A Deep Ship Trajectory Clustering Method Based on Feature Embedded Representation Learning

High-quality ship trajectory clustering is an important fundament for learning meaningful knowledge from massive historical trajectories, which are commonly extracted from the shipborne Automatic Identification System (AIS) data. However, clustering of massive trajectory data easily suffers from two limitations, i.e., low accuracy and high computational cost. In this work, we propose to develop an efficient ship trajectory clustering method based on the virtual maritime traffic network, which is constructed using the massive AIS data. In particular, we first adopt the popular Douglas-Peucker (DP) algorithm to simplify the ship trajectories to obtain the first type of feature points, where the moving ships change the course significantly. The second type of features points, where exist the large number of ship encounters, is also obtained through estimating the ship density map. The density-based clustering method is employed to remove the unwanted outliers in these feature points, contributing to the extraction of important waypoints. The virtual maritime traffic network is then constructed using the extracted way-points. To obviously shorten the computational cost, the virtual network is finally used to cluster the massive ship trajectories by introducing the concepts of base cluster and stream cluster. Two famous metrics, i.e., Silhouette Coefficient (SC) and Davies-Bouldin Index (DBI), will be exploited to evaluate the trajectory clustering results for our method and several competing methods. Experimental results on realistic AIS data have demonstrated the superior clustering performance of our method in terms of computational efficiency, SC and DBI metrics.

Exploring urban travel patterns can analyze the mobility regularity of residents to provide guidance for urban traffic planning and emergency decision. Clustering methods have been widely applied to explore the hidden information from large-scale trajectory data on travel patterns exploring. How to implement soft constraints in the clustering method and evaluate the effectiveness quantitatively is still a challenge. In this study, we propose an improved trajectory clustering method based on fuzzy density-based spatial clustering of applications with noise (TC-FDBSCAN) to conduct classification on trajectory data. Firstly, we define the trajectory distance which considers the influence of different attributes and determines the corresponding weight coefficients to measure the similarity among trajectories. Secondly, membership degrees and membership functions are designed in the fuzzy clustering method as the extension of the classical DBSCAN method. Finally, trajectory data collected in Shenzhen city, China, are divided into two types (workdays and weekends) and then implemented in the experiment to explore different travel patterns. Moreover, three indices including Silhouette Coefficient, Davies–Bouldin index, and Calinski–Harabasz index are used to evaluate the effectiveness among the proposed method and other traditional clustering methods. The results also demonstrate the advantage of the proposed method.

Recently, deep clustering methods have achieved remarkable results compared to traditional clustering approaches. However, its performance remains constrained by the absence of annotations. A thought-provoking observation is that there is still a significant gap between deep clustering and semi-supervised classification methods. Even with only a few labeled samples, the accuracy of semi-supervised learning is much higher than that of clustering. Given that we can annotate a small number of samples in a certain unsupervised way, the clustering task can be naturally transformed into a semi-supervised setting, thereby achieving comparable performance. Based on this intuition, we propose ClusMatch, a unified positive and negative pseudo-label learning based semi-supervised learning framework, which is pluggable and can be applied to existing deep clustering methods. Specifically, we first leverage the pre-trained deep clustering network to compute predictions for all samples, and then design specialized selection strategies to pick out a few high-quality samples as labeled samples for supervised learning. For the unselected samples, the novel unified positive and negative pseudo-label learning is introduced to provide additional supervised signals for semi-supervised fine-tuning. We also propose an adaptive positive-negative threshold learning strategy to further enhance the confidence of generated pseudo-labels. Extensive experiments on six widely-used datasets and one large-scale dataset demonstrate the superiority of our proposed ClusMatch. For example, ClusMatch achieves a significant accuracy improvement of 5.4% over the state-of-the-art method ProPos on an average of these six datasets.

Trajectory association and clustering are two key problems in motion analysis. While association links the points of interest to form trajectories, clustering discovers motion patterns of these trajectories and group them into clusters. Despite mutually related, the two problems have been typically studied separately in the literature. In this letter, we formulate them as a unified optimization problem and take the advantage of high-order information to capture the interrelations for Motion Segmentation by Trajectory Association and Clustering (MSTAC). To solve this unified problem, we propose an alternating optimization strategy to improve the association and clustering in each iteration. Specifically, a tensor-based multidimensional assignment method with high-order motion context information is proposed for trajectory association; and a minimum cost multicut-based trajectory clustering method is introduced for trajectory clustering. While the association process provides incomplete trajectories to clustering, the clustering method presents high-order context information to improve the performance of association; and thus they benefit from each other. Experiments on the Hopkins 155 dataset and a realistic airport sequence demonstrate that the proposed MSTAC framework obtains high accuracy on both trajectory association and clustering.

