Evaluation of Machine Learning Models for Forecasting Water Consumption: A Case Study in New York City

To research forecasting water consumption in New York City, this paper uses daily water consumption (based on million gallons per day) from ۱۹۷۹ to ۲۰۱۷ as sample data, and so there are ۳۹ observations in this data set. Firstly, the paper applies two machine learning models, Simple Linear Regression model and Polynomial Regression model. In both models, the water consumption is predicted by the feature “Year”. As a result, the evaluation of these two models shows that the polynomial regression model is more accurate on this data set and fits more better on real data values. Secondly, the Multiple Linear Regression is examined considering two independent features, “Year” and “Population”. This model acts well on real test values too. Lastly, Hierarchical Clustering is quoted in this research. Regarding five clusters, the proportion of each year in water consumption is analyzed.

کلیدواژه ها:

Forecasting ، Linear Regression ، Water Consumption ، Machine Learning Models