Machine Learning-based Prediction of Optimal Rubber Tapping Days: A Case Study at Dartonfield Estate, Sri Lanka.

Abstract

In Sri Lanka, rubber is a key agricultural industry, and improving its yield is crucial for sustainability. Rubber harvest depends on both the quality and quantity of the yield, which are influenced by environmental factors such as temperature, humidity, and rainfall. Identifying optimal tapping days is essential, as they directly impact on yield and dry rubber content (DRC), enhancing both productivity and economic efficiency. However, due to fluctuating environmental conditions, accurately predicting the best tapping days remains challenging. This study proposes a machine learning-based approach to model the relationship between rubber yield, DRC, and environmental factors. The predictive model estimates optimal tapping days by analyzing historical weather data, tapping schedules, and yield performance. Machine learning techniques, including regression models and time series forecasting, are utilized to enhance prediction accuracy, achieving a reliability of 77.27%. The primary objectives are to determine the best tapping days and improve yield volume and DRC predictions. The implementation involves calculating daily average weather parameters from historical data, which are then fed into the model to forecast DRC and yield volume. By detecting weather and yield patterns, the model recommends optimal tapping days within a given week, minimizing wastage and maximizing efficiency. Furthermore, this research contributes to improved crop management and resource allocation, ensuring higher production and sustainability in rubber farming. The proposed model is specifically applicable to Dartonfield Estate, Sri Lanka, providing valuable insights for farmers to optimize tapping schedules and reduce wasteful tapping practices.