Researchers have defined a new machine learning-based methodology that is reported to reduce customer acquisition costs by about 15% or $0.07/Watt. It is based on an adapted version of the XGBoost algorithm and takes into account factors such as summer bills, household income, and homeowner age.
An international research team utilized a machine learning algorithm known as XGBoost (eXtreme Gradient Boosting) to predict PV adoption among homeowners. The algorithm consists of the Variance Gradient Boosted Decision Trees (GBDT) machine learning library and helps to accurately predict a target variable by combining an ensemble of estimates from a set of simple and weak models. .
“We delve further into the modeling details of XGBoost and decompose its enhanced predictive performance for logistic regression into two factors: variable interaction and nonlinearity,” said the scientist. Finally, she demonstrates XGBoost’s potential to reduce customer acquisition costs, and then he demonstrates the ability to identify new market opportunities for PV companies. ”
According to them, the new methodology could help solar companies reduce customer acquisition costs and other soft costs associated with residential solar businesses.
They compared the performance of their proposed algorithm with a logistic regression approach. The logistic regression approach is the most commonly used method to analyze the differences between PV adopters and non-adopters, the researchers explained. “Our logistic regression model, with nine unique and highly visible home features, successfully predicts 71% of out-of-sample PV adoption,” they further explain. Did. “The model correctly identified 66% of recruits and her 75% non-adopters.”
According to the research group, adaptive algorithms could provide better results than logistic regression in predictive performance. “The predictive model correctly predicted 87% of the two PV adoption statuses compared to 71% for the logistic regression,” they added. “The correct answer rate increased from 66% to 87%, and the incorrect answer rate increased from 75% to 88%.”
They say the machine-learning-based approach’s superior performance integrates complex nonlinearities and variable interactions, specifically considering factors such as summer bills, household income, and homeowner age. I believe it is due to facts.
“The advantage of using these variables is that they are highly accessible, allowing PV companies to collect the data at very little cost,” they said. “Another reason to explain XGBoost’s improved performance is its potential to recover important latent information embedded in the data.” , which improves the predictive accuracy of logistic regression to some extent.”
The research group estimates that the new methodology will help PV companies reduce customer acquisition costs by about 15%, or $0.07/Watt. He also explained that data mining and machine learning could also help reduce soft costs around contract cancellations, supply chain management, labor allocation, and permitting and inspection issues.
I described the new methodology in a study published in , “Machine Learning Reduces Soft Costs of Residential Solar Power”. scientific reportThe research group is formed by scientists from the U.S. Department of Energy’s National Renewable Energy Laboratory (NREL), Lawrence Berkeley National Laboratory, Florida State University, University of Wisconsin-Madison, and Renmin University of China.
This content is copyrighted and may not be reused. If you would like to collaborate with us and reuse some of our content, please contact us at editors@pv-magazine.com.
