Class imbalance is a pervasive challenge in real-world machine learning (ML) applications, where the minority class, often the class of interest, is significantly underrepresented. This imbalance can degrade model performance, result in misleading evaluation metrics, and complicate validation processes. Two prominent data-augmentation techniques to address class imbalance are the Synthetic Minority Oversampling Technique (SMOTE) and Generative Adversarial Networks (GAN). However, both techniques have inherent limitations, motivating the emergence of novel variants designed to overcome these challenges. While previous reviews have typically focused on specific domains, conventional methodologies, or broad strategy overviews, this review presents a unified taxonomy that outlines the causes, types, and implications of class imbalance across diverse ML tasks. It further examines emerging trends in the application of SMOTE and GAN techniques, their limitations, and hybrid adaptations. By categorising imbalance types and analysing models, metrics, datasets, and comparative approaches, this review provides actionable insights and identifies future research directions for practitioners and researchers working to address class imbalance in real-world ML tasks.

This study explores the fabrication and characterisation of 3D-printed polylactic acid (PLA) scaffolds reinforced with calcium hydroxyapatite (cHAP) for bone tissue engineering applications. By varying the cHAP content, we aimed to enhance PLA scaffolds’ mechanical and thermal properties, making them suitable for load-bearing biomedical applications. The results indicate that increasing cHAP content improves the tensile and compressive strength of the scaffolds, although it also increases brittleness. Notably, incorporating cHAP at 7.5% and 10% significantly enhances thermal stability and mechanical performance, with properties comparable to or exceeding those of human cancellous bone. Furthermore, this study integrates machine learning techniques to predict the mechanical properties of these composites, employing algorithms such as XGBoost and AdaBoost. The models demonstrated high predictive accuracy, with R² scores of 0.9173 and 0.8772 for compressive and tensile strength, respectively. These findings highlight the potential of using data-driven approaches to optimise material properties autonomously, offering significant implications for developing custom-tailored scaffolds in bone tissue engineering and regenerative medicine. The study underscores the promise of PLA/cHAP composites as viable candidates for advanced biomedical applications, particularly in creating patient-specific implants with improved mechanical and thermal characteristics.

The investigation of the mechanical properties of additively manufactured (AM) composite has been the focus of several research over the past decades. However, testing constraints of time and cost have encouraged the exploration of more pragmatic methods such as machine learning (ML) for predicting these characteristics. This study builds on experimental investigations of the flexural, tensile, compressive, porosity, and hardness properties of 3D printed carbon fibre-reinforced polyamide (CF-PA) and carbon fibre-reinforced acrylonitrile butadiene styrene (CF-ABS) composites, proposing the application of ML for predicting these mechanical properties. A comprehensive comparative analysis of various machine learning approaches was executed, with a resultant accuracy ranging between 80 and 99%. The results unveiled the superior predictive performance of ensemble tree learners and the K-NN regressor algorithms when temperature and porosity are selected (based on correlation analysis) as predictors for material hardness and strength in tension, com- pression, and flexion. In particular, the model built on the extra-tree regressor algorithm demonstrated a remarkably robust fit, with R-squared evaluation scores of 0.9993 and 0.9996 for CF-PA and CF-ABS, respectively. This work develops a ML model that relates porosity to the other mechanical properties of AM composites and the prediction models’ exceptional accuracy, along with their precise alignment with experimental data, provide invaluable insights for the autonomous control and data-driven optimization of the structures.

The use of fibre-reinforced plastics (FRPs) in various industrial applications continues to increase thanks to their good strength-to-weight ratio and impact resistance, as well as the high strength that provides engineers with advanced options for the design of modern structures subjected to a variety of out-of-plane impacts. An assessment of the damage morphology under such conditions using non-destructive techniques could provide useful data for material design and optimisation. This study investigated the damage mechanism and energy-absorption characteristics of E-glass laminates and sandwich structures with GFRP face sheets with PVC cores under quasi-static indentation with conical, square, and hemispherical indenters. An acoustic emission (AE) technique, coupled with a k-means++ pattern-recognition algorithm, was employed to identify the dominant microscopic and macroscopic damage mechanisms. Additionally, a post-mortem damage assessment was performed with X-ray micro computed tomography and scanning electron microscopy to validate the identified clusters. It was found that the specific energy absorption after impact with the square and hemispherical indenters of the GFRP sandwich and the plain laminate differed significantly, by 19.29% and 43.33%, respectively, while a minimal difference of 3.5% was recorded for the conical indenter. Additionally, the results obtained with the clustering technique applied to the acoustic emission signals detected the main damaged modes, such as matrix cracking, fibre/matrix debonding, delamination, the debonding of face sheets/core, and core failure. The results therefore could provide a methodology for the optimisation and prediction of damage for the health monitoring of composites.

Multilevel quadrature amplitude modulation (M-QAM) combined with DC-bias in optical orthogonal frequency division multiplexing (DCO-OFDM) offers a spectrally efficient solution and adaptive transmission rates for indoor light-fidelity (Li-Fi) systems. However, a significant challenge posed by the DCO-OFDM scheme is the additional power of the DC bias required to ensure that the amplitudes of the transmitted signals are nonnegative. These biased signals are clipped according to optical power constraints, imposing clipping noise that affects the transmission bit error rate (BER). This performance degradation is conditioned by the adjustments made to the DC bias, which requires continuous modification to support adaptive transmission. Therefore, simultaneously addressing DC bias optimization and clipping mitigation is essential to provide reliable and power-efficient transmissions. This paper proposes a machine learning (ML) approach to predict the optimum DC bias based on the statistical properties of the OFDM signal and system characteristics. A robust ML regressor selection process using LazyPredict algorithm (LPA) was employed to identify the optimal regressors for developing the predictive model. The model demonstrated significant prediction accuracy for DC bias across a wide range of transmission settings. In particular, the models built on variants of gradient boosting regressor (GBR) and support vector regressor (SVR) demonstrated superior performance, with R-squared evaluation scores of 0.9792 and 0.9225, respectively, for two different set of features. Furthermore, the BER performance of our adaptive DC bias approach was compared to a fixed DC bias in adaptive DCO-OFDM transmission, demonstrating the superiority of our approach in effectively mitigating clipping noise at high transmission rates while maintaining power efficiency at lower rates. These results provide a promising solution for the future practical deployment of Li-Fi systems in indoor applications.