The concept of data spaces has emerged as a structured, scalable solution to streamline and harmonize data sharing across established ecosystems. Simultaneously, the rise of AI services enhances the extraction of predictive insights, operational efficiency, and decision-making. Despite the potential of combining these two advancements, integration remains challenging: data spaces technology is still developing, and AI services require further refinement in areas like ML workflow orchestration and energy-efficient ML algorithms. In this paper, we introduce an integrated architectural framework, developed under the Green.Dat.AI project, that unifies the strengths of data spaces and AI to enable efficient, collaborative data sharing across sectors. A practical application is illustrated through a smart farming use case, showcasing how AI services within a data space can advance sustainable agricultural innovation. Integrating data spaces with AI services thus maximizes the value of decentralized data while enhancing efficiency through a powerful combination of data and AI capabilities.

Abstract Predictive analytics over mobility data is of great importance since it can assist an analyst to predict events, such as collisions, encounters, traffic jams, etc. A typical example is anticipated location prediction, where the goal is to predict the future location of a moving object, given a look-ahead time. What is even more challenging is to be able to accurately predict collective behavioural patterns of movement, such as co-movement patterns as well as their course over time. In this paper, we address the problem of Online Prediction of Co-movement Patterns. Furthermore, in order to be able to calculate the accuracy of our solution, we propose a co-movement pattern similarity measure, which facilitates the comparison between the predicted clusters and the actual ones. Finally, we calculate the clusters’ evolution through time (survive, split, etc.) and compare the cluster evolution predicted by our framework with the actual one. Our experimental study uses two real-world mobility datasets from the maritime and urban domain, respectively, and demonstrates the effectiveness of the proposed framework.

Forecasting vessel locations is of major importance in the maritime domain, with applications in safety, logistics, etc. Nowadays, vessel tracking has become possible largely due to the increased GPS-based data availability. This paper introduces a novel Vessel Location Forecasting (VLF) framework, based on Long-Short Term Memory (LSTM) Neural Networks, aiming to perform effective location forecasting in time horizons up to 60 minutes, even for vessels not recorded in the past. The proposed VLF framework is specially designed for handling vessel data by addressing some major GPS-related obstacles including variable sampling rate, sparse trajectories, and noise contained in such data. Our framework also learns by incorporating a novel trajectory data augmentation method to improve its predictive power. We validate VLF framework using three real-word datasets of vessels moving in different sea areas, comparing with various methods, and examining several aspects. Results prove VLF framework’s generic nature, robustness regarding parameter changes, and superiority against state of the art in terms of prediction accuracy (higher than 30%) and computational effort.

Maritime transport systems are essential to human mobility. A vital part of the maritime transport systems is the accurate vessel route forecasting (VRF). However, accurate VRF is a challenging task due to the fact that maritime traffic conditions are complex and dynamic. Machine learning (ML) methods can leverage from the ”explosion” of vessel surveillance information in order to encourage, enable deeper digitalization in the shipping industries and tackle the VRF problem. In this paper we investigate some of the most popular ML methods to address the VRF problem and we present an overview of these methods through an experimental testbed based on real vessel surveillance data. This work results in introducing baseline ML models for VRF purposes.

Within the last years the shipping industry invest-ments continue to grow to improve maritime transport systems. A vital part of the maritime transport systems is the accurate Vessel Traffic Flow Forecasting (VTFF). In this paper, we approach the VTFF problem from two different perspectives: a) indirect - as a vessel route forecasting application via employing predicted vessels locations in the future, and b) direct - as a flow sequence forecasting problem. In both strategies, machine learning methods are employed because they can leverage from the massive vessel surveillance information to enable deeper digitalization in the shipping industry. This work performs an experimental comparative study between the two approaches over a real dataset from the maritime domain.

The wide spread of Automatic Identification System (AIS) and tools based on it has motivated several maritime analytics operations. One of the most critical operations for the purpose of maritime safety is the so-called Vessel Collision Risk Assessment (VCRA). Accurate VCRA is a challenging task as maritime traffic is quite volatile, often affected by external factors, such as weather, etc. Addressing this problem by using complex models introduces a trade-off between accuracy quality and responsiveness. On the other hand, Machine Learning (ML) methods can better address this tradeoff. In this paper, we study the VCRA problem from the ML perspective, by proposing an architecture based on the Multi-Layered Perceptron (MLP) model. Our preliminary experimental study over a large-scale AIS dataset shows that the proposed methodology outperforms the kinematic equations-based approach.

The i4sea research project provides effective and efficient big data integration, processing and analysis technologies to deliver both real-time and historical operational snapshots of fishing vessels activity in national sea areas. This paper presents the architecture of the i4sea big data platform for sea area monitoring and analysis of fishing vessels activity and demonstrates the operation of some use-case pilot scenarios.

We present a big data framework for the prediction of streaming trajectory data, enriched from other data sources and exploiting mined patterns of trajectories, allowing accurate long-term predictions with low latency. To meet this goal, we follow a multi-step methodology. First, we efficiently compress surveillance data in an online fashion, by constructing trajectory synopses that are spatio-temporally linked with streaming and archival data from a variety of diverse and heterogeneous data sources. The enriched stream of trajectory synopses is stored in a distributed RDF store, supporting data exploration via SPARQL queries. The enriched stream of synopses along with the raw data is consumed by trajectory prediction algorithms that exploit mined patterns from the RDF store, namely medoids of (sub-) trajectory clusters, which prolong the horizon of useful predictions. The framework is extended with offline and online interactive visual analytics tool to facilitate real world analysis in the maritime and the aviation domains.

This brief presents a novel learning scheme for categorical data based on radial basis function (RBF) networks. The proposed approach replaces the numerical vectors known as RBF centers with categorical tuple centers, and employs specially designed measures for calculating the distance between the center and the input tuples. Furthermore, a fast noniterative categorical clustering algorithm is proposed to accomplish the first stage of RBF training involving categorical center selection, whereas the weights are calculated through linear regression. The method is applied on 22 categorical data sets and compared with several different learning schemes, including neural networks, support vector machines, naïve Bayes classifier, and decision trees. Results show that the proposed method is very competitive, outperforming its rivals in terms of predictive capabilities in the majority of the tested cases.

This paper presents a novel scheme for the estimation of large earthquake event occurrence based on radial basis function (RBF) neural network (NN) models. The input vector to the network is composed of different seismicity rates between main events, which are easy to calculate in a reliable manner. Training of the NNs is performed using the powerful fuzzy means training algorithm, which, in this case, is modified to incorporate a leave-one-out training procedure. This helps the algorithm to account for the limited number of training data, which is a common problem when trying to model earthquakes with data-driven techniques. Additionally, the proposed training algorithm is combined with the Reasenberg clustering technique, which is used to remove aftershock events from the catalog prior to processing the data with the NN. In order to evaluate the performance of the resulting framework, the method is applied on the California earthquake catalog. The results show that the produced RBF model can successfully estimate interevent times between significant seismic events, thus resulting to a predictive tool for earthquake occurrence. A comparison with a different NN architecture, namely, multilayer perceptron networks, highlights the superiority of the proposed approach.

This paper presents a novel algorithm for training radial basis function (RBF) networks, in order to produce models with increased accuracy and parsimony. The proposed methodology is based on a nonsymmetric variant of the fuzzy means (FM) algorithm, which has the ability to determine the number and locations of the hidden-node RBF centers, whereas the synaptic weights are calculated using linear regression. Taking advantage of the short computational times required by the FM algorithm, we wrap a particle swarm optimization (PSO) based engine around it, designed to optimize the fuzzy partition. The result is an integrated framework for fully determining all the parameters of an RBF network. The proposed approach is evaluated through its application on 12 real-world and synthetic benchmark datasets and is also compared with other neural network training techniques. The results show that the RBF network models produced by the PSO-based nonsymmetric FM algorithm outperform the models produced by the other techniques, exhibiting higher prediction accuracies in shorter computational times, accompanied by simpler network structures.