Fig 1.
Process followed in this study.
Fig 2.
Geographic coordinates are 2.796162372495068, -75.29639700197488.
Fig 3.
Pond variables measured every 3 hours during 48 hours.
(A) Dissolved oxygen and pH. (B) Environmental and pond temperature. (C) Ammonia, ammonium and nitrites. (D) Electric conductivity and alkalinity. Note that the maximum and minimum values tend to be reached at dawn and sunset.
Fig 4.
Distribution of measured variables.
(A) Environmental temperature, (B) pond temperature, (C) dissolved oxygen, (D) pH, (E) electric conductivity, (F) ammonia NH3, (G) ammonium NH4, (H) nitrites, and (I) alkalinity.
Table 1.
Performance indicators for measurement estimation based on other variables.
Fig 5.
Weighted variable importance (arcs) and R2 (nodes) for measurement estimation based on other variables using the random forest model.
Nodes with lighter colors indicate higher R2. Lighter colors of an arc from node i to node j indicate higher importance of variable i to estimate variable j.
Table 2.
Summary of results for variable estimation.
Table 3.
Performance indicators for variable forecasting.
Indicator PO refers to the number of past observations of all variables required to forecast each variable.
Fig 6.
Weighted variable importance for forecasting of each parameter and R2 of that forecasting using the random forest algorithm.
Lighter color of nodes refers to a higher R2. Similarly, lighter color of an arc that goes from i to j refers to a higher importance of that parameter i to forecast variable j.
Fig 7.
Forecasting of on-site measurements for a time horizon of 14-time steps (one week).
Note that, in general, the error for the first two predicted measurements (one day) for each variable, is low.
Table 4.
Summary of results for variable forecasting.
Table 5.
Performance indicators for variable forecasting based only in the most important variables using random forest algorithm.
Table 6.
Performance indicators of an information system for fish farmers implemented using amazon web services.