Integrated Artificial Intelligence in Data Science

Artificial Intelligence (AI) is increasingly pervading everyday life since it can be used to solve high-complexity problems, as well as determine optimal solutions, in various domains and for numerous applications. From this perspective, it has the potential to create a better society. The benefits of AI in science, medicine, technology, and the social sciences have already been clearly proven. Data science, also referred to as pattern analytics and mining, can be used to retrieve useful and meaningful information from databases, which subsequently allows us to make decisions efficiently and build strategies in different domains. In particular, as a result of the exponential growth of data in recent years, the dual concept of big data and AI has given rise to many research topics, such as scale-up behaviour from classical algorithms.

In addition, a recent critical challenge that has emerged is the integration of multiple AI technologies from different fields (e.g., vision, security, control, bioinformatics) to develop efficient and robust systems that interact in the real world. Despite the tremendous progress in core AI technologies in recent years, the integration of such capabilities into larger systems that are reliable, transparent, and maintainable is still in its infancy. There are numerous open issues from both theoretical and practical perspectives.

This Special Issue aimed to collect and present all breakthrough research in this emerging field. A total of 22 papers have been accepted for publication in this Special Issue. In [1], neural language processing (NLP) and clustering techniques were used as the basis for a job advertisement analysis algorithm. In [2], a fuzzy logic-based dynamic ensemble model was discussed and employed for malware detection. The authors of [3] proposed an approach based on convolutional neural network (CNN) dual-attention unit (DAU) and selective kernel feature synthesis (SKFS) for the enhancement of dark images under low-light conditions. The topic of [4] is a new training algorithm for graph convolutional networks (GCNs) and the graph attention model (GAT). In [5], a fuzzy multicriteria decision-making (MCDM) system based on Z numbers was developed to support order-picking process management. Large-scale building information modeling was the application case studied in [6]. The authors used machine learning (ML) for the automated classification of elements and the detection of anomalies based on geometric inputs. In [7], a red fox optimizer for a deep learning-enabled microarray gene expression classification model was investigated. The article [8] deals with temporal fuzzy utility itemset mining, proposing a one-phase tree-structure method to find the high temporal fuzzy utility itemsets in a temporal database. In [9], several ML techniques were compared to assess the appropriateness of surgical prophylaxis. In [10], a deep learning (DL) model was developed for the early prediction of the learning performance of students, as well as for the analysis of student behaviour in a virtual learning environment. A comparative analysis of ensemble models for road traffic congestion was discussed in [11]. Advanced genetic Bollinger Bands and the trading strategy of correlation coefficient-based pairs have been proposed in [12]. In [13], the authors incorporate a crowdsourcing-based approach for annotated data acquisition in order to enhance the Active Learning procedure. In [14], an enhanced image captioning model for the automatic textual description of images is discussed. A hybrid DL model was proposed in [15] for distributed denial of service (DDoS) attacks. The article [16] illustrates the means with which DL can be used to enhance an end-to-end table detection model. The article [17] likewise concerns a DL scheme, but is properly improved for pneumonia image classification. The authors of [18] presented an adaptive online task scheduling algorithm based on a variant of Q-learning for the efficient management of virtual machine resources in cloud computing. In [19], a variant of the data particle geometrical divide was proposed to tackle classification problems in the presence of imbalanced data. The article [20] deals with group multi-objective optimization, and the specific case study is related to portfolio optimization. In [21], a DL-based approach combined with natural language description was discussed and incorporated into a smart surveillance system. Finally, [22] presents a more general implementation of the Monte Carlo Tree Search using neural networks when there is an insufficient knowledge of the problem.