Enterprise supply chain network optimization algorithm based on blockchain-distributed technology under the background of digital economy

Abstract

Demand forecasting is essential for streamlining supply chain operations in the digital economy and exceeding customer expectations. On the other hand, traditional forecasting techniques cannot frequently present real-time data and respond to dynamic changes in the supply chain network, leading to less-than-ideal decision-making and higher costs. This research aims to create a technique for optimizing the supply chain network based on blockchain-distributed technology (SCN-BT) to overcome these drawbacks and fully utilize the potential of the digital economy. The suggested framework uses the hybridized LSTM network and Grey Wolf Optimization (GWO) algorithm to examine demand forecasting in the supply chain network for inventory planning. The SCN-BT framework develops a safe and productive, enabling precise and flexible demand by combining blockchain with optimization techniques. A thorough case study utilized information collected from an enterprise supply chain that operates in the digital economy to show the efficiency of the suggested framework. Compared to conventional approaches, the results show considerable gains in demand forecasting precision, responsiveness of the supply chain, and cost-effectiveness. In the context of the digital economy, demand sensing and prediction enable firms to react to changes swiftly, shorten turnaround times, optimize inventory levels, and improve overall supply chain performance. The results highlight how blockchain technology has the potential to enhance collaboration, trust, and transparency inside intricate supply chain networks working in the digital economy. The experimental results show the proposed to achieve prediction rate of demand prediction rate of 128.93, demand forecasting accuracy ratio of 92.18%, optimum efficiency of 94.25%, RMSE rate of 1841.25, MAE rate of 260.74, and sMAPE rate of 0.1002 compared to other methods.

Keywords

Digital economy demand forecasting supply chain network optimization blockchain technology LSTM grey wolf optimization

1. Introduction

Demand forecasting in the enterprise digital economy refers to predicting inventory demand for products or services in a business environment driven by digital technologies and online interactions [1]. It involves analyzing historical time-series sales data, market trends, customer behavior, and other relevant factors to estimate future demand accurately. In the digital economy, businesses operate within a highly dynamic and interconnected environment characterized by rapid customer preferences and changes in market conditions. Therefore, demand forecasting becomes crucial for enterprises to optimize their supply chain operations, plan production and inventory levels, allocate resources efficiently, and meet customer expectations [2]. However, traditional forecasting techniques often struggle to provide information and adapt to the dynamic changes within the supply chain network [3]. These limitations hinder the decision-making process and result in suboptimal outcomes for enterprises in the digital economy.

To fully leverage the potential of the digital economy and overcome the limitations of traditional forecasting methods, there is a need for innovative approaches that integrate cutting-edge technologies [4]. These may include machine learning algorithms [5, 6], statistical models [7], and predictive analytics [8, 9]. By leveraging these tools, businesses can analyze vast amounts of data from various sources, such as online transactions, customer reviews, sales history, and market research, to uncover patterns and trends influencing demand [10]. One such technology is blockchain, which offers secure, transparent, decentralized data sharing and collaboration capabilities [11, 12]. By applying blockchain to supply chain operations, businesses can capture and share information among relevant stakeholders, leading to more accurate demand forecasting and improved decision-making [13].

The primary motivation of this research is to develop a technique for optimizing the supply chain network using blockchain-distributed technology for demand forecasting. The proposed framework combines the LSTM network [14] and GWO [15] algorithm to address the challenges associated with demand forecasting in the supply chain network. The reason for using LSTM is that it is ideal for predicting tasks due to its ability to capture temporal dependencies and patterns in time series data efficiently. GWO is used to streamline and enhance forecasting accuracy. The SCN-BT framework combines these methods to boost the accuracy and flexibility of demand forecasting, improving supply chain performance and raising end-user satisfaction. The proposed framework provides a secure and productive platform for demand forecasting by integrating blockchain with the supply chain network. Blockchain technology ensures data integrity, traceability, and confidentiality, fostering collaboration and transparency among supply chain partners [16, 17]. This collaborative environment enables precise demand forecasts and facilitates efficient decision-making processes. The results demonstrate significant improvements in demand forecasting precision and supply chain responsiveness, with minimal error rates compared to conventional approaches.

