This study investigated cardaba banana succinylation modification, which slowly improves the activity of the digestible starch fraction. Non-linear (response surface methodology [RSM] and artificial neural network [ANN] and partial least squares [PLS] models were used and their predictions were compared. The results showed that all simulation methods were accurate in predicting the experimental performance. The RSM values optimized for slow digestible starch fraction (SDS) were 4% OSA concentration, 47.49 min reaction time and pH 10 with a predicted SDS value of 44.64%. In simulation methods, ANN is known as a predictive model for increasing SDS output. Green banana is a common herbaceous plant in the genus Musa (Arvanitoyannis et al., 2008). It is native to Southeast Asia. However, many varieties have been commercialized decades ago, leading to banana cultivation in more than 100 countries around the world. Among the banana varieties grown in Africa, especially in southwestern Nigeria, is the Kardaba banana (Musa ABB). Kardaba banana (Musa ABB) is mainly used in food processing such as fried chips. This is a rarely used group of bananas that spoil quickly after harvest. 
Their susceptibility to deterioration is due to their extremely high moisture content as well as their metabolic and metabolic characteristics. This high starch content makes cardaba bananas an important raw material for starch production. Starch, an essential component of the human diet, is composed of homoglucose (amylose and amylopectin) and falls into three major categories: resistant starch, SDS, and rapidly digestible starch (RDS). RDS is the fraction of starch that is digested within 20 minutes after ingestion in the human gastrointestinal tract, and RDS-containing foods/starches are generally considered high glycemic index foods. Consequently, starch digestion is slower, resulting in slower release of glucose into the bloodstream. Other benefits of SDS include improved overall blood glucose control in diabetic patients, reduced oxidative stress, and reduced serum cholesterol in patients with hyperlipoproteinemia. Resistant starch refers to the portion of starch that inhibits digestion in the small intestine but is fermented by the gut microbiota to produce short-chain fatty acids (SCFAs). These SCFAs provide the body with extra energy as well as high levels of butyrate; an effective drug that helps reduce the incidence of colon cancer. Foods containing SDS and RS are associated with a low glycemic index, and hence the demand for the production of low-glycemic starch foods has increased recently. However, starch in its native state contains a high proportion of rapidly digestible starch, resulting in a high response to glucose and hence the need for native starch replacement to overcome this problem. 
In general, starch modified using physical, chemical, enzymatic and genetic methods has been extensively investigated and used in the food industry as an alternative approach to improve the slow-digesting starch fraction. Response Surface Methodology (RSM), a statistical and mathematical modeling technique used to design the experiment, build the model, analyze the cross-correlation between the test points, and determine the optimal operating conditions for enhance or decrease test recovery. In fact, these are optimization models, after which nonsignificant independent variables were removed. Partial least squares regression analysis (PLS), unlike RSM, is a non-linear supervised classification, statistical and mathematical modeling techniques used to generate and establish a relationship between two sets of experimental data. An artificial neural network (ANN) is a machine learning tool that simulates the human brain. It contains synchronous neurons and hence can solve the complex nonlinear model of two sets of experimental data. Its advantage over other nonlinear modeling tools such as RSM is that no prior knowledge of the experimental data design is required prior to simulation. Many researchers have used it in conjunction with other modeling tools in their modeling and found it to be superior to other tools.
Therefore, this study aimed to optimize the content of slowly digestible starch fraction in cardaba banana starch by modeling and optimizing process variables (OSA concentration, reaction time and substrate pH) using response surface methodology (RSM).) and compare updated values. obtained using RSM with partial least squares regression (PLS-R) and artificial neural network (ANN). CARDABA BANANA STARCH ISOLATION Isolation of starch from cardaba banana was carried out according to the method described (Olavoye and Gbadamosi, 2020b). Briefly, bananas were washed and cut under water to prevent the banana slices from browning. Sliced cardaba bananas were crushed on a Stephan mill (Stephan universal mill, Germany). After the trip, the starch pellet was mixed with water, and then the starch layer was filtered through a sieve (200 μm). The resulting starch solution was left overnight, after which the supernatant was discarded. The remaining starch was washed three times with distilled water. The resulting starch was dried at 45°C for 12 h. The dried starch was packed in a sealed container and stored at room temperature until analysis. PRODUCTION OF OSA STARCH Isolated cardaba banana starch (native starch) was converted using the method of Han and BeMiller (2007) modified by Olavoye and Gbadamosi (2020b) to produce OSA starch. 
Cardaba native banana starch (100 g, dry weight) was dispersed in a 500 ml conical flask containing distilled water with constant stirring on a magnetic stirrer. The pH of the starch paste was adjusted by magnetic stirring to various pH values from 8 to 10, as indicated in the experimental setup, using 1 M NaOH and 1 M HCl. After adjusting the pH of the substrate, varying concentrations of octenyl succinic anhydride (3–5%) were added and the stirring continued at room temperature (≈26 ± 2°C), while the pH was maintained with 1 M NaOH or HCl at different temperatures reaction times ranging from 30 to 60 mi. At the end of each reaction time, the starch pellet was neutralized at pH 7.0 with 1M HCl or 1M NaOH. The modified starch was collected by centrifugation, washed three times with water, and then dried at 45° C. for 12 h to obtain OSA starch. The dried starch was stored in a sealed container until further analysis.
Your comment submitted.