Computing Open

Changing a database schema is a fact of life in information systems, as a response to changes inside the enterprise (e.g., new users’ requirements, correction of errors in the current database schema) or outside it (e.g., new regulations, new partners’ requirements). In the database research field, a well-known technique has been proposed for managing schema changes, called schema evolution. It allows the database to survive schema changes by adapting existing data to conform to the new schema. A lot of research efforts addressed the topic of schema evolution, in both conventional (i.e., relational) and advanced (e.g., XML, stream, NoSQL) databases, providing a plethora of heterogeneous approaches and solutions making up a quite large literature. Since there is no research work that extensively deals with different proposals and compares them, the purpose of this paper is to fill this gap by reviewing the available schema evolution literature. For that, first we collected and summarized the contributions of research papers dealing with database schema evolution. Then we organized their presentation in a chronological order, also giving a historical perspective on the topic development. Finally, we defined a list of six comparison criteria (database model, implementation, schema change semantics, schema change propagation, integrity constraints, and software evolution) that have helped us to categorize and compare the different database schema evolution proposals. In sum, our paper (i) provides an overview of the state-of-the-art research approaches on database schema evolution, with tables that compare such approaches based on some proposed criteria, (ii) studies the support of schema evolution in commercial DataBase Management Systems (DBMSs), and (iii) points out some possible future research directions in this area.

Changes involving the database schema are unavoidable events that sometimes occur in data-centric systems, due to many reasons like evolution of user requirements, corrective maintenance, improvement of the current schema with the addition of new integrity constraints, implementation of new business rules, or compliance to new regulations. In order to manage schema changes, researchers have proposed a technique named “schema versioning”, in which all versions of the database schema are kept along with their underlying data, preventing loss of information and obsolescence of legacy applications. This technique has been widely studied in both conventional and emerging databases. Despite lot of interesting research efforts that have dealt with database schema versioning and the importance of this technique for database maintenance, current Database Management Systems (DBMSs) do not provide sufficient support for its implementation. This paper (i) gives an overview of state-of-the-art research approaches to schema versioning in databases, based on the study of more than 270 papers and providing a historical perspective on the development of the topic, (ii) illustrates the support of schema versioning in commercial DBMSs, and (iii) points out some possible future research directions in this area.

The application of the combination of unmanned aerial vehicles (UAVs) and artificial intelligence is a hot topic in the intelligent inspection of substations, and meter reading is a very challenging task. This paper proposes a method based on the combination of YOLOv6n object detection and Deeplabv3$+$ image segmentation and performs post-processing on the segmented images to obtain meter readings. First, YOLOv6n is used to detect the meter area of the aerial image and classify the meters. Second, the detected meter images are fed into the image segmentation model. The backbone network of the Deeplabv3$+$ algorithm is improved by using the MobileNetv3 network, which not only effectively extracts pointers and scales, but also makes the model more lightweight. Third, License Plate Recognition Network (LPRNet) is used to recognize digital meter images. In order to solve the problem of inaccurate pointer meter readings, to begin with, the segmented image is corroded; in addition, the circular dial area is flattened into a rectangular area by concentric circle sampling method. Finally, the meter reading is calculated by the position of the pointer, the scale and the total range of the meter. The post-processing part uses numba to optimize the inference speed. The experimental results show that in two datasets, The mean average precision of 50 (mAP50) accuracy of the YOLOv6n model using this method reached 99.71% and 98.60%, respectively, and the inference speed of a single image was 17.1ms and 13.2ms, respectively. The mean intersection over union (mIoU) of the image segmentation model reached 82.00%, 74.73%, 73.50%, 82.26% and 73.20%, respectively, and the single segmentation speed reached 33.7ms. The LPRNet model has a recognition accuracy of 99.17% and a single image inference speed of 14.7ms. At the same time, several mainstream object detection and semantic segmentation algorithms are compared. The experimental results show that the method in this paper greatly improved the accuracy and efficiency of intelligent inspection of substation meter readings.

Currency recognition is an important field in the area of pattern recognition. Recent improvements in the field of deep learning have increased its ability to recognize complex features from images. The goal of this research was to create a first-ever Bhutanese dataset and train using transfer learning techniques. This paper proposes three different pre-trained models to train the Bhutanese currency datasets. A parameter fine-tuning was applied to get better accuracy from the custom dataset. The currency images were gathered using crowdsourcing techniques and then the data augmentation was performed to generate 1000 images per class. The datasets were split into the train, test and validation sets with the ratio of 70:20:10. The train and test sets were used while training the model and after the training, it was validated using the validation set. After training the model for certain epochs, VGG16 architecture outperformed other models with a training accuracy of 99.82%, a test accuracy of 99.12% and a validation accuracy of 95.5%. In the future, a greater number of images need to be included in the datasets and trained using other pre-trained models.

