Cloud computing offers cost efficiency, flexibility, and scalability, but efficient resource managemen, especially dynamic scaling, remains a key challenge. This paper presents a predictive autoscaling strategy using transformer-based workload prediction model integrated with Kubernetes-based Event Driven Autoscaler (KEDA) to dynamically scale an ingress controller based on anticipated workloads. Trained on real-time data from Alibaba Cloud, the transformer model was deployed in a cloud-native Kubernetes environment. Performance was evaluated through load testing against standard autoscaling methods, including Horizontal Pod Autoscaler (HPA). Results show that the proposed model maintains stable response times and improves resource utilization, with further gains possible through hyperparameter tuning. The study demonstrates that transformer-based predictive autoscaling can reduce resource wastage and mitigate performance degradation during traffic surges, offering a promising direction for future cloud resource management solutions.
Anomaly detection is essential for ensuring the safe and efficient operation of industrial systems like smart grids. Smart grid stations handle sensitive data and are often hesitant to share it with third-party servers for centralized anomaly detection. Federated Learning (FL) offers a viable solution to this issue by enhancing anomaly detection in smart grids without compromising data privacy. We present a method for developing an unsupervised anomaly detection system using FL applied to a synthetic dataset that mimics a real-world smart grid system's behavior. We focus on utilizing FL's long short-term memory autoencoder in short, LSTM-AE for anomaly detection. However, there are concerns about potential privacy breaches in the FL system. Hence, to address this issue, we propose to integrate differential privacy (DP) with FL for anomaly detection by adding artificial noise to parameters at the client side before aggregation. This method ensures data privacy while maintaining the convergence of federated learning algorithms. Moreover, this research determines the optimal privacy level to balance noise scale and model accuracy. Our findings suggest a criterion for selecting the right privacy budget of DP based on the requirement of the system to provide good level of privacy in the system while maintaining the f1-score of FL-based anomaly detection system greater than 90%.
Latency, jitter, and reliability are three KPIs in QoS of modern edge and cloud computing systems. Their analytical insights can be well gained through queuing theory. In this paper, we develop a queuing model with two parallel heterogeneous servers with a randomly-initial-dispatch discipline. This model generalizes several existing models, yet simple enough as it can be configured by a couple of parameters. With this simplicity, we analytically prove the optimality of several standard measures of effectiveness for the latency and jitter.
In the realm of Database-as-a-Service (DBaaS), multi-tenancy has emerged as a key concept, enabling a single database instance to be shared among multiple tenants. Within a multi-tenant parallel DBMS, at any point, many tenants may submit SQL queries concurrently, each associated with a performance Service-Level Objective (SLO). Any violation to these SLOs requires the provider to pay fines. Given the instability of query workload in these systems, along with the need to satisfy the strict performance objectives, innovative query optimization strategies are required. These strategies should be well adapted to the specific needs of multi-tenant environments while ensuring profitability for the service provider. In this paper, we design a workload prediction management system (or Prediction Manager in short) inside the query optimizer. The main objective of this manager is to continuously predict the upcoming workload and make this information available to the optimizer. This enables the query optimizer, along with the performance SLOs, to decide when and which execution plan to select. Experimental results showed high efficiency in terms of the provider's long-term profit relatively to those in existing literature.
The rapid increase in data volume over the past years has given rise to big data science. Distributed File Systems (DFS) have become widely employed to handle this vast amount of data, such as Google File System and Hadoop Distributed File System (HDFS). The primary objective of a DFS is to ensure data availability and system reliability in the event of failure. Data availability and system reliability are achieved by replicating files across multiple locations, which, however, results in the consumption of storage space and other resources. The significance of these files varies based on their frequency of use within the system. As a result, specific files are deemed less critical and do not warrant extensive replication, as they hold little importance in the overall system. This paper presents a novel approach called "Optimizing HDFS Replication Policies with Predictive Categorization" for storage efficiency. This approach aims to minimize storage consumption while ensuring data availability and system reliability. Experimental results using the Spotify Hit Predictor Dataset (1960-2019) showcase significant improvements in storage utilization and throughput. This approach not only enhances system performance but also positions organizations for substantial cost efficiencies in managing their data-intensive workloads.
Cloud and big data applications increasingly rely on the softwarization of underlying infrastructure to achieve highly flexible reconfiguration. Software Defined Networking (SDN) stands as a vital component, offering centralized control and efficient orchestration of network resources. This paper underscores the critical importance of security within cloud and big data environments and highlights SDNs role in this context, particularly through the implementation of Transport Layer Security (TLS) encryption. TLS establishes secure communication channels between SDN controllers and switches. While prior studies have noted potential latency impacts of TLS, this paper presents contrasting findings within the context of SDN switches widely used in cloud infrastructure. Through a comprehensive examination, this study shows the pivotal role TLS plays in fortifying the integrity and security of SDN communications. Technical insights into TLS integration, its impact on performance, and practical case studies offer valuable guidance for enhancing the security of SDN networks which play a key role in cloud infrastructures and big data applications.
Wind forces on structures have the potential to cause significant damage. A database involving the distance from the ridge, enclosure classification, surface type, elevation above ground level, wind direction, basic wind speed, presence of wall/surface openings, and effective net wind pressure (ENWP) was created using computation fluid dynamics (CFD). This paper focuses on the development of a model for predicting ENWP using a backpropagation-artificial neural network (BP-ANN). Utilizing the Levenberg-Marquardt algorithm (LMA) and hyperbolic tangent sigmoid function (HTSF) as the model hyperparameters, the study investigated several network structures and the simulations revealed that the 7-20-1 is the best model among the topologies observed in this study. The results showed an R value of 0.99868, MSE and MAPE of 0000749 and 5.036%, respectively. Additionally, the Akaike Information Criterion (AIC) was used as another layer of metric to measure the effectiveness of the model. The least was observed in the 7-20-1 network structure indicating that this is the best among the topologies observed in this study. Moreover, a sensitivity analysis (SA) through Garson's Algorithm (GA) was performed to determine the relative contribution (RC) of the input parameters (IP) including the distance from the ridge, enclosure classification, surface type, elevation above ground level, wind direction, basic wind speed, and presence of wall/surface opening to the effective net wind pressure. The findings presented that the basic wind speed is the most significant parameter to the effective net wind pressure value. The results of this study can be utilized in considering appropriate configuration to minimize the effects of wind pressure in structures.
With the rapid advancements in medical imaging technologies such as CT, MRI, PET, and ultrasound, these modalities have become pivotal for precise clinical diagnoses and treatment planning. Notably, they facilitate accurate patient evaluations and expedite early disease detection and targeted interventions. However, the voluminous nature of medical image data, coupled with stringent privacy requirements, poses significant challenges to efficient and secure storage and sharing within the realm of big data management. In light of these complexities, this paper introduces a blockchain-centric framework that innovatively integrates InterPlanetary File System (IPFS), cloud storage, AES and blockchain technologies. The framework is meticulously designed to address the dual objectives of secure storage and streamlined sharing of medical image data. A key novelty lies in the strategic differentiation of cloud service roles: the cloud's “chain-on” nodes participate in blockchain's validation and consensus processes without engaging in direct data management or operations, whereas “chain-off” nodes are tasked with establishing transient channels for data transmission upon successful validation of user requests by their chain-on counterparts. In practice, medical image owners encrypt the data using IPFS and create indexes derived from examination reports, facilitating keyword-based searches for data users. Moreover, a Hybrid Encryption scheme with Multiple Public Keys is implemented, necessitating users to utilize their blockchain management system-generated private keys to unlock the AES decryption key for image data access. This multi-faceted technology integration approach not only fortifies the security of storage and ensures privacy preservation but also streamlines retrieval processes and bolsters permission management for medical image data. Therefore, it presents a pioneering solution for medical image data storage and sharing.
Data has become an asset for companies, originating from various sources, such as IoT paradigms. It is crucial to safeguard its life cycle using suitable, scalable, and effective technologies, like those enabled by cloud computing models. However, in order to extract value from this data, complementary processes of collection, refinement, cleaning, or modeling, among many others, are required. Furthermore, organizations greatly vary in their methodologies and approaches to handling data, which further emphasizes the need for standardized techniques. In this regard, data management methodologies promote the adoption of the various dimensions of data quality in order to ensure the reliability of data across different systems and processes. The main contribution of this manuscript is the proposal of a new data quality dimension, coined purity, to measure the importance of the data in a processing pipeline topology. As a result, organizations can better guarantee the quality of their datasets in order to raise the success of data-driven endeavors within organizations. The proposed methodology is validated in an urban mobility use case.
Childhood obesity is a growing public health concern worldwide, with significant implications for long-term health outcomes and healthcare costs. Early identification of children at risk of obesity is essential for implementing effective prevention and intervention strategies. In recent years, machine learning techniques have emerged as powerful tools for predicting childhood obesity risk based on various predictors such as demographic factors, dietary habits, physical activity levels, and genetic predisposition. This paper explores the use of temporal deep-learning models, specifically recurrent neural networks and long short-term memory, to predict obesity childhood stages using longitudinal datasets. The long short-term memory model emerges as the top performer, achieving an accuracy of 85% and an F1-score of 0.85, surpassing existing models. To enhance model transparency, model-agnostic methods are employed, revealing critical insights into model predictions. Results indicate that BMI at age 11 is the most influential feature in predicting obesity. These findings underscore the importance of applicability models and explainable machine learning methods in understanding and predicting childhood obesity progression.
Melanoma is a lethal skin cancer that is increasingly threatening the public health system due to increased incidence rates and mortality rates. Early detection of the disease is vital for improved outcomes and the reduction of mortality rates. Skin cancer classification remains a challenging task in the field of dermatology. While self-attention mechanisms and large language models have gained traction in skin cancer detection research, there is still insufficient evidence demonstrating their superior performance compared to CNNs. Thus, further exploration of this area is warranted. Where the quest for the optimal CNN pretrained model persists. In this study, we address this gap by assessing various pretrained models to determine the most effective one for skin cancer classification. Additionally, we introduce a novel approach that leverages transfer learning to develop a multi-task model capable of providing more comprehensive prediction information from dermatological images. Unlike conventional single output classification tasks that rely solely on label prediction, our proposed model utilizes transfer learning techniques to extract valuable features from pretrained models, enhancing its ability to predict multiple tasks simultaneously. This novel approach not only advances the field of dermatology by improving classification accuracy but also meets the growing demand for more informative predictions in clinical settings.
Dementia is a disease that imposes medical, social, and economic challenges on medical professionals, caregivers, and the patients themselves. Dementia monitoring and prognosis are critical factors besides dementia diagnosis. However, recent studies on dementia prognosis involve people with diagnosed dementia and not non-demented with a high risk of being demented, like people with cognitive difficulties. The aim of this paper is to use Machine Learning (ML) algorithms to predict patients with the risk of deterioration from medical histories containing clinical, cognitive, and profile data collected from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. The best model was the Random Forest model with a sensitivity of 0.79, accuracy of 0.77, specificity of 0.76, F1-score of 0.78, and an AUROC of 0.83. Moreover, the model was interpreted through permutation importance. Using the permutation importance tool, the study highlighted the strong effect of diagnosis information and specific symptoms like muscle pain for dementia deterioration prediction.
The digitization of production is an essential component of Industry 4.0. The collection of data in the production process is a decisive factor and a prerequisite for the use of artificial intelligence. At machine level, groundbreaking successes have been achieved in the field of predictive maintenance using machine learning, which is directly reflected in increased productivity and cost savings. In contrast, digitization at item level, i.e. the digital tracking of all items in the process context that ultimately lead to the finished product, is much more complex. A non-structured collection of data at item level prevents the use of artificial intelligence, for example in the quality testing of products. As part of the RECAST project, the extent to which existing non-networked production facilities can be digitized with this objective was investigated. As part of the work, a framework was created using exclusively open source components, which allows a digital representation of various real existing complex production processes in the aerospace industry to be created. With this digital twin, the system data can be recorded in real time on a item-specific basis. Thanks to the generic approach to process mapping, the results can be applied to a wide range of application fields in mechanical engineering and data-intensive administrative tasks.
This study compared and analyzed the usability, performance, and cost-efficiency of deploying Machine Learning (ML) models in two ML-AI platforms in Google Cloud: BigQuery ML and Vertex AI. Through the experiments with two separate cases, the analysis was conducted with MIMIC-IV datasets of hospitalized patients to deploy regression models on each platform to predict mortality and progression of diseases. The documentation, learning curve, and resource suitability of the platforms were evaluated to access their usability. The study evaluated the total running times and resource utilizations, including storage and compute, to analyze their performance and cost efficiency. The analysis results showed that BigQuery ML offers good usability with easy-to-follow documentation and a moderate learning curve for cloud users, making it more suitable for SQL-savvy users and large-scale data analytics tasks. It also showed efficient resource management and deployment despite its higher initial processing times during the training. Vertex AI incurred higher costs due to longer training times and specific resource allocations. The findings indicate that BigQuery ML seems to be more efficient, particularly in terms of processing time and cost for the experimented clinical dataset and regression models, emphasizing its suitability for large-scale data processing tasks where efficiency is essential.
The stock market's ascent typically mirrors the flourishing state of the economy, whereas its decline is often an indicator of an economic downturn. Therefore, for a long time, significant correlation elements for predicting trends in financial stock markets have been widely discussed, and people are becoming increasingly interested in the task of financial text mining. The inherent instability of stock prices makes them acutely responsive to fluctuations within the financial markets. In this article, we use deep learning networks, based on the history of stock prices and articles of financial, business, technical news that introduce market information to predict stock prices. We illustrate the enhancement of predictive precision by integrating weighted news categories into the forecasting model. We developed a pre-trained NLP model known as FinBERT, designed to discern the sentiments within financial texts. Subsequently, we advanced this model by incorporating the sophisticated Long Short Term Memory (LSTM) architecture, thus constructing the innovative FinBERT-LSTM model. This model utilizes news categories related to the stock market structure hierarchy, namely market, industry, and stock related news categories, combined with the stock market's stock price situation in the previous week for prediction. We selected NASDAQ-100 index stock data and trained the model on Benzinga news articles, and utilized Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Accuracy as the key metrics for the assessment and comparative analysis of the model's performance. The results indicate that FinBERT-LSTM performs the best, followed by LSTM, and DNN model ranks third in terms of effectiveness.
ABSTRACT The occurrence of financial crisis in a corporation is a gradual and cumulative process. The use of short-duration data does not allow visualization of the development process of financial crises. This paper aim to propose a financial trajectory dynamic tracking algorithm for quantifying the financial status process of a corporation. Including clustering module for calculating financial signal scores and signal cluster probabilities, feature module for attention assignment of financial indicators, and trajectory tracking module for dynamic prediction of financial crises, which deriving an N-step ahead algorithm based on the Kalman filter to achieve dynamic prediction of corporate financial crises. The finding shows that the N-step ahead Kalman filter can perform real-time dynamic prediction of financial trajectories. A prediction accuracy of 93.8% was achieved in the first 4 years of the financial crisis. N-step ahead filter does not need to store a large amount of historical data, which can realize real-time update of financial signals. The empirical demonstration shows the accuracy and sophistication of the algorithm, which provides a new idea in the field of financial crisis dynamics prediction.
Artificial Intelligence is revolutionizing the healthcare business, thanks to the rising availability of structured and unstructured data and the rapid advancement of analytical methodologies. With artificial intelligence's expanding importance in healthcare, there are growing issues about a lack of transparency and explainability, as well as potential bias in model predictions. The goal of this paper is to use interpretable ML to provide a better understanding of the lifestyle factors that influence the model's predictions of the weight levels, as well as to identify the most critical features for the classification task. We aim to create more accurate and effective predictive models for weight management, which could eventually help individuals make more informed decisions about their diet and lifestyle choices. Our machine learning model got an accuracy of 76% using XG-boost and Random Forest and we tried to interpret the result using a model-agnostic method including, the permutation importance method and partial dependence plot. The results showed that the number of main meals, frequent consumption of vegetables, and time using technology are the most important features. Further explanation was performed using a partial dependence plot which interprets the relationship between these important features and the model behavior.
Skin cancer is one of the most common types of cancer, with research now increasingly focused on the use of deep learning algorithms to perform diagnosis in experimental settings. Deep neural networks can be used to assist early detection; however, accuracy can be highly reliant on aspects such as dataset quality and class distribution. This study investigates the impact on melanoma classification when using images that are visually similar from the publicly available ISIC 2019 dataset. The negative effect of image duplication is well known in deep learning; however, the effect of image similarity is an under-researched topic. In this work, we used an open source image similarity algorithm to identify similar images in the ISIC 2019 dataset. We identify groups of similar images at different similarity thresholds and investigate the effect of removing each threshold on a classification model. We then evaluate the best performing model on the ISIC 2019 datatest. Our results show that the best performing model was DenseNet201 when trained using the 100% similarity threshold images, and InceptionResNetV2 when trained using the 95% similarity threshold images. These results indicate that highly similar images present in the ISIC 2019 training set result in performance degrading bias, and that their removal shows in a boost to model performance.
Recent AI research has significantly reduced the barriers to apply AI, but the process of setting up the necessary tools and frameworks can still be a challenge. While AI-as-a-Service platforms have emerged to simplify the training and deployment of AI models, they still fall short of achieving true democratization of AI. In this paper, we aim to address this gap by comparing several popular AI-as-a-Service platforms and identifying the key requirements for a platform that can achieve true democratization of AI. Our analysis highlights the need for self-hosting options, high scalability, and openness. To address these requirements, we propose our approach: the "Open Space for Machine Learning" platform. Our platform is built on cutting-edge technologies such as Kubernetes, Kubeflow Pipelines, and Ludwig, enabling us to overcome the challenges of democratizing AI. We argue that our approach is more comprehensive and effective in meeting the requirements of democratizing AI than existing AI-as-a-Service platforms.
In recent years, the prevalence of obesity and its related co-morbidities have been increasing significantly. Therefore, it is an important challenge to pursue an early prediction of obesity risk that could help in reducing the pace of obesity rise when appropriate interventions are placed, accordingly. The prediction and classification of obesity depend on different factors such as body mass index (BMI) and lifestyle aspects, including eating habits. By focusing on these lifestyles and eating habit factors, we can develop a more holistic approach to weight management and prevention of obesity. The aim of this paper is to propose a machine-learning model that can classify weight levels using lifestyle variables without relying on BMI which enables us to investigate how lifestyle factors affect different levels of weight categorization. Although BMI is the most widely used estimation of obesity, there are other factors that can contribute to gaining weight such as lifestyle factors. The accuracy of our lifestyle-based model reached 75% excluding weight, height, and family history. Our model could serve as a starting point for using an interpretable machine learning model to better understand the effect of lifestyle factors on obesity levels.
