Predicting Monthly Household Water Consumption

The monthly household water consumption prediction is an interesting and practical problem in predicting the volume of water use of a residential area. Therefore, this study develops a hybrid approach between a Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory networks (Bi-LSTM) named CBLSTM model for predicting the monthly household water consumption. The experiments are conducted on a dataset collected in Can Tho city, Vietnam in three years from 2018 to 2020 (named by MWC-CT dataset). We compared the proposed model with the two state-of-the-art models for time series prediction which are the LSTM and Stacked LSTM models. Evaluation of experimental results on the MWC-CT dataset indicates that CBLSTM performs better than the comparative models.