The rapid expansion of Wireless Local Area Networks (WLANs) has introduced significant performance challenges, particularly due to the increasing number of mobile and connected devices. Traditional static network management techniques are inadequate for handling the dynamic and complex nature of modern WLAN environments, often resulting in latency, congestion, and interference. This study systematically examines the potential of machine learning (ML) approaches, including advanced algorithms such as Q-learning and Support Vector Machines (SVM), for WLAN performance optimisation. By enabling predictive traffic analysis, adaptive configuration, and intelligent resource allocation, ML techniques offer opportunities to enhance throughput and minimise delay. The study aims to: (1) identify and categorise ML algorithms addressing key WLAN challenges such as latency reduction, interference mitigation, and load balancing; (2) analyze the performance metrics used across studies using standardised formulations; (3) evaluate the generalisability of simulation results to real-world deployments; and (4) identify computational, scalability, and dataset limitations affecting real-time implementation. Despite promising laboratory results, challenges persist due to the scarcity of large, high-quality, real-world datasets required for robust training. The paper highlights the critical role of data efficiency and advocates for open-source WLAN datasets and methods such as transfer learning and few-shot learning to reduce data dependence. Ultimately, this study emphasises that overcoming data constraints is key to realising adaptive, real-time, and scalable ML-driven WLAN optimisation for future wireless communication systems.
This study highlights the importance of radio as a major means of communication, particularly in French-speaking Africa. Despite the rise of digital media, radio remains widely used due to its accessibility, reliability, and ability to reach rural areas or regions with limited internet access. It also plays an important social and cultural role by fostering the exchange of information and community engagement. In the university setting, particularly at Gamal Abdel Nasser University in Conakry, the lack of an internal media outlet limits the dissemination of information. The creation of a university frequency modulation (FM) radio station is therefore proposed as a solution to improve internal communication, while also serving as an educational tool for students. We then described the technical design of a simple, low-cost FM transmitter capable of covering a distance of about 200 to 300 meters. We explained how it works, based on an oscillator, electronic components (transistors, capacitors, coils), and a microphone that converts sound into an electrical signal. Details are provided on construction, frequency tuning, and operating conditions. A discussion highlights certain technical limitations, such as interference, frequency instability, and range variability depending on the environment. Solutions are proposed to improve performance, including the use of appropriate antennas and amplifiers. Finally, we have emphasized that establishing a university radio station is technically feasible and beneficial. It serves as an effective tool for communication, learning, and promoting academic activities, while also contributing to the integration of communication technologies into higher education.
Voluntary customer churn constitutes a persistent financial risk for telecommunications operators, particularly within enterprise customer segments where high-value accounts administer complex, multi-subscription portfolios. Industry data indicate that acquiring a new account costs between five and seven times more than retaining an existing one. Despite heightened industry awareness, the majority of operational retention platforms remain reactive, detecting departure only after the event has occurred. This investigation constructs and evaluates a machine learning pipeline engineered to identify enterprise customer churn risk proactively, drawing on authentic operational records extracted from a business-tobusiness telecommunications environment. The study follows the Cross-Industry Standard Process for Data Mining (CRISP-DM) lifecycle. A dataset of 8,454 unique business accounts, characterised by 14 raw attributes and enriched to a final 22-variable feature set, underpins the empirical work. Pronounced class imbalance, churned accounts representing approximately 6.5minority ratio of 14.3:1, necessitated specialised resampling prior to classifier training. Five oversampling strategies were benchmarked; SVMSMOTE produced the largest gain in minority-class sensitivity and was adopted for all subsequent training cycles. Ten classifier families were trained and assessed, including EasyEnsembleClassifier, RUSBoostClassifier, XGBoost, LightGBM, CatBoost, Histogram Gradient Boosting, Balanced Bagging, a multilayer perceptron, a soft-voting ensemble, and a stacking ensemble. EasyEnsembleClassifier emerged as the leading model, attaining an F1-score of 0.129 and a recall of 38.242 of 110 churned accounts. Post-hoc explainability analysis through SHAP and LIME identified active subscriber rate, geographic billing zone, and engineered interaction terms as the dominant predictive signals. The framework was operationalised within a FastAPI-based application supporting realtime individual scoring, batch CSV prediction, and retention campaign monitoring. The projected annual revenue protection under conservative assumptions exceeds 74,000 currency units. The study illustrates that interpretable, explainability-augmented machine learning frameworks can bridge the gap between quantitative model output and managerial action, offering a replicable blueprint for data-driven churn governance in both emerging and mature telecommunications markets.