This study investigates the sectoral impact of Donald Trump’s election and subsequent policy implementations on the stock performance of LQ45 companies in Indonesia. By leveraging unsupervised learning techniques, we aim to cluster the price performance of 45 listed companies to identify underlying market behavior patterns during this geopolitical shift. Three clustering methods are employed to capture different structural characteristics of the data: K-Means (partition-based), DBSCAN (density-based), and Agglomerative Hierarchical Clustering (AHC). The resulting clusters are evaluated using internal cluster evaluation such as Silhouette Coefficient, Davies–Bouldin Index, and Calinski–Harabasz Index to ensure robustness and interpretability. The analysis reveals that most LQ45 stocks grouped in Cluster 0 experienced significant foreign capital outflows. In contrast, Cluster 1 contained a single outlier stock (GOTO), which exhibited strong foreign inflows and high transaction volume, indicating strategic investor interest likely tied to expectations of benefiting from trade reallocation or domestic resilience.These findings offer valuable insights into market segmentation under external political shocks and highlight the importance of sector-specific resilience in Indonesia's capital market.

<span lang="EN-US">Automatic identification system (AIS) is a vessel radio navigation equipment that has been determined by international maritime organization (IMO). Historical AIS data can be utilized for anomaly detection, trajectory prediction, and vessel trajectory planning. These benefits can be achieved by identifying the vessel's trajectory pattern through trajectory clustering. However, more effort is needed in trajectory clustering using AIS data due to their large volume and the significant number of deficiencies. In addition, trajectory clustering cannot be directly applied to trajectory data, which also applies to vessel trajectory. Therefore, we propose a trajectory clustering framework by combining douglas peucker (DP), longest common subsequence (LCSS), multi-dimensional scaling (MDS), and density-based spatial clustering of applications with noise (DBSCAN). Our experiments, carried out with AIS data for the Lombok Strait, Indonesia, showed that the trajectory compression with DP significantly accelerates the similarity measurement process. Moreover, we found that the LCSS is the optimal algorithm for similarity measurement of vessel trajectories based on AIS data. We also applied the right combination of MDS and DBSCAN in density-based clustering. The proposed framework can distinguish trajectoriess in different directions, identify the noise, and produce good quality clusters in relatively fast total processing time.</span>

Oceanic trajectories frequently exhibit multiple periodic patterns across various time intervals, e.g., tidal variations, mesoscale eddies, and El Niño events correspond to diurnal, seasonal, and interannual fluctuations in environmental factors. To explore hidden spatiotemporal multiple periodic behaviors in noisy ocean data, we propose a novel trajectory clustering method, namely DTID-STFC. It first identifies dense time intervals (DTIs) in which trajectories occur frequently. Subsequently, within each DTI, it utilizes spectral embedding to project trajectories onto a latent subspace and proposes three-way fuzzy clustering to obtain results. We evaluate the proposed method on simulated datasets and compare it with traditional and state-of-the-art trajectory clustering approaches. Experimental results indicate that it outperforms other methods across all five metrics. Moreover, when applying the DTID-STFC method to the analysis of mesoscale cyclonic eddies in the South China Sea and vessel data, it demonstrates more discernible results than traditional methods, and it aligns well with physical oceanographic processes. This proposed method offers valuable insights into identifying periodic behaviors from complex and noisy spatiotemporal oceanic trajectory data.

We present a scene understanding strategy for video sequences based on clustering object trajectories. In this chapter, we discuss a set of relevant feature spaces for trajectory representation and we critically analyze their relative merits. Next, we examine various trajectory clustering methods that can be employed to learn activity models, based on their classification into hierarchical and partitional algorithms. In particular, we focus on parametric and non-parametric partitional algorithms and discuss the limitations of existing approaches. To overcome the limitations of state-of-the-art approaches we present a soft partitional algorithm based on non-parametric Mean-shift clustering. The proposed algorithm is validated on real datasets and compared with state-of-the-art approaches, based on objective evaluation metrics.

Trajectory clustering has played an essential role in trajectory mining tasks. It serves in a wide range of real-life applications, including transportation, location-based services, behavioral study, and so on. To support trajectory clustering analytics, a plethora of trajectory clustering methods have been proposed, which mainly extend traditional clustering algorithms by using spatio-temporal characteristics of trajectories. However, existing traditional trajectory clustering approaches based on raw trajectory representation highly rely on hand-craft similarity metrics, and can not capture hidden spatial dependencies in trajectory data, which is inefficient and inflexible for clustering analysis. To this end, we propose an end-to-end deep trajectory clustering framework via self-training, termed as E2DTC, inspired by the data-driven capabilities of deep neural networks. E2DTC does not require any additional manual feature extraction operations, and can be easily adapted for trajectory clustering analytics on any trajectory dataset. Extensive experimental evaluations on three real-life datasets show that our framework E2DTC achieves superior accuracy and efficiency, compared with classical clustering methods (i.e., K-Medoids) and state-of-the-art neural-network based approaches (i.e., t2vec).

Based on a data set (100 biographies) created by the authors through content analysis of biographical material about outstanding scientists of the 19th and 20th centuries in the humanities and natural sciences, the clustering of professional trajectories of gifted individuals was carried out. Neural network analysis based on self-organizing Kohonen maps was used as a clustering method. The professional trajectories were formed within the framework of the behavioral model of the linear-stage approach to studying life cycles. Within this approach, career and professional self-realization are understood as a sequence of evolutionary stages fixed in their order of occurrence. Each stage was encoded, and the biographies were transformed into a vector system. In turn, the task of clustering consisted in dividing a hundred vectors into typical groups with several real-valued coordinates. The criteria for the quality of clustering were the minimum sum of quantization errors and the silhouette coefficient. As a result of the study, seven professional trajectories of gifted individuals were identified and interpreted. The analysis of trajectories was carried out from the point of view of the speed of success (average age of success) and those factors and conditions of the life path that could affect either rapid or slow achievement of professional goals and self-realization. This example demonstrates the possibilities and limitations of using neural network analysis for solving similar research tasks, especially when working with complex cluster forms and finding their optimal number.

Clustering is one of the most fundamental unsupervised learning tasks with numerous applications in various fields. Clustering methods based on neural networks, called deep clustering methods, leverage the representational power of neural networks to enhance clustering performance. ClusterGan constitutes a generative deep clustering method that exploits generative adversarial networks (GANs) to perform clustering. However, it inherits some deficiencies of GANs, such as mode collapse, vanishing gradients and training instability. In order to tackle those deficiencies, the generative approach of implicit maximum likelihood estimation (IMLE) has been recently proposed. In this paper, we present a clustering method based on generative neural networks, called neural implicit maximum likelihood clustering, which adopts ideas from both ClusterGAN and IMLE. The proposed method has been compared with ClusterGAN and other neural clustering methods on both synthetic and real datasets, demonstrating promising results.

The research objective was to apply unsupervised learning techniques by comparing several clustering methods in data mining to determine the best method for cell phone use cases in Indonesia. The data source comes from the Central Bureau of Statistics with the website url https://www.bps.go.id/. The data category used is the percentage of the population using cellular phones by province from 2012&#x2013;2018 (34 records). The attributes used are the name of the province, the percentage (&#x0025;) of the population using cellular telephones in urban areas and the percentage (&#x0025;) of the population using cellular phones in rural areas. The unsupervised learning methods compared were k-medoids and k-means using the T - Test validity test based on the Davies Bouldin Index (DBI) to evaluate the cluster model. The analysis process uses Rapid Miner software. The mapping label is in the form of the cluster used, namely the high cluster (K1) for cellular phone use and the low cluster (K2) for cellular phone use. The best mapping result for the use of cell phones in urban areas is the k-medoids method with the T - Test result = 0.526 (DBI with <tex>$\mathbf{k}=2$</tex>) where 26 provinces <tex>$(\mathbf{K}1=\mathbf{cluster}_{-}1)$</tex> and 8 provinces <tex>$(\mathbf{K}2\ =\mathbf{cluster}_{-}0)$</tex> &#x2022; Meanwhile, the best mapping result for the use of cellular phones in rural areas is the k-medoids method with the results of T - Test <tex>$=0.431$</tex> (DBI with k <tex>$=2)$</tex> where 31 provinces <tex>$(\mathbf{K}1=\mathbf{cluster}_{-}0)$</tex> and 2 provinces (K2 <tex>$= \mathbf{cluster}_{-}1)$</tex>. Based on the mapping results, the mastery of cellular telephones in rural areas (2 &#x0025;) and in urban areas (24 &#x0025;) means that the understanding of the use of cellular telephones in Indonesia is very good. Overall, the k-medoid method is better at mapping cases of cell phone use in Indonesia. Based on the results, it is expected to be information for the government considering that Indonesia has the opportunity to grow very fast and large with the support of the government so that Indonesia&#x0027;s digital industry can overcome the lag behind other countries.

ObjectiveCluster analysis of spatio-temporal event-related potential (ERP) data is a promising tool for exploring the measurement time window of ERPs. However, even after preprocessing, the remaining noise can result in uncertain cluster maps followed by unreliable time windows while clustering via conventional clustering methods. MethodsWe designed an ensemble deep clustering pipeline to determine a reliable time window for the ERP of interest from temporal concatenated grand average ERP data. The proposed pipeline includes semi-supervised deep clustering methods initialized by consensus clustering and unsupervised deep clustering methods with end-to-end architectures. Ensemble clustering from those deep clusterings was used by the designed adaptive time window determination to estimate the time window. ResultsAfter applying simulated and real ERP data, our method successfully obtained the time window for identifying the P3 components (as the interest of both ERP studies) while additional noise (e.g., adding 20 dB to −5 dB white Gaussian noise) was added to the prepared data. ConclusionCompared to the state-of-the-art clustering methods, a superior clustering performance was yielded from both ERP data. Furthermore, more stable and precise time windows were elicited as the noise increased. SignificanceOur study provides a complementary understanding of identifying the cognitive process using deep clustering analysis to the existing studies. Our finding suggests that deep clustering can be used to identify the ERP of interest when the data is imperfect after preprocessing.

Text mining is a powerful modern technique used to obtain interesting information from huge datasets. Text clustering is used to distinguish between documents that have the same themes or topics. The absence of the datasets ground truth enforces the use of clustering (unsupervised learning) rather than others, such as classification (supervised learning). The “no free lunch” (NFL) theorem supposed that no algorithm outperformed the other in a variety of conditions (several datasets). This study aims to analyze the k-means cluster algorithm variations (three algorithms (k-means, mini-batch k-means, and k-medoids) at the clustering process stage. Six datasets were used/analyzed in chapter Al-Baqarah English translation (text) of 286 verses at the preprocessing stage. Moreover, feature selection used the term frequency–inverse document frequency (TF-IDF) to get the weighting term. At the final stage, five internal cluster validations metrics were implemented silhouette coefficient (SC), Calinski-Harabasz index (CHI), C-index (CI), Dunn’s indices (DI) and Davies Bouldin index (DBI) and regarding execution time (ET). The experiments proved that k-medoids outperformed the other two algorithms in terms of ET only. In contrast, no algorithm is superior to the other in terms of the clustering process for the six datasets, which confirms the NFL theorem assumption.

Stunting has been widely known as the highest case of malnutrition suffered by toddlers in the world and has a bad impact on children's future. In 2018, Indonesia was ranked the 31st highest stunting in the world and ranked 4th in Southeast Asia. About 30.8% (roughly 3 out of 10) of children under 5 years suffer from stunting in Indonesia. To support the government policy making in handling stunting, it is undoubtedly necessary to classify the levels of stunting handling in regions in Indonesia. In this work, the hierarchical agglomerative and non-hierarchical clustering is compared and evaluated to perform clustering on stunting data. The agglomerative hierarchical cluster uses Single Linkage, Average Linkage, Complete Linkage, and Ward Method, while the non-hierarchical cluster uses K-Means, K-Medoids (PAM) Clustering, and Fuzzy C-Means. This study uses data from 12 IKPS indicators in 34 provinces in Indonesia in 2018. Based on the results of the evaluation using the Connectivity Coefficient, Dunn Index, Silhouette Coefficient, Davies Bouldin Index, Xie &amp; Beni Index, and Calinski-Harabasz Index, the results show that the Average Linkage is the best cluster method with the optimal number of clusters is four clusters. The first cluster is a cluster with a good level of stunting management which consists of 28 provinces. The second cluster consists of only one province, DI Yogyakarta with a very good level of stunting handling. The third cluster consists of four provinces with poor stunting handling rates. Finally, the last cluster consisting of one province, Papua, has a very poor level of stunting handling.