

Penelitian ini membandingkan kinerja dua algoritma klasifikasi machine learning, yaitu Decision Tree dan Naive Bayes, dalam mengklasifikasikan jenis jamur sebagai layak konsumsi (edible) atau beracun (poisonous) berdasarkan dataset kategorikal dari UCI Machine Learning Repository yang terdiri dari 8124 entri dengan 22 fitur dan satu label target. Evaluasi dilakukan melalui variasi pembagian data latih dan uji (80:20, 70:30, dan 60:40), validasi silang stratified 5-fold, serta eksperimen dengan lima fitur terpenting. Hasil penelitian menunjukkan bahwa Decision Tree mencapai akurasi sempurna (100%) pada seluruh skenario, baik saat menggunakan seluruh fitur maupun lima fitur utama, menunjukkan keunggulan dalam menangani data kategorikal kompleks serta kemampuan generalisasi yang tinggi. Sebaliknya, Naive Bayes mencatat akurasi stabil di kisaran 92% namun mengalami penurunan performa signifikan saat jumlah fitur dikurangi. Temuan ini mengindikasikan bahwa Decision Tree lebih sesuai untuk klasifikasi data kategorikal yang saling bergantung, sedangkan Naive Bayes tetap relevan untuk kondisi dengan keterbatasan komputasi dan data yang lebih sederhana.

The proliferation of Android devices has resulted in a rise in complex malware specifically designed for these platforms, requiring higher detection techniques beyond conventional static and dynamic analyses. In this study, the Artificial Bee Colony (ABC) algorithm for feature selection is integrated with the eXtreme Gradient Boosting (XGBoost) and Random Forest (RF) classifiers to provide a novel method for Android malware detection. The ABC algorithm, which draws inspiration from honeybee foraging behavior, improves the performance of classifiers by balancing exploration and exploitation within feature subsets. Evaluation of the suggested approach on the Debrin Android malware dataset showed significant enhancements in detection accuracy and decreased false positives. The experimental findings demonstrated that both RF and XGBoost classifiers showed excellent performance, with RF slightly surpassing XGBoost in accuracy, precision, recall, and ROC-AUC metrics. The results highlight the efficacy of integrating metaheuristic feature selection with strong classifiers to enhance Android malware detection and tackle the difficulties presented by progressing threats.

Dalam setiap jenjang pendidikan, perpustakaan menjadi salah satu pusat informasi pembelajaran. Melihat perpustakaan sebagai pusat informasi pembelajaran, maka harus disesuaikan dengan ekspektasi penggunanya, ditata dan diatur dengan sistem tertentu sehingga memudahkan pengguna dalam mengakses informasinya (pelayanan). Kepuasan pengguna merupakan alat ukur keberhasilan suatu perpustakaan. Dalam ilmu statistik, regresi mempelajari hubungan antara satu variabel dengan variabel yang lain. Misalnya dalam menentukan hubungan kepuasan pengguna dengan pelayanan perpustakaan. Regresi dimanfaatkan selain untuk menentukan hubungan kuantitatif antar variabel juga untuk meramalkan nilai dari variabel dimasa yang akan datang. Analisis regresi linier adalah teknik statistika yang dapat digunakan untuk menjelaskan pengaruh variabel bebas (independent variable) terhadap variabel tak bebas (dependent variable). Adapun K-NN, merupakan algoritma yang memberikan pengenalan ke k-terdekat sebagai nilai prediksi dari query instance yang baru, yang dasar pendekatan unsupervised k-Nearest Neighbor atau KNN, yang mana dalam regresi adalah untuk menganalisis nilai output. Berdasarkan data yang dikumpulkan akan diketahui kualitas pelayanan perpustakaan terhadap kepuasan pengguna. Penelitian ini bertujuan untuk mengetahui prediksi kepuasan pengguna perpustakaan menggunakan kombinasi regresi linier berganda dan K-Nearest Neighbor(K-NN).