The rapid growth in wireless communications and the increasing scarcity of spectrum necessitate intelligent and adaptive technologies for efficient utilization of available resources. Dynamic Spectrum Access (DSA) in Cognitive Radio Networks (CRNs) offers a promising solution to these challenges. However, achieving real-time spectrum agility and interference mitigation remains a technical hurdle. This paper presents a novel artificial intelligence (AI)-assisted adaptive beamforming scheme based on reinforcement learning (RL) to dynamically steer antenna beams toward legitimate users while suppressing interference. An 8-element Uniform Linear Array (ULA) operating at 2.4 GHz is modeled in MATLAB, and a Q-learning algorithm is employed to learn optimal beamforming weights through spectrum feedback. Simulation results demonstrate that the RL-based approach achieves a 4.9 dB improvement in Signal-to-Interference-plus-Noise Ratio (SINR) and 38% faster convergence compared to classical Least Mean Squares (LMS) algorithms. Unlike conventional adaptive beamforming methods, the proposed scheme does not require prior knowledge of the interference environment or channel statistics, enabling autonomous adaptation in highly dynamic spectrum conditions. Moreover, the system exhibits robustness to user mobility and Signal-to-Noise-Ratio (SNR) variations, making it suitable for cognitive base stations, Unmanned Aerial Vehicles (UAV) communications, and Spectrum-sharing Internet of Things (IoT) environments. These results indicate that reinforcement learning–driven beam control can serve as a practical enabler for real-time spectrum intelligence in next-generation wireless systems. This work underscores the potential of intelligent beamforming for next-generation wireless systems and sets the stage for future enhancements using deep RL and hybrid beamforming architectures.
Motor Imagery (MI)-based electroencephalography (EEG) signal classification plays a pivotal role in enhancing Brain-Computer Interface (BCI) systems, particularly in medical diagnostics, neurorehabilitation, and assistive technologies. However, EEG signals are inherently non-stationary and often contaminated with noise and artifacts, making accurate classification a significant challenge. Additionally, the high dimensionality of raw EEG data further complicates the feature extraction and classification process. To address these issues, we present an optimized deep learning approach that integrates a Deep Neural Network (DNN) with Teacher Learning-Based Optimization (TLBO). This hybrid model is designed to enhance the quality of feature selection, reduce irrelevant information, and improve overall classification performance. The proposed method involves a three-stage pipeline: discrete wavelet transforms (DWT)-based feature extraction, TLBO for selecting the most informative features and mitigating noise, and a DNN for classification. The TLBO algorithm, inspired by the pedagogical process between teachers and students, facilitates global optimization in the feature space. This integration ensures that only the most discriminative EEG features are used for training, thereby improving the robustness and generalization ability of the classification model. Extensive experimental validation has been performed using benchmark datasets from BCI Competition III and IV. The results demonstrate that the proposed approach significantly outperforms traditional classifiers and other deep learning baselines in terms of accuracy, precision, sensitivity, and specificity. For example, classification accuracy improved up to 97.05% in certain frequency bands, surpassing conventional methods such as BN and EBL. These findings highlight the potential of the proposed method for real-time BCI applications, such as motor control for prosthetic devices, wheelchair navigation, and communication systems for individuals with severe motor impairments. This work contributes to the advancement of intelligent BCI systems by offering a scalable, accurate, and computationally efficient solution for MI-EEG signal classification.
