The increasing demand for mobile usage raises many challenges for service providers. Satisfying subscribers by providing a better quality of service (QoS) is one of the major concerns of an operator. One way to solve this problem is to adopt an efficient radio resource use mechanism. In this context, smart network planning management requires the estimation and enhancement of capacity in terms of the number of subscribers within a cell for better QoS provisioning through radio resource usage optimization. The model proposed in this article explicitly uses some long term evolution (LTE)-Advanced (LTE-A)-specific enhancements such as carrier aggregation (CA) and channel quality indication (CQI)-based resource allocation. Further, this study proposes a CQI-based clustering approach with packet bundling and CA to optimize radio resource utilization and enhance LTE-A cell capacity. LTE-A cell is logically divided into clusters such as the Silver class, Platinum class, Gold class, and Diamond class based on the CQI from the user end to eNodeB (eNB). Further, cell capacity (CCa) estimation algorithms are proposed in a simplistic scenario as well as in each cluster using packet bundling factor ? considering CA. From the result analysis, it is found that appropriate modulation and voice codecs can be used in appropriate clusters to enhance the cell capacity. Furthermore, it is observed that the packet bundling factor helps in improving the radio resource usage and thereby improving the capacity of a cell. The research work proposed in this article can be extended further to estimate the user capacity in the context of the 5th generation cellular network.

Over the past few decades, heart disease has seen significant growth among all ages and early prediction became necessary. Data mining and machine learning techniques are used to solve the prediction problem utilizing new approaches to supervised learning. The Internet of Medical Things (IoMT) emerged from the combination of multiple fields and machine learning. The goal of this research is to develop an adaptive model for predicting cardiac disease. We provide a ranking-based hybrid feature selection method for identifying essential characteristics. The model proposed in this paper employs a clustering method in conjunction with support vector machine (SVM) to save training time and eliminate classification errors, hence boosting the model’s performance and increasing its efficiency.

Diabetes, often regarded as a chronic illness, is a condition that occurs due to high blood sugar for a prolonged period of time. The risk of obtaining diabetes can be reduced by precise early prediction and analysing factors such as hereditary involvement and several other factors. Although advanced techniques came into existence, we can observe that the risk of developing diabetes is substantially higher among adults due to modern life. Timely treatment and diagnosis are required to prevent the outbreak and the advancement of diabetes. The lack of robustness in the precise early prediction of diabetes is a rigid task due to the size of the dataset and deficient labelled data. In this literature, we propose an architectural framework for the early prediction of diabetes disease where data pre-processing, outlier detection and avoidance, K-fold cross validation, and distinct predictive machine learning (ML) and deep learning (DL) classifiers (Decision Tree, Logistic Regression, and Neural Network) are appointed. In this literature, the ensembling of various machine learning and deep learning classifiers are used as a method of enhancing diabetes prediction, utilising K-fold cross validation as a validation strategy. The base classifiers are hypertuned using the grid search approach by considering numeric hyperparameters. The experiments conducted in this literature were conducted under similar conditions using the benchmark PIMA Indian Diabetes (PID) dataset. As a substitute for the conventional approach of testing the proposed approach, we have chosen the chronic kidney disease (CKD) dataset from the University of California (UCI) machine learning repository as a comparative study.

Heart disease prediction is important in healthcare because it enables timely identification and intervention of actual condition of the patient. However, the task of accurately predicting disease remains a challenging task. In this paper, we have proposed a framework for heart disease prediction using explainable artificial intelligence (XAI) based Machine Learning (ML) models such as Decision Tree (DT), Random Forest (RF), k-nearest neighbors (KNN), AdaBoost, Logistic Regression (LR), Naive Bayes (NB), and Neural Network (NN). The efficiency of those models were evaluated using MCC, accuracy, precision, recall, and AUC. Finally, it is observed that, DT emerges as the most effective model offering interpretability. This study underscores the importance of transparent models in healthcare and advocates in order to incorporate XAI to enhance interpretability and medical decision-making.

