Update README.md

README.md

@@ -1,15 +1,8 @@
 # Water-Demand-Forecasting Characterizing demand non-stationarity in the water system.
 
-Abstract: The assumption of per-capita demand stationarity plagues municipal water systems.
-Identifying sources of uncertainty in projecting water demand and representing their influence in quantitative estimates will help improve water system planning and operations. 
-To address these needs, this paper has two objectives A) characterize the impacts of demand stationarity on forecast accuracy and B) develop a coherent machine learning framework to improve the accuracy of seasonal demand estimates.
-Using the Salt Lake City Department of Public Utilities service area, we benchmark stationary demand forecasting performance and limitations with observed demands from historical periods representing drought, average, and surplus supply conditions.
-Using an ensemble of ML tools, the ridge regression-based Salt Lake City Water Demand Model (SLC-WDM) mitigates the stationary limitations to produce climate-driven season demand forecasts, expected to be more accurate.
-Performance evaluation of the SLC-WDM forecasts compared with multi-layered perceptron (MLP) and Random Forest regression (RFR) models indicates performance-interpretability trade-offs.
-We find the stationary assumption leads to over-estimates of monthly drought-condition demands up to 9\% and seasonally up to 40\%, a result of an extended irrigation season and regional drought awareness.
-In these conditions, SLC-WDM forecasts are within 7\%  and 0.2\% of the observed monthly and season values, respectively.
-The SLC-WDM also minimized error in all supply scenarios, exceeding those from the stationary,  MLP, and RFR forecasts (\textit{MAPE} = 8.4\%, 31.0\%, 9.0\%, and 10.5\%, respectively).
-As a result, we demonstrate stationarity is a critical limitation in demand forecasting and a coherent machine learning framework can capture climate-driven demand non-stationarity.
+Abstract: Comprehensive seasonal water system assessments in the western U.S. are confounded by per-capita water demand stationarity, owing to increased system uncertainty from inter-annual demand variability not represented in industry forecasting methods. Using Salt Lake City's Department of Public Utilities observed demand variability from recent episodes of drought, average, and surplus supply conditions, we measure the prediction accuracy of industry demand forecasting methods embedded with per-capita stationarity assumptions. To address the observed variability in intra- and inter-annual water demands, an ensemble of machine learning variable selection and optimization tools, hydro-climate and exogenous service area factors, and regression algorithms are leveraged to form the Climate-Supply-Development Water Demand Model (CSD-WDM).
+A multi-layered perceptron (MLP) and Random Forest regression (RFR) model are also developed to evaluate model complexity vs. accuracy trade-offs with the CSD-WDM.
+Industry forecasting methods exhibited high prediction errors in all climate scenarios, which we attribute to insufficient model complexity. These errors peak during drought conditions where over-predictions near 90\% and 40\% of the observed monthly and seasonal demands, respectively. The mean absolute percent error of CSD-WDM (8.4\%) shows improvement over the MLP (9.0\%), RFR (10.5\%), and industry models (31.0\%), accurately projecting demands in all climate scenarios. Overall, the CSD-WDM reduced seasonal water demand uncertainty by 6.0\%, 40\%, and 30\%  during surplus, drought, and average climate conditions when compared with industry methods.
 
 
 This repository contains the data (in the Data branch) and modeling scripts (also containing data analysis and figure development) for the Salt Lake City Water Demand Model (SLC-WDM), multi-layered perceptron (MLP), and Random Forest (RFR).