Data mining techniques for the reduction of costs in the automation of the tilapia cultivation processdata mining techniques for the reduction of costs in the automation of the tilapia cultivation process.

Abstract

The economic investment in technology hinders the automation of the processes of people engaged in agricultural, livestock or fishing activities, mainly due to the high price of some sensors used. In the facilities of the Centro de Bachillerato Tecnológico Agropecuario No. 14, the automation of the tilapia cultivation process was carried out in order to evaluate the gain in the weight and size of this species using a mechanized system versus the traditional system. The physical and chemical variables that influence the biological development of these fish are diverse; however, dissolved oxygen (DO) is the most important factor in the cultivation process. Unfortunately, the cost of the sensor required for this process is expensive. For this reason, data mining techniques were used in order to determine a model that allows predicting the levels of dissolved oxygen from the values obtained by turbidity, temperature and hydrogen potential (pH) sensors. The techniques used for this study were multilayer perceptron algorithms, M5P and linear regression. The mathematical model obtained will allow predicting the degree of the DO without the need to have the corresponding sensor. The impact is that the model could be extrapolated to other similar automation projects, thereby reducing its investment costs.