The shipping industry deals with an amount of information much of which is not properly stored or secured and ends up getting lost over time. However these data become crucial in times of incidents. In the future ships will require Big Data Analytics for purposes such as condition monitoring, auto-piloting, freight tracking and shipbuilding. The progress in Big Data will enable ships to communicate with each other through condition monitoring systems and engines. As a result Big Data Analytics enhances both the safety and efficiency of the industry. It is important to store the data from classification societies and shipbuilders, for references and advancements where Big Data Analytics plays a significant role. Temporal data mining is a field that focuses on analyzing ordered data streams with interdependencies. In this work, the goal is to detect anomalies in maritime vessel data, particularly sudden speed changes and unusual course deviations. This study presents an evaluation of the detection accuracy and identify the most effective algorithm for anomaly detection in the context of maritime activities. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2024.

Traffic prediction in5G is important for effective deployment and operation of Internet of Things (IoT) ecosystems. It enables resource management and optimization, guaranteeing that the network can handle unpredictable traffic volumes with-out experiencing traffic jams. This helps to ensure high quality of service and low latency for applications such as autonomous automobiles and virtual reality. Predictive traffic management further enhances user experience by keeping services consistent and reliable, particularly during busy hours. There are various approaches to traffic prediction in 5G networks, and each has advantages and disadvantages of its own. The choice of model will depend on how precise, adaptable, and computationally demanding the network must be. The model proposed in this paper integrates lightweight convolution with temporal attention to deliver accurate and efficient traffic prediction for 5G networks that may further be useful for developing IoT ecosystem

Cloud computing operates similarly to a utility, providing users with on-demand access to various hardware and software resources, billed according to usage. These resources are primarily virtualized, with virtual machines (VMs) serving as critical components. However, task allocation within VMs presents significant challenges, as uneven distribution can lead to underloading or overloading, causing system inefficiencies and potential failures. This study addresses these issues by proposing a novel hybrid task allocation algorithm that combines the strengths of the Artificial Bee Colony (ABC) algorithm with Particle Swarm Optimization (PSO). Our approach aims to enhance resource utilization and reduce the risks of VM overload or underload. We conduct a comprehensive evaluation of the proposed hybrid algorithm against traditional ABC and PSO algorithms, focusing on their effectiveness in managing diverse task loads. The results of our empirical analysis indicate that our hybrid approach outperforms the conventional algorithms, leading to better resource utilization and more accurate task allocation. These findings have significant implications for optimizing task allocation in cloud computing environments, and we suggest potential avenues for future research to further refine these strategies.

The critical features of atomic or molecular scalar fields can be captured by mapping their topography. The topography maps, viz., molecular electron density (MED) and molecular electrostatic potential (MESP), are successfully used to characterize the weak interactions that exist in systems ranging from small molecular clusters to large supramolecular systems. Topography mapping mainly involves identifying and characterizing their critical points (CPs). These CPs are extremely useful for describing the nature of weak interactions and quantifying the strength of interaction. The first part of the chapter delineates fundamental concepts related to MED and MESP. Later, the usefulness of MED and MESP analyses for understanding weak interactions is demonstrated by taking various molecular complexes. © 2024 John Wiley & Sons Ltd.

In today’s competitive landscape, accurately assessing customer satisfaction is crucial for business success. This paper presents an innovative methodology for call centers, leveraging advanced speaker diarization, sentiment analysis, and speech emotion recognition (SER) to gain a nuanced understanding of customer needs and preferences. By integrating these techniques, our approach predicts customer satisfaction scores (CSAT), empowering businesses to refine products, optimize services, and make data-driven decisions. Experimental validation demonstrates the effectiveness of our methodology in uncovering actionable intelligence from call center interactions, enabling businesses to enhance customer experience and drive continuous improvement

Coral reefs, which are, among the valuable ecosys-tems on Earth are currently facing unprecedented threats from climate change, pollution and human activities. To address the concerning decline of reefs there is a need for innovative and effective restoration methods. Machine learning algorithms have shown abilities in analyzing ecological data and offering valuable insights to aid decision making processes. In the realm of coral reef restoration, Computer Vision technology can play a role in tackling challenges such as identifying suitable restoration sites efficiently, optimizing deployment strategies and improving monitoring of coral health. Integrating machine learning into efforts to restore reefs shows potential for rehabilitating and conserving these highly endangered ecosystems. By leveraging ML techniques we can gain insights, improve decision making processes, refine restoration strategies and ultimately contribute to the long term resilience and sustainability of reefs. This study delves into the exploration of using YOLO V5 for detecting crown of thorns starfish, on reefs.

