Integrating Deep Learning and Statistical Models for Purchasing Prediction in Inventory Management

Effective inventory management is an essential component of all corporate operations, particularly for sustaining equilibrium between demand and supply. Efficient inventory management strategies can markedly improve product availability, decrease storage expenses, and increase firm profitability. Businesses frequently encounter hurdles, including market demand volatility and difficulties in forecasting evolving trends, rendering precise buy prediction in inventory management crucial. The suggested ARI-RNN model employs a three-phase strategy to tackle these challenges: Preprocessing, Feature Selection, and Model Training. Data preprocessing guarantees quality, whereas feature selection use chi-square and ANOVA techniques to discern principal predictors. The approach combines two linear methodologies—the ARI average and exponential smoothing models—with a non-linear Recurrent Neural Network (RNN) model. A unique regression technique is presented to produce optimised training data for the RNN. Experimental findings indicate that the RNN model substantially surpasses linear models in producing precise predictions. The optimised ARI-RNN model attained a remarkable 99.21% accuracy and a 99.65% F1-score during testing, surpassing conventional classification methods. The findings demonstrate that the ARI-RNN methodology offers a reliable and precise solution for buy forecasting in inventory management, hence assisting organisations in enhancing their inventory planning and operational efficiency.