Hierarchical clustering that takes advantage of both density-peak and density-connectivity

Zhu, Ye; Ting, Kai Ming; Jin, Yuan; Angelova, Maia

doi:10.1016/j.is.2021.101871

Cited by 15 publications

(5 citation statements)

References 30 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Players were considered to be within congestion if identified as core points, while players were deemed outside congestion if recorded as noise. Clusters of congestion were originally assigned output labels of 0- n , while -1 was assigned to all points clustered as noise (see Fig 1 ) [ 20 , 21 ]. These labels where converted to provide a practical description of congestion and to differentiate between separate clusters of congestion if more than one cluster was identified for a unique time interval (see Fig 1 ).…”

Section: Methodsmentioning

confidence: 99%

Quantifying congestion with player tracking data in Australian football

Alexander

Jackson²,

Bedin³

et al. 2022

PLoS ONE

View full text Add to dashboard Cite

With 36 players on the field, congestion in Australian football is an important consideration in identifying passing capacity, assessing fan enjoyment, and evaluating the effect of rule changes. However, no current method of objectively measuring congestion has been reported. This study developed two methods to measure congestion in Australian football. The first continuously determined the number of players situated within various regions of density at successive time intervals during a match using density-based clustering to group players as ‘primary’, ‘secondary’, or ‘outside’. The second method aimed to classify the level of congestion a player experiences (high, nearby, or low) when disposing of the ball using the Random Forest algorithm. Both approaches were developed using data from the 2019 and 2021 Australian Football League (AFL) regular seasons, considering contextual variables, such as field position and quarter. Player tracking data and match event data from professional male players were collected from 56 matches performed at a single stadium. The random forest model correctly classified disposals in high congestion (0.89 precision, 0.86 recall, 0.96 AUC) and low congestion (0.98 precision, 0.86 recall, 0.96 AUC) at a higher rate compared to disposals nearby congestion (0.72 precision, 0.88 recall, 0.88 AUC). Overall, both approaches enable a more efficient method to quantify the characteristics of congestion more effectively, thereby eliminating manual input from human coders and allowing for a future comparison between additional contextual variables, such as, seasons, rounds, and teams.

show abstract

Section: Methodsmentioning

confidence: 99%

Quantifying congestion with player tracking data in Australian football

Alexander

Jackson²,

Bedin³

et al. 2022

PLoS ONE

View full text Add to dashboard Cite

show abstract

“…Since the introduction of the DPC, there have been many hierarchical clustering algorithms that incorporate DPC. For example, DC-HDP [25] combines the strengths of DPC and DBSCAN to achieve the goal of identifying clusters with complex structures and density variations. HCFS [26] aims to address the distribution issue of data points within clusters.…”

Section: Hierarchical Clusteringmentioning

confidence: 99%

Hierarchical clustering algorithm based on natural local density peaks

Cai,

Feng,

Yang

et al. 2024

Preprint

View full text Add to dashboard Cite

Natural neighbor is known for its ability to adaptively identify the nearest neighbors in a dataset, but it sometimes can not accurately cluster in datasets with density variations. To overcome this drawback, this paper proposes a hierarchical clustering algorithm based on natural local density peaks (HC-NLDP). HC-NLDP mainly consists of partition and merge. Firstly, HC-NLDP searches natural local density peaks to form sub-clusters. This effectively solves the problem that points in low-density regions are difficult to cluster accurately. Secondly, HC-NLDP computes the shared similarity between every pair of sub-clusters. The shared similarity describes the relevance between sub-clusters. When confronted with datasets exhibiting density variations, the shared similarity can ensure that HC-NLDP achieves superior clustering performance. Numerical experiments based on synthetic and real datasets demonstrate the effectiveness and superiority of HC-NLDP.

show abstract

“…Hierarchical clustering methods are the simplest and most fundamental methods and famously play an important role in statistical data analysis (Johnson, 1967). They are relatively fast and easy to understand and implement (Zhu et al, 2022). The autoencoder has the ability to reduce the dimension of the input in a hierarchical way, leading to high-quality reconstructions of data (Tewari et al, 2017;Zhao et al, 2017).…”

Section: The Pollution Area Using Machine Learningmentioning

confidence: 99%

An analysis of finding the best strategies of water security for water source areas using an integrated IT2FVIKOR with machine learning

Zamri

Azman

Pairan

et al. 2023

Front. Environ. Sci.

View full text Add to dashboard Cite

Worldwide, water security is adversely affected by factors such as population growth, rural–urban migration, climate, hydrological conditions, over-abstraction of groundwater, and increased per-capita water use. Water security modeling is one of the key strategies to better manage water safety and develop appropriate policies to improve security. In view of the growing global demand for safe water, intelligent methods and algorithms must be developed. Therefore, this paper proposes an integrated interval type-2 Fuzzy VIseKriterijumska Optimizcija I Kompromisno Resenje (IT2FVIKOR) with unsupervised machine learning (ML). This includes IT2FVIKOR for ranking and selecting a set of alternatives. Unsupervised machine learning includes hierarchical clustering, self-organizing map, and autoencoder for clustering, silhouette analysis and elbow method to find the most optimal cluster count, and finally Adjusted Rank Index (ARI) to find the best comparison within two clusters. This proposed integrated method can be divided into a two-phase fuzzy-machine learning-based framework to select the best water security strategies and categorize the polluted area using the water datasets from the Terengganu River, one of Malaysia’s rivers. Phase 1 focuses on the IT2FVIKOR method to select five different strategies with five different criteria using five decision makers for finding the best water security strategies. Phase 2 continues the unsupervised machine learning where three different clustering algorithms, namely, hierarchical clustering, self-organizing map, and autoencoder, are used to cluster the polluted area in the Terengganu River. Silhouette analysis is applied along with the clustering algorithms to estimate the number of optimal clusters in a dataset. Then, the ARI is applied to find the best comparison within the original data with hierarchical clustering, self-organizing map, and autoencoder. Next, the elbow method is applied to double-confirm the best clusters for each clustering algorithm. Last, lists of polluted areas in each cluster are retrieved. Finally, this 2-phase fuzzy-Machine learning–based framework offers an alternative intelligent model to solve the water security problems and find the most polluted area.

show abstract

Hierarchical clustering that takes advantage of both density-peak and density-connectivity

Cited by 15 publications

References 30 publications

Quantifying congestion with player tracking data in Australian football

Quantifying congestion with player tracking data in Australian football

Hierarchical clustering algorithm based on natural local density peaks

An analysis of finding the best strategies of water security for water source areas using an integrated IT2FVIKOR with machine learning

Contact Info

Product

Resources

About