Abstract

To overcome Climate Change, countries are turning to greener transportation systems. Therefore, the use of Electric Vehicles (EVs) is leveraging substantially since they present multiple advantages, like reducing hazardous emissions. Recently, the demand for EVs has increased, which means that more charging stations need to be available. By the year 2030, 15 million EVs will be accessible, and since the number of charging stations is limited, the charging needs should be defined for better management of the charging infrastructure. In this research, we aim to tackle this problem by efficiently predicting the energy consumption of EVs. We proposed a Genetic Algorithm (GA) based Three HyperParameter optimization of Deep Long Short Term Memory (GA3P-DLSTM), which is an optimized LSTM model that incorporates a GA for Hyperparameter Tuning. After experimenting our methodology and performing a comparative analysis with previous studies from the literature, the obtained results showed the efficiency of our novel model, with Mean Squared Error (MSE) equals to 0.000112 and a Determination Coefficient (R) equals to 0.96470. It outperformed other models of the literature for predicting energy use based on real-world data collected from the campus of Georgia Tech in Atlanta, USA.

Abstract

This research presents a machine learning-based context-driven collaborative filtering approach with three
steps: contextual clustering, weighted similarity assessment, and collaborative filtering. User data is clustered
across 3 aspects, and similarity scores are calculated, dynamically weighted, and aggregated into a normalized
User-User similarity matrix. Collaborative filtering is then applied to generate contextual recommendations.
Experiments on the LDOS-CoMoDa dataset demonstrated good performance, with RMSE and MAE rates of
0.5774 and 0.3333 respectively, outperforming reference approaches.

Abstract

Vehicle-as-a-Service (VaaS) refers to on-demand rides and the sharing/offering of various kinds of intelligent transportation facilities (e.g., smart buses, electric vehicles, autonomous cars) to move from a source to a destination across one or several regions. Coupled with smart transportation systems—which are critical for addressing road network issues such as traffic congestion, parking shortages, and safety concerns—VaaS is increasingly being adopted in smart cities.
For example, a user may specify their needs in terms of source and destination stations, time and cost constraints, as well as preferred transportation modes and services (e.g., only connected buses and cars). Such a user profile is evaluated against available VaaS options under current and anticipated urban network conditions. However, current solutions do not support the customization of VaaS compositions and often treat user requests as traditional vehicle routing problems. In a smart mobility context, however, processing VaaS requests involves not only finding the optimal transportation path that meets user constraints (e.g., time, cost, transfer stations) but also selecting the top-rated combination of available VaaSs (e.g., a sequence of smart buses) with respect to the user profile (e.g., connectivity needs, specific facilities) and the quality of smart urban services.
To address these challenges, the goal of this paper is to develop a multi-modal recommender system that enables the personalized selection, composition, and scheduling of VaaS services while optimizing trip constraints (e.g., minimizing trip time and cost, and maximizing VaaS availability and reputation).
As a multi-population technique, Composite Particle Swarm Optimization (CPSO) is applied to aggregate the optimal set of high-quality and high-coverage VaaSs with respect to the requested trip. The regions composing the trip are explored using a modified A* search algorithm to find the optimal local (partial) path in each traversed region of the smart urban network. Comparative experiments involving two metaheuristics, a greedy algorithm, and a fuzzy clustering technique demonstrate the efficiency and superiority of our CPSO-based approach, achieving approximately a 28% improvement over its closest competitors.

Abstract

Abstract

As the adoption of unmanned aerial vehicles by both consumers and companies is growing rapidly, the use of drones is nowadays leading the way various types of packages (e.g., food, medication, supplies) are delivered. Drone services have increased companies’ benefits and saved money on their shipping costs, which resulted in a reduced delivery time and cost for consumers. However, the delivery tasks either achieved by single or swarming drones are facing several challenges, which are mainly related to the modeling of drones’ skyway network, the uncertainty of flight conditions and available resources, the consumers’ trust and quality of experience, the privacy concerns caused by shared user information (e.g., GPS, camera). Among these issues, we focus on the modeling and resource availability issues. To cope with complex delivery requests, this paper takes advantage of the service computing paradigm to propose a service composition approach, in which multiple drone services can participate in the delivery plan. We first propose a graph-based modeling of the skyway network and flights history. This latter, is fed into a proposed frequent subgraph mining algorithm, and is processed to extract relevant patterns from the previously generated delivery paths, based on consumers’ positive feedback. We also adopt Long Short-Term Memory to propose a model that forecasts the overload of charging stations, with the goal of tuning the mined (i.e., frequently traversed) paths with the low-congestion stations. The prediction and mining results are, finally, exploited, in the selection of the appropriate drone formation that will achieve the delivery mission. Experimental studies based on real-world datasets have confirmed the efficiency of our approach, compared to three graph-based approaches.

