A New Approach to Classify Growth Patterns Based on Growth Function Selection and K-means Method

Abstract

While every growth behaviour can be described by a growth function with longitudinal independent variables, modeling longitudinal growth behaviour is complex and requires a strategic search for an appropriate growth function. Existing literature shows that several growth functions have been derived and applied to suitable growth behaviour. This implies that there is always the possibility of the existence of a unique growth function that fits a given growth data. In this paper, we propose a new statistical procedure to seek the "best" growth function as well as the "best" clusters of growth patterns. This new procedure involves a mixture of growth function selection and k-means method. Although our experiments were limited, using a real data set of sugi (Cryptomeria japonica) tree growth and hypothetical observations, our results show that the proposed method is promising. It allowed us to choose the "best" growth function that matches the "best" growth patterns at the same time.