Final Clustering Research Articles

Mapping the discursive dimensions of the reproducibility crisis: A mixed methods analysis.

To those involved in discussions about rigor, reproducibility, and replication in science, conversation about the "reproducibility crisis" appear ill-structured. Seemingly very different issues concerning the purity of reagents, accessibility of computational code, or misaligned incentives in academic research writ large are all collected up under this label. Prior work has attempted to address this problem by creating analytical definitions of reproducibility. We take a novel empirical, mixed methods approach to understanding variation in reproducibility discussions, using a combination of grounded theory and correspondence analysis to examine how a variety of authors narrate the story of the reproducibility crisis. Contrary to expectations, this analysis demonstrates that there is a clear thematic core to reproducibility discussions, centered on the incentive structure of science, the transparency of methods and data, and the need to reform academic publishing. However, we also identify three clusters of discussion that are distinct from the main body of articles: one focused on reagents, another on statistical methods, and a final cluster focused on the heterogeneity of the natural world. Although there are discursive differences between scientific and popular articles, we find no strong differences in how scientists and journalists write about the reproducibility crisis. Our findings demonstrate the value of using qualitative methods to identify the bounds and features of reproducibility discourse, and identify distinct vocabularies and constituencies that reformers should engage with to promote change.

Screening of Key Drought Tolerance Indices for Cotton at the Flowering and Boll Setting Stage Using the Dimension Reduction Method.

Drought is one of the main abiotic stresses that seriously influences cotton production. Many indicators can be used to evaluate cotton drought tolerance, but the key indicators remain to be determined. The objective of this study was to identify effective cotton drought tolerance indicators from 19 indices, including morphology, photosynthesis, physiology, and yield-related indices, and to evaluate the yield potential of 104 cotton varieties under both normal and drought-stress field conditions. Combined with principal component analysis (PCA) and a regression analysis method, the results showed that the top five PCs among the 19, with eigenvalues > 1, contributed 65.52, 63.59, and 65.90% of the total variability during 2016 to 2018, respectively, which included plant height (PH), effective fruit branch number (EFBN), single boll weight (SBW), transpiration rate (Tr) and chlorophyll (Chl). Therefore, the indicator dimension decreased from 19 to 5. A comparison of the 19 indicators with the 5 identified indicators through PCA and a combined regression analysis found that the results of the final cluster of drought tolerance on 104 cotton varieties were basically consistent. The results indicated that these five traits could be used in combination to screen cotton varieties or lines for drought tolerance in cotton breeding programs, and Zhong R2016 and Xin lu zao 45 exhibited high drought tolerance and can be selected as superior parents for good yield performance under drought stress.

Abstract 627: Molecular validation of the 8th edition AJCC cancer staging system for resectable pancreatic cancer: Proposal of integrative translational staging system

Abstract Introduction: Even though the 8th edition AJCC cancer staging system for pancreatic cancer has validated major clinicopathologic factors in multiple clinical cohorts, there is still an unmet need for integrative consideration using multi-omics data to stratify the patients with pancreatic cancer elaborately. Methods: We performed a comprehensive analysis and profiling using genomic, transcriptomic, and proteomic data from TCGA-PAAD and other translational cohorts (4 cohorts, n=340). Molecular features and major subtypes were analyzed mutually with clinical and pathologic factors, especially the 8th AJCC staging system. Results: Aggressive molecular subtypes, basal-like and squamous subtype, were significantly associated with a higher nodal stage, but tumor size didn't show a clear association with molecular features. The activated stroma of the pancreatic cancer microenvironment was significantly correlated with poor differentiation and large tumor size. The mutational pattern of KRAS and several transcriptomic pathways, such as epithelial-mesenchymal transition and DNA repair, were differently presented in each clinical stage from the 8th AJCC TNM staging system. The optimal algorithm was identified to show significantly higher performance for the prediction for cancer relapse and cancer-specific survival in discovery and validation cohorts. The in-silico prediction for molecular target agents and immunotherapy was performed for final clusters from the optimal stratification system revealed from the integrative analysis. Conclusions: Our comprehensive multi-omics analysis reveals clear needs for the combination of clinical staging and molecular profiling and provides crucial evidence for precision strategy in patients with resectable pancreatic cancer. Citation Format: Sung Hwan Lee, Sunyoung Lee, Ju-Seog Lee. Molecular validation of the 8th edition AJCC cancer staging system for resectable pancreatic cancer: Proposal of integrative translational staging system [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2021; 2021 Apr 10-15 and May 17-21. Philadelphia (PA): AACR; Cancer Res 2021;81(13_Suppl):Abstract nr 627.

