Yield Prediction for Lac Cultivation in Tropical Region using Moving Average LSTM Reinforcement Algorithms

Lac, a resin secreted by the lac insect, holds significant value across various industries, including food, pharmaceuticals, cosmetics, and varnishes. Efficient management and accurate prediction of lac production are critical for optimizing supply chains, meeting market demands, and enhancing profitability. This study evaluates advanced machine learning techniques, specifically Moving Average Long Short-Term Memory (MA-LSTM), Convolutional Neural Networks (CNN), and Random Forest (RF) models, for predicting lac yield in tropical regions. The MA-LSTM model combines Exponential Moving Average (EMA) for smoothing short-term data fluctuations with LSTM networks to capture long-term dependencies. Results show that the MA-LSTM model surpasses traditional LSTM and other machine learning approaches in predictive accuracy, achieving lower Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) values. Enhanced prediction capabilities from the MA-LSTM model could benefit stakeholders in the lac industry by supporting more efficient production planning, inventory management, and strategic decision-making. This research underscores the potential of machine learning to advance agricultural productivity and sustainability through improved predictive analytics.