Inventory model using Machine Learning for demand forecast with imperfect deteriorating products and partial backlogging under carbon emissions

In today's environment, organizations utilise Machine Learning based-models to keep stocks depending on the demand for a particular type of product. This article develops an inventory model considering the imperfect deteriorating product in a fuzzy environment. The shortages are allowed and partially backlogged. Since the deterioration rate and defective percentage in quantity in the received lot may not be predicted precisely because it depends on many uncertain situations, therefore both are considered fuzzy variables. This study aims to determine the optimal ordering quantity and replenishment period to optimize (minimize) the average overall cost with carbon emission cost. The defuzzification process is done using the sign distance approach method. A methodology based on Machine Learning is used to demand forecast seasonally. Some numerical examples are taken to validate the proposed mathematical model. The findings demonstrate the generation of direct month-wise predicted demand for deteriorating products based on the input of the month value, enabling organizations to optimize their inventory management according to forecasted demand. A comparative analysis is conducted between fixed and month-wise forecasted demand by highlighting the advantages of machine learning-based forecasting approaches. Sensitivity analysis performs to examine the behaviour of several parameters on an optimal solution and provides some managerial insights.