When several pictures overlap, we can merge them into one high-resolution image which is called image stitching. To tackle this problem effectively within the images, SIFT algorithm that is robust to such noisy elements was used to identify the matching keypoints which eventually enhances the overall picture standards significantly - advancing panoramic photography as well as medical imaging alongside satellite imagery domains through improving their technical aspects vastly with potential practical use cases across various industries like healthcare, remote sensing or even general professional photography services too. Our proposed procedure shall subsequently undergo rigorous testing phases using varied datasets for efficiency analysis purposes towards validating its soundness while providing robust enough to handle all possible inputs thrown at it without faltering under any circumstances. The results of this project could have practical implications in industries such as healthcare, remote sensing, and photography.

Cardiovascular disease, often treated as a cardiac illness, is a condition that occurs due to a change in the blood flow among the arteries in the heart. Although advanced techniques came into existence, we can observe that the mortality rate is increasing substantially among adults. Timely treatment and diagnosis are required to prevent heart failure. The lack of robustness in the accurate prediction of heart disease is a tough task due to insufficient data and the existence of outliers in the datasets. Several machine learning (ML) classifiers have been predominantly used in solving critical tasks and have proven their versatility by showing significant results. In this work, we have proposed a novel framework for early prediction of cardiovascular disease, where data pre-processing, outlier detection, predictive ML classifiers such as Naïve Bayes (NB), Support Vector Machine (SVM), k-Nearest Neighbours (k-NN), XGBoost (XGB)) were employed. In this work, the ensembling of various ML classifiers is also used as a method for improving heart disease prediction using k-fold cross validation (k-CV). The base classifiers are hypertuned using the grid search approach by considering numeric hyper parameters. The benchmark Cleveland heart disease dataset from the University of California (UCI) repository is consider for experiment. From the experiment it is found that, the proposed method outperforms the standard results in various evaluation metrics.

Geomembrane (GM) liners anchored in the trenches of municipal solid waste (MSW) landfills undergo pull-out failure when the applied tensile stresses exceed the ultimate strength of the liner. The present study estimates the tensile strength of GM liner against pull-out failure from anchorage with the help of machine-learning (ML) techniques. Five ML models, namely multilayer perceptron (MLP), extreme gradient boosting (XGB), support vector regression (SVR), random forest (RF) and locally weighted regression (LWR) were employed in this work. The effect of anchorage geometry, soil density and interface friction were studied with regards to the tensile strength of the GM. In this study, 1520 samples of soil–GM interface friction were used. The ML models were trained and tested with 90% and 10% of data, respectively. The performance of ML models was statistically examined using the coefficients of determination (R, R) and mean square errors (MSE, RMSE). In addition, an external validation model and K-fold cross-validation techniques were used to check the models’ performance and accuracy. Among the chosen ML models, MLP was found to be superior in accurately predicting the tensile strength of GM liner. The developed methodology is useful for tensile strength estimation and can be beneficially employed in landfill design.

The exponential growth of retail transactions with a variety of customers having different interests makes the pattern mining problem trivial. Hence this paper proposes a novel model for mining frequent patterns. As per the proposed model the frequent pattern discovery is carried out in three phases. In the first phase, dataset is divided into n partitions based on the time stamp. In the second phase, clustering is performed in each of the partitions parallelly to classify the customers as HIG, MIG, and LIG. In the third phase, proposed algorithm is applied on each of the classified groups to obtain frequent patterns. Finally, the proposed model is validated using a sample dataset and experimental results are presented to explain the capability and usefulness of the proposed model and algorithm. Further, the proposed algorithm is compared with the existing algorithm and it is observed that the proposed algorithm performs better in terms of time complexity.

The Apriori algorithm is one of the association rule mining algorithms which is commonly used in retail data analysis to find patterns in itemsets. However, as the amount of data generated by retailers continues to increase, traditional approaches may no longer be sufficient. To address this challenge, In this paper, we have proposed a distributed architecture implementation of the Apriori algorithm in the Hadoop ecosystem using Spark. The approach involves three phases: categorising and partitioning customer data based on seasons, clustering customers based on behaviour and preferences, and applying the Apriori algorithm to obtain frequent itemset patterns and association rules. By implementing this approach in a distributed architecture, we are able to efficiently analyse large datasets and make accurate predictions about customer needs and preferences, which has important implications for retail strategy and we anticipate that this approach will be particularly useful in the context of retail data analysis where large amounts of data must be processed quickly and accurately.

