An Application of Geospatial Clustering for Assets Optimization in Forest Harvesting
Usman, Muhammad (2023)
2023
Master's Programme in Automation Engineering
Tekniikan ja luonnontieteiden tiedekunta - Faculty of Engineering and Natural Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2023-12-31
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-2023123111242
https://urn.fi/URN:NBN:fi:tuni-2023123111242
Tiivistelmä
In the context of a prevailing global mandate for sustainable and renewable practices, the forest industry is undergoing a transformative revolution led by digitalization and Machine Learning (ML). This research investigates the pivotal role of unsupervised machine learning (Clustering) in the geospatial clustering of vegetation polygons within the forest landscape. Different state-of-the-art clustering algorithms are employed to categorize the polygons based on their centroids and vertices.
The study assessed the relative quality of clustering utilizing diverse matrices including Silhouette Scoring, Davis Bouldin Index (DBI) scoring, determination of the optimal number of clusters, average Silhouette, identification of outliers, and accuracy in cluster assignments. These matrices have collectively served as an effective indicator, ensuring the precision and potency of clustering. The evaluation criteria identified K-means clustering using polygons vertices resulted in the best-quality clusters followed by Mean Shift and Hierarchical Clustering based on vertices. Algorithms using polygon vertices resulted in better quality clusters in general compared to centroids-based clustering.
A primary objective of asset optimization in this research is achieved by scrutinizing pair-wise distances between cluster centroids and evaluating homogeneity within clusters. The most effective clustering strategy for effective resource allocation is identified as Fuzzy C-means clustering based on polygon vertices followed by K-means clustering using polygons centroids and vertices respectively. This approach enhances operational efficiency in forestry and contributes to global sustainability goals by reducing environmental impact.
The research has successfully demonstrated the importance and effectiveness of unsupervised machine learning in geospatial clustering by emphasizing optimal strategies such as Fuzzy C-means and K-means clustering for enhanced operational efficiency and resource allocation as a contribution to the digital-driven and sustainable future of the forest industry.
The study assessed the relative quality of clustering utilizing diverse matrices including Silhouette Scoring, Davis Bouldin Index (DBI) scoring, determination of the optimal number of clusters, average Silhouette, identification of outliers, and accuracy in cluster assignments. These matrices have collectively served as an effective indicator, ensuring the precision and potency of clustering. The evaluation criteria identified K-means clustering using polygons vertices resulted in the best-quality clusters followed by Mean Shift and Hierarchical Clustering based on vertices. Algorithms using polygon vertices resulted in better quality clusters in general compared to centroids-based clustering.
A primary objective of asset optimization in this research is achieved by scrutinizing pair-wise distances between cluster centroids and evaluating homogeneity within clusters. The most effective clustering strategy for effective resource allocation is identified as Fuzzy C-means clustering based on polygon vertices followed by K-means clustering using polygons centroids and vertices respectively. This approach enhances operational efficiency in forestry and contributes to global sustainability goals by reducing environmental impact.
The research has successfully demonstrated the importance and effectiveness of unsupervised machine learning in geospatial clustering by emphasizing optimal strategies such as Fuzzy C-means and K-means clustering for enhanced operational efficiency and resource allocation as a contribution to the digital-driven and sustainable future of the forest industry.