HARVESTING OPEN DATA FOR HOUSEHOLD CLASSIFICATION

Research Highlights

- Free geographic data from OpenStreetMap helps to predict household characteristics for residential electricity customers

- Semi-structured geographic data was modelled such that machine learning algorithms can effectively process the data

- Public statistical data and data from real estate portals only lead to marginal improvements

 

Challenge

Open Data emerged as a promising source of information freely available to firms. Such data can come in the form of Volunteered Geographic Information (VGI), e.g., from OpenStreetMap (OSM) or real estate portals. Such data might be especially helpful in combination with customer data that companies already have available, e.g., to improve classification models for targeted marketing or energy efficiency campaigns.

 

Approach

We tested to what extend VGI data can support the classification of households based on location data and annual electricity consumption data for 3,905 electricity customers in Germany and Switzerland. For this purpose, we derived 60 features from semi-structured OSM data and included information from Eurostat and from over 100,000 advertisements from real estate portals.

Results

The results show that the identification of customers with specific characteristics (e.g., living alone, having large dwellings or electric heating systems) is possible based on data available to energy retailers. The accuracy of the results significantly improved when geographic features are included in the analysis. However, public statistical data and real estate advertisements did not significantly improve the prediction performance of the classifiers under study.

Selected publications

Hopf, K. (2018). Mining Volunteered Geographic Information for Predictive Energy Data Analytics. Energy Informatics 1:4

 

Hopf, K., Sodenkamp, M., & Kozlovskiy, I. (2016). Energy data analytics for improved residential service quality and energy efficienc, presented at 24. European Conference on Information Systems (ECIS), in ECIS 2016 Proceedings. Istanbul, Turkey: AIS electronic library.

 

Funding

This project has been funded by Swiss Federal office of Energy (Grant numbers SI/501053-01, SI/ 501202-01) and the European Union (EUROSTARS Grant number E!9859 - BENgine II).

Date: 2015-2017

 

Team

Konstantin Hopf, Mariya Sodenkamp, Thorsten Staake.  


All projects on Machine Learning for Customer Insights

Harvesting Open Data for Household Classification

Testing the Transferability of Household Classifiers

Identifying Target Customers for Up-Selling: The Case of Biogas


CONTACT US

For further information or if you have any questions please do not hesitate to contact us.