Predicting Prosumption Survey Response using Machine Learning

The prosumer attitude is more and more common nowadays. A prosumer is a person who is a consumer and a producer at the same time, e.g., by giving feedback to the producer from the consumer’s point of view. Surveying prosumer attitudes is costly and time-consuming, so an attempt was made to use a machine learning technique, a decision tree, to predict survey answers based on respondents’ characteristics. The trees were built, validated, and tested on data from a survey conducted on a representative group of prosumers. Two factors influencing prosumer attitude were predicted. The validation accuracy of both trees was 58.1%, and the prediction accuracy on new data was 57.1% and 57.4%, respectively. The results show that it is possible to predict the prosumer survey answers with machine learning techniques on quite a satisfactory level. In further research, other machine learning methods will be employed to check if they give better results.