@masterthesis{Bueltemann, type = {Bachelor Thesis}, author = {B{\"u}ltemann, Marlene}, title = {Sustainable Learning Analytics: Measuring and Understanding the Drivers of Energy Consumption of AI in Education}, school = {Hochschule f{\"u}r Technik und Wirtschaft Berlin}, pages = {43}, abstract = {AIED (AIED) is used increasingly to achieve higher learning outcomes through adaptive personalized learning paths. However, its environmental impact has been disregarded thus far. In this thesis, local energy and resources consumption of the prediction in four AIED models are measured using two different measurement approaches. The results suggest that boolean or mathematically explainable models such as decision tree and logistic regression have a low energy demand. On the other hand, in models that use a complex optimization process or have mulitple interconnected layers like support vector machine (SVM) or multilayer perceptron (MLP), the resource demand rises significantly. A decision tree model is recommended as the model of choice to deploy on the learning platform as it reveals the lowest emissions in both measurements and shows high performance prediction values. The thesis concludes with recommendations for AIED developers and policy ecommendations and pledges for higher attention on the environmental impact of AIED.}, language = {en} }