Penentuan Klaster Koridor TransJakarta dengan Metode Majority Voting pada Algoritma Data Mining

The Covid-19 pandemic has made many changes in the patterns of community activity. Large-Scale Social Restrictions were implemented to reduce the number of transmission of the virus. This clearly affects the mode of transportation. The mode of transportation makes new regulations to reduce the number of passenger capacities in each fleet, for example, TransJakarta services. This study will categorize the TransJakarta corridors before and during the Covid-19 pandemic. The clustering method of K-Means and K-Medoids is used to obtain accurate calculation results. The calculations are performed using Microsoft Excel, Rapid Miner, and Python programming language. The clustering results obtained that using K-Means algorithm before Covid-19 pandemic, an optimum number of clusters is 3 clusters with DBI (Davies Bouldin Index) value is 0.184, and during Covid-19 pandemic, the optimum number of clusters is 2 clusters with DBI value is 0.188. Meanwhile, when using the K-Medoids algorithm before the Covid-19 pandemic, an optimum number of clusters is 3 clusters with the DBI value is 0.200, and during the Covid-19 pandemic, an optimum number of clusters is 4 clusters with the DBI value is 0.190. The final cluster is determined using the majority voting approach from all the tools used.

An ensemble solution for multivariate time series clustering

Technologies such as Big Data and IoT have shown the need for intelligent unsupervised processing of Multivariate Time Series (MTS), MTS clustering among them. The challenges in MTS clustering includes not only the selection of the algorithm but also the MTS representation and the similarity measurement among the instances. This study proposes an ensemble of MTS clustering methods that merges different MTS representations and distance functions, aggregating them to obtain a similarity measurement. Furthermore, a proposal for prior knowledge representation is propose to balance the aggregation of the distances. The final clustering is performed either using k-means or hierarchical clustering. The experimentation set up includes the implementation of the ensemble with either 4 or 5 different methods, including an MTS extension of k-Shape. The results show that the ensemble is biased towards the best methods, which helps the clustering practitioner in the selection of the most suitable prototypes. Moreover, the evaluation of the ensemble with the number of clusters set to the number of labels shows that metrics, such as the sensitivity and specificity, must drive the rule of the elbow; alternatively, this value represents the most interesting prior knowledge bit in MTS clustering. Further work includes the study of digital markers to compare MTS representations and distance functions and the use of external metrics to balance the aggregation of the methods.

Adaptive Total-Variation Regularized Low-Rank Representation for Analyzing Single-Cell RNA-seq Data.

High-throughput sequencing of single-cell gene expression reveals a complex mechanism of individual cell's heterogeneity in a population. An important purpose for analyzing single-cell RNA sequencing (scRNA-seq) data is to identify cell subtypes and functions by cell clustering. To deal with high levels of noise and cellular heterogeneity, we introduced a new single cell data analysis model called Adaptive Total-Variation Regularized Low-Rank Representation (ATV-LRR). In scRNA-seq data, ATV-LRR can reconstruct the low-rank subspace structure to learn the similarity of cells. The low-rank representation can not only segment multiple linear subspaces, but also extract important information. Moreover, adaptive total variation also can remove cell noise and preserve cell feature details by learning the gradient information of the data. At the same time, to analyze scRNA-seq data with unknown prior information, we introduced the maximum eigenvalue method into the ATV-LRR model to automatically identify cell populations. The final clustering results show that the ATV-LRR model can detect cell types more effectively and stably.

