Key Phases Research Articles

A machine learning-based framework using the particle swarm optimization algorithm for credit card fraud detection

The detection of fraudulent activities in credit cards transactions presents a significant challenge due to the constantly changing and unpredictable tactics used by fraudsters, who take advantage of technological advancements to evade security measures and cause substantial financial harm. In this paper, we suggested a machine learning based methodology to detect fraud in credit cards. The suggested method contains four key phases, including data normalization, data preprocessing, feature selection, classification. For classification artificial neural network, decision tree, logistic regression, naive bayes, random forest while for feature selection particle swarm optimization is employed. With the use of a dataset created from European cardholders, the suggested method was tested. The experimental results show that the suggested method beats the other machine learning techniques and can successfully classify frauds with a high detection rate.

Redesign of Military Patient Handoff Training for Emergency Medical Services

We conducted a qualitative research inquiry to facilitate the redesign of a military patient handoff training program for a new population of civilian Emergency Medical Services (EMS) learners. Emergency Medical Service providers provided insights about the organizational and interpersonal factors that contribute to patient handoffs, which were used to tailor the existing military patient handoff training for a civilian population. Results revealed four key phases in the patient handoff journey and three topics regarding improvements to the EMS handoff process: importance of interpersonal communication, dedicated time for handoffs, and standardized communication.

Comparison of Structure and Reactivity of Hydrothermally Prepared Bi−Mo−Co−Fe−O Catalysts in Selective Propylene and Isobutene Oxidation

AbstractThe elucidation of structure‐activity correlations in selective oxidation of propylene and isobutene over mixed metal oxides (MMO) is attractive for knowledge‐based catalyst design and process optimisation. Particularly, 4‐component Bi−Mo−Co−Fe−O catalysts need to be studied since their complex metal oxide phase mixture leads to higher activity and selectivity than 2‐component Bi−Mo−O. Hence, three Bi−Mo−Co−Fe‐oxides with different metal ratios were prepared by hydrothermal synthesis and compared during selective oxidation tests with propylene and isobutene. The active phases after several days on stream were investigated by synchrotron X‐ray diffraction (XRD) and Raman spectroscopy, while the structural evolution under reaction conditions was followed by operando Raman spectroscopy, synchrotron XRD, and multi‐edge X‐ray absorption spectroscopy. Similar structural transformations were observed during selective oxidation of propylene and isobutene, with similar influence on catalytic performance. A phase mixture of β‐CoMoO4/β‐Co0.7Fe0.3MoO4, γ‐Bi2MoO6, Fe3O4 and Bi3FeMo2O12 was observed, whereby high amounts of β‐CoMoO4/β‐Co0.7Fe0.3MoO4 increased selectivity to acrolein/methacrolein. In contrast, high amounts of γ‐Bi2MoO6 and Fe3O4 favoured total oxidation to CO and CO2. The simultaneous presence of β‐CoMoO4/β‐Co0.7Fe0.3MoO4, Bi3FeMo2O12 and Fe2O3 increased selectivity to methacrolein in isobutene oxidation, whereas no comparative increase in acrolein selectivity was observed in propylene oxidation. This suggests different key active phases in both reactions.

Low and very low birthweight disadvantage in compulsory education achievement and the transition to upper secondary education in the Finnish birth cohorts of 1987 to 1997.

