Efficient Value Stream Mapping in Perishable Materials Supply Chain Using Linear Programming-Based Recurrent Neural Method

The challenges faced in the supply chain of perishable goods are significantly impacted by deterioration of the product over time, volatility of demand, limited shelf life and inefficiency caused by logistics and inventory coordination; these factors together increase cost and waste while reducing the quality of service. This research study presents an innovative Linear Programming based Recurrent Neural Model (LPbRNM) for optimising the mapping process of value streams and the optimisation of decisions. The linear programming (LP) aspect of the model seeks to minimise total operational cost (production costs, storage and transportation costs) given demand, capacity, and shelf life limitations, while the Recurrent Neural Network (RNN) portion of the model forecasts time-based demands based on historical supply chain data and the deterioration of product quality. The LP component of the model provides an optimal approach to the allocation of resources relative to demand and capacity constraints. In contrast, the RNN component provides the ability to identify temporal patterns in order to dynamically adjust inventory and fulfilment decisions to create an adaptive and predictive optimisation framework. In addition, activities are classified into value added or non-value-added activities in order to reduce waste and improve throughput. The experimental results demonstrate that LPbRNM provides high levels of predictive accuracy (0.957), low levels of prediction error (MAE 0.052; RMSE 0.081; MAPE 5.63%), high levels of diversity between potential solutions (Maximum Spread 0.529; SNS 0.5736) and stable levels of performance in terms of computational efficiency, thereby providing a method for improving cost efficiency; reducing spoilage; and improving operational reliability with perishable supply chain operations.