The Particle Swarm Optimisation (PSO) algorithm was inspired by the social and biological behaviour of bird flocks searching for food sources. In this nature-based algorithm, individuals are referred to as particles and fly through the search space seeking for the global best position that minimises (or maximises) a given problem. Today, PSO is one of the most well-known and widely used swarm intelligence algorithms and metaheuristic techniques, because of its simplicity and ability to be used in a wide range of applications. However, in-depth studies of the algorithm have led to the detection and identification of a number of problems with it, especially convergence problems and performance issues. Consequently, a myriad of variants, enhancements and extensions to the original version of the algorithm, developed and introduced in the mid-1990s, have been proposed, especially in the last two decades. In this article, a systematic literature review about those variants and improvements is made, which also covers the hybridisation and parallelisation of the algorithm and its extensions to other classes of optimisation problems, taking into consideration the most important ones. These approaches and improvements are appropriately summarised, organised and presented, in order to allow and facilitate the identification of the most appropriate PSO variant for a particular application.
Noncontact injuries are prevalent among professional football players. Yet, most research on this topic is retrospective, focusing solely on statistical correlations between Global Positioning System (GPS) metrics and injury occurrence, overlooking the multifactorial nature of injuries. This study introduces an automated injury identification and prediction approach using machine learning, leveraging GPS data and player-specific parameters. A sample of 34 male professional players from a Portuguese first-division team was analyzed, combining GPS data from Catapult receivers with descriptive variables for machine learning models—Support Vector Machines (SVMs), Feedforward Neural Networks (FNNs), and Adaptive Boosting (AdaBoost)—to predict injuries. These models, particularly the SVMs with cost-sensitive learning, showed high accuracy in detecting injury events, achieving a sensitivity of 71.43%, specificity of 74.19%, and overall accuracy of 74.22%. Key predictive factors included the player’s position, session type, player load, velocity and acceleration. The developed models are notable for their balanced sensitivity and specificity, efficiency without extensive manual data collection, and capability to predict injuries for short time frames. These advancements will aid coaching staff in identifying high-risk players, optimizing team performance, and reducing rehabilitation costs.
The Cyclic Alternating Pattern (CAP) is a periodic activity detected in the electroencephalogram (EEG) signals. This pattern was identified as a marker of unstable sleep with several possible clinical applications; however, there is a need to develop automatic methodologies to facilitate real-world applications based on CAP assessment. Therefore, a deep learning-based EEG channels’ feature level fusion was proposed in this work and employed for the CAP A phase classification. Two optimization algorithms optimized the channel selection, fusion, and classification procedures. The developed methodologies were evaluated by fusing the information from multiple EEG channels for patients with nocturnal frontal lobe epilepsy and patients without neurological disorders. Results showed that both optimization algorithms selected a comparable structure with similar feature level fusion, consisting of three electroencephalogram channels (Fp2–F4, C4–A1, F4–C4), which is in line with the CAP protocol to ensure multiple channels’ arousals for CAP detection. Moreover, the two optimized models reached an area under the receiver operating characteristic curve of 0.82, with average accuracy ranging from 77% to 79%, a result in the upper range of the specialist agreement and best state-of-the-art works, despite a challenging dataset. The proposed methodology also has the advantage of providing a fully automatic analysis without requiring any manual procedure. Ultimately, the models were revealed to be noise-resistant and resilient to multiple channel loss, being thus suitable for real-world application.
The unique geographical and topographical features of Madeira International Airport in Portugal significantly influence flight safety, primarily due to variable wind patterns. In this study, a machine learning approach is developed to predict runway operational statuses at Madeira International Airport, focusing on addressing wind-related challenges. To tackle this issue, a Deep Learning model is utilized. This model undergoes a particle swarm optimization process, resulting in one optimized model for each timestep, to provide minute-resolution predictions within a 20-minute timeframe. The training, validation, and testing phases for the optimized models were conducted using high-frequency wind data from Madeira International Airport. The main objective is to accurately predict the runway operational statuses, specifically whether the airport is open or closed for landing, take-off, or both. The models exhibit high performance, particularly in identifying operational conditions, reaching 99.93% precision, and a top accuracy of 94.35% predicting all runway status, underscoring their potential to enhance decision-making processes and operational efficiency under challenging weather conditions.
Finding arbitrary roots of polynomials is a fundamental problem in various areas of science and engineering. A myriad of methods was suggested to address this problem, such as the sequential Newton’s method and the Durand–Kerner (D–K) simultaneous iterative method. The sequential iterative methods, on the one hand, need to use a deflation procedure in order to compute approximations to all the roots of a given polynomial, which can produce inaccurate results due to the accumulation of rounding errors. On the other hand, the simultaneous iterative methods require good initial guesses to converge. However, Artificial Neural Networks (ANNs) are widely known by their capacity to find complex mappings between the dependent and independent variables. In view of this, this paper aims to determine, based on comparative results, whether ANNs can be used to compute approximations to the real and complex roots of a given polynomial, as an alternative to simultaneous iterative algorithms like the D–K method. Although the results are very encouraging and demonstrate the viability and potentiality of the suggested approach, the ANNs were not able to surpass the accuracy of the D–K method. The results indicated, however, that the use of the approximations computed by the ANNs as the initial guesses for the D–K method can be beneficial to the accuracy of this method.