We compared the educational achievements of very low-birthweight (VLBW) and low-birthweight (LBW) adolescents (ages 16 to 19) to those of their normal-birthweight (NBW) peers in the complete Finnish birth cohorts of 1987 to 1997. We focused on three key phases of the education process: the end of compulsory education (9th-grade completion), and the transition to and the completion of upper secondary-level education. We used register data on grades, educational transitions and completed education. We employed multiple indicators on the progression of the education process and estimated population-level and within-families linear probability (LPM) models with robustness checks at the population level using logistic regression. We tested whether parental education and the child's sex modify the association between (V)LBW and educational achievement. Results of both descriptive analysis and the population-level and within-family LPM models indicate that (V)LBW is associated with an increased risk of not being able to keep up with the normative education process and to compete for upper secondary education study places at the end of compulsory education. The modifying effect of parental education was robust, whereas that of the child's sex was not. Among (V)LBW students who were able to keep up with the normative education process, (V)LBW was not associated with a lower grade point average or with a meaningfully lower probability of completing upper secondary education by the normative age. The upper secondary-level educational choices and achievements of the children born with (V)LBW who managed to complete the standard compulsory education curriculum and complete the transition to upper secondary-level education within the expected time did not, in essence, differ from those of the NBW children. Some specific characteristics of the Finnish education system likely contributed to these results, such as the grading at compulsory education being only relatively loosely standardized.

Transfer Learning with EfficientNetV2S for Automatic Face Shape Classification

The classification of human face shapes, a pivotal aspect of one’s appearance, plays a crucial role in diverse fields like beauty, cosmetics, healthcare, and security. In this paper, we present a multi-step methodology for face shape classification, harnessing the potential of transfer learning and a pretrained EfficientNetV2S neural network. Our approach comprises key phases, including preprocessing, augmentation, training, and testing, ensuring a comprehensive and reliable solution. The preprocessing step involves precise face detection, cropping, and image scaling, laying a solid foundation for accurate feature extraction. Our methodology utilizes a publicly available dataset of female celebrities, comprising five face shape classes: heart, oblong, oval, round, and square. By augmenting this dataset during training, we magnify its diversity, enabling better generalization and enhancing the model’s robustness. With the EfficientNetV2S neural network, we employ transfer learning, leveraging pretrained weights to optimize accuracy, training speed, and parameter size. The result is a highly efficient and effective model, which outperforms state-of-the-art approaches on the same dataset, boasting an outstanding overall accuracy of 96.32%. Our findings demonstrate the efficiency of our approach, proving its potential in the field of face shape classification. The success of our methodology holds promise for various applications, offering valuable insights into beauty analysis, cosmetic recommendations, and personalized healthcare.

Norwegian physical education teachers’ assessment after the introduction of a new curriculum – LK20

ABSTRACT In 2020, a new curriculum, Kunnskapsløftet 2020 (LK20), was introduced in Norwegian schools. This study investigated how physical education (PE) teachers in upper secondary schools have changed their assessment practice following the introduction of LK20. To achieve this, nine individual in-depth interviews were conducted with PE teachers from six upper secondary schools in Norway. Using NVivo 12, the teachers’ experiences were analysed, and corresponding categories of meaning were created. The results indicated that there were changes in their assessment practice after the introduction of LK20. The PE teachers found it more difficult in LK20 to distinguish between grades, and greater room for interpretation was identified. The results also suggested that PE teachers have become more concerned with the pupils’ development over time, rather than with achieving a specific, predetermined goal, and that LK20 better facilitates this aim. PE teaching also seemed to be characterised by a greater amount of play after the introduction of LK20, which teachers viewed as being positive for most pupils. Finally, the analysis pointed to a need in the teachers for more professional support outside of their own immediate circle of colleagues. The results are discussed in relation to Fullan (2016) and one of his key phases according to education policy-driven reforms – the implementation phase. Our findings, and those of other research, identify a need for skills development work and/or courses for PE teachers in assessment.

AMP-activated protein kinase activation suppresses leptin expression independently of adipogenesis in primary murine adipocytes.

