A Comparative Study of Various Traditional and Hybrid Cryptography Algorithm Models for Data Security

<title>Abstract</title> Background The crucial process of choosing viable embryos for in vitro fertilization (IVF) has always depended on the subjective evaluation of embryologists. Automating this process using deep learning may improve pregnancy outcomes and make the evaluation more reliable. Methods A 3D Convolutional Neural Network (3D-CNN), CNN + LSTM hybrid model, and TimeDistributed CNN + GRU were among the deep learning architectures that we investigated. We also used transfer learning using pre-trained models like EfficientNet and ResNet50. To improve model generalization, frame selection tactics and data augmentation techniques were also used. Accuracy, precision, recall, and F1-score metrics on embryo time-lapse films were used to assess performance. Results The 3D-CNN model stood out by not misclassifying any non-viable embryos, which makes it highly reliable in avoiding poor embryo choices. Hybrid models and those based on transfer learning reached up to 82% accuracy and showed a good balance in identifying both viable and non-viable embryos. The TimeDistributed CNN + GRU model, especially when combined with frame selection and augmentation, improved the ability to detect viable embryos even further. Conclusion The proposed models, particularly the 3D-CNN, show strong potential for automating embryo selection with high accuracy, making them suitable for clinical use. Using frame selection and data augmentation helped the models better capture how embryos develop over time. To further improve performance, future research could explore combining multiple models, finding the right balance between recall and specificity, and incorporating more clinical data.

The rapid advancement of generative artificial intelligence has led to the widespread availability of highly photorealistic AI-generated images, making it increasingly difficult to distinguish synthetic content from real images using manual inspection. This raises serious concerns related to misinformation, digital forgery, and loss of trust in visual media. To address this challenge, this paper presents a comparative study between two deep learning approaches for AI-generated image detection: a baseline Convolutional Neural Network (CNN) operating in the spatial domain and a hybrid CNN model incorporating Fast Fourier Transform (FFT) based frequency-domain features. A large and heterogeneous dataset consisting of real images and AI-generated images from multiple generative sources was constructed and carefully preprocessed to simulate real-world conditions such as resizing and compression. The dataset was split into training, validation, and test sets using strict separation to prevent data leakage. Both models were trained under identical experimental conditions for 15 epochs using the same optimization strategy. Experimental results show that the spatial-domain CNN achieved a test accuracy of approximately 85.3%, outperforming the CNN+FFT hybrid model, which achieved an accuracy in the range of 76–78%. The findings demonstrate that, contrary to common assumptions, frequency-domain features do not necessarily enhance AIgenerated image detection under realistic preprocessing constraints. This study highlights the robustness of spatial-domain CNNs and emphasizes the importance of empirical evaluation when designing deepfake detection systems.

Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, and Gated Recurrent Units (GRUs) have gained significant popularity in time series forecasting across diverse domains including healthcare, astronomy, and engineering. However, the inherent variability in model performance due to random weight initialization raises questions about the reliability and consistency of these architectures for time series analysis. This study addresses this concern by conducting a comprehensive benchmark evaluation of nine neural network architectures: vanilla RNN, LSTM, GRU, and six hybrid configurations (RNN-LSTM, RNN-GRU, LSTM-RNN, GRU-RNN, LSTM-GRU, and GRU-LSTM). Performance evaluation was conducted using Monte Carlo simulation with 100 iterations across three real-world datasets: sunspot activity, Indonesian COVID-19 cases, and dissolved oxygen concentration measurements. Statistical analysis employed the Friedman test to assess performance differences across architectures. Results showed no statistically significant differences among the nine architectures. Despite the lack of statistical significance, consistent performance patterns emerged favoring LSTM-based hybrid architectures. The LSTM-GRU and LSTM-RNN configurations demonstrated superior performance across multiple evaluation metrics, with LSTM-RNN excelling in sunspot and dissolved oxygen forecasting, while standalone LSTM showed optimal performance for COVID-19 prediction. These findings provide evidence-based guidance for architecture selection in time series forecasting applications, suggesting that while statistical equivalence exists among architectures, LSTM-based hybrids offer practical advantages in terms of consistency and robustness across diverse temporal patterns.

