This work introduces Lean Adaptive XGBoost for Regression (LAX‐Reg), a novel, high‐efficiency algorithm designed for intelligent processing of nonstationary data streams. The rising volume and velocity of data generated by connected systems necessitate models that can adapt in real time to concept drift. While data stream classification has been extensively studied, data stream regression, particularly using boosting‐based methods, remains comparatively underexplored. LAX‐Reg builds on XGBoost while avoiding the alternating‐model paradigm commonly adopted in existing stream adaptations, which can lead to abrupt performance degradation during model replacement and increased training overhead. Instead, it maintains a single, continuously updated ensemble with bounded complexity through dynamic tree management: outdated trees are selectively removed, remaining trees are updated, and new trees are added using either a FIFO strategy or a Target strategy driven by individual ADWIN drift detectors. The proposed approach is evaluated using prequential evaluation on nine real and synthetic data streams, considering mean squared error (MSE) alongside execution times and memory consumption and compared against five state‐of‐the‐art baselines, including ARF‐Reg and AFXGB. Experimental results show that LAX‐Reg variants achieve the best average predictive rankings and belong to the leading statistical group according to the Nemenyi post hoc test, while delivering substantial efficiency gains: up to 112 × faster execution and 456 × lower memory usage compared with ARF‐Reg. These results highlight LAX‐Reg as a competitive and lightweight solution for adaptive regression in data streams, while also motivating future work on alternative drift detectors, recurring drift scenarios, and the incorporation of temporal features.

In the current economy, companies are increasingly interested in optimizing their logistics operations to reduce costs and increase agility. Transport logistics is one of the processes to be optimized, since companies have a limited fleet of heterogeneous vehicles, with particular capacities and costs, and have to attend to their customers within restricted periods. These features characterize the problem as a Heterogeneous Fixed Fleet Vehicle Routing Problem with Time Window (HFVRPTW). To solve this problem, this work proposes a method based on the Adaptive Large Neighborhood Search (ALNS) metaheuristic particularly focused on selecting vehicles that reduce the costs of the used fleet. The experiments showed that the proposed method improved\(69.6\%\)of the benchmark instances compared with the literature state-of-the-art, with\(0.44\%\)of average reduction in the total cost. Besides that, the implemented ALNS algorithm was around\(35\)times faster to run than the most relevant compared work.

Several works in GIScience propose approaches to identify general motion patterns through the analysis of objects’ trajectories. However, they are not suitable to identify functional relationships in scenarios where domain-dependent motion behaviors exist. In this work, we explore the identification of a particular pattern found in team-based invasion sports. We propose a method to identify defensive assignments between players of opposite teams based on the analysis of their trajectories. A defensive assignment happens when a defensive player blocks or hinders the progress of an opponent player inside his/her field. The defensive assignment can be classified as a behavioral pattern because it combines other behavioral patterns, such as pursuit, evasion, attack and defense. The identification of the assignments takes into account the following aspects of the players’ trajectories: proximity, position, speed and direction. The assessments pointed out that the method provides promising results, achieving 84% of success rate when compared with the analysis of a human specialist.

The amount of spatiotemporal data collected by gadgets is rapidly growing, resulting in increasing costs to transfer, process and store it. In an attempt to minimize these costs several algorithms were proposed to reduce the trajectory size. However, to choose the right algorithm depends on a careful analysis of the application scenario. Therefore, this paper evaluates seven general purpose lossy compression algorithms in terms of structural aspects and performance characteristics, regarding four transportation modes: Bike, Bus, Car and Walk. The lossy compression algorithms evaluated are: Douglas-Peucker (DP), Opening-Window (OW), Dead-Reckoning (DR), Top-Down Time-Ratio (TS), Opening-Window Time-Ratio (OS), STTrace (ST) and SQUISH (SQ). Pareto Efficiency analysis pointed out that there is no best algorithm for all assessed characteristics, but rather DP applied less error and kept length better-preserved, OW kept speed better-preserved, ST kept acceleration better-preserved and DR spent less execution time. Another important finding is that algorithms that use metrics that do not keep time information have performed quite well even with characteristics time-dependent like speed and acceleration. Finally, it is possible to see that DR had the most suitable performance in general, being among the three best algorithms in four of the five assessed performance characteristics.

Inexperienced drivers usually use the most-known paths to move inside the cities, while drivers with a better knowledge of the road network normally taken alternative routes that are shorter, faster or safer. This knowledge about roads usage, when shared with other drivers, could offer more paths options to distribute the traffic load across the city by suggesting alternative routes. However, the problem lies in how to suggest alternative route directions for ordinary drivers considering knowledge gathered from experienced drivers. In order to try to solve this problem, it is proposed an algorithm, named TODS - Trajectory Outlier Detection and Segmentation, to group and segment car road trajectories in standard and alternative routes based on city roads usage in different day times periods. After that, the segmentation results are suggested as driving directions for ordinary drivers. To evaluate the results was performed a qualitative comparison with TRA-SOD algorithm considering the segmentation process. The tests were executed using two trajectories datasets collected by drivers in San Francisco - USA and Joinville - Brazil. The results assessment indicate that TODS is superior to TRA-SOD due to its segmentation characteristics. Besides that, it has been observed that the time period of the day influences how routes are used along the day.

Coding at the sample level in still image watermarking takes advantage of avoiding a non-optimum initial diversity stage, used in many watermarking systems for tailoring Gaussian channels with desired properties. Nevertheless, the low signal-to-noise ratio (SNR) encountered, due to the perceptual constraints imposed to the watermark, makes necessary the use of very powerful coding techniques as it happens in deep-space communications.We study the application of turbo coding to sample-level discrete cosine transform (DCT) domain watermarking. This near-optimum codification strategy is possible thanks to the existence of statistical models in the DCT domain, that also permit the computation of theoretical bounds for performance.