Abstract

Massive Open Online Courses (MOOCs) face high
dropout rates, often exceeding 80%, undermining their educational
potential. This study presents a comparative evaluation
of Machine Learning (ML) and Deep Learning (DL) models for
early dropout prediction using the KDDCup2015 dataset, with a
dual focus on predictive performance and model interpretability
through eXplainable AI (XAI) techniques. Among traditional
ML models, the Decision Tree (DT) achieves the highest
performance (90.18% AUC-PR by Week 4), outperforming
Logistic Regression (LR) and Support Vector Machine (SVM).
In the ensemble category, AdaBoost leads with 90.35% AUCPR.
The hybrid CNN-LSTM outperforms standalone CNN and
LSTM models, reaching up to 93,76% AUC-PR. XAI analysis
reveals that frequent platform access, navigation patterns and
problem solving activities are key predictors of dropout. These
insights support early interpretable interventions to improve
learner retention while maintaining model transparency.

Abstract

Biclustering is an important data analysis technique
that simultaneously groups coherent rows and columns of a
matrix. It has broad applications, especially in bioinformatics,
where it helps identify groups of genes with similar behavior under
specific experimental conditions, potentially revealing shared
biological functions. Although many algorithms have been developed,
including evolutionary approaches, most rely on statistical
criteria that may not reflect true biological significance, making
interpretation difficult for biologists. To address this limitation,
we propose integrating biological knowledge directly into an
evolutionary algorithm, particularly during the selection and
crossover phases. We also employ a global alignment technique
with a binary mask to guide crossover operations more effectively.
Experiments conducted on real Saccharomyces cerevisiae data
confirm that integrating biological knowledge yields biologically
meaningful and non-trivial biclusters, demonstrating the effectiveness
of our approach.

Abstract

The rapid development of devices on the Internet of Things
(IoT) and the diversity of their applications have made them
ubiquitous. However, deploying these devices in large-scale
networks presents several challenges, including limited energy
capacity, security concerns, unreliable links, and transmission
delays. This paper, proposes a multi-objective optimization
approach for wireless IoT networks based on machine learning
techniques. Specifically, a clustering scheme is developd by
using an improved k-means algorithm. This is combined
with a dynamic routing strategy based on multi-objective
Q-learning using parallel Q-tables. This approach leads to
measurable gains in energy efficiency, transmission latency,
and reliability. Compared to existing approaches in similar
contexts, such as weighted sum, the proposed solution achieves
significant improvements in overall network performance.

Abstract

Artificial Intelligence (AI) is revolutionizing information security, influencing both attack and defense strategies. Attackers leverage AI to automate cyberattacks and exploit vulnerabilities, while defenders utilize it for anomaly detection, predictive threat modeling, and automated responses. Federated Learning (FL), a privacy-preserving training method, remains vulnerable to inference attacks. To address this, we propose the Rényi Differential Privacy (RDP) based federated learning (RDP-FL) framework, which incorporates moment accounted noise scaling to dynamically regulate the privacy budget, achieving an optimal balance between privacy and utility. This method minimizes unnecessary noise addition while maintaining strong privacy guarantees, thereby preserving data integrity and enhancing model performance. Experimental validation on the Medical-MNIST and CIFAR-10 datasets demonstrates the effectiveness of RDP-FL, showing its ability to safeguard data privacy while ensuring high classification accuracy. This work advances the ongoing efforts to enhance cybersecurity in an AI-driven landscape.

Abstract

This paper tackles the Green Vehicle Routing Problem (GVRP), where vehicles with limited driving range must visit customers while recharging at Alternative Fuel Stations (AFSs). We propose a Multi-Start Tabu Search with Set Partitioning (MSTS-SP) approach structured in two phases. In the first phase, MSTS-SP uses a new constructive heuristic, Randomized Sectoring with Repair, to generate diverse initial solutions, which are then improved through multiple independent tabu search runs. The high-quality routes found during these runs are collected into a global pool. In the second phase, an exact set partitioning model is applied to this pool to select the best combination of routes. Computational experiments on 52 GVRP benchmark instances show that MSTS-SP matches 46 known best solutions (88%) and improves upon the best known solution for one large instance. These results demonstrate that MSTS-SP offers a competitive balance between solution quality and computational efficiency compared to state-of-the-art methods.