Adipogenesis, defined as the development of mature adipocytes from stem cell precursors, is vital for the expansion, turnover and health of adipose tissue. Loss of adipogenic potential in adipose stem cells, or impairment of adipogenesis is now recognised as an underlying cause of adipose tissue dysfunction and is associated with metabolic disease. In this study, we sought to determine the role of AMP-activated protein kinase (AMPK), an evolutionarily conserved master regulator of energy homeostasis, in adipogenesis. Primary murine adipose-derived stem cells were treated with a small molecule AMPK activator (BI-9774) during key phases of adipogenesis, to determine the effect of AMPK activation on adipocyte commitment, maturation and function. To determine the contribution of the repression of lipogenesis by AMPK in these processes, we compared the effect of pharmacological inhibition of acetyl-CoA carboxylase (ACC). We show that AMPK activation inhibits adipogenesis in a time- and concentration-dependent manner. Transient AMPK activation during adipogenic commitment leads to a significant, ACC-independent, repression of adipogenic transcription factor expression. Furthermore, we identify a striking, previously unexplored inhibition of leptin gene expression in response to both short-term and chronic AMPK activation irrespective of adipogenesis. These findings reveal that in addition to its effect on adipogenesis, AMPK activation switches off leptin gene expression in primary mouse adipocytes independently of adipogenesis. Our results identify leptin expression as a novel target of AMPK through mechanisms yet to be identified.

Can Machine Learning Predict Running Kinematics Based on Upper Trunk GPS-Based IMU Acceleration? A Novel Method of Conducting Biomechanical Analysis in the Field Using Artificial Neural Networks

This study aimed to investigate whether running kinematics can be accurately estimated through an artificial neural network (ANN) model containing GPS-based accelerometer variables and anthropometric data. Thirteen male participants with extensive running experience completed treadmill running trials at several speeds. Participants wore a GPS device containing a triaxial accelerometer, and running kinematics were captured by an 18-camera motion capture system for each trial. Multiple multilayer perceptron neural network models were constructed to estimate participants’ 3D running kinematics. The models consisted of the following input variables: 3D peak accelerometer acceleration during foot stance (g), stance time (s), running speed (km/h), participant height (cm), leg length (cm), and mass (kg). Pearson’s correlation coefficient (r), root mean squared error (RMSE), and relative root mean squared error (rRMSE) showed that ANN models provide accurate estimations of joint/segment angles (mean rRMSE = 13.0 ± 4.3%) and peak segment velocities (mean rRMSE = 22.1 ± 14.7%) at key gait phases across foot stance. The highest accuracies were achieved for flexion/extension angles of the thorax, pelvis, and hip, and peak thigh flexion/extension and vertical velocities (rRMSE < 10%). The current findings offer sports science and medical practitioners working with this data a method of conducting field-based analyses of running kinematics using a single IMU.

High-resolution species distribution modelling reveals spatio-temporal variability of habitat suitability in a declining grassland bird

Abstract Context Species distribution models (SDMs) may provide accurate predictions of species occurrence across space and time, being critical for effective conservation planning. Objectives Focusing on the little bustard (Tetrax tetrax), an endangered grassland bird, we aimed to: (i) characterise the drivers of the species distribution along its key phenological phases (winter, breeding, and post-breeding); and (ii) quantify spatio-temporal variation in habitat suitability across phenological phases and over the years 2005–2021. Methods Combining remotely sensed metrics at high temporal resolution (MODIS) with long-term (> 12 years) GPS telemetry data collected for 91 individuals at one of the species’ main strongholds within the Iberian Peninsula, we built SDMs (250 m resolution) for the species key phenological phases. Results The use of both dynamic and static predictors unveiled previously unknown ecological responses by little bustards, revealing a marked change in the spatial distribution of suitable habitat among phenological phases. Long-term habitat suitability trends showed considerable fluctuations, mainly in the breeding and post-breeding phases. Overall, SDM projections into the past revealed that while the species’ winter and post-breeding habitats apparently increased since 2005, suitable habitat during the species’ most critical phenological phase, breeding, apparently reduced in area over time. Conclusions Our findings show that matching remotely sensed data with GPS tracking data results in accurate habitat suitability predictions throughout the yearly cycle. Additionally, our findings stress the importance of quantifying habitat loss and its potential impact on little bustard decline over nearly 20 years. Spatio-temporal variations in habitat suitability are also identified in this work, which can help prioritize conservation areas, particularly the breeding areas that have remained stable over time, as this is a key requirement for little bustard lek breeding system.