The mathematical skeleton of a complex form has been essential for a variety of scientific fields and of great interest to many researchers for decades. It is based on several concepts such as the reconstruction of forms and image processing. This paper aims to develop a novel mathematical algorithm to approximate the skeleton of a non-polygonal shape and to compare it to the most used methods. The mathematical technique of skeletonization is used as a reference to validate and compare the proposed method to the most used ones. The crux of the proposed technique is to Cartesianize the shape (polygonize in 2D), then skeletonize it. Moreover, this novel method is grounded upon the construction of bisectors on the simplex of the corresponding Cartesianized shape. Python is used to implement the algorithm proposed and test it on multiple shapes. The comparison of the results generated by the proposed algorithm and the Python predefined function skeletonize() shows that the proposed method is more effective and could be adjusted through the rate of Cartesianization of the target shape. The major contributions of this novel technique include the mitigation of some issues of existing methods, simplification, and optimization of the processing performance mainly in terms of algorithm complexity.

The BCM learning rule, named for Elie Bienenstock, Leon Cooper, and Paul Munro, was first proposed to measure the selectivity of neurons in the primary visual cortex and its dependency on neuronal inputs. We show that an artificial BCM neuron has the ability to detect clusters in a dataset. By exploring the qualitative behaviors of an underlying system of differential equations, we present a rigorous mathematical analysis of this cluster-detecting property. While the focus of this work is not to develop a robust state-of-the-art clustering method, we also analyze and discuss the performance of a resulting preliminary clustering algorithm.

Smart cities leverage advanced technologies to enhance urban living through the real-time collection, processing, and analysis of contextual information. The potential to improve residents’ outdoor experiences in these cities increases dramatically as smart technologies are integrated into urban environments, making them more interconnected. It helps to explore the pivotal role of real-time data in optimizing various aspects of city management, focusing on key domains such as traffic, public transportation, emergency response, waste management, and environmental monitoring. A variety of datasets, such as those on the weather, air quality, traffic patterns, event schedules, and user activity patterns, are gathered and analyzed as part of the methodology. This data are processed and interpreted using machine learning algorithms, which find correlations, trends, and patterns that affect outdoor activities. Suggestions for appropriate outdoor activities can be generated in real time based on contextual information, past behavior, and user preferences. This model addresses the dynamic and context-aware nature of Smart Cities by proposing a novel framework for real-time contextual information prediction and personalized outdoor activity suggestions for users. Leveraging the vast amount of data generated by Smart City infrastructure, this study integrates advanced data analysis techniques with deep learning models to enhance the urban living experience. A variety of datasets, such as those on the weather, air quality, traffic patterns, event schedules, and user activity patterns, are gathered and analyzed as part of the methodology. This data are processed and interpreted using machine learning algorithms, which find correlations, trends, and patterns that affect outdoor activities. Suggestions for appropriate outdoor activities can be generated in real time based on contextual information, past behavior, and user preferences. The framework begins by collecting and processing diverse datasets from sensors, Internet of Things (IoT) devices, and other urban sources to create a comprehensive understanding of the current city context. Deep learning models, such as recurrent neural networks (RNNs) and convolutional neural networks (CNNs), are employed to analyze this data and predict real-time contextual information, including weather conditions, traffic patterns, and social events. It contributes to the growing field of Smart Cities by introducing a scalable and adaptable framework that harnesses the power of deep learning to improve urban living. The result shows that the proposed air pollution model predicted 96.06700 PM2.5 concentration levels, subsequently the temperature model predicted 14.06800^∘C. The integration of real-time contextual information prediction and personalized outdoor activity suggestions showcases the potential for creating more engaging and user-centric Smart City ecosystems. This research attempts to provide personalized recommendations that are in line with users’ preferences, the state of the environment at the time, and other pertinent contextual factors by utilizing data from multiple sources, including IoT devices, mobile applications, and environmental sensors.

With the advent of microblogging platforms like Twitter, there has been a substantial shift toward digital media for getting acquainted with ongoing global issues. Although Twitter is an incredible source of information for real-time news, the information is widely scattered, opinionated, and unorganized, making it tedious for users to apprise themselves of the latest issues. Therefore, this study proposes a framework to automatically generate short news compositions from tweets utilizing state-of-the-art artificial intelligence techniques. The proposed framework scrapes tweets from authentic news Twitter handles, semantically analyzes and clusters them, predicts sentence ordering of the formed clusters, and summarizes the text of the clusters to produce structured compositions automatically. The generated compositions are further augmented with their corresponding sentiment scores to provide an overall perspective to the end-user toward the news topic in consideration. Evaluating the automatically generated compositions shows that the proposed framework is 77.5% efficient in generating quality compositions.

