Multivariate Analysis of Clustering Algorithms for Marine Data

An innovative communication ecosystem called the Internet of Underwater Things (IoUT) was created to link underwater things in marine and undersea habitats. The IoUT devices generate massive volume, variety and velocity of data thus classifying marine data as Big Marine Data (BMD). Clustering these...

Celý popis

Uloženo v:
Podrobná bibliografie
Vydáno v:2024 3rd International Conference on Sentiment Analysis and Deep Learning (ICSADL) s. 31 - 38
Hlavní autoři: Vincent, Agnes Nalini, Sakthidasan, K., Laloo, Nassirah, Bagurubumwe, Uhoze
Médium: Konferenční příspěvek
Jazyk:angličtina
Vydáno: IEEE 13.03.2024
Témata:
ant colony optimization
Clustering
Clustering algorithms
Correlation
Genetics
grouped KMeans
Internet of Underwater Things
KMeans
marine big data
Mauritius
Measurement
particle swarm algorithm
practical salinity
Prediction algorithms
predictive modelling
sea surface temperature
self organising map
Self-organizing feature maps
Sentiment analysis
On-line přístup:Získat plný text
Tagy: Přidat tag
Žádné tagy, Buďte první, kdo vytvoří štítek k tomuto záznamu!
Popis
Shrnutí:An innovative communication ecosystem called the Internet of Underwater Things (IoUT) was created to link underwater things in marine and undersea habitats. The IoUT devices generate massive volume, variety and velocity of data thus classifying marine data as Big Marine Data (BMD). Clustering these BMD and studying their characteristics promotes safeguarding the oceans. This leads to working toward the fourteenth Sustainable Development Goals of United Nations - 'Life Below Water' thereby leading to ocean sustainability. This paper, proposes a comparative analysis of different clustering algorithms such as KMeans, Genetic KMeans, Group K-means (Birch) Algorithm, Self-Organizing Map (SOM), Particle Swarm Optimization Algorithm (PSO) and Ant Colony Optimization Algorithm (ACO) based on their intrinsic evaluation metrics such as Silhouette Score, Calinski Harabasz Score and Davies-Bouldin Score. The winning algorithm with optimal value is determined and this algorithm is further proposed for prediction analysis of clustered data. The data used for this purpose is big marine data of Flic en Flac Region of Republic of Mauritius. The independent variables of this research were the physio chemical parameters or stressors such as Sea Surface Temperature (SST), pH, Practical Salinity (PCU), and Chemical Oxygen Demand (CoD). The dependent variables considered in this research for clustering were mean percentage cover of Hard Corals (HC) and fish assemblages such as Pomacentridae, Chaetodontidae, and Acanthuridae. The research determined that of various clusters obtained by various clustering algorithms, Genetic KMeans clusters could be more homogeneous as it produced a better evaluation metrics and cluster quality. Correlation analysis of variables was conducted to determine the impact of stressors on the dependent variables which serves as one of essential findings for the informed decision making in the big marine data of the Flic en Flac region of Mauritius towards Ocean Sustainability. The correlation analysis confirmed that there exists a strong association between Sea Surface temperature, pH, PSU and the dependent marine community structure.
DOI:10.1109/ICSADL61749.2024.00011