Adaptive Neuro-Fuzzy Inference System on Aquaphotomics Development for Aquaponic Water Nutrient Assessments and Analyses

Abstract

Water quality monitoring, assessment, and analysis in an aquaponics system are vital procedures in maintaining a productive and an efficient ecosystem for cultivars being cultured. However, these require labor-intensive, long-standing, and high-priced laboratory methods, as water quality and its nutrients are dependent on micro-biological and physio-chemical variables. To reduce the use of costly sensors and the time consumed for expensive calculations, an Adaptive Neuro-Fuzzy Inference System (ANFIS)-based aquaphotomics approach is conducted on this study. Water samples were collected from a pond water of an aquaponics system (AP) where species of fish are cultivated. The samples went through aquaphotomics with the aid of spectrophotometer and was applied on to near infrared, visible light, and ultraviolet (NIR-Vis-UV) spectrum with wavelength range of 100 to 1000 nm. Spectrometry was utilized to determine three significant nutrient compounds which are the nitrate, phosphate, and potassium. Temperature, power of hydrogen (pH), and electrical conductivity sensors (EC) were used simultaneously to serve as data attributes in predicting the three compounds assigned as targets. Feature selection algorithms such as Minimum Redundancy Maximum Relevance (MRMR) and Univariate Feature Ranking for Regression Using F-Tests (UFT) were used to determine the two most significant predictor relative to a specific target. Results showed that MRMR with ANFIS is best used for predicting Phosphate with R2 value of 0.8284. The UFT with ANFIS produced the best performance for regressing Nitrate and Potassium with R2 values of 0.9321 and 0.9961 respectively.