Classification of organometallic reactions using machine learning

Automated evaluation of argumentative writing has emerged as an essential field in educational technology, providing systematic feedback on student essays to improve critical thinking and writing quality. This study examines two complementary approaches: the PERSUADE model (Personalized Evaluations and Recommendations for Students Using Argumentation Data and Evidence), which relies on annotated discourse elements, and a multi-document clustering (MDC) model implemented with PyTorch and transformer embeddings. The MDC model, when implemented with Bidirectional Encoder Representations from Transformers (BERT)-base, achieved 78 percent accuracy and clustering metrics indicating strong separation (Silhouette Score 0.789, Davies–Bouldin Index 0.304, Calinski–Harabasz Score 3,529.2). To test the impact of richer embeddings, the MDC framework was extended with Robustly Optimized BERT Pretraining Approach (RoBERTa) and Decoding-enhanced BERT with disentangled attention (DeBERTa). The results show consistent performance improvements: RoBERTa reached 89 percent accuracy with higher clustering stability, while DeBERTa-v3-large achieved the strongest performance at 91 percent accuracy, with the best clustering metrics (Silhouette Score 0.817, Davies–Bouldin Index 0.276, Calinski–Harabasz Score 3,788.5). These findings confirm that the choice of encoder significantly influences clustering coherence and classification effectiveness. The comparative analysis highlights complementary strengths: PERSUADE excels in micro-level discourse evaluation, while MDC, enhanced by advanced embeddings, offers macro-level organization of argumentative structures. Together, these approaches demonstrate the potential for integrated frameworks that capture both discourse depth and cross-document clustering.

Affinity Propagation Method it is necessary to modify the algorithm by using Principal Component Analysis (PCA). PCA method is used to reduce the attributes or characteristics that are less influential on the data so that the most influential attributes are obtained to then be carried out the clustering process with Affinity Propagation. The comparison results of the PCA + AP grouping model have better performance than the conventional AP grouping model. This is justified because the number of iterations and clusters produced by the PCA + AP clustering model does not change and converges when there are 8 optimal cluster clusters. While the performance of conventional clustering models produces an optimal number of clusters from 14 clusters with a significant number of iterations. So it can be concluded that the PCA + AP grouping model is suitable for the Air Quality dataset because it produces an optimal number of clusters and iterations of 8 clusters. The comparison results of the PCA + AP grouping model have better performance than the conventional AP grouping model. This is justified because the number of iterations and clusters produced by the PCA + AP clustering model does not change and converges when the optimal number of clusters is 5 clusters. While the performance of conventional clustering models produces a suboptimal number of 10 clusters with a significant number of iterations. So it can be concluded that the PCA + AP grouping model is suitable for the Water Quality Status dataset because it produces an optimal number of clusters and 5 cluster repetitions.

The paper presents Centroid Ranking Optimized Clustering (CROC), a novel approach aimed at enhancing educational teaching and resource optimization through advanced clustering techniques. CROC integrates centroid-based clustering with ranking optimization strategies to segment students or educational resources into meaningful clusters, facilitating personalized learning experiences and targeted interventions. Experimental analysis conducted on real-world educational datasets demonstrates that CROC outperforms traditional clustering algorithms in terms of clustering quality, accuracy, and interpretability. The findings suggest that CROC holds significant promise for informing data-driven decision-making in educational settings, enabling educators to better understand student performance patterns, identify at-risk students, and tailor instructional strategies to meet diverse learning needs. Experimental analysis on real-world educational datasets reveals that CROC achieves a Silhouette Score of 0.75 and a Davies-Bouldin Index of 0.40 on the MathScores dataset, outperforming traditional algorithms such as k-means (Silhouette Score: 0.68, Davies-Bouldin Index: 0.53), Hierarchical (Silhouette Score: 0.62, Davies-Bouldin Index: 0.57), and DBSCAN (Silhouette Score: 0.45, Davies-Bouldin Index: 0.78). Similarly, on the EnglishTest dataset, CROC attains a Silhouette Score of 0.82 and a Davies-Bouldin Index of 0.35, surpassing k-means (Silhouette Score: 0.74, Davies-Bouldin Index: 0.47), Hierarchical (Silhouette Score: 0.68, Davies-Bouldin Index: 0.52), and DBSCAN (Silhouette Score: 0.53, Davies-Bouldin Index: 0.68). These findings underscore the effectiveness of CROC in clustering educational data, enabling personalized learning experiences and data-driven decision-making in educational settings.

