Acceptance Rate Prediction of Blockchain in Automotive Supply Chain Management With a Bayesian Distributive Gradient Integrated BiLSTM

• Published in: IEEE access 2024, Vol.12, p.171777-171789
• Main Authors: Albuloushi, Ahlaam, Alzubi, Ahmad, Oz, Tolga
• Format: Article
• Language: English
• Subjects: Acceptance; Adaptation models; Bayes methods; Bayesian analysis; Bayesian optimization; BILSTM; Blockchain; Blockchains; Computational modeling; Efficiency; gradient boosting machine; Long short term memory; Mathematical models; Mean square errors; Optimization; Predictions; Predictive models; Root-mean-square errors; supply chain management; Supply chains; Training
• ISSN: 2169-3536; 2169-3536
• DOI: 10.1109/ACCESS.2024.3500793

Maintaining the integrity of products and processes in a multi-party supply chain is a complex task. Consequently, companies are employing data-driven technology to gain the capacity to overcome these challenges. Numerous existing solutions suffer from various protocol regulations across multiple distributions and processes, data fragmentation, and unreliable provenance. Blockchain has become a dominant technology among other solutions because it offers low-cost IT solutions with safe traceability and control, immutability, and data confidentiality. Blockchain is having a significant impact in many industries, but supply chains are facing numerous obstacles to its general implementation. By using a Bayesian distributive gradient integrated bidirectional long short-term memory (BDG-BiLSTM), the research aimed to improve the acceptance rate prediction of blockchain. The research's use of data normalization enhances the benefit of compiling blockchain data into a defined range. By preventing the model from leaning toward a broad class, the Gradient Boosting Machine (GBM) and Bidirectional long short-term memory (BiLSTM) incorporation enhances a forecasting model's efficiency even for large datasets. The BDG-BiLSTM model improves acceptance rate prediction by offering a thorough comprehension of the variety of variables included in blockchain information. The BDG-BiLSTM's integration of Bayesian optimization has the advantage of finding the best solution, which raises the predictive efficiency and is evaluated in terms of 94.76% accuracy, 94.63% precision, 94.94% recall, 2.75 Mean Square Error (MSE), and 1.66 Root Mean Square Error (RMSE) when its efficiency is assessed using the TP 90.