GoT: a Growing Tree Model for Clustering Ensemble

The clustering ensemble technique that integrates multiple clustering results can improve the accuracy and robustness of the final clustering. In many clustering ensemble algorithms, the co-association matrix (CA matrix), which reflects the frequency of any two samples being partitioned into the same cluster, plays an important role. However, generally, the CA matrix is highly sparse with low value density, which may limit the performance of an algorithm based on it. To handle these issues, in this paper, we propose a growing tree model (GoT). In this model, the CA matrix is firstly refined by the shortest path technique so that its sparsity will be mitigated. Then, a set of representative prototype examples is discovered. Finally, to handle the low value density of the CA matrix, the prototypes gradually connect to their neighborhood, which likes a set of trees growing up. The rationality of the discovered prototype examples is illustrated by theoretical analysis and experimental analysis. The working mechanism of the GoT is visually shown on synthetic data sets. Experimental analyses on eight UCI data sets and eight image data sets show that the GoT outperforms nine representative clustering ensemble algorithms.

A Folded Concave Penalty Regularized Subspace Clustering Method to Integrate Affinity and Clustering

Most sparse or low-rank-based subspace clustering methods divide the processes of getting the affinity matrix and the final clustering result into two independent steps. We propose to integrate the affinity matrix and the data labels into a minimization model. Thus, they can interact and promote each other and finally improve clustering performance. Furthermore, the block diagonal structure of the representation matrix is most preferred for subspace clustering. We define a folded concave penalty (FCP) based norm to approximate rank function and apply it to the combination of label matrix and representation vector. This FCP-based regularization term can enforce the block diagonal structure of the representation matrix effectively. We minimize the difference of l1 norm and l2 norm of the label vector to make it have only one nonzero element since one data only belong to one subspace. The index of that nonzero element is associated with the subspace from which the data come and can be determined by a variant of graph Laplacian regularization. We conduct experiments on several popular datasets. The results show our method has better clustering results than several state-of-the-art methods.

Hybrid fuzzy logic and artificial Flora optimization algorithm-based two tier cluster head selection for improving energy efficiency in WSNs

Wireless Sensor Network (WSN) is the one of the hot area of research in which energy stability and network lifetime are considered to be the twin challenges during its application. Clustering is the optimum energy efficiency strategy that organizes the sensor nodes into potential groups for the objective of attaining energy stability and network lifetime. In this energy potent clustering process, cluster head selection is determined to be highly significant in order to balance energy among the nodes sensor nodes. Moreover, two-tier cluster head selection that includes temporary and final cluster head is identified to be challenging in WSNs. In this paper, Hybrid Fuzzy Logic and Artificial Flora Optimization Algorithm (FL-AFA)-based Two Tier Cluster Head Selection is proposed for improving energy efficiency and prolog network lifetime. This FL-AFA scheme achieved the cluster head selection in two stages, such as, i) Temporary Cluster Head (TCH) selection using FL and, ii) Final Cluster Head (FCH) selection using AFA. In the first stage, the concept of fuzzy logic applied over the input parameters of residual energy (RE), distance to BS (DTBS), and node degree (NDE). In the second stage, the benefits of AFA is employed for computing the fitness function through distance to nearby nodes (DNN), cluster compactness estimation factor (CCEF), and position estimation (PE). Simulation experiments of the proposed FL-AFA scheme and the benchmarked schemes are conducted based on the evaluation metrics of energy efficiency, network lifetime, average delay, and packet delivery ratio (PDR) under the impact of different sensor nodes. .

Dual-granularity weighted ensemble clustering

Ensemble clustering is one of the research hotspots of data mining in recent years. The selection of high-quality and large-diversity base clustering results plays a key role in the quality of the final result. Traditional ensemble clustering selection algorithms usually treat each base clustering result as a whole which ignores the difference between the clusters in the same clustering result. It may cause the validity of the final clustering result to be affected. Aiming at this problem, inspired by the measurement method of uncertainty in the rough set theory, a dual-granularity weighted ensemble clustering model is proposed. The main contribution of this paper is shown as follows: (1) the evaluation of the reliability of clusters is transformed into an uncertainty measurement problem in the rough set; (2) in a finer-grained level, a sample local similarity measurement method is designed; (3) a weighted co-association matrix elements generation method based on global cluster reliability and local sample pair similarity is proposed, then the fusion function is used to get the final clustering result. Experimental results show that the proposed method is not sensitive to the size and diversity of base clustering members which has good robustness and stability. The final result obtained by this model is closer to the actual distribution of data sets.

