•Ensemble Machine Learning predicts Asiaticoside in Centella Asiatica.•Utilization of LSTM, GRU, Conv-LSTM, A-LSTM models.•Effective combination of machine learning results using the DE algorithm.•Factors impacting Asiaticoside in CAU: CO2, light, exposure time, cultivar, spectra. This study proposes a novel heterogeneous ensemble machine learning methodology to predict the concentration of asiaticoside in Centella asiatica (CA-CA) in the context of the lack of an effective prediction method capable of accurately estimating its quantity based on various growing environmental factors. The accurate prediction of the asi-aticoside concentration in CA-CA holds great significance in optimizing cultivation practices and improving the efficacy of the derived medicinal products. The presented approach aims to address this crucial need by employing a diverse ensemble of machine learning techniques. The proposed model integrates several machine learning tech-niques, including the standard long short-term memory (LSTM), gated recurrent unit (GRU), convolutional long short-term memory (ConvLSTM), and attention-based LSTM, by utilizing a differential evolution algorithm to optimize the ensemble model's weights. The developed model is called the heterogeneous ensemble machine learning model (He-ML). Experimental results demonstrate that the He-ML achieves an im-pressive root-mean-square error (RMSE) value of 4.76, which is up to 12.48 % lower than the RMSE. The findings highlight the advantages of employing an ensemble model over a single model, as the ensemble model achieves an RMSE value that is 14.67 % lower than that of the individual machine learning model. The utilization of differential evolution as the decision fusion strategy provides a notable improvement over the unweighted average approach. As a result, the RMSE value achieved is 8.46 % lower than that obtained with the unweighted average (UWA) technique.