A Data Mining Approach to Understanding UK Fuel Poverty

K.J. Brazier, W. Wang, G. Richards, and C. Waddams (UK)


Fuel poverty, data mining and survey analysis


Fuel poverty is broadly defined as the placement of an unreasonable demand on a household’s income by its need to maintain adequate levels of essential fuel use. Research is described into the application of data mining techniques to assess the impact of fuel poverty policy. In particular, survey data are used to induce and validate decision tree models, and the salience of surveyed factors, quantifying their relevance to fuel poverty, is assessed. Less than 300 of the original 1422 factors are shown to be relevant in the construction the induced fuel poverty models, around 100 making a substantial contribution. Further decision tree models are induced using only these most salient factors. In terms of the identification of positive cases of fuel poverty these outperformed the models generated using all surveyed factors by more than 10%. The results show that relationships modelling the incidence of fuel poverty to a high degree of accuracy may be abstracted and suggest substantial efficiency savings in the monitoring of fuel poverty, with a potential for improvement to policy implementation.

Important Links:

Go Back