Repair Strategies Employed by Korean Speakers in Producing English Final Consonant Clusters

This study attempted to investigate how Korean speakers produce English final consonant clusters by using repair strategies. Based on the preservation principle in interphonology, three main points were examined. The first point focused on the most used repair strategy by Korean speakers, such as deletion, insertion, and substitution. The second point was whether any age difference existed in using repair strategies when pronouncing various final consonant clusters. The third point concentrated on the specific patterns of repair strategies used in certain syllable structures. For the production test, twenty-four target words were pronounced by eight Korean speakers and the data were statistically analyzed by 3-way mixed ANOVA to see if there were any variable interactions. The major findings of the study were exhibited as follows: First, most Korean speakers employed deletion for repair strategies, and the highest error rates were presented in fricative-fricative combined clusters. Second, the younger group made fewer errors in stop-# and nasal-# combined clusters. Third, all kinds of the repair strategies were used in fricative-fricative clusters, and insertion was primarily used by the older group in fricative-stop clusters. According to the findings, it could be assumed that productions of English clusters by the younger group have been in progress by maintaining certain English phonetic features.

Semisupervised Adaptive Symmetric Non-Negative Matrix Factorization.

As a variant of non-negative matrix factorization (NMF), symmetric NMF (SymNMF) can generate the clustering result without additional post-processing, by decomposing a similarity matrix into the product of a clustering indicator matrix and its transpose. However, the similarity matrix in the traditional SymNMF methods is usually predefined, resulting in limited clustering performance. Considering that the quality of the similarity graph is crucial to the final clustering performance, we propose a new semisupervised model, which is able to simultaneously learn the similarity matrix with supervisory information and generate the clustering results, such that the mutual enhancement effect of the two tasks can produce better clustering performance. Our model fully utilizes the supervisory information in the form of pairwise constraints to propagate it for obtaining an informative similarity matrix. The proposed model is finally formulated as a non-negativity-constrained optimization problem. Also, we propose an iterative method to solve it with the convergence theoretically proven. Extensive experiments validate the superiority of the proposed model when compared with nine state-of-the-art NMF models.

A local search algorithm for k-means with outliers

k-Means is a well-studied clustering problem that finds applications in many fields related to unsupervised learning. It is known that k-means clustering is highly sensitive to the isolated points (called outliers). Such outliers can significantly influence the final cluster configuration and should be removed to obtain quality solutions. In this paper, we study the k-means with outliers problem. Given a set of data points, the problem is to remove a set of outliers such that the k-means clustering cost of the remaining points is minimized. Designing efficient algorithms for this problem remains an active area of research due to its important role in dealing with noisy data. We consider a relaxed objective function and propose a local search algorithm for k-means with outliers. It is shown that the algorithm has better performance guarantees than previously implemented methods. In particular, it yields a constant-factor bi-criteria approximation solution to the problem. Moreover, we show experimentally that the algorithm performs much better than its provable guarantee and dominates other state-of-the-art methods for the problem.

Radar Jamming Effect Analysis Based on Bayesian Inference Network With Adaptive Clustering

Radar jamming effect analysis (RJEA) is important for improving the performance of radar sensors in jamming environments and one of the most popular methods for RJEA is Bayesian inference network (BIN). Notice the features in RJEA are usually continuous, so they should be discretized for BIN, where the feature discretization approach determines the node sizes and the discretization error, which further seriously affect the training time and inference precision of BIN, respectively. In this paper, we put forward a adaptive clustering approach for the features in RJEA to reduce the node sizes of BIN in the premise of keeping or even improving its inference precision. To do so, a modified K-means algorithm is designed, where the initial clustering centers are preset according to the density of samples, and the number of final clustering centers, not given in advance, is estimated with their locations simultaneously via iterative optimization. Finally, the performance of BIN with the adaptive clustering is validated in the experiments on pulse Doppler radar sensor with radial frequency noise, where 20 features are extracted for RJEA. Compared with uniform discretization, the proposed method can reduce the training time of BIN from 19.52 second to 4.39 second and improve its inference precision from 71 % to 82% simultaneously. Compared with K-means discretization, the proposed method results in similar inference precision but less training time of BIN.

