Machine learning techniques for water demand modeling

The codes contained here implement the methods of the paper "Urban water demand modeling using machine learning techniques: Fortaleza – Brazil case study", accepted for publication in the Journal of Water Resources Planning and Management. The code includes the R implementation of the Iterative Input Selection (IIS) algorithm proposed by Galelli and Castelletti (2013). The original Matlab implementation of the IIS algorithm (Galelli and Castelletti; 2013) can be found here.

Galelli, S., and A. Castelletti (2013), Tree-based iterative input variable selection for hydrological modeling, Water Resour. Res., 49(7), 4295-4310.

Requirements

Before running the codes, you need to install the following packages:

install.packages(randomForest)
install.packages(RSNNS)
install.packages(kohonen)
install.packages(Metrics)
install.packages(dplyr)

Description

The codes and data contained here are as follows:

• performance_measures: compute the desired performance measure
• variable_ranking_rf: rank the variables according to the IncMSE and the RF algorithm
• som_clustering: performs SOM based clustering
• miso_ann: runs ANN with multiple inputs/single output
• siso_ann: runs ANN with single input/single output
• iterative_input_selection: runs the IIS algorithm with RF and ANN

Infographic

License

These codes are free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published bythe Free Software Foundation, either version 3 of the License, or any later version.