Racket sports such as table tennis involve a wide range of three-dimensional complex spatial movements of the human body and the racket. Novice players might benefit from the evaluation of the motion profile of the racket to facilitate better adoption of more expert movement. Computer-based evaluation of such novice vs. expert play behavior characteristics includes reducing the required multiple human interactions and easy applicability for subsequent automation to accurately differentiate the motion profile of a novice player from that of an expert. This study has, for the first time, applied the widely used support vector machine (SVM) classification technique for the development of a table tennis player movement evaluation model. The model was trained using an existing dataset of displacements and velocities from various important anatomical landmarks across the body and points on the racket. These were obtained and evaluated for table tennis forehand strokes for two subgroups of expert and novice ability levels, respectively. Different combinations of variables were selected for model input from the same dataset with the outcomes being noted for each. The resulting SVM classification model exhibited good/noteworthy performance ($>90$% accuracy) in distinguishing racket motion between expert and novice players.

Diabetes is a global health concern that affects people of all races. With different uncertainties in human lifestyles, it is difficult to predict diabetes while assuming that the risk patterns are the same for all. The likelihood of diabetes in a patient is mostly predicted using machine learning (ML) models on features explicitly available in datasets, while the intrinsic relationship between features viz-a-viz their potential relevance to the presence of diabetes is oftentimes neglected. In this work, we explored feature importance and correlation to derive the top 15 feature pairs from a dataset of 263,882 samples of anonymized patient information. These top-15 feature pairs were fed into five different ML models (decision tree (DT), neural networks (NN), random forest (RF), support vector machine (SVM) and extreme gradient boosting (XGB)) for predicting the likelihood of diabetes, while also feeding the direct features (without correlated pairing) separately into the same 5ML models. The models’ performances were evaluated using accuracy, precision, recall and F1-score and NN presented the best performance overall achieving an F1-score of 85% for the correlated feature pairs (CF) and 75% for the direct feature pairs. The results confirm the importance of the correlation/relationship between features in predicting the likelihood of diabetes in patients more accurately.

Medical image segmentation has always been a challenging task. This paper proposes a new LUneXt medical image segmentation model based on the characteristics analysis of medical image data sets and testing of different nonlinear activation units. Both normalized activations for the original negative input image activation have good optimization capabilities for the tokenization module parameters proposed in the original UneXt model. Using different activation coefficients for different foreground and background areas has achieved better results. The experimental results of this paper on the Breast Ultrasound Images (BUSI) data set reached an intersection over union (IoU) value of 62.64%, a Dice value of 76.12%, and a single inference speed of 807.57ms. The experimental IoU value of the International Skin Imaging Collaboration (ISIC 2018) data set reached 82.95%, and the Dice value reached 90.50%. The single inference speed reached 842.58ms. The LUneXt model is more robust than other models. While improving model performance, it does not introduce higher computational complexity and does not have a major impact on the processing speed of a single image.

The purpose of this study is to extend the idea of complex intuitionistic fuzzy set (CIFS) with a new notion referred to as a complex picture fuzzy set (CPFS). CPFS is a generalized version of CIFS since it includes a neutral membership degree in the concept. The capacity to cover a wider range of information with the aid of neutral membership, non-membership, and membership makes this new theory distinctive. The unit disc of complex plane has been used to cover the range of values of membership degrees. Pertaining to Frank t-norm and t-conorm operations and a few aggregation methods, we establish some fundamental CPFS aggregation operators and attributes and use them to investigate multi-attribute decision-making (MADM) problems. Then, we offer various operators for the purpose of aggregating the CPF data. These are complex picture fuzzy Frank weighted averaging (CPFFWA), complex picture fuzzy Frank ordered weighted averaging (CPFFOWA), complex picture fuzzy Frank hybrid averaging (CPFFHA), and complex picture fuzzy Frank weighted geometric averaging (CPFFWGA), complex picture fuzzy Frank ordered weighted geometric averaging (CPFFOWG), and complex picture fuzzy frank hybrid geometric averaging (CPFFHGA) operators, which benefit from the basic Frank operations and averaging, geometric aggregation techniques. Furthermore, an algorithm for solving multi-attribute decision-making MADM problems has been presented under the framework of CPFSs by using CPFFWA and CPFFWG operators. Finally, in order to depict the potential applicability of our proposed technique, a numerical problem aiming at finding the best alternative has been solved and outcomes have been well compared with some existing techniques.

Closeness is an important measure of network centrality. In this paper, we will calculate the closeness of graphs, created by using operations on graphs. We will prove a formula for the closeness of shadow graphs. We will calculate the closeness of line graphs of some well-known graphs (like path, star, cycle, and complete graphs) and the closeness of line graphs of two of these graphs, connected by a bridge (like lollipop, tadpole, broom, and bistar graphs).