Optimizing Text Categorization for Indonesian Text Using Clustering Label Technique

Text Categorization plays an important role for clustering the rapidly growing, yet unstructured, Indonesian text in digital format. Furthermore, it is deemed even more important since access to digital format text has become more necessary and widespread. There are many clustering algorithms used for text categorization. Unfortunately, clustering algorithms for text categorization cannot easily cluster the texts due to imperfect process of stemming and stopword of Indonesian language. This paper presents an intelligent system that categorizes Indonesian text documents into meaningful cluster labels. Label Induction Grouping Algorithm (LINGO) and Bisecting K- means are applied to process it through five phases, namely the pre-processing, frequent phrase extraction, cluster label induction, content discovery and final cluster formation. The experimental result showed that the system could categorize Indonesian text and reach to 93%. Furthermore, clustering quality evaluation indicates that text categorization using LINGO has high Precision and Recall with a value of 0.85 and 1, respectively, compare to Bisecting K-means which has a value of 0.78 and 0.99. Therefore, the result shows that LINGO is suitable for categorizing Indonesian text. The main contribution of this study is to optimize the clustering results by applying and maximizing text processing using Indonesian stemmer and stopword.

Star-based learning correlation clustering

Correlation clustering (CC) is a clustering method using a signed graph as input without specifying the number of clusters a priori. It had been widely used in real applications, such as social network and text mining. However, its exact optimization or approximate algorithms often give unsatisfactory results, especially for large-scale signed graphs. This paper tackles this problem and proposes a novel CC algorithm, termed star-based learning correlation clustering (SL-CC). The proposed SL-CC contains two phases. The first is a scale reduction for signed graphs. We propose a special motif, called a star structure, for reducing the scale of signed graphs. We assign the vertices within a star structure to have the same cluster label and then merge these vertices as a new vertex in the graph so we can shrink a large-scale graph to a much small-scale one. The second is to give a learning schema for the local search on the reduced graphs. It can discover some important stars as seeds of clusters according to the graph structure, and then justify whether the other stars need to be merged with seeds or not. We also construct a new integer linear programing (ILP) model based on cycle inequalities to perform the local search with final clustering results. The experiments and comparisons of the proposed SL-CC with some existing CC methods on synthetic and real data sets with variant scale structures of signed graphs demonstrate the efficiency and usefulness of the SL-CC algorithm.

Opinion dynamics in modified expressed and private model with bounded confidence

In a social network, an individual may express an opinion against his/her own private opinion. Moreover, during the evolution of opinion dynamics, an individual may only accept parts of opinions that are similar to his/her own opinion. These phenomenons bring new challenges to the study of the opinion dynamics. In this paper, we propose a modified expressed-private-opinion (MEPO) model with bounded confidence. In the MEPO model, two categories of neighborhood relationships, namely, the communication neighbor and the opinion neighbor, are proposed. More specifically, the communication neighbor represents the information flow among individuals while the opinion neighbor, which is a subset of the communication neighbor, implies whether or not the corresponding opinions will be accepted to be incorporated into the individual’s opinion updating process. Furthermore, the scope of the opinion neighbor is determined by the confidence level such that the whole group is divided into open-minded, moderate-minded, and closed-minded individuals accordingly. As a result, an individual’s private opinion in the proposed MEPO model evolves by his/her own private opinion and the stubbornness value, and also the impacts from the expressed opinions of the opinion neighbors.Specifically, we focus on the design of the stubbornness parameter λ, i.e., one’s stubbornness to his/her initial private opinion. By different scenarios, three expression formats are taken into consideration including (1) λ with constant value 1; (2) λ monotonously increasing to 1; and (3) λ determined by the rate of opinion neighbor number over communication neighbor number. A series of experiments are designed to analyze the relation between the stubbornness parameter λ and the final opinion cluster number. The experimental results show that the final cluster number is mainly determined by both of the rate of closed-minded individuals and λ: it is inversely proportional to the rate of closed-minded individuals as well as the growth rate of λ.

