The Improvement Of K-Means Algorithm And Its Application In Forest Health Assessment

Aiming at some shortcomings of K-Means algorithm and combined with the actual situation of forest health evaluation index selection, the classical K-Means algorithm is improved in some aspects. The improved K-Means algorithm is applied to forest health assessment, which has a strong theoretical and practical significance to improve accuracy of forest health evaluation results. The main research content and research conclusion are mentioned below:(1)Through the analysis and comparison of various clustering methods and the typical process and clustering theory of the classical K-Means algorithm, the paper analyses the advantages and defects of the classical K-Means algorithm.(2)Summarizing the defects of classical K-Means clustering algorithm, an improved K-Means algorithm which bases on the variance of dimension ordering partitioning is proposed to initialize clustering center. At the same time, it determines the clustering number by the minimum distance cost function. So it can increase the rationality of algorithm on the initial clustering center selection and improve the accuracy of clustering results.(3)Using four international date sets---Iris,Glass, Wine and Ecoli as testing datasets, and test this algorithm which bases on the variance of dimension ordering partitioning by numerical simulation. The experiments show that the initial clustering centers chosen by the proposed optimization method are very close to the clustering centers received by the ultimate convergence. Compared with the classical K-Means clustering algorithm, the clustering results have been more stable as well.(4) Combining the actual circumstance of the research of Da Wei mountain Nature Reserve in Hunan Province with the scientific theory, the paper selects 13 indicators, such as stand crown density, to build a forest health assessment indicator system of Da Wei Mountain area, and adopts the method of determining the weight of each index weight, which is determined based on knowledge granularity. The improved K-Means algorithm which bases on the variance of dimension ordering partitioning is applied to the forest health assessment. The quite ideal results of the evaluation were obtained. And the evaluation results were analyzed. Compared with the traditional comprehensive value evaluation and analysis calculated by health formula method, it is more practically significant and applicable.The selection of evaluation method is the key step of forest health assessment process, which will affects the final evaluation results directly. The improved K-Means algorithm which bases on the variance of dimension ordering partitioning is applied to forest health assessment, which increased the objectivity and rationality of assessment standard, and improved the accuracy of the assessment results to some extent as well.