An Adaptive Approach to Quantify Plant Features by Using Association Rule-Based Similarity

Abstract

In elementary school, the first step to learning about plants is observing plant features. In this study, we developed a plant search system that allows users to do a search even when they do not know the plant name simply by observing plant characteristics. The system consists of a total of 12 plant features, searches for the features according to the input features, and returns the top 10 plants with the best match. Furthermore, we adopted three different calculation methods for feature calculation, including the fuzzy function, association rule-based similarity (ARBS), and a combination of the two. The calculated results of the three aforementioned approaches are analyzed in case of a single feature input error or n features were input incorrectly; then the Top 1 and Top 3 accuracy are explored. According to the results of this study, the accuracy of ARBS is significantly higher than that of the other two calculation methods, thus proving that calculating the similarity through association rules can greatly increase accuracy. If future researchers were to expand to other features, the various features, even those that are hard to quantify, can also be quantified using the ARBS method easily.