Subtractive Gradient Boost Clustering for Mobile Node Authentication in Internet of Things Aware 5G Networks

The 5G mobile wireless network systems faces a lot of security issues due to the opening of network and its insecurity. The insecure network prone to various attacks and it disrupts secure data communications between legitimate users. Many works have addressed the security problems in 3G and 4G networks in efficient way through authentication and cryptographic techniques. But, the security in 5G networks during data communication was not improved. Subtractive Gradient Boost Clustered Node Authentication (SGBCNA) Method is introduced to perform secure data communication. The subtractive gradient boost clustering technique is applied to authenticate the mobile node as normal nodes and malicious nodes based on the selected features. The designed ensemble clustering model combines the weak learners to make final strong clustering results with minimum loss. Finally, the malicious nodes are eliminated and normal mobile nodes are taken for performing the secured communication in 5G networks. Simulation is carried out on factors such as authentication accuracy, computation overhead and security level with respect to a number of mobile nodes and data packets. The observed outcomes clearly illustrate that the SGBCNA Method efficiently improves node authentication accuracy, security level with minimum overhead than the state-of-the-art-methods.

Customer Choice Modelling: A Multi-Level Consensus Clustering Approach

Customer Choice Modeling aims to model the decision-making process of customers, or segments of customers, through their choices and preferences identified by the analysis of their behaviors in one or more specific contexts. Clustering techniques are used in this context to identify patterns in their choices and preferences, to define segments of customers with similar behaviors, and to model how customers of different segments respond to competing products and offers. However, data clustering is an unsupervised learning task by nature, that is the grouping of customers with similar behaviors in clusters must be performed without prior knowledge about the nature and the number of intrinsic groups of data instances, i.e., customers, in the data space. Thus, the choice of both the clustering algorithm used and its parameterization, and of the evaluation method used to assess the relevance of the resulting clusters are central issues. Consensus clustering, or ensemble clustering, aims to solve these issues by combining the results of different clustering algorithms and parameterizations to generate a more robust and relevant final clustering result. We present a Multi-level Consensus Clustering approach combining the results of several clustering algorithmic configurations to generate a hierarchy of consensus clusters in which each cluster represents an agreement between different clustering results. A closed sets based approach is used to identified relevant agreements, and a graphical hierarchical representation of the consensus cluster construction process and their inclusion relationships is provided to the end-user. This approach was developed and experimented in travel industry context with Amadeus SAS. Experiments show how it can provide a better segmentation, and refine the customer segments by identifying relevant sub-segments represented as sub-clusters in the hierarchical representation, for Customer Choice Modeling. The clustering of travelers was able to distinguish relevant segments of customers with similar needs and desires (i.e., customers purchasing tickets according to different criteria, like price, duration of flight, lay-over time, etc.) and at different levels of precision, which is a major issue for improving the personalization of recommendations in flight search queries.

Optimization of K-means Algorithm Base on MapReduce

In the context of big data, the K-means clustering algorithm is sensitive to the initial clustering center, the detection and removal of outliers are difficult, and the data is tilted. The KMEANS-BRMS (K-means algorithm based on range mean and sampling) algorithm based on range mean and sampling is proposed. First, for the initial data set, the “range mean value method”(Mean Range method, MRM) is proposed to obtain the initial cluster centers. In order to eliminate the influence of outliers, a range threshold is set above and below the range mean, and all points under this threshold constitute an initial center. The set of points to be selected, in which the maximum and minimum distance criteria are adopted to select the initial clustering center, which effectively avoids the problem of the sensitivity of the initial center point and the influence of outliers caused by the random selection of the initial clustering center. Next, the BSA(Based on pond sampling and first adaptation algorithm) strategy of adapting algorithm for the first time to deal with the data skew problem in the MapReduce stage, and improve the clustering efficiency. Finally, combined with the MapReduce framework model, the data cluster center is mining in parallel to generate the final clustering result. Experiments show that the clustering results of the KMEANS-BRMS algorithm are more stable and effective, and at the same time it can more effectively improve the efficiency of parallel computing in the big data environment.