Pre-processing techniques using a machine learning approach to improve model accuracy in estimating oil palm leaf chlorophyll from portable chlorophyll meter measurement

Abstract

Chlorophyll is essential for plants because it absorbs and adjusts solar energy as an energy source in photosynthesis. Thus, chlorophyll content can be used as an indicator of plant performance. However, laboratory analysis should usually be conducted to measure chlorophyll content, which requires a lot of tools and time. Therefore, Chlorophyll Meter SPAD (Soil Plant Analysis Development) is a portable tool to estimate relative leaf chlorophyll rapidly. Still, it needs to be modeled and validated to get accurate results as laboratory analysis. Therefore, appropriate data pre-processing techniques are required before entering the model development stage. The pre-processing techniques included data collection, identification, and anomalies check. In addition, a machine-learning method was employed to do K-Means clustering during checking data anomalies. The results showed that the pre-processing stages increased the R-value from 0.588 to 0.912 and reduced the Mean Absolute Percentage Error (MAPE) value from 58.91% to 39.86%.