This research contributes to the study by showcasing how integrating blockchain technology and optimization techniques can enhance demand forecasting and improve supply chain operations in the digital economy. The findings highlight the potential of blockchain to enhance collaboration, trust, and transparency within complex supply chain networks. By adopting the proposed framework, businesses can make data-driven decisions, quickly adapt to dynamic market conditions, and optimize their supply chain operations, ultimately leading to enhanced customer satisfaction and a competitive advantage in the digital economy.

The main contribution of this research:

i)
The proposed SCN-BT framework increases the efficiency of the supply chain network through demand forecasting and improved manufacturing and inventory scheduling.
ii)
By combining the LSTM-GWO methodology in the SCN-BT platform with blockchain smart contract technology, businesses can improve the precision and adaptability of demand forecasting.
iii)
The SCN-BT system eliminates intermediaries, automates contractual processes, and enables confidential and transparent recording of demand projections to enhance the supply chain network.

The remainder of the script is divided into the following sections. Section 2 highlights related research in the supply chain network and blockchain technology. The SCN-BT system with the LSTM-GWO algorithms is discussed in Section 3. Section 4 analyses and discusses the results of the various models and algorithms. Finally, we conclude the script in Section 5.
2. Related study

Predicting demand relies on multi-layer LSTM networks, a class of recurrent neural networks well-known for detecting patterns and long-term dependencies in sequential data [18]. With performance metrics Root Mean Square Error (RMSE) and Symmetric Mean Absolute Percentage Error (sMAPE) of 2595.96 and 0.1085, respectively, the experimental findings show the proposed strategy to be superior in performance. Variational mode decomposition and a grey wolf optimizer-based LSTM are combined in a hybrid model termed VMD-GWO-LSTM [19]. In comparison to the other models, it shows improved accuracy with significant decreases in RMSE and Mean Absolute Percentage Error (MAPE) of 77.95% and 75.57%, respectively, and increases in R (correlation coefficient) and CE (coefficient of efficiency) of 81.67% and 397.93%. Recurrent Neural Networks/Long-Short Term Memory (RNN/LSTM) with a modified Adam optimizer to forecast the demand for spare parts [20]. In comparison to Simple Exponential Smoothing, Syntetos-Boylan Approximation, Croston, Teunter-Syntetos-Babai, and Modified SBA methods, the performance of the proposed method consistently outperforms them in terms of forecast inventory management metrics, with the MSE (mean square error) and ME (mean error) being 0.455 and 30.167, respectively.

Multi-layer Bi-LSTM and an LSTM model combined in the RNN-structured deep learning method to measure and predict future gas consumption [21]. Even if the prediction horizon is extended to 10 steps ahead, the Mackey Glass series-based forecasting maintains a precision level of about 99%. The data-driven strategy combines machine learning forecasting of demand and simulation-based optimization to synchronize supply and demand in omnichannel retail supply chains [22]. The scenario resulted in a 97% and 98% decline in orders, respectively, as those that went from scenario 1.2 to situations 2.2 and 3.2. LSTM networks and random forest (RF) are combined innovative forecasting approaches for multi-channel retail demand forecasting [23]. The recommended approach outperforms ARIMAX, artificial neural networks, multiple regression, LSTM networks, and RF with a 95% accuracy rate when tested on a multivariate real-world dataset from a multi-channel retailer. The study recognizes that alternative forecasting models would be needed when demand and resources from various channels are fully incorporated in an omnichannel retail setting.

A supply chain network model must be used to overcome communication problems, unshared information, and supply-demand mismatches in businesses [24]. The PSO algorithm and the network neural commodities demand forecast approach were coupled in the model. The MAPE and RMSE values of the MLP-LSTM are 117.342 and 2.334, respectively. The best performance was attained by the IPSO with the Autoregressive Integrated-Mixture Density Networks (AR-MDN) model combination, with a total assessment score of 67.41. A yield prediction model based on the Back Propagation Neural Network (BPNN) algorithm to address issues with aquaculture production and export scale prediction [25]. The BPNN is optimized by a series of algorithms built into the model. The input neuron’s dimensionality is decreased using the Johnson technique, which also identifies the neural network’s hidden layer. Finally, the model’s fit to the actual value, 0.96571, satisfied the demand.

