Prediction of Shelf Life of Pearl Millet Flour Based on Rancidity and Nutritional Indicators Using a Long Short-Term Memory Network Model
Harshvardhan Pande, Shrikant Kapse, Veda Krishnan, Shankar B. Kausley, C. Tara Satyavathi, Beena Rai
Abstract
Millets have garnered less attention as compared to other grains despite being versatile and highly nutritious crops due to the poor shelf life of millet-based products such as millet flour. The shelf life constraints stem from significant concerns about the rancidity and degradation of nutritional value in millet products over time. Conventionally, the shelf life of millet-based products has been estimated through experimental analysis of rancidity indicators. Prediction of shelf life of food and beverages has been facilitated by the advancements in artificial intelligence and computational learning techniques. This study employs a long short-term memory (LSTM) network architecture model to predict shelf life based on nutritional and rancidity indicators. The performance of the LSTM network was compared with the feed forward neural network for six different types of pearl millet variants. The LSTM network demonstrated better performance with R 2 > 0.96 for all of the predicted variables. Additionally, the study found that the nutritional value limits the shelf life of low-rancidity variants to 14–16 days, while high-rancidity variants have a shorter shelf life of 4–7 days due to higher rancidity.