Big data analytics using optimized LSTM for carbon cost reduction in the construction industry

Reducing carbon emissions is a major challenge for the building sector and is essential for environmental sustainability. Because existing approaches cannot handle complex, time-dependent data, they frequently fail to anticipate carbon prices accurately. Our method improves forecast accuracy by utilizing optimized LSTM networks and large data analytics, allowing for more efficient carbon cost reduction plans. This method improves real-time decision-making by incorporating temporal interdependence and overcoming constraints in existing models. Initially, historical data on energy, emissions, investments, and economics of the Chinese residential construction industry were given by the China-Building-Energy-and Emission-Database (CBEED) during 2000 to 2015. Z-score normalization and missing value relevant data were used in the data prepossessing process to normalize the data's scale. Recursive feature elimination with linear regression model (RFE-LR) is a wrapper technique that expands the set of features for the LSTM model by repeatedly removing the least significant features based on model performance. This reduces overfitting and increases computing performance. Features were gathered from huge quantities of data. Our proposal of the hybrid harmony search algorithm with red fox optimized (HHSARFO) to reduce carbon costs in the construction industry by utilizing long short term memory (LSTM) for hyper parameter optimization. The proposed improved performance over other similar models in terms of precision (99%), carbon emission (150ton) RMSE (26%), MAE (30%), and R2 (67%) indicates the resultant carbon costs reduction in the building industry. The concluded a great deal of promise for lowering carbon emissions, improving sustainability, and encouraging environmentally friendly practices in the building sector.