According to earlier research, most studies have concentrated on applying machine learning techniques and predictive analytics to forecast demand; however, there are limits in terms of scalability, applicability to larger retail settings, and assessment of optimization methodologies. By utilizing the LSTM-GWO hybrid algorithm and stressing the usage of blockchain technology, this study fills in these gaps to increase the precision and adaptability of demand forecasting in the digital business economy.

3. System model

The proposed system model SCN-BT, shown in Fig. 1, is intended for use in a supply chain network optimization study on demand forecasting. The supply chain network is optimized, and precise demand forecasting is achieved using the hybridized LSTM and GWO algorithm. Additionally, the suggested methodology uses the advantages of blockchain technology to improve contractual transactions, trust, and transparency in the company’s supply chain processes.

Figure 1.

System structure of the SCN-BT framework.

3.1 System structure of the SCN-BT model

The Supply Chain Network Optimization using Blockchain Technology for the Demand Forecasting system structure model shown in Fig. 1 consists of several essential elements that work together to improve the effectiveness and precision of supply chain operations.

•
Customer behavior patterns are the key factor in demand forecasting and inventory planning. The consumer behavior analysis, which generates the historical time series data of sales, offers perspectives on customer behavior, views, and purchasing habits, enabling precise forecasts of future demand.
•
The supply chain network component represents the interconnected flow of goods, information, and resources among various supply chain participants. Suppliers, producers, distributors, sellers, and other intermediaries are included. The framework considers the intricate network structure and dynamics to optimize supply chain activities.
•
Demand forecasting is an essential component of LSTN-GWO. The LSTN-GWO uses past time series data to forecast demand in the future. The supply chain network’s many features are optimized via the optimization component. It also seeks to resolve problems like resource usage, inventory allocation, and production scheduling as best as possible.
•
The blockchain technology component provides the system’s foundation, offering a decentralized and unchangeable ledger for recording and verifying transactions, data transfers, and contracts. Secure data exchange, continuous product visibility, and traceability lower the possibility of fraud, forgery, and unauthorized alterations. The model produces the best results by merging the outputs of demand forecasts, optimization algorithms, and blockchain-enabled processes. These findings serve as a roadmap for decision-making, empowering supply chain managers to make well-informed decisions on production scheduling, inventory control, purchasing, and distribution.
•
Lastly, inventory planning is a crucial component that uses the demand value to choose the right inventory levels and restocking tactics.

The SCN-BT system aims to increase productivity, decrease delivery times, and boost customer satisfaction to optimize these variables. Besides, blockchain technology is utilized in the SCN-BT to centralize and uphold the conditions and limitations of agreements between supply chain stakeholders, and therefore, smart contracts become a crucial component of this system. These self-executing contracts do away with the necessity for intermediaries and improve participant confidence, security, and transparency. The SCN-BT system strives to save storage expenses, minimize stockouts, and sustain a good equilibrium between supply and demand by matching the accessible inventory with the projected demand. Ultimately, this system structure model uses smart contracts, demand forecasts, optimization methods, and blockchain technology to improve decision-making, increase supply chain efficiency, and promote collaboration among supply chain actors.
3.2 Blockchain distributed technology for SCN-BT system

The SCN-BT system uses smart contracts in conjunction with distributed blockchain technology. The system’s key component, smart contracts, is made possible by blockchain technology. They are necessary for centralizing and enforcing the terms and restrictions of agreements between different supply chain participants.

A decentralized, immutable database that records and validates transactions, data transfers, and contracts is provided by blockchain technology. On the blockchain, smart contracts automate and uphold the agreed-upon terms and activities within the supply chain network. Smart contracts are self-executing contracts with established rules and conditions. By eliminating the need for mediators and guaranteeing that the predetermined criteria are satisfied, these smart contracts promote confidence and transparency among the parties involved.

