Abstract
The objective of this paper is to analyze the use of nonlinear models to predict the CPU frequency that reaches the lowest power consumption of a smartphone based on Android OS context variables. Artificial neural networks (ANNs) and k-nearest neighbors (k-NN) techniques are investigated, and their results are compared to those obtained by the linear method (LM). Experimental results indicate the k-NN technique is the best option in terms of model accuracy and performance when compared to the other prediction models.