Multiplatform information system for monitoring and predicting physicochemical variables in the inland fish farming industry

According to FAO, global fish farming has decreased by thirty points the biological capacity to sustain supply against population demand, i.e. from 90% to 60% in 2018, which implies recognizing that there is a lack of good responsible practices that ensure the integrity of the environment and the resources needed to produce. According to the 2030 sustainability agenda, it is evident the need to incorporate low-cost technological aids that impact water quality processes, which are guarantors of the quality of the final product. The department of Huila-Colombia currently produces about 68,000 tons of protein by 2021 and is a source of work for small and medium-sized producers, who have achieved formalization, but lack the resources to reinvest in technology, as they assume very high production costs.This research project endorsed and funded by the Ministry of Science, Technology and Innovation of Colombia in its call 808 of 2018 - challenges for peace in alliance with the University of the Andes, the University Corporation of Huila - CORHUILA and ASOPISHUILA, where it was possible to characterize, collect, analyze, generate a predictive model integrated with the information system tailored and low cost to predict physicochemical crop variables such as pH, temperature, ammonium, nitrates among others in an observation sale of seven days. This makes it possible to take preventive decisions in case the permissible values are above or below the established values.Finally, it was possible to obtain a random forest prediction algorithm with an R2 of 0.62, much higher than those processed as linear regression and neural networks. Indicating the high probability of prediction in temperature, pH, and dissolved oxygen, but not so efficient for nitrates and nitrites.