ABSTRAK
Di era digital yang terus berkembang, kebutuhan akan sistem pengambilan keputusan otomatis semakin meningkat di berbagai sektor, seperti bisnis, teknologi, dan industri lainnya. Sistem cerdas yang mampu menganalisis data secara efisien dan akurat menjadi kunci dalam meningkatkan kualitas dan kecepatan proses pengambilan keputusan. Salah satu pendekatan yang menjanjikan untuk mencapai tujuan ini adalah dengan mengembangkan sistem hybrid berbasis kombinasi algoritma machine learning.
Penelitian ini bertujuan untuk merancang dan membangun sistem cerdas hybrid yang mengintegrasikan beberapa algoritma machine learning melalui pendekatan modular. Pendekatan modular dipilih karena memungkinkan setiap algoritma beroperasi secara independen, sehingga memudahkan integrasi dan fleksibilitas sistem. Sistem yang dihasilkan diharapkan mampu menghasilkan keputusan secara lebih cepat dan akurat dibandingkan penggunaan algoritma tunggal. Berdasarkan hasil pengujian, sistem menunjukkan performa yang efektif dan kemudahan implementasi untuk berbagai skenario pengambilan keputusan otomatis.

To improve fraud detection in the banking industry, the study investigates integrating neural networks with the Harmony Search Algorithm (HSA). Advanced solutions are required since traditional procedures are frequently insufficient in the face of sophisticated fraud techniques. HSA is combined with neural networks, which are renowned for their capacity to recognize intricate patterns, to effectively optimize the settings. This study shows that fraud detection accuracy and reliability are greatly increased when neural networks' learning flexibility and HSA's optimization capabilities are combined. Findings indicate that models with near-perfect accuracy, including Decision Tree Classifier and Sequential models, have the potential to transform fraud prevention.

Amaç: Bu çalışmanın amacı, Borsa İstanbul İmalat Sanayiinde işlem gören şirketlerin finansal başarısızlık
tahmininde güvenilir bir model geliştirmektir. Bunun için Borsa İstanbul’da kayıtlı 100 üretim şirketinin 2005-
2019 yılları arasında finansal tablolarından elde edilen 18 değişken analizde kullanılmıştır.
Yaklaşım: Finansal başarısızlık tahmininde optimizasyon algoritmaları içinde yer alan karınca kolonisi
algoritması tabanlı bir yaklaşım (antminer +) kullanılmıştır.
Bulgular: Yapılan çalışma sonucunda finansal başarısızlık tahmininde en yüksek başarım oranına sahip kurallar
tespit edilmiştir. Bazı değişkenlerin finansal başarısızlık riski ile doğrudan ilişkili olduğu belirlenmiştir. Ayrıca
oluşturulan kuralların başarı oranlarının %82-%93 arasında olduğu belirlenmiş olup, ortalama başarı oranı ise%88
olarak tespit edilmiştir. Bu nedenle kullanılan algoritmanın literatüre göre geçerli olduğu tespit edilmiştir.
Özgünlük: Literatürde finansal başarısızlığın tespiti için karınca koloni algoritmasının (AntMiner+) kullanıldığı
az sayıda çalışma mevcuttur. Türkiye’de ise finansal başarısızlık tespitinde bu algoritmayla yapılan çalışma
bulunmamaktadır.

ABSTRAK Penelitian ini mengulas pemanfaatan strategi ensemble learning dalam pengembangan sistem cerdas hibrida yang mampu beroperasi secara andal di tengah tantangan data yang kompleks dan berisik. Sistem cerdas hibrida dirancang dengan mengombinasikan metode kecerdasan buatan berbasis simbolik dan pembelajaran mesin guna meningkatkan ketepatan dan fleksibilitas. Pendekatan metodologis mencakup studi pustaka dan eksperimen empiris, dengan pengujian terhadap berbagai algoritma ensemble seperti bagging, boosting, dan stacking yang diterapkan pada model dasar seperti Decision Tree, SVM, serta Neural Network. Hasil eksperimen menunjukkan bahwa model ensemble secara konsisten melampaui performa model tunggal, baik dari sisi akurasi, ketahanan terhadap noise, maupun kemampuan generalisasi. Selain itu, integrasi pendekatan neuro-symbolic dan pembelajaran penguatan multi-agen (MARL) menambah tingkat adaptabilitas dan kejelasan interpretasi sistem. Dengan demikian, ensemble learning memiliki peran vital dalam pengembangan kecerdasan buatan yang kuat dan dapat diandalkan, khususnya dalam aplikasi-aplikasi yang menuntut keputusan presisi tinggi.