In this study we investigate the extent to which the integration of psychological mechanisms from attitude theory into conventional analytical approaches can advance our understanding of travel behaviour. Three models, which explain volitions (intentions) in different ways, are specified and discussed: a customary model (CM) that directly links attitude and external variables (eg socio-demographics) to volition; a simplified version of the extended model of goal-directed behaviour (EMGB) that links attitude to volition via behavioural desire; and a hybrid model (HM) that integrates the two former models. Using survey data about the volition to buy media products (books, music, and DVDs, for example) online and in-store collected in four locations in the Utrecht region, the Netherlands, we find that shopping behaviour is reasonably well explained by the simplified EMGB. Past behaviour, perceived behavioural control, and subjective norms all have a statistically significant impact on the volition to shop online, while goal desire and perceived behavioural control significantly affect the volition to shop in-store. The results of the HM indicate that it is important to take external variables, such as access to physical stores and gender, into account when explaining shopping behaviour.

Snow depth is a critical parameter for characterizing snow dynamics and water resources, and its accurate inversion is essential for hydrological processes, climate studies, and disaster prevention in cold regions. Based on long-term daily ground meteorological observation data from the hydrological years 1961 to 2015 at two meteorological stations in Mohe and Mishan, Heilongjiang Province, China, this study integrates physical parameters of snow density and snow albedo from the ERA5-Land reanalysis data to systematically compare the performance of three machine learning and three deep learning models in retrieving daily snow depth. Four feature combination schemes were designed to evaluate the contributions of meteorological factors, lagged snow depth terms, and snow physical parameters. The results indicate that, for both machine learning and deep learning models, the first-order lagged value of snow depth is the most important variable determining prediction accuracy. In terms of model performance, machine learning methods excelled, with XGBoost performing particularly outstandingly, achieving optimal prediction accuracy and stability under the best feature combination (coefficient of determination, R2, reaching 0.989; root mean square error, RMSE, of 1.19 cm). Among deep learning methods, 1D CNN demonstrated strong local feature extraction capabilities, achieving prediction accuracy comparable to the best-performing machine learning model (R2 of 0.9878, RMSE of 1.26 cm). Notably, models specifically designed for time-series data, such as LSTM (R2 of 0.9848, RMSE of 1.41 cm), and the more complex 1D CNN-LSTM hybrid model (R2 of 0.9803, RMSE of 1.60 cm) did not show significant advantages in this study. This indicates that model complexity and predictive performance are not simply positively correlated. Through comprehensive analysis of data from both stations, this study demonstrates that a prediction framework centered on historical snow depth as the core driving factor, combined with key meteorological elements, is highly robust. Although the inclusion of ERA5-Land snow physical parameters did not significantly improve model accuracy, it provides important insights for the future development of hybrid models that integrate physical mechanisms with data-driven approaches. The findings offer an effective solution for reconstructing long-term snow depth time series and hold significant application value for simulating cryospheric hydrological processes and studying climate change.

In order to save unnecessary energy losses, power companies need to accurately predict electricity consumption. Various classification methods are used for electricity consumption forecasting: traditional time series models, deep neural network models based on Artificial Intelligence, and hybrid models. Most scholars use a single view to forecast electricity consumption. This article predicts electricity consumption from three views: traditional models, deep neural network models, and hybrid models. First, the electricity consumption time series is preprocessed. Then, from three views, models that have achieved good performance in some time series are selected, and then the forecast analysis of the model is given. Finally, the performance of different models in electricity consumption forecasting is evaluated and compared. The results show that the expected performance of the hybrid model is not as good as the traditional model and the deep neural network model in electricity consumption forecasting.

