Samarkand State University

Despite substantial airline industry investment in personalized services, limited empirical research examines how generational cohorts differ in their intentions toward personalized in-flight offerings. We investigate how Generation Z, Millennials, Generation X, and Baby Boomers differ in their intentions across three service dimensions—seating, in-flight entertainment, and meal selection—while examining flight duration as a moderating variable. A survey of 476 airline travelers revealed findings that challenge conventional generational assumptions. We have found that personalized services contradict expectations that only younger passengers would embrace such innovations. However, specific preferences varied: Millennials showed strongest intentions toward customized seating, Generation X exhibited unexpectedly high preferences for entertainment and seating personalization, and Baby Boomers demonstrated stronger entertainment intentions than anticipated. Flight duration significantly influenced personalization intentions across all generations, with long-haul flights amplifying preferences regardless of age, though no interaction effect emerged between generation and flight duration, indicating contextual factors operate independently of generational characteristics

Abstract The rapid development of Machine Learning (ML) and Deep Learning (DL) methods in recent years has made them popular among the scientific community. These methods have been applied across various fields, including radiation detection and the nuclear industry. These techniques have demonstrated the potential to enhance the performance and efficiency of automated radiation detection practices and nuclear safety and security practices. This paper will discuss incorporating deep learning techniques, Fully Connected Artificial Neural Networks (FCANN), and Convolutional Neural Networks (CNN) to detect radioactive isotopes with low-resolution gamma spectral data. Further, our work aims to enhance the performance of identification while exploring the explainability of these models. Using saliency mapping methods, we visualize the networks' regions of interest for a better understanding of the underlying decision-making processes. Both CNN and FCANN models demonstrated high accuracy while saliency mapping highlighted the relevant spectral features for underlying desision making process of each type of model.

Today, the main signal used to detect and diagnose any cardiovascular changes in the medical field is the electrocardiogram. However, various background noises that are added to the signal interfere with accurate diagnosis. In order to solve these problems, the development of a long-term ECG monitoring system, remote monitoring of patients, the widespread use of new technologies, and the collection of ECG data are of great importance for the early detection of heart diseases. In this scientific study, a personal cloud platform for data storage and a portable gadget for long-term monitoring of heart activity and ECG data collection were created. Wavelet Transform was used to filter and extract features to detect arrhythmia diseases from the collected ECG data. Highfrequency, short-duration signals (QRS complex and RR interval) were filtered and features extracted using Symlet4 wavelet transform, while low-frequency, long-duration signals (ST segment, P-wave, T-wave, PR interval, and QT interval) were filtered and features extracted using Coiflet2 wavelet transform. The findings showed that, compared with other components, Symlet4 produced higher values for QRS complex and RR interval in SNR and MSE calculations, while Coiflet2 produced higher values for ST segment, T-wave, P-wave, PR interval, and QT interval. Compared with other components, Symlet4 showed lower values for QRS complex and RR interval in RMSE calculations, while Coiflet2 showed lower values for ST segment, T-wave, P-wave, PR interval, and QT interval.

In this study, artificial intelligence models were selected for preparing and diagnosing a dataset of diabetes mellitus. The effectiveness of artificial intelligence algorithms in the early detection of diabetes was analyzed.To achieve good efficiency results, the presence of NaN values and categorical values in the dataset, as well as ensuring a balanced dataset, led to an improvement in classification results. For this purpose, data preprocessing, feature selection, and testing of various classification models were performed based on real medical data. Using genetic algorithms, the NaN values of the most diagnostically significant factors were filled in. Based on the prepared data, classification models with high accuracy were created. These models enable the automation of medical diagnostic processes based on Logistic Regression, K-NN, Decision Tree, and Naive Bayes algorithms, as well as assist patients in early detection of disease risks. This research may contribute to improving the efficiency of the healthcare system through early detection of diabetes.

This research addresses the development of algorithms for intelligent analysis of economic data based on artificial intelligence models. The study focuses on identifying delays in sales data through comprehensive analysis. A classification problem is formulated using a neural network-based model to enable intelligent data analysis. To solve the problem, the dataset was first prepared, and the mathematical model of the LSTM neural network was developed. Furthermore, optimization of the LSTM model was conducted using the gradient method. An intelligent algorithm based on the LSTM neural network model was proposed to solve the classification task. The performance of the developed model and algorithm was evaluated using specific assessment criteria.

In 2022–2023, fragments of figurative wall paintings were discovered in the Royal Palace at Sanjar-Shah, a Sogdian site near Panjikent in Tajikistan. The paintings depict a procession of priests approaching a large fire altar—this offers a rare insight into religious imagery and a representation of fire worship in Sogdian murals.

Recent results on tandem solar cells involving halide perovskites look very promising in obtaining economic solar cells with efficiencies >30%. Current research in the literature is about finding and optimizing materials for tandem solar cells, but only a few works focus on the tandem architecture. In the case of monolithic integration, in series connection of subcells is usually considered. However, such architectures come with a current matching condition, which narrows the choice of absorbers and questions device performances under real-world conditions where spectral variations are observed. In this work, we investigate the possibility of in parallel connection considering two absorbers with close band gap energies (i.e., nonideal configuration for current-matching condition): CH3NH3PbI3 (1.59 eV) and PM6:Y6 (1.36 eV). We show experimentally that if the currents are summed, the tandem open-circuit voltage (Voc) tends to be reduced to the lowest single junction Voc. We perform a simulation based on the single-diode equivalent circuit model to investigate the voltage matching condition and the impact of subcell series resistances on tandem performance. Our findings show that nearly matching open-circuit voltages is preferable to avoid any loss in power efficiency. Finally, we offer a solution that allows the combination of absorbers with different band gap energies: association of cells to control subcells' Voc while maintaining the same active area. With such a strategy, an almost 21% power efficiency nonmonolithic tandem is demonstrated in parallel connection architecture, opening the path toward efficiency in parallel connection monolithic tandem.

ABSTRACTIn this study, the Boussinesq problem is modeled using the half‐space model of the soil foundation to determine the collapse of the ground surface under a point load applied to the surface when the elastic modulus of the ground changes as a continuous function of depth. Analytical solutions are developed for the cases where the elastic modulus increases depending on the depth coordinate. The solutions are presented as the product of the classical Boussinesq solution and a dimensionless function that considers the heterogeneity with depth. Finally, a comparative analysis of the subsidence at dimensionless distances from the point load for different heterogeneity functions is performed compared to the classical solution.

The study examined the issue of applying artificial intelligence (AI) technologies to job search processes in employment systems. Based on the literature analysis, the main concepts of using AI in human resource management were presented. It was found that in recruitment, there is a relationship between AI replicating human cognitive processes and imitating human learning mechanisms. As a result, 3 main skills for mastering and using AI were presented. The general algorithm and description of the application of AI technologies to the labor market were presented, and the algorithms and methods of the Bat algorithm, the Gaussian distribution approach used in its implementation were studied. A method was proposed in which the optimization process was improved by adding an approach based on the Gaussian distribution to the Bat algorithm. This provides effective training data for vector machines and has been proven to give good results in predicting labor force employment.

In this research work, ensemble methods for analyzing classification problems were analyzed. Currently, various methods and algorithms are proposed and implemented based on machine systems to increase the accuracy of results. In this study, XGBoost, Random Forest, LightGBM, and AdaBoost algorithms were applied, and their results were evaluated based on model performance criteria including Accuracy, Precision, Recall, and F1-score. The research results showed that the XGBoost model achieved better results than ensemble models, which achieved a 92% accuracy result. This confirms its potential for widespread application in the field of medical diagnostics. High results were also achieved with the LightGBM and AdaBoost algorithms. LightGBM and AdaBoost showed high recall performance. Ensemble methods possess higher accuracy compared to traditional classifiers and help effectively address issues related to data class imbalance characteristics. The results of this research demonstrate that ensemble methods provide high reliability and adaptability in multi-class classification problems. The research results confirm the high effectiveness of ensemble methods in the process of early diagnosis and prediction of diabetes mellitus. This suggests that it could be applied to detect other diseases in the future.