Coefficients search for a regression curve of a neonatal incubator using evolutionary computing techniques

A regression model to correct temperature readings of an infrared sensor has been implemented, the target was to predict the known inner object temperature readings using the infrared temperature readings of the object’s surface, the infrared sensor die temperatures and light intensity data collected from a 35 hour experiment with the object lying inside a neonatal incubator. We have applied different evolutionary computing techniques like Genetic Algorithms (GA), Particle Swarm Optimization (PSO) and Differential Evolution (DE), to obtain the best fit to the data and in this way correct the temperature reading. The best prediction was achieved with differential evolution with a mean square error metric of 0.044404 followed by 0.044410 obtained with PSO and 0.046037 with GA. Also, with the best results obtained, the error was calculated considering the measured light intensity signal to see if this variable helps to further decrease the error.