Energy modelling: Determining the key factors that differentiate household electricity consumption - A South African study
This thesis describes the results obtained from an investigation focused on determining household features that differentiate the amount of electricity consumed within a household. In analysing the data, Classification and Regression Trees (CART) and Generalised Linear Models (GLM) were used. This report also compares the competence of the CART method and the variable selection in GLM in identifying the key variables.
The household data used was originally obtained from a survey and from monitoring electric meters per household. The data used had 131 independent variables and only 179 households. CART and GLM methods both showed that the as number of appliances, time length since electrification of household and the number of people per household increase so does the amount of electricity. Households that use of coal for cooking used less electricity. Households from some residential areas used less due to cultural differences and the average socio-economic status in the respective residential areas.
With the structure of the data, CART was more competent in identifying the household features that differentiate household electricity consumption.