Regresar

A New Approach to Supervised Data Analysis in Embedded Systems Environments: A Case Study

Abstract:

Nowadays, the implementation of embedded systems with sensors for massive data collection has become widely used for their flexibility and improvement in decision making. However, this process can be affected by errors in reading, attrition of systems, among others. For this, a selection approach of supervised algorithms with a prototypes selection criterion is presented, which allows an adequate embedded system performance. To do that a quality measure was established which compromises between the data reduction of the training set, algorithm processing time and the classification performance. As a result, it was determined that the algorithm for the data selection is Condensed Nearest Neighbors (CNN) and the classification algorithm is k-Nearest Neighbour (k-NN).