<![CDATA[Purpose: This study evaluated an inventory system integrating LSTM forecasting and Hyperledger Fabric blockchain to improve prediction accuracy and transaction integrity. Design/Methods: A design-and-development approach used 12,450 inventory records from Retail Company X (January 2021-December 2023), split chronologically into 70% training, 15% validation, and 15% testing subsets. The LSTM used two hidden layers, 128 units per layer, dropout 0.2, Adam optimizer, learning rate 0.001, batch size 64, and 100 epochs. Blockchain used Hyperledger Fabric with Raft consensus. Evaluation included forecasting benchmarks, 50 stock-modification simulations, and 45 purposively recruited users after hands-on prototype interaction. Findings: LSTM achieved MAE 3.2% and RMSE 4.5%, outperforming Moving Average and Exponential Smoothing. A two-tailed paired-samples t-test across 62 matched testing windows against Exponential Smoothing confirmed significant improvement (t(61) = -5.34, p < 0.001, Cohen's dz = 0.68). Blockchain detected 48 of 50 unauthorized stock modifications, producing a 96% detection rate with two missed detections (4%) and 120 ms latency. User evaluation was positive across forecast accuracy, security, transparency, ease of use, and intention to use. Implications: The prototype can support inventory planning, auditability, and secure transaction records. Originality: The study empirically combines AI forecasting, permissioned blockchain integrity, and user acceptance in one inventory workflow.]]>
Copyrights © 2026