Smart contracts are utilized in the SCN-BT system to automate and streamline many supply chain activities. They enable vendors, manufacturers, distributors, and other participants to engage securely and transparently. Order fulfillment, payment processing, ensuring quality, and compliance verification are just a few of the processes that smart contracts may automate. The SCN-BT system uses smart contracts to increase efficiency, cut costs, and lower the supply chain network’s risk of fraud and legal issues.

Overall, the SCN-BT system’s smart contracts and blockchain technology adoption create a transparent and safe environment for supply chain activities. Removing intermediaries and streamlining procedures helps stakeholders work more successfully, make better decisions based on reliable data, and save money. Smart contracts and distributed blockchain technology increase supply chain management practices by enhancing network accuracy, efficiency, and transparency.

3.3 Prediction and optimization in SCN-BT system

The proposed SCN-BT framework outlines specific research goals to boost supply chain networks’ efficacy and efficiency through precise demand forecasting. The SCN-BT system strongly emphasizes using LSTM models, renowned for their aptitude for handling complex time series data and capturing temporal dependencies. The supply chain network is optimized using GWO. Grey wolves’ hunting style inspired GWO, which imitates how people naturally look for and choose the best answers. The suggested technique improves overall performance by utilizing GWO to minimize costs, shorten delays, and optimize inventory levels within the supply chain network. Figure 2 illustrates the function of the Hydrid LSTM-GWO algorithm for demand forecasting and supply chain network optimization in the proposed SCN-BT system.

The first step entails gathering historical time series data on demand, including information on client sales, market trends, and other pertinent elements. The SCN optimization and demand forecast methods use this data as input. The leveraging methodology preprocesses the collected data for optimum prediction. Data cleansing is not required because the dataset lacks no values. However, min-max normalization turns the dataset’s values into a consistent scale, allowing for fair comparisons between variables, enhancing convergence, and supplying robustness against potential outliers. These advantages improve supply chain optimization and demand forecasting processes regarding accuracy and dependability.

Figure 2.

Demand prediction and optimization techniques.

3.3.1 LSTM-GWO hybridization for SCN-BT system

The SCN-BT system uses a hybridized strategy that combines the GWO algorithm and the LSTM model. This hybridization aims to improve supply chain network optimization and demand forecasting performance. The effectiveness of LSTM models in handling time series data and capturing temporal dependencies is well established. The SCN-BT system uses LSTM models to examine past data and patterns, providing precise demand forecasting. The system can efficiently capture complicated linkages and variations in demand by utilizing the capabilities of LSTM models, increasing forecasting accuracy.

A. Demand Prediction

The LSTM model, a recurrent neural network renowned for its capacity to capture and evaluate sequential input, is fed the preprocessed data after being split into training and test sets. LSTM examines and learns from historical demand trends to forecast future demand accurately. The LSTM equation can be expressed as follows when used for supply chain optimization and demand prediction:

The LSTM cell generates the anticipated demand value, $y_{t}$ , at time step $t$ using the historical data $x_{t}$ , as input. The following are the LSTM equations:

The input gate $i_{t}$ , the forget gate $f_{t}$ , and the output gate $o_{t}$ , are the first three gates that the LSTM cell calculates, according to Eqs (1)–(3), respectively. These gates regulate the information flow inside the LSTM cell.

$\displaystyle i_{t}=\sigma({W_{i}\ast x_{t}+U_{i}\ast h_{\{{t-1}\}}+b_{i}})$ (1) $\displaystyle f_{t}=\sigma({W_{f}\ast x_{t}+U_{f}\ast h_{\{{t-1}\}}+b_{f}})$ (2) $\displaystyle o_{t}=\sigma({W_{o}\ast x_{t}+U_{o}\ast h_{\{{t-1}\}}+b_{o}})$ (3)

As shown in Eq. (4), the LSTM cell determines the cell state $c_{t}$ , which is in charge of remembering and losing information over time.

$\displaystyle{c}_{t}=f_{t}\ast c_{\{{{t}-1}\}}+i_{t}\ast\text{tanh}({{W}_{c}% \ast x_{t}+{U}_{c}\ast h_{\{{{t}-1}\}}+b_{c}})$ (4)

Finally, the LSTM cell is represented in Eq. (5) and generates an outcome value $h_{t}$ based on the cell state $c_{t}$ , and the output gate $o$ .