Diabetic Retinopathy (DR) is a debilitating ocular complication of diabetes that results from prolonged exposure of the retina to elevated levels of blood glucose. This exposure can lead to progressive microvascular changes and neuronal injury, resulting in a spectrum of visual impairments ranging from mild vision changes to severe vision loss and blindness. DR typically manifests as structural changes in the blood vessels of the retina, including capillary non-perfusion, microaneurysms, retinal hemorrhages, and new vessel formation. DR is challenging to diagnose and treat due to the gradual onset of symptoms and the lack of early warning signs. Therefore, regular eye exams are critical for early detection and management of DR. A human ophthalmologist would take a significant amount of time, based on their ability and experience, to go through the fundus image and diagnose DR. Despite advancements in DR management, it remains a significant public health issue, and further research is essential to improve the understanding of DR in order to overcome the existing complications. This paper proposes a solution for improving retinal fundus images by creating more precise computerized image analysis medical diagnosis with fewer computational requirements as the images are grayscaled so that irrespective of the imaging apparatus the features of the images are enhanced without loss of information. The results of the proposed framework are assessed using entropy, contrast improvement index and structural similarity index measure.

Voice over long term evolution (VoLTE) in long term evolution-advanced (LTE-A) is gaining more and more popularity these days. However, the increasing demands of subscribers raise new challenges in wireless network to support large number of concurrent active users while maintaining the desired delay requirements. To realize the issue, this paper proposes an LTE-A capacity or eNodeB (eNB) capacity estimation model, i.e. defined as the number of simultaneous users for location-dependent diverse radio conditions. This proposed model incorporates different bundling mechanisms to utilize the radio resources efficiently. Then, optimal transmission parameters ?, ?? using packet bundling, ?, ?? using transmission time interval (TTI) bundling, and ?, ?? for mixed radio condition is derived to enhance the accommodation of a large number of users within an LTE-A cell. There has been almost no consideration in the existing research on LTE-A capacity estimation based on these aspects. Therefore this paper presents novel algorithms based on the proposed models that enable capacity estimation and optimization under diverse radio conditions. The LTE-A cell capacity obtained using the proposed algorithms can be further used as an important parameter for designing call admission control (CAC) in packet-switched wireless networks.

Association rule mining is one of the common way to analyze market basket data, and it provides knowledge to the decision-makers that help them to make strategic decisions. In the literature, many techniques have been studied for association rule mining but the exponential growth of data from various sources and the changing nature of data with respect to time and zone makes the analysis task trivial. In this paper, we propose a novel partition-based frequent pattern mining algorithm in order to generate robust and useful patterns from dataset in a more efficient way. The frequent patterns found from the proposed algorithm are used to generate interesting association rules. This paper provides an optimized method, which split the dataset into multiple loads on the basis of a particular attribute depending upon the number of cores available in the system, and these individual loads will get executed in parallel using our proposed algorithm. We show experimental results using datasets from retail sector to validate the capability and usefulness of our proposed algorithm.

A system model is proposed to ascertain Chronic Kidney Disease (CKD) using Computed To-mography (CT) scan images and blood samples. The proposed system model is tested using a dataset of 400 patients. To annotate the CT/Magnetic Resonance Imaging (MRI) scan images, the edge detection technique is used in this paper. The system model is proposed considering four units such as edge detection, prediction, virtual assistant and book reader. Edge detection is an image processing technique that is used to identify points or edges in a digital image which are scanned images of CT/MRI. K-Nearest Neighbours (KNN) algorithm is used in this model for disease prediction. Further, virtual assistant and book reader is used to assist the doctor. The proposed model in this paper may be useful for doctor as well as patients for CKD detection. The Machine Learning (ML) approach proposed in this paper may save time and costs for diagnostic screening.