The development of inclusive competencies of the future specialist in physical culture and sports in the pedagogical process of the university

The relevance of the research is conditioned by the need for a detailed study of the theory and practice of the development of inclusive competencies in the professional training of school teachers. In this sense, this work aims to identify the main stages of the development of inclusive competencies among teachers, develop methodological recommendations aimed at facilitating the process of developing inclusive competencies among teachers, determine the stages of implementation of the technology for the development of inclusive competencies for the training of school teachers, describe theoretical and practical aspects in the development and implementation of competencies, as well as describe the priorities of using inclusive approaches in working with children. The main research methodology of this work is a comprehensive analysis, the purpose of which is to study the theoretical and practical characteristics of the development of inclusive competencies among teachers in the context of a complex structure of higher pedagogical education and the school educational system. Furthermore, the following research methods were involved in writing this article: logical reasoning, comparative, synthesis and deduction, classification. The article details the findings of the analysis, points out the key phases in the evolution of inclusive competencies in teachers, offers methodological advice to improve this development, establishes the phases of application of technologies in the formation of inclusive competencies for educators, addresses theoretical and practical aspects of these competencies, and highlights the importance of inclusive approaches when working with children. The materials of the document are of practical value for teachers and students of pedagogical universities, teachers and other employees of secondary schools and schools for children with special needs. Keywords: inclusive competencies; specialist; physical Culture; sports; pedagogical process; university.

Circular e‐waste ecosystems in necessity‐driven contexts: The impact of formal institutional voids

AbstractRecent studies suggest an ecosystem view is most appropriate for understanding the impact of institutional voids on entrepreneurial activities. Expanding the focus to the entrepreneurial ecosystem is crucial for tackling complex low‐ and middle‐income countries' (LMICs) environmental challenges, as the transition to circular e‐waste management, which are heavily impacted by the institutional environment. Moreover, most entrepreneurs in LMICs engage in circular practices out of economic necessity, rather than environmental reasons. However, scholars have overlooked the effects of institutional voids on entrepreneurial ecosystems' potential to grow and realize environmental benefits. This study investigates how formal institutional voids impact the evolution of circular ecosystems in necessity‐driven contexts in LMICs. We study the evolution of a circular e‐waste ecosystem in Kenya and uncover two key phases: emergence and growth. We show that formal institutional voids' impact differs across phases; regulatory and contract enforcement voids create opportunities for informal and private actors to establish e‐waste collection, repair, remanufacture, and recycle businesses, thereby facilitating the emergence of the circular ecosystem. However, labor market, capital market, product market, regulatory, and contract enforcement voids hamper its growth. Specifically, the lack of skills, equipment, guidelines, and financial incentives hinders advanced repairs and remanufactures and limits recycling opportunities, while the lack of formal e‐waste infrastructure increases illegal dumping and negatively impacts human health and the environment. We develop propositions and a phase model to explain the impact of formal institutional voids on the evolution of circular e‐waste ecosystems and the realization of environmental benefits in necessity‐driven contexts.

Assessment of the challenges influencing the adoption of smart building technologies

Introduction: Over the past few decades, there has been an increasing focus on Smart Building Projects (SBP) and the technologies associated with them. Numerous studies have been conducted globally to define smart building technologies (SBT), identify challenges, and explore areas for improvement. This study aims to examine the concept and terminology of SBT and the expertise and experience of participants in SBP in the Arab Gulf countries, specifically Saudi Arabia. The study also investigates the challenges faced by SBT throughout its life cycle.Methods: To identify and assess the challenges affecting the adoption of smart building technologies. This approach included a literature review, pilot-testing, and a questionnaire survey. The survey targeted a sample of 90 architects/engineers, managers, and contractors.Results: A total of 55 challenges were identified and categorized into four groups, corresponding to the key phases of the project life cycle. These phases include the programming and feasibility analysis phase, design phase, installation and commissioning phase, and operation and maintenance phase. The findings of this research expand the body of knowledge by providing architects/engineers, managers, and contractors in the architecture, engineering, construction, and facility management (AEC/FM) industry with insights into the influential challenges related to the adoption of SBT. In conclusion, this study sheds light on the concept and terminology of smart building technologies and explores the challenges faced by SBT during its life cycle. By identifying and categorizing these challenges, the study provides valuable information to AEC/FM practitioners, enabling them to overcome obstacles and improve the adoption of SBT.

Mineralisation of CO2 in wood biomass ash for cement substitution in construction products

This study extends our exploration of the potential of biomass ashes for their CO2-reactivity and self-cementing properties. The ability of three hardwood-based biomass ashes to mineralise CO2 gas and partially replace CEM I in mortars was investigated. The three hardwoods were English oak (Quercus rober), English lime (Tilia x europaea), and beech (Fagus sylvatica). The woody biomass wastes were incinerated at 800°C to extract their key mineral phases, which are known to be reactive to CO2 gas to form carbonates. The selected biomass ashes were analysed for their CO2-reactivity, which was in the range of 32–43% (w/w). The ashes were used to replace CEM I at 7 and 15% w/w and this “binder” was mixed with sand and water to produce cylindrical monolithic samples. These monoliths were then carbonated and sealed cured over 28 days. The compressive strength, density and microstructure of the carbonate-hardened monoliths were examined. The ash-containing monoliths displayed mature strengths comparable to the cement-only reference samples. The CO2 uptake of oak containing monoliths was 7.37 and 8.29% w/w, for 7 and 15% ash substitutions, respectively. For beech and English lime they were 4.96 and 6.22% w/w and 6.43 and 7.15% w/w, respectively. The 28 day unconfined compressive strengths for the oak and beech ashes were within the range of ~80–94% of the control, whereas lime ash was 107% of the latter. A microstructural examination showed carbonate cemented sand grains together highlighting that biomass ash-derived minerals can be very CO2 reactive and have potential to be used as a binder to produce carbonated construction materials. The use of biomass to energy ash-derived minerals as a cement replacement may have significant potential benefits, including direct and indirect CO2 emission savings in addition to the avoidance of landfilling of these combustion residues.

ELSSIE: A compact stereo spectral imager for planetary surface morphology and composition

Here we describe the design, prototyping, testing, and simulations that were conducted to demonstrate the technology for a concept of the next generation landed planetary spectral imager, the Europa Lander Stereo Spectral Imaging Experiment (ELSSIE). The concept was developed originally for a Europa Lander mission, but the design is applicable, with simplifications, to any ocean world of the outer solar system or to non-icy bodies, including Enceladus, the Moon, Mars, or the surface of Ceres. ELSSIE's design consists of two subassemblies. A Sensor melds a high-resolution, 20-filter, 0.4–3.65 μm, adjustable-focus multispectral stereo imager with a 0.8–3.6 μm point spectrometer, sharing a radiation-shielded single Teledyne H2RG 2048 × 2048 pixel focal plane array (FPA). Each camera includes two 6-position filter wheels with 5 filters and a blank position, providing 10 bandpasses for each of the 2 stereo eyes, and uses 700 × 700 pixels of the FPA. The point spectrometer uses a 6 ×350 pixel strip of the FPA. The Sensor provides stereo and imaging/spectroscopic measurements of reflected light from visible to medium wave-infrared (MWIR) wavelengths to characterize surface morphology, search for pyroclastic plumes, search for organics, identify salts and possible biominerals, characterize crystalline vs. amorphous ice and ice grain sizes, and map the distributions of key phases. In addition to addressing important geologic questions, these measurements support selection of a site for in situ sampling and analysis. A Data Processing Unit (DPU) performs mitigation of radiation that penetrates the shielding using sets of same-filter image frames or spectra of a single spot by removing image spatial pixels with radiation hits, and coadding the remainder for the same spatial pixel, improving signal-to-noise ratio (SNR). The DPU also performs onboard calibration of imager and spectrometer data, co-registration of multispectral images, and calculation of spectral index (“summary parameter”) images for efficient use of lander downlink. Co-registered multispectral image sets and spectra are retained onboard and can be downlinked upon query.

A ResNet-LSTM hybrid model for predicting epileptic seizures using a pretrained model with supervised contrastive learning

In this paper, we propose a method for predicting epileptic seizures using a pre-trained model utilizing supervised contrastive learning and a hybrid model combining residual networks (ResNet) and long short-term memory (LSTM). The proposed training approach encompasses three key phases: pre-processing, pre-training as a pretext task, and training as a downstream task. In the pre-processing phase, the data is transformed into a spectrogram image using short time Fourier transform (STFT), which extracts both time and frequency information. This step compensates for the inherent complexity and irregularity of electroencephalography (EEG) data, which often hampers effective data analysis. During the pre-training phase, augmented data is generated from the original dataset using techniques such as band-stop filtering and temporal cutout. Subsequently, a ResNet model is pre-trained alongside a supervised contrastive loss model, learning the representation of the spectrogram image. In the training phase, a hybrid model is constructed by combining ResNet, initialized with weight values from the pre-trained model, and LSTM. This hybrid model extracts image features and time information to enhance prediction accuracy. The proposed method’s effectiveness is validated using datasets from CHB-MIT and Seoul National University Hospital (SNUH). The method’s generalization ability is confirmed through Leave-one-out cross-validation. From the experimental results measuring accuracy, sensitivity, and false positive rate (FPR), CHB-MIT was 91.90%, 89.64%, 0.058 and SNUH was 83.37%, 79.89%, and 0.131. The experimental results demonstrate that the proposed method outperforms the conventional methods.

Contagion effects of permissionless, worthless cryptocurrency tokens: Evidence from the collapse of FTX

This paper investigates the price discovery relationships between FTT Token, issued by the cryptocurrency exchange FTX, and a set of assets and liabilities held by FTX amid a period of catastrophic financial decline by applying novel information flow measurement techniques. Results indicate that during key phases associated with the collapse of FTX, FTT Token had an informational lead over multiple assets, including cryptocurrencies such as Ethereum. Furthermore, we identify significant interactions between the FTT Token and both Robinhood shares and the token Serum, raising concerns about the direct influence of permissionless, technically valueless tokens on other assets and the potential challenges to market stability and investor protection. Our findings underscore the need for stronger policy-making, regulatory, and ethical considerations in cryptocurrency markets.

Intelligent Prediction of Network Security Situations based on Deep Reinforcement Learning Algorithm

The limitations of traditional network security assessment methods characterized by manual definitions and measurements, data overload, poor performance, and non-negligible drawbacks are addressed in this research. A novel network security system employing a deep learning algorithm is proposed to overcome these challenges. The research unfolds in three key phases. First, a deep self-encoding model is developed to distinguish various network attacks effectively. Subsequently, the creation of missing measurement weights enhances pattern detection, even when dealing with a limited number of training samples. Finally, the model assesses and computes attack issues, assigns impact scores to each attack, and determines the overall network security value. Experimental results demonstrate that the deep auto encoder-based deep neural network (DAEDNN), in conjunction with the proposed unique oversampling weighting (UOSW) algorithm, significantly outperforms traditional methods such as decision trees (DT), support vector machines (SVM), and long short-term memory (LSTM) models. The F1 score of UOSW surpasses these models by approximately 2.77, 10.5, and 5.2, respectively. The deep self-encoding model employed in the proposed system offers superior accuracy and recall rates, leading to more precise and efficient measurement results.

Tackling Noise: Technology Integration for Improved Noise Pollution Monitoring.

Noise pollution derived from motor vehicles, especially those with noisy engines, constitutes a significant challenge in urban environments, impacting quality of life and public health. In this context, the present study focuses on the implementation of an advanced technological solution: a web-based system specifically designed to monitor and manage noise pollution in Tarapoto, Peru. The execution of this project was structured in 4 key phases. The first phase involved extensive data collection and definition of the project scope. In the second phase, the system architecture was established based on initial observations. The third phase focussed on evaluating and improving system functionalities, while the fourth involved rigorous testing and corrections. The successful implementation of this system led to a significant improvement in the noise pollution monitoring process, reducing reporting times and improving data-driven decision making. This project not only provides a valuable tool for authorities in the management and mitigation of noise pollution, but also encourages community awareness and active participation in the sustainable management of the urban acoustic environment.

3D kinematic analysis of salto backward in acrobatic rock and roll: a case study

Background and Study Aim. The aim of the study was to determine and compare basic kinematic characteristics in the key phases of the acrobatic element Salto backward tucked. Material and Methods. The research sample was created from two couples of the acrobatic rock and roll category Main class free style. The time, spatial, and time-spatial characteristics for both couples were determined based on a three-dimensional kinematic analysis. Measurements were carried out with Simi Motion 3D. The acrobatic element Salto backward tucked was chosen based on the fact that it is usually learned as the first overturn acrobatic element with the airborne phase. Results. In the time duration of the acrobatic element, a very similar duration was observed for both couples (CZ couple: 3.59 s and SK couple 3.55 s). The maximal height of the center of gravity the partner from the Czech Republic reach 3.54 m and partner from the Slovak Republic 3.35 m. The biggest difference noticed was between the male partners 14.59° in knee angle in the staff phase. Among the female partners, a difference was noted in the angle between the torso and the vertical plane at the moment of the end of the vertical displacement, when SK female partner has twice as much deviation of the torso, by exactly 19.49°. In the staff phase, differences in acceleration were also observed, when the male partner (16.56 m.s-2) and female partner (25.08 m.s-2) from the Slovak Republic achieved higher maximum acceleration. In the airborne phase was the biggest difference between female partners in the culmination point, namely in the angle in the knee joint (35.98°) and the angle between the torso in the vertical plane at the moment of opening the somersault (25.52°). Conclusions. The findings showed similarities and differences in various aspects of the performance, providing valuable insights into the kinematic structure of this acrobatic element in acrobatic rock and roll. Our study should contribute to the extension of knowledge of the kinematic structure of acrobatic element in acrobatic rock and roll.

BISINDO Sign Language Recognition: A Systematic Literature Review of Deep Learning Techniques for Image Processing

This study uses the Systematic Literature Review (SLR) method to examine the application of deep learning techniques focusing on BISINDO (Bahasa Isyarat Indonesia) image recognition. This is crucial for enhancing communication accessibility for the hearing-impaired community. The SLR process involves three stages: planning, conducting, and reporting. During the planning stage, research topics, questions, and search criteria are established, while the conducting stage involves comprehensive article retrieval and rigorous filtering. In the reporting stage, the study highlights the significance of various deep learning methodologies, including the implementation of several algorithms that ace in image recognition. For example, Convolutional Neural Network (CNN), Long Short Term Memory (LSTM), and a combination of the two methods, each with unique advantages and limitations. Along with that, this paper aims to find gaps in previous research and act as a guide for future deep learning model development. Moreover, the research outlines the development of a high-performance model, emphasizing key phases such as image augmentation and data preprocessing, as well as model optimization. These efforts contribute to a better understanding of BISINDO image recognition, offering valuable insights for researchers and practitioners aiming to support easier accessibility and communication for the hearing-impaired community through advanced deep learning approaches.