Modelling the impact of ISO 14001 on environmental performance: A comparative approach

Hybrid Modeling for Soft Sensing of Molten Steel Temperature in LF

Ecological models are useful in modeling organizations and their competition over resources. However, the traditional approaches, particularly Blau space models, are restrictive in their dependence on a continuous space. In addition, these models are susceptible to indicating competition in sparsely populated areas of an ecology, resulting in competition being indicated where there are no resources to compete over. To deal with these problems we reconceptualize Blau space into the Hybrid Blau space model, using both a cellular structure to model a wider number of variable types, and probabilistic urn models to simulate competition between organizations. We briefly review the basic concepts of Blau space, demonstrate the issues with traditional Blau space modeling, present a new model referred to as the Hybrid model, and propose several new metrics to describe the behavior of organizations in this new model. A novel data source, attribute data from Parliament Members of the Ukrainian Parliament, are used to illustrate the Hybrid Blau space model.

Hybrid algorithm for short-term forecasting of PM2.5 in China

Cleft lip and palate (CLP) is a prevalent congenital anomaly associated with persistent speech disorders, including articulation errors, resonance imbalances, and nasal emission, despite surgical repair. Access to specialized speech-language pathologist care remains limited, particularly in remote and underserved regions. Telehealth has emerged as a scalable solution, yet systematic evidence on its efficacy, technological reliability, and implementation in CLP-specific speech therapy is lacking. This systematic review critically evaluates the effectiveness of telehealth-based speech interventions in improving articulation, resonance, nasal emission, and intelligibility in children with repaired CLP, while examining technological modalities, feasibility, data security, and barriers to adoption. Following PRISMA 2020 guidelines, we searched PubMed, Scopus, Web of Science, Cochrane Library, and Google Scholar from inception to 30 December 2025 using structured Boolean terms combining CLP, speech therapy, and telepractice. Interventional studies in English reporting paediatric speech outcomes were included. Data were extracted on study design, sample characteristics, intervention delivery, outcomes, and risk of bias using RoB 2.0, ROBINS-I, and SCED tools. Narrative synthesis was applied due to heterogeneity. Eleven studies demonstrated consistent improvements in articulation accuracy (e.g., PCC increases of 15%-30%), resonance, and intelligibility via synchronous (Zoom, WhatsApp) and hybrid platforms. AI-assisted feedback and acoustic optimization enhanced fidelity. High caregiver satisfaction, reduced costs, and continuity during restrictions were key resilience factors. Connectivity issues, audio distortion, and small sample sizes were common limitations. Risk of bias was moderate overall. Telehealth is a clinically effective, feasible, and secure modality for CLP speech rehabilitation, comparable to in-person therapy when technologically optimized. Hybrid models and caregiver integration are recommended. Large-scale RCTs are needed to confirm long-term efficacy and cost-effectiveness WHAT THIS PAPER ADDS: What is already known on this subject Telehealth has been used for speech therapy in various paediatric populations, with evidence of comparable outcomes to in-person care during the COVID-19 pandemic. However, prior to this review, no systematic synthesis existed specifically evaluating telepractice for speech disorders in children with repaired cleft lip and/or palate, despite their unique needs for targeted articulation and resonance intervention. What this paper adds to existing knowledge This review demonstrates that telehealth-based speech therapy consistently improves articulation accuracy, resonance, and intelligibility in children with repaired cleft lip and palate, with gains equivalent to in-person therapy across synchronous and hybrid models. It identifies platform-specific acoustic fidelity (e.g., Google Meet, optimized hardware) and caregiver engagement as critical enablers, while highlighting connectivity and audio distortion as manageable barriers. These findings establish telepractice as a viable, scalable standard for cleft-related speech rehabilitation. What are the potential or actual clinical implications of this study? Clinicians can confidently implement hybrid telepractice models with encrypted platforms and external microphones to maintain treatment continuity, especially in underserved areas. Routine integration of caregiver training and AI-assisted feedback is recommended to enhance adherence and outcomes. Health systems should prioritize low-bandwidth protocols and standardized acoustic calibration to ensure equitable access.

BACKGROUND AND AIM: Recent advances in mobile monitoring offer excellent opportunities to explore the hyperlocal variation of ambient air pollution. One example is Google Street View (GSV) car equipped with high-quality instruments. Here, some challenges are translating on-road pollution levels to the building façades and the scalability of such mapping at large spatial scales, e.g. national scale. Since &amp;#x22;traditional” deterministic modelling has been a reliable method for pollution assessment and mapping, this talk aims to explore whether deterministic modelling, via hybrid modelling, can help address the challenges mentioned above and leverage mobile monitoring for improved pollution mapping. METHODS: Three GSV cars measured hyperlocal levels of nitrogen dioxide (NO2), black carbon (BC) and Ultrafine particles (UFP) on all streets of Amsterdam (N = 46664) and Copenhagen (N = 28499) from October 2018 to March 2020. The measurements were corrected and, among others, compared with pollution estimates from national prediction models, the Danish DEHM-UBM-AirGIS, and the Dutch NSL (National Collaborative Air Quality Programme). Further, model estimates are incorporated with GSV measurements to test and apply hybrid modelling approaches using statistical (e.g. kriging) and machine learning techniques. RESULTS: Overall, Amsterdam&amp;#x27;s measured pollution levels were relatively higher (e.g. median NO2 = 24 µg/m3) than in Copenhagen (median NO2 = 13 µg/m3). In addition, GSV NO2 measurements correlated moderately (Spearman’s r = 0.50) (N = 7004) with the NSL estimates in Amsterdam, whereas in Copenhagen, the Spearman’s correlation (r) was in the range 0.45 – 0.67 (N = 97 and 58234). CONCLUSIONS: High-quality mobile monitoring offers a great way to study the hyperlocal variation of air pollution. Since the modelled vs measured correlation was moderate to slightly high, combining both datasets may better predict external data. The presentation will reflect on hybrid model development. KEYWORDS: Mobile measurements, Google Street View, deterministic modelling, AirGIS, hybrid model

The paper considers a hybrid model to analyze and forecast time series data based on an empirical mode decomposition (EMD) that accommodates complex characteristics of time series such as nonstationarity and nonlinearity. We aggregate IMFs using the concept of cumulative energy to improve the interpretability of intrinsic mode functions (IMFs) from EMD. We forecast aggregated IMFs and residue with a hybrid model that combines the ARIMA model and an exponential smoothing method (ETS). The proposed method is applied to forecast KOSPI time series and is compared to traditional forecast models. Aggregated IMFs and residue provide a convenience to interpret the short, medium and long term dynamics of the KOSPI. It is also observed that the hybrid model with ARIMA and ETS is superior to traditional and other types of hybrid models.

Leasing has been increasingly seen as a viable alternative to traditional business models. In this paper, we consider a manufacturer making decisions on green product design by accounting for the trade‐off between traditional and environmental qualities under three business models, including a pure selling, a pure leasing, and a hybrid model with both selling and leasing. Under leasing, there exists the pooling effect that allows a manufacturer to meet consumer needs with fewer products. Since the pooling effect decreases the marginal cost of production, leasing produces positive incentives to increase product quality. However, the cannibalization effect within the product line distorts the incentives so that the pooling effect only increases the traditional quality rather than the environmental quality. As a result, leasing may have a negative impact on the average environmental quality of products. The manufacturer should make business model choices depending on some factors, including the types of markets, the usage cost, and the pooling effect. In general, when the pooling effect is strong, the manufacturer prefers a leasing or hybrid model to selling but designs products with lower environmental quality than selling. When the pooling effect is weak, the optimal decision should be made depending on the types of markets and the usage cost: in the high‐end (low‐end) market, the manufacturer should adopt a leasing or hybrid model only when the usage cost is high (low); the adoption of leasing or hybrid model can improve the average environmental quality.

Text classification is a classic task in the field of natural language processing. however, the existing methods of text classification tasks still need to be improved because of the complex abstraction of text semantic information and the strong relecvance of context. In this paper, we combine the advantages of two traditional neural network model, Long Short-Term Memory(LSTM) and Convolutional Neural Network(CNN). LSTM can effectively preserve the characteristics of historical information in long text sequences, and extract local features of text by using the structure of CNN. We proposes a hybrid model of LSTM and CNN, construct CNN model on the top of LSTM, the text feature vector output from LSTM is further extracted by CNN structure. The performance of the hybrid model is compared with that of other models in the experiment. The experimental results show that the hybrid model can effectively improve the accuracy of text classification.