Artificial Neuro-Fuzzy Inference System (ANFIS) based validation of laccase production using RSM model

Abstract Numerous reports exist in recent literature where the One Variable at Time (OVAT) based optimization of medium components achieved for the laccase production enhancement by fungi and bacteria. OVAT strategy is not a suitable for the cost-effective production of enzymes in lieu of modern statistical and mathematical techniques like Artificial Neural Network (ANN), Artificial Neuro-Fuzzy Inference System (ANFIS) and Response Surface Methodology (RSM). By using RSM maximum laccase yield was achieved 7.4 × 104 nkat L−1 from γ-proteobacterium JB in best combination of the factors, pH 8.0, 210 rpm, 100 µM, CuSO4 after 60 h of incubation time. In this paper an ANFIS was designed and trained by inputting 75% of the total combinations of factors (pH, rpm, CuSO4 and incubation time) along with their respective laccase yield as produced by the conventional system of experimentation. The trained system was tested on 25% of the total combinations of factors (pH, rpm, CuSO4 and incubation time) along with their respective laccase yield as produced by the conventional system of experimentation. The training phase and testing phase error reported by the ANFIS is 0.084573 and 0.12647 respectively which are quite tolerable while dealing with the limited actual experiment results. The ANFIS laccase yield prediction results are in consonance with those produced by the RSM system and in fact are closer to the actual laccase yield.