$\displaystyle{h}_{t}={o}_{t}\ast\text{tanh}({{c}_{t}})$ (5)

Here, $x_{t}$ represents the input at the time step $t$ , ${h}_{\{{{t}-1}\}}$ , represents the hidden state (output) of the previous time step $t-1$ , and ${h}_{\{{{t}-1}\}}$ , represents the hidden state of the current time step $t$ . $W_{i}$ , $W_{f}$ , $W_{o}$ , ${W}_{c}$ , $U_{i}$ , $U_{f}$ , $U_{o}$ , ${U}_{c}$ , $b_{i}$ , $b_{f}$ , $b_{o}$ and $b_{c}$ , are the weight matrices and bias vectors associated with the LSTM cell.

Hyperparameters Setting

It takes thorough testing and refining to determine the best values for demand forecasts utilizing historical time series sales data in a supply chain network. Here are some suggested values as a place to start:

Amount of LSTM units: Start with a respectable number of units, such as 50–100, and adapt according to the intricacy of the sales patterns and the quantity of data available.

ii)

Use a 3D tensor with the following dimensions: batch_size, timesteps, and input_features.

iii)

Activation methods: In LSTM networks, the input gate, forget gate and output gate is frequently activated using the sigmoid method.

iv)

Dropout regularization: Use this technique to stop overfitting. Consider using the Xavier initialization method for weight initialization.

Learning rate: Start with a low learning rate, like 0.001, then adjust it throughout training.

The demand forecast is produced by the LSTM algorithm using past data. This forecast shows the predicted demand levels for various products and periods. It aids supply chain managers in making defensible choices on resource allocation, inventory control, and production planning.

B. System Optimization

The SCN-BT system incorporates the GWO algorithm to enhance demand forecasting. A naturalistic optimization algorithm, GWO, imitates grey wolves’ social structure and hunting methods. It optimizes the parameters of the LSTM model and boosts their predicting capabilities. The SCN-BT system employs the GWO algorithm to improve the accuracy and efficiency of demand forecasting, resulting in enhanced decision-making and overall supply chain optimization. The hybridized approach in the SCN-BT system delivers a reliable and effective solution for demand forecasting and optimization in the supply chain network by combining the strengths of LSTM models with GWO algorithms. It uses LSTM models’ capacity to capture intricate patterns and temporal relationships, but the GWO approach sharpens the models to increase forecasting precision. As depicted in Fig. 2, demand forecasting, supply chain optimization, and monitoring are repeatedly carried out in an iterative process. This makes it possible for continuous improvement and adaptation to changes in the market environment, consumer tastes, and other factors that impact the supply chain.

Parameters Setting

Demand forecasting and supply chain optimization are two examples of optimization or machine learning algorithms where parameter setting is essential. The algorithm operates well and produces the expected results when correctly specifying parameters. The performance of the algorithm and its capacity to identify optimal or nearly optimal solutions are affected by the following optimization parameters in the LSTM-GWO.

The population size (N) is set to 50, representing a variety of possible designs or solutions for the supply chain network.

ii)

The Maximum number of iterations (Max_Iter) is fixed at 200, indicating a maximum number of iterations the algorithm will do to update the population of candidate solutions.

iii)

The exploration factor is adjusted to 1 (a $=$ 1) under the desired exploration-exploitation trade-off. It is looking through many solutions to find new and perhaps better ones. It facilitates the finding of global optima and prevents being trapped in local optima.

iv)

The threshold or halting condition is defined as A maximum number of iterations without noticeably improving the objective function. The objective is to find the best demand prediction that maximizes customer service while minimizing the stockout penalty. In light of this, the objective function can be expressed in Eq. (6).

$\displaystyle\textit{Obj{\_}func}=\alpha*\textit{CSL}+\beta*SP$ (6)

Where the variables $\alpha$ and $\beta$ are weighting factors that determine the relative importance of customer service level and stockout penalty; CSL is the customