Application of ANN and traditional ML algorithms in modelling compost production under different climatic conditions

Plentiful and diverse Organic wastes such as green, food, pruning and landscaping waste necessitate upon effective and efficient recycling strategies such as composting. In this article, for the first time, machine learning (ML) models have been applied to model compost generation rate as a function of climatic parameters and organic waste content. The data size contained approximately 864 sample points records of the meteorological parameters (Modern-Era Retrospective Analysis), organic waste (Central Pollution Control Board) and compost yields (Open Government Data) data for 2010–2021. The modelling efforts involved the consideration of MLP and traditional ML algorithms namely k-nearest neighbour (kNN), gradient boosting (GB) and random forest (RF) for prediction and autoregressive integrated moving average (ARIMA) model supplemented ML models for long-term forecasting of the compost generation rate. Model validation resulted in an RMSE of 0.757 and R² of 0.99 for GB model, and a correlation index of 0.68 between observed and predicted values to thereby outperform all other models. However, forecasted data for ten years after the predicted outcomes resulted in the best performance of ARIMA–MLP model with a standard error of 21.1152 and a CP yield of 74,958 kg. Thereby, the findings affirm upon the evidence for the limitations of the broader application of the empirical approaches and the feasibility of ML algorithms as a potential reconstruction technique for developing robust and accurate region-specific compost prediction and forecasting models to assist integrated circular agricultural system development for a sustainable global future.