Hospitals are the most common option for health checks, illness diagnosis, and treatment for sick people. This practice is followed by almost everyone in the world. But there is a drawback with this method of getting diagnosed. There are a lot of patients with various diseases/viruses which have a potential to spread in the hospital premises. People never considered the diseases/viruses present in the hospital atmosphere. People are aware of the risk of viral transmissions in hospital environments, post COVID era. Getting diagnosed and going through the reports with an efficient accuracy takes time and some people in emergency may not have enough time to perform the conventional procedures. Users have a necessity of an online website which can help them diagnose their health problems at the comfort of their homes. This would benefit people as they don't have to travel to the hospitals and reduce their risks of transmitting hospital acquired infections. This paper presents an interactive interface that functions as a virtual therapist which accepts input in the form of text, voice, or video. Data is pushed into the machine learning pipeline that generates results. The end result of this model is a report containing root cause of the disease, a tentative prescription, and any estimated treatment expenses. This model helps to prevent hospital-acquired infections, reduces the costs of treatment as users would be able to diagnose earlier and would prefer frequent testing, reducing surgeries and also reduces the tasks of doctors.

Big Data is huge in volume, diverse in information, and growing at flourishing rates. The major distributed file systems in the current market in Big Data Analysis includes Apache Hadoop, Storm, Cassandra, Flink, Cloudera, and many more. Hadoop is an open-source framework divided into Hadoop Distributed File System (HDFS) and Map-Reduce. Hadoop plays a leading role in storing and processing Big Data in contemporary society as it is cost- effective and can manage large volumes of data in low-cost commodity hardware. HDFS is a type of Data Warehouse which is scalable and has fast access to the information. Metadata is the information about the data, such as which block is storing on what datanode, how many replications are that particular block has, and on which datanodes those replications reside. In HDFS, this metadata is stored at a fixed place in namenode, and attackers can access the metadata and modify it without notice. Also, the metadata is mutable, which means that the attacker can erase his presence easily. To resolve this issue, in this paper we have provided a mechanism using blockchain technology that follows a decentralized architecture against the centralized architecture followed by HDFS. Hyperledger Fabric (HLF) is the blockchain proposed to be effective and trusted for such a purpose. HLF is a private blockchain with a distributed immutable ledger. The metadata will be stored in the ledger. If an attacker tries to modify the data, he cannot erase his presence as the ledger is immutable, unlike HDFS. Further, the work proposed in this paper can be extended in real-time HDFS with the secure ledger and multiple nodes.

Cardiovascular disease also known as heart disease is commonly found among all the ages since many years. Prediction of this disease has become a critical task in the field of medical analysis. As there is a significant improvement in the health industry but early prediction is necessary rather than making it worse and it's important to identify at the earlier stages. Recent studies reflect the use of hybrid approaches subjected to machine learning and deep learning techniques based on sensing technology in numerous applications, mainly involving complex tasks such as the collection of patient data and transforming them into Electronic Health Records (EHR). In this paper, we present a ranking-based intelligent feature selection method for identifying the optimal set of features for developing an adoptive model. The predictive classifier used to construct the model is a clustering-based Support Vector Machine (SVM) for the prediction of heart disease.

Frequent pattern mining (FPM) is one of the most important areas in the field of data mining. Several FPM algorithms have been proposed in the literature by many researchers. In most of the approaches, data set is scanned repeatedly in almost every steps of the algorithm that leads to high time complexity. That is why, processing huge amount of data using those algorithms may not be a suitable option. Hence, a novel FPM algorithm is proposed in this paper that improves efficiency by decreasing the time complexity as compared to classical frequent pattern mining algorithm. The proposed FPM algorithm converts the real-world data set into a binary matrix in a single scan, then join operation is performed to obtain the candidate itemsets. Further, AND operation is performed on the candidates to obtain frequent itemsets. Further more, using our proposed algorithm, interesting association rules can be derived.

Distributed Frequent Pattern Mining (DFPM) is a well known technique of Distributed Data Mining (DDM) that deals with finding interesting frequent patterns from a huge dataset in a distributed environment. Many algorithms do exist for mining such huge dataset in distributed environment but still it is an interesting problem because of the exponential growth of variety of data from various sources. In this paper, we have proposed a novel distributed frequent pattern mining algorithm using load-matrix. This algorithm split the dataset vertically into multiple loads which are assigned to the available cores in the system for parallel execution. From the experimental results it is observed that the proposed algorithm outperforms the existing apriori algorithm.