Urban transportation systems encounter significant challenges in extracting meaningful traffic patterns from extensive historical datasets, a critical aspect of smart city initiatives. This paper addresses the challenge of analyzing and understanding these patterns by employing various clustering techniques on hourly urban traffic flow data. The principal aim is to develop a model that can effectively analyze temporal patterns in urban traffic, uncovering underlying trends and factors influencing traffic flow, which are essential for optimizing smart city infrastructure. To achieve this, we applied DBSCAN, K-Means, Affinity Propagation, Mean Shift, and Gaussian Mixture clustering techniques to the traffic dataset of Aarhus, Denmark's second-largest city. The performance of these clustering methods was evaluated using the Silhouette Score and Dunn Index, with DBSCAN emerging as the most effective algorithm in terms of cluster quality and computational efficiency. The study also compares the training times of the algorithms, revealing that DBSCAN, K-Means, and Gaussian Mixture offer faster training times, while Affinity Propagation and Mean Shift are more computationally intensive. The results demonstrate that DBSCAN not only provides superior clustering performance but also operates efficiently, making it an ideal choice for analyzing urban traffic patterns in large datasets. This research emphasizes the importance of selecting appropriate clustering techniques for effective traffic analysis and management within smart city frameworks, thereby contributing to more efficient urban planning and infrastructure development.

This study aims to detect pulmonary Tuberculosis (TB) endemic areas in Aceh Utara Regency based on altitude, population density, and the number of TB cases. The Agglomerative Hierarchical Clustering (AHC) algorithm was used to cluster 27 subdistricts, each measured by these three factors. The clustering results divided the subdistricts into three main clusters with distinct characteristics. Cluster 1 consists of subdistricts with low altitude, high population density, and relatively high numbers of TB cases, identifying this area as having the highest risk of TB endemicity. Cluster 2 includes areas with moderate population density and TB case numbers, while Cluster 3 consists of subdistricts at higher altitudes with fewer TB cases. The clustering results were evaluated using three key metrics: Silhouette Score, Davies-Bouldin Index, and Dunn Index, which indicated that the clustering model performed well, although some subdistricts were positioned near the cluster boundaries. This research provides valuable information for health authorities to prioritize interventions and allocate resources to areas most in need of TB management.

In the field of mechanical fault diagnosis, in order to improve the diagnosis accuracy with high feature dimension, a method of features linear fusion method based on AP(Affinity propagation clustering) and AASC(Affinity Aggregation for Spectral Clustering) algorithm was proposed. Low dimension space was obtained by AP clustering, then using AASC algorithm to fuse this selected features, and gained a new feature, meanwhile which can be used independently as a new feature. In the phase of experiment, this method was compared with respectively LPP and PCA, it’s result indicates that this method is excellent in improving diagnosis accuracy.

In this study, a new intelligent fault diagnosis approach based on composite multi-scale dimensionless indicators (CMDIs) and affinity propagation (AP) clustering is proposed to identify working conditions of mechanical components. For this goal, CMDIs as fault features are for the first time proposed to search more informative fault information hidden in the original vibration signals and useful variational mode decomposition signals. Next, two feature evaluation methods, Fisher criterion (FC) and self-weight criterion (SW), are introduced to search the top ranked CMDIs, respectively. Subsequently, AP clustering is applied to select the important CMDIs. The most important CMDI of each cluster is chosen as the final fault feature. Finally, these important CMDIs are fed into the following supervised learning classifiers to identify working conditions of mechanical components: k-nearest neighbors, linear discriminant analysis, naive Bayesian and random forests. Two experimental examples for different fault types of mechanical components are conducted. Experimental comparison results demonstrate that the combination diagnosis model based on CMDIs-FC-AP with naive Bayesian classifier would be most suitable for complex samples with a high dimensional feature set.

Automatic mode tracing of dispersion relations for guided waves in elastic waveguides via physics-driven affinity propagation (AP) clustering

The data-driven characterization of discretionary lane-changing behaviors has traditionally been hindered by the scarcity of high-resolution data that can precisely record lateral movements. In this study, we conducted an exploratory investigation leveraging the Third Generation Simulation (TGSIM) dataset to advance our understanding of discretionary lane-changing behaviors. In this paper, we developed a discretionary lane-changing extraction pipeline and scrutinized crucial factors such as gaps and relative speeds in leading and following directions. A dynamic time warping (DTW) analysis was performed to quantify the difference between any pair of lane-changing behaviors, and an affinity propagation (AP) clustering, evaluated on normalized DTW distance, was conducted. Our results yielded five clusters based on lead and lag gaps, enabling us to categorize lane-changing behaviors into aggressive, neutral, and cautious for both leading and following directions. Clustering based on relative speeds revealed two distinct groups of lane-changing behaviors, one representing overtaking and the other indicative of transitioning into a lane with stable and homogenous speed. The proposed DTW analysis, in conjunction with AP clustering, demonstrated promising potential in categorizing and characterizing lane-changing behaviors. Additionally, this approach can be readily adapted to analyze any driving behavior.

The primary issue in clustering analysis of multivariate data is the low accuracy resulting from a mismatch between the Distance Metric used and the characteristics of the data. This study aims to comprehensively evaluate the effect of eight Distance Metric in the KMeans algorithm integrated with the Principal Component Analysis (PCA)dimension reduction technique. The analysis process was conducted by transforming the data into two principal components using PCA, then applying K-Means to each Distance Metric. Performance evaluation was conducted based on five internal metrics: Silhouette Score, Davies-Bouldin Index, Sum of Squared Errors, Calinski-Harabasz Index, and Dunn Index. The results show that the Bray-Curtis formula provides the best performance, with a Silhouette Score of 0.4291 and SSE of 30.3673. This is followed by Euclidean and Minkowski, which yield the highest Calinski-Harabasz Index value of 2239.85 and Dunn Index of 0.0108, respectively. In contrast, Hamming’s formula yielded the lowest performance across all metrics, with a Silhouette Score of 0.0000 and an SSE of 1996.00. The ANOVA test revealed significant differences between the Distance Metric, with a p-value of ¡0.000 for all metrics, which was further supported by the Tukey HSD follow-up test results. The implications of these findings confirm the importance of selecting an appropriate Distance Metric in the clustering process to ensure the validity, efficiency, and interpretability of multivariate data analysis results.

Pregnant women's health is an important aspect of the public health system, where grouping disease data can help in risk identification and better treatment planning. However, traditional clustering methods such as K-Means often face challenges in optimal separation between clusters, especially when the attributes used are irrelevant. This study aims to optimize the K-Means method in disease clustering in pregnant women by applying Random Forest-based attribute selection. Of the six available attributes (age, weight, height, gestational age, systole, and diastole), the three main attributes namely systole, diastole, and gestational age were selected based on the Importance Score from Random Forest. The test results showed that the use of these three attributes increased the Silhouette Score by 0.21 (from 0.23 to 0.44), indicating better cluster separation, and lowered the Davies-Bouldin Index by 0.69 (from 1.50 to 0.81), indicating a more compact and well-separated cluster. Clustering visualization using Principal Component Analysis (PCA) supports these results. In addition, the calculation of the Elbow method shows the optimal number of clusters at k=3, reinforcing the conclusion that the selection of the right attributes and the number of clusters improves the quality of clustering. Overall, this study proves that the selection of Random Forest-based features is able to optimize the K-Means method in disease clustering in pregnant women, which is expected to improve the effectiveness of diagnosis and treatment planning.Kesehatan ibu hamil merupakan aspek penting dalam sistem kesehatan masyarakat, di mana pengelompokan data penyakit dapat membantu dalam identifikasi risiko dan perencanaan perawatan yang lebih baik. Namun, metode clustering tradisional seperti K-Means sering kali menghadapi tantangan dalam pemisahan yang optimal antar cluster, terutama ketika atribut yang digunakan tidak relevan. Penelitian ini bertujuan untuk mengoptimalkan metode K-Means dalam clustering penyakit pada ibu hamil dengan menerapkan seleksi atribut berbasis Random Forest. Dari enam atribut yang tersedia (usia, berat badan, tinggi badan, usia kehamilan, sistole, dan diastole), tiga atribut utama yaitu sistole, diastole, dan usia kehamilan dipilih berdasarkan Importance Score dari Random Forest. Hasil pengujian menunjukkan bahwa penggunaan tiga atribut ini meningkatkan Silhouette Score sebesar 0,21 (dari 0,23 menjadi 0,44), yang mengindikasikan pemisahan cluster yang lebih baik, serta menurunkan Davies-Bouldin Index sebesar 0,69 (dari 1,50 menjadi 0,81), menunjukkan cluster yang lebih kompak dan terpisah dengan baik. Visualisasi clustering menggunakan Principal Component Analysis (PCA) mendukung hasil ini. Selain itu, perhitungan metode Elbow menunjukkan jumlah cluster optimal pada k=3, memperkuat kesimpulan bahwa pemilihan atribut dan jumlah cluster yang tepat meningkatkan kualitas clustering. Secara keseluruhan, penelitian ini membuktikan bahwa seleksi fitur berbasis Random Forest mampu mengoptimalkan metode K-Means dalam clustering penyakit pada ibu hamil, yang diharapkan dapat meningkatkan efektivitas diagnosis dan perencanaan perawatan.

With the fast development of internet of things (IoT), a large amount of missing data is produced in the process of data collection and transmission. We call these data incomplete data. Many traditional methods use imputation or discarding strategy to cluster incomplete data. In this paper, we propose an improved incomplete affinity propagation (AP) clustering algorithm based on K nearest neighbours (IAPKNN). IAPKNN firstly partitions the dataset into complete and incomplete dataset, and then clusters the complete data set by AP clustering directly. Secondly, according to the similarity, IAPKNN extends the responsibility and availability matrices to the incomplete dataset. Finally, clustering algorithm is restarted based on the extended matrices. In addition, to address the clustering efficiency of large scale dataset, we give a distributed clustering algorithm scheme. Experiment results demonstrate that IAPKNN is effective in clustering incomplete data directly.

Customer segmentation is crucial for optimizing marketing strategies. This study applies and compares the effectiveness of three common clustering algorithms: K-Means, Hierarchical Clustering, and Gaussian Mixture Models (GMM) to classify customers based on shopping behavior and demographics (age, gender, total spending). Utilizing three retail datasets (two from Kaggle, one from Sling Academy), the research performs data preprocessing, applies the clustering algorithms, and evaluates their performance using Silhouette Score, Davies-Bouldin Index, and Calinski-Harabasz Index. The results indicate that GMM performs most effectively for segmenting based on total spending and gender, creating distinct clusters. Hierarchical Clustering proves suitable for detailed age-based analysis on specific datasets, while K-Means offers a balanced solution, particularly effective when cluster structures are clear or rapid results are needed. The study recommends selecting appropriate algorithms based on specific business objectives and data characteristics, enabling businesses to develop more effective personalized marketing strategies.

Clustering is an important phase in data mining. The grouping method commonly used in data mining concepts is using K-Means. Choosing the best value of k in the k-means algorithm can be difficult. In this study the technique used to determine the value of k is the silhouette score. Then, to evaluate the k-means model uses the Davies Bouldin Index (DBI) technique. The best DBI value is close to 0. The parameters used are total consumer income and spending. Based on the results of this study it can be concluded that the silhouette score method can provide a k value with optimal results. For mall customer data of 200 data, the most optimal silhouette score is obtained at K = 5 with a DBI = 0.57.

Flood is one of the most destructive natural disasters in Malaysia. To understand flood characteristics, we investigate the effectiveness of unsupervised machine learning techniques to identify clusters of flood events based on historical flood data provided by the Department of Irrigation and Drainage Sabah. The dataset used in this study consists of hydrological features such as average recurrence interval, flood duration, flood depth, and affected areas. Due to the nature of raw dataset such as missing data, varying scales, and outliers, we generated six different preprocessed datasets using an extensive data preprocessing which includes handling of missing values, feature normalization, and outlier removal. We then applied five clustering algorithms, which are K-means, K-medoids, hierarchical clustering, density-based spatial clustering of applications with noise (DBSCAN) and its extended version, HDBSCAN, to derive natural clustering in the flood events. The clustering quality was carefully assessed using internal validation metrics such as silhouette score, Davies-Bouldin index, and two-dimensional visualization by principal component analysis. We found that K-means, K-medoids, and hierarchical clustering outperformed DBSCAN and HDBSCAN when a predefined number of clusters was required. The best clustering quality was achieved using datasets with imputation and normalization. Our study provides meaningful insights into the characteristics of flood events.