Inter-cluster Separation Research Articles

Topic Modeling of the 2024 Election Using the BERTopic Method on Detik.com News Articles

In 2024, Indonesia will hold simultaneous general elections dominated by the participation of young people, especially Generation Z and Millennials, who seek political information primarily through the Internet, highlighting the crucial role of digital media in shaping public opinion. Detik.com is actively reporting on the 2024 elections, evidenced by a special election subchannel. However, the lack of topic categorization in this subchannel makes it difficult for readers to find in-depth information, and tracking and analyzing the large volume of news articles published daily is a significant challenge. This study employs Topic Modelling techniques, specifically the BERTopic method, to analyze topics related to the 2024 elections from Kompas.com news articles. The dataset, sourced from the detik.com election sub-channel, was collected via scraping from September 1, 2023, to February 14, 2024, totalling 15,019 articles. The text preprocessing involves case folding, cleaning, tokenizing, and stopword removal. Topic modelling using BERTopic includes embeddings with sentence-transformers "distiluse-base-multilingual-cased-v1," dimensionality reduction with UMAP, clustering with K-Means using optimal k=5 value evaluated by Elbow, tokenizer with CountVectorizer, and weighting scheme using c-TF-IDF. Based on the Silhouette Score of 0.566 and Silhouette plot results, the clustering results using the K-Means model with a value of k equal to 5 produce good clustering with clear inter-cluster distances. For other evaluations, the SSE value of 70223.257 provides an overview of the cluster distribution, the Davies-Bouldin Index of 0.758 shows that the cluster has a relatively good level of inter-cluster separation with good closeness within the cluster, the Calinski-Harabasz Index of 20083.489 shows good and compact inter-cluster separation, and the Dunn Index of 0.003 shows outliers that cause overlapping clusters and lack of clear separation. The evaluation results show that implementing the K-Means model with a value of k equal to 5 again emphasizes that the clustering results are good. The modelling results show an average topic coherence value of 0.0902 and produce five main topics in the 2024 election news on Detik.com topic 0: about presidential and vice presidential candidates (5,215 articles) with the representation of the words 'ganjar', 'prabowo', 'anies' and 'imin', topic 1: about general elections and related surveys (3,191 articles) with the representation of the words '2024', 'pemilu', 'pilpres' and 'suara', topic 2: news about Joko Widodo President (2,604 articles), topic 3: news about presidential and vice presidential debates (2046 articles) with representations of the words 'presiden', 'jokowi', 'demokrat' and 'politik' and topic 4: news about the figure Gibran Rakabuming Raka and related issues (1963 articles) with representations of the words 'raka', 'rakabuming', 'nomor' and 'urut'. Using the results of this research, readers can gain insights into the most discussed issues and the attention given to key figures in the 2024 election news on the detik.com news portal.

Energy efficient cluster-based routing protocol for WSN using multi-strategy fusion snake optimizer and minimum spanning tree

In recent years, the widespread adoption of wireless sensor networks (WSN) has resulted in the growing integration of the internet of things (IoT). However, WSN encounters limitations related to energy and sensor node lifespan, making the development of an efficient routing protocol a critical concern. Cluster technology offers a promising solution to this challenge. This study introduces a novel cluster routing protocol for WSN. The system selects cluster heads and relay nodes utilizing the multi-strategy fusion snake optimizer (MSSO) and employs the minimum spanning tree algorithm for inter-cluster routing planning, thereby extending the system’s lifecycle and conserving network energy. In pursuit of an optimal clustering scheme, the paper also introduces tactics involving dynamic parameter updating, adaptive alpha mutation, and bi-directional search optimization within MSSO. These techniques significantly increase the algorithm convergence speed and expand the available search space. Furthermore, a novel efficient clustering routing model for WSN is presented. The model generates different objective functions for selecting cluster heads and relay nodes, considering factors such as location, energy, base station distance, intra-cluster compactness, inter-cluster separation, and other relevant criteria. When selecting cluster heads, the fuzzy c-means (FCM) algorithm is integrated into MSSO to improve the optimization performance of the algorithm. When planning inter-cluster routing, the next hop node is selected for the relay node based on distance, residual energy, and direction.The experimental results demonstrate that the proposed protocol reduces energy consumption by at least 26.64% compared to other cluster routing protocols including LEACH, ESO, EEWC, GWO, and EECHS-ISSADE. Additionally, it increases the network lifetime of WSN by at least 25.84%, extends the stable period by at least 52.43%, and boosts the network throughput by at least 40.99%.

DEA-based internal validity index for clustering

Internal validity indices are crucial in evaluating the quality of clustering results, serving as valuable tools for comparing various clustering algorithms and determining the optimal number of clusters for datasets. Most existing internal validity indices use the worst-case scenario to represent the overall validity. Moreover, some indices assign equal weights to distances among different clusters, even when these distances might have varying degrees of influence on the overall validity. Data envelopment analysis (DEA) is an effective technique for evaluating the performance of decision-making units through the computation of the ratio of the weighted sum of outputs to the weighted sum of inputs. The weight assigned to each indicator signifies its degree of influence on efficiency. Furthermore, DEA can be viewed as a multiple-criteria evaluation methodology, wherein inputs and outputs are two sets of performance criteria. We propose a DEA-based internal validity index (DEAI) to evaluate the validity of the clustering results. In this approach, intra-cluster compactness and inter-cluster separation are employed for determining the input(s) and output(s). The DEAI is then applied to the artificial datasets and empirical examples. Experimental results illustrate that DEAI outperforms six classic internal validity indices in accurately identifying the optimal cluster across all 10 datasets.

Study on cluster centers optimization of max-min distance k-means clustering algorithm based on inter-cluster separation measure

Clustering is an essential unsupervised technique when category information is not available. Although K-means and Max-min distance K-means clustering algorithms are widely used, they have some disadvantages such as dependence on the initial centers, sensitivity to outliers caused by using only distance as the clustering criterion. To overcome the problems, this paper proposes SMM-K-means algorithm which overcomes the dependence on the initial cluster centers and the initial number of clusters and the sensitivity to the outliers. First, the initial value K of the optimal cluster number is determined by the elbow method, and K-means is used for initial clustering. A new inter-cluster separation measure is then constructed based on the idea of q-nearest neighbors, which is constructed by comprehensive considering the separation between clusters and the distribution compactness of clusters themselves. Finally, the two sample points with highest degree of separation are brought into Max-min distance K-means algorithm as new initial centers for clustering. The definite determining method of cluster centers eliminates the complicated iterative calculation, and the construction of inter-cluster separation measure overcomes the sensitivity of clustering results to noise points and isolated points, and has good applicability and generalization. In addition, this algorithm is not limited by the shape and size of the clusters and has better flexibility. The experimental results show that the SMM-K-means algorithm has higher CH values, resulting in a better clustering effect and stability.

Dual Contrast-Driven Deep Multi-view Clustering.

Consensus representation learning is one of the most popular approaches in the field of multi-view clustering. However, most of the existing methods cannot learn discriminative representations with a clustering-friendly structure since these methods ignore the separation among clusters and the compactness within each cluster. To tackle this issue, we propose a new deep multi-view clustering network with a dual contrastive mechanism to learn clustering-friendly representations. Specifically, our method employs dual contrasting losses: a dynamic cluster diffusion loss to maximize the distance between different clusters and a reliable neighbor-guided positive alignment loss to enhance compactness within each cluster. Our approach includes several key components: view-specific encoders to extract high-level features from each view, and an adaptive feature fusion strategy to obtain consensus representations across multiple views. The dynamic cluster diffusion module ensures inter-cluster separation by maximizing distances between different clusters in the consensus feature space. Simultaneously, the reliable neighbor-guided positive alignment module improves within-cluster compactness through a pseudo-label and nearest neighbor structure-driven contrastive loss. Experimental results on several datasets show that our method can acquire clustering-friendly representations with both good properties of inter-cluster separation and within-cluster compactness, and outperforms the existing state-of-the-art approaches in clustering performance.

Feature selection for binary classification based on class labeling, SOM, and hierarchical clustering

Feature selection plays an important role in algorithms for processing high-dimensional data. Traditional pattern classification and information theory methods are widely applied to feature selection methods. However, traditional pattern classification methods such as Fisher Score, Laplacian Score, and relief use class labels inadequately. Previous information theory based feature selection methods such as MIFS ignore the intra-class to tight inter-class to sparse property of the samples. To address these problems, a feature selection algorithm for the binary classification problem is proposed, which is based on class label transformation using self-organizing mapping neural network (SOM) and cohesive hierarchical clustering. The algorithm first converts class labels without numerical meaning into numerical values that can participate in operations and retain classification information through class label mapping, and constitutes a two-dimensional vector from it and the attribute values to be judged. Then, these two-dimensional vectors are clustered by using SOM neural network and hierarchical clustering. Finally, evaluation function value is calculated, that is closely related to intra-cluster to tightness, inter-cluster separation, and division accuracy after clustering, and is used to evaluate the ability of alternative attributes to distinguish between classes. It is experimentally verified that the algorithm is robust and can effectively screen attributes with strong classification ability and improve the prediction performance of the classifier.

An improved collaborative filtering model based on time weighted correlation coefficient and inter-cluster separation

An improved collaborative filtering model based on time weighted correlation coefficient and inter-cluster separation

Unsupervised anomaly analysis-based manufacturing quality test and grading method for combine harvesters

Manufacturing quality tests mainly depend on manual methods, and most quality evaluations rely on the quality evaluation intervals, which lacks quantitative means for grading. As a consequence, it is challenging for the test process to find hidden quality problems early. In this paper, we conduct research on the test and grading method for combine harvester manufacturing quality and propose a hybrid method of “end-of-line test & secondary grading”, including a LabVIEW®-based test system and LOF (Local Outlier Factor)-based secondary grading method. The developed test system collects data from multiple working systems in a combine harvester and constructs the dataset. We further investigate eight unsupervised anomaly detection methods for grading harvester quality. Taking the evaluation indices of intracluster compactness (CP), intercluster separation (SP), and anomaly hit amount, we comprehensively assess these methods for different system performances in the harvester. The assessment results validate the LOF as being better among the eight anomaly detection methods, which obtains smaller CP integral values of 241.49, 433.51, and 309.93 and larger SP integral values of 383.82, 565.07, and 543.80 for the electrical system, hydraulic system, and braking performance datasets, respectively. Meanwhile, the LOF also obtains a good anomaly hit amount. Finally, the end-of-line test system integrates the advanced LOF-based grading method to quantitatively assess the harvester performance and identify potential quality-violated parts over time. The developed system proposes a solution for the automation and optimization demand in agricultural machinery manufacturing evaluation and provides technical guidance for product operation and maintenance management.

A new deep clustering method with application to customer selection for demand response program

Demand response (DR) is regarded as a promising solution to the problem of renewable energy integration, while it remains one of the key barriers for DR to target the right customers, i.e. the customers with high potential in peak reduction. Exploring the electricity consumption behaviors of customers can help DR program practitioners evaluate customers’ peak reduction potential. Clustering has been the most common approach to describe customers’ consumption behaviors in the literature. However, the existing clustering methods suffer from either unscalability or poor performance when handling a large number of daily load curves. To overcome these issues, a new deep learning-based clustering method is proposed in this paper, which integrates both intra-cluster compactness and inter-cluster separation into the objective function of clustering for consideration. Based on the clustering results, a new customer selection framework is developed to identify the potential good candidates for DR programs, which takes the stability of consumption behaviors and the characteristics at peak hours into account. Case studies on a London dataset demonstrate the superiority of the proposed clustering method over classical K-means and two state-of-the-art deep clustering methods in terms of clustering validity indexes. Furthermore, the simulations performed on an actual price-based pilot show that the customers selected by the proposed customer selection framework contribute to 59 % of total peak reduction with only accounting for 23 % of population.

Deep Reinforcement Clustering

Deep clustering has attracted plentiful attention in various domains owning to the superior performance. However, the previous deep clustering methods are guided by pre-specified clustering strategies that lack sustained explorations of data structures, degrading recognition of intrinsic patterns hidden in data. To address this challenge, deep reinforcement clustering (DRC) is proposed to learn an adaptive partition policy for pattern mining, which can fully explore structure knowledge of data in an adaptive manner. DRC is defined as a Markov decision process of data partitions, which chooses the optimal cluster prototype for data via maximizing the cumulative reward in state transition of environment. To implement the definition, a Bernoulli action prototype is devised to capture decision distributions in the transition of states, where the heavy-tailed Cauchy distribution precisely measures the structure divergences of data. Furthermore, a reward maximizing policy is designed to guide sustained explorations of data structures, which ensures intra-cluster compactness and inter-cluster separation of data partitions. Finally, extensive experiments are conducted on eight benchmark datasets, and the results demonstrate that DRC outperforms the state-of-the-art baseline methods.

A graph based approach for mining significant places in trajectory data

Significant place mining in spatiotemporal trajectory data is a key task for mobile pattern mining, useful for supporting location-aware services. State-of-the-art trajectory clustering algorithms utilize a density-based distance measure. However, some major problems with this approach are that (1) results are often inaccurate, especially on data of varying density, (2) the user must fine-tune many thresholds that are unintuitive to set, and (3) boundary points between clusters are often assigned to the wrong locations. Performance is also a major issue as many state-of-the-art algorithms have a very high time complexity. Motivated by these issues, this paper proposes an approach inspired by the data field theory and community detection. It is a graph-based significant place mining algorithm, called GB-SPM, for capturing and characterizing the essence of similarity between nodes. GB-SPM first applies a novel low index neighborhood velocity point filtration method to extract characteristic points. Then, a characteristic point index neighborhood is used to map them to graph nodes. In this way, the original problem is transformed into a community detection problem in complex community networks. Finally, a new edge weight metric is proposed to capture and characterize the nature of similarity between nodes. To evaluate clustering quality, we used the Silhouette (SI) for unannotated data to value inter-cluster separation and intra-cluster homogeneity. To evaluate mining effectiveness, we used Matthew’s correlation coefficient (MCC) for annotated data. Numerous experiments were carried out on real world datasets, and the accuracy and performance of the designed algorithm was compared with the state-of-the-art algorithms. Results show that GB-SPM improves on average SI by 13.9%, MCC by 20.7%, and runtime by 5.15 times.

Evolution of Labor Relations in the Development of Human Resources Based on Improved Genetic Algorithm

This paper systematically constructs a multi-project and a multi-objective human resource scheduling mathematical model in construction projects and puts forward an improved genetic algorithm to solve it by aiming at the problems existing in the process of human resource scheduling and optimization. Specifically, first, the basic mathematical models of GPRs and resource-constrained construction project human resource scheduling problem (RCWSP/GPRs) are established, and the multi-project equilibrium problem is extended. Then, an improved inter-cluster separation (ICS) algorithm is proposed and used to solve the RCWSP/GPRs problem. Finally, on this basis, the mathematical model of multi-project and multi-objective human resource scheduling problem and the solution method based on multi-objective-integrated circuit are proposed. At the same time, the resource-constrained multi-project and multi-skill human resource scheduling problem and the integer programming mathematical model under the generalized priority relationship are proposed. Also, the simulation results verify the accuracy of the proposed algorithm and model.

Shadowed-set-based three-way clustering methods: An investigation of new optimization-based principles

Shadowed set approximation form a cornerstone for the explainable decision advice provided by shadowed C means (SCM) clustering in unsupervised learning. Due to its advantage of managing uncertainty in fuzzy clustering, it has been used in data classification. Existing SCM clustering method requires the overall amount of uncertainty associated with a fuzzy cluster ck to be preserved in its boundary region. This requirement may suffer serious risk of having high number of unclassified patterns, especially when the uncertainty in ck is very high. Consequently, the ensuing clustering may not adequately maximize the inter-cluster separation necessary for achieving optimum cluster validity results. To tackle this problem, this paper considers new SCM clustering methods arising from (i) trade-off between uncertain and certain regions, which is necessary for refraining from making uncertain classification as much as possible, (ii) measure of sharpness balance, which helps to leverage on the location of a pattern from borderline and identify included or excluded patterns by means of their location from borderline, (iii) measure of gradualness balance, which exploits the degree of transition of a pattern out of or into ck. Each method comes with some advantages. For instance, the first and third methods may minimize the overall amount of unclassified patterns. To provide an overall evaluation of the performance of the proposed methods, a comparative study with some other shadowed set-based optimization methods are involved by considering some data sets from UCI Machine Learning repository. Friedman testing followed by Holm-Bonferroni testing are also exploited to provide statistical analysis on the performance significance of the compared methods.

ScNAME: neighborhood contrastive clustering with ancillary mask estimation for scRNA-seq data.

The rapid development of single-cell RNA sequencing (scRNA-seq) makes it possible to study the heterogeneity of individual cell characteristics. Cell clustering is a vital procedure in scRNA-seq analysis, providing insight into complex biological phenomena. However, the noisy, high-dimensional and large-scale nature of scRNA-seq data introduces challenges in clustering analysis. Up to now, many deep learning-based methods have emerged to learn underlying feature representations while clustering. However, these methods are inefficient when it comes to rare cell type identification and barely able to fully utilize gene dependencies or cell similarity integrally. As a result, they cannot detect a clear cell type structure which is required for clustering accuracy as well as downstream analysis. Here, we propose a novel scRNA-seq clustering algorithm called scNAME which incorporates a mask estimation task for gene pertinence mining and a neighborhood contrastive learning framework for cell intrinsic structure exploitation. The learned pattern through mask estimation helps reveal uncorrupted data structure and denoise the original single-cell data. In addition, the randomly created augmented data introduced in contrastive learning not only helps improve robustness of clustering, but also increases sample size in each cluster for better data capacity. Beyond this, we also introduce a neighborhood contrastive paradigm with an offline memory bank, global in scope, which can inspire discriminative feature representation and achieve intra-cluster compactness, yet inter-cluster separation. The combination of mask estimation task, neighborhood contrastive learning and global memory bank designed in scNAME is conductive to rare cell type detection. The experimental results of both simulations and real data confirm that our method is accurate, robust and scalable. We also implement biological analysis, including marker gene identification, gene ontology and pathway enrichment analysis, to validate the biological significance of our method. To the best of our knowledge, we are among the first to introduce a gene relationship exploration strategy, as well as a global cellular similarity repository, in the single-cell field. An implementation of scNAME is available from https://github.com/aster-ww/scNAME. Supplementary data are available at Bioinformatics online.

Tuning the electrical and magnetic properties in multifunctional composite materials based on the PEDOT:DBS conducting polymer and magnetite nanoparticles

Dipolar and RKKY interactions, tuned by intercluster separation, play a significant role in the magnetic behaviour of PEDOT:DBS–Fe3O4 composites.

Grid‐Based Whole Trajectory Clustering in Road Networks Environment

In the data mining of road networks, trajectory clustering of moving objects plays an important role in many applications. Most existing algorithms for this problem are based on every position point in a trajectory and face a significant challenge in dealing with complex and length‐varying trajectories. This paper proposes a grid‐based whole trajectory clustering model (GBWTC) in road networks, which regards the trajectory as a whole. In this model, we first propose a trajectory mapping algorithm based on grid estimation, which transforms the trajectories in road network space into grid sequences in grid space and forms grid trajectories by recognizing and eliminating redundant, abnormal, and stranded information of grid sequences. We then design an algorithm to extract initial clustering centers based on density weight and improve a shape similarity measuring algorithm to measure the distance between two grid trajectories. Finally, we dynamically allocate every grid trajectory to the best clusters by the nearest neighbor principle and an outlier function. For the evaluation of clustering performance, we establish a clustering criterion based on the classical Silhouette Coefficient to maximize intercluster separation and intracluster homogeneity. The clustering accuracy and performance superiority of the proposed algorithm are illustrated on a real‐world dataset in comparison with existing algorithms.

Analysis of the plasmonic excitations in assemblies of three-dimensional electron clusters

In the quest to built novel metamaterials with unique optical properties, three-dimensional assemblies of metal clusters and nanoparticles are gathering significant attention. Organized geometries, such as tetrahedra and icosahedra, can be built, for example, by using DNA strands or virus capsids as templates. Here we use the jellium model and time-dependent density functional theory to study the plasmonic resonances in different arrangements of eight-electron clusters from the electronic perspective. A charge transfer ratio index based on the induced transition densities is used to quantify the charge transfer nature of the excitations at different energies. We vary the size, shape, and intercluster separation, building systems of four-cluster tetrahedra, 12-cluster icosahedra and cuboctahedra, and 20-cluster dodecahedra. All the studied systems are found to have charge transfer plasmon-type excitations at low energies. Analysis of the electron-hole transitions contributing to the transition dipole moment is further used to characterize these excitations, showing that they have significant screening contributions unlike the higher-energy excitations. The understanding gained for the optical response of these simple model systems can help in interpreting the properties of real, complex cluster systems.

A parameter-free affinity based clustering

Several methods have been proposed to estimate the number of clusters in a dataset; the basic idea behind all of them has been to study an index that measures inter-cluster separation and intra-cluster cohesion over a range of cluster numbers and report the number which gives an optimum value of the index. In this paper, we propose a simple parameter-free approach that is more like human cognition of clusters, where closely lying points are easily identified to form a cluster and the total number of clusters is revealed. To identify closely lying points, the affinity of two points is defined as a function of distance and a threshold affinity is identified, above which two points in a dataset are likely to be in the same cluster. Well separated clusters are identified even in the presence of outliers, whereas for a not well-separated dataset, the final number of clusters is estimated from the detected clusters. And they are merged to produce the final clusters. Experiments performed with several large dimensional synthetic and real datasets show good results with robustness to noise and density variation within a dataset.

Meta-Heuristic Optimization Based Two-stage Residential Load Pattern Clustering Approach Considering Intracluster Compactness and Inter-cluster Separation

This article proposes a meta-heuristic optimization-based two-stage residential load pattern clustering (LPC) approach to address two main issues that exist in the most current LPC methods: 1) unreasonable typical load pattern (TLP) extraction; 2) a good clustering should achieve a good balance between the compactness and separation of the formed clusters. However, few clustering algorithms integrate both of these two aspects into the objective function of clustering for consideration. In the first stage, an adaptive density-based spatial clustering of applications with noise (DBSCAN) is proposed to automatically detect the uncommon load curves and obtain the TLP of each individual customer. In the second stage, LPC is formulated as an optimization problem in which clustering validity index (CVI) considering both compactness and separation is used as the objective function. Gravitational search algorithm (GSA) is adopted to solve this optimization problem. Four different CVIs are investigated to find the most appropriate one for LPC. A comparative case study using the real load data from 208 households from the U.K. verified the effectiveness of the proposed approach.

Label-Based Trajectory Clustering in Complex Road Networks

In the data mining of road networks, trajectory clustering of moving objects is of particular interest for its practical importance in many applications. Most of the existing approaches to this problem are based on distance measurement, and suffer from several performance limitations including inaccurate clustering, expensive computation, and incompetency to handle high dimensional trajectory data. This paper investigates the complex network theory and explores its application to trajectory clustering in road networks to address these issues. Specifically, we model a road network as a dual graph, which facilitates an effective transformation of the clustering problem from sub-trajectories in the road network to nodes in the complex network. Based on this model, we design a label-based trajectory clustering algorithm, referred to as LBTC, to capture and characterize the essence of similarity between nodes. For the evaluation of clustering performance, we establish a clustering criterion based on the classical Davies-Bouldin Index (DB), Akaike Information Criterion (AIC), and Bayesian Information Criterion (BIC) to maximize inter-cluster separation and intra-cluster homogeneity. The clustering accuracy and performance superiority of the proposed algorithm are illustrated by extensive simulations on both synthetic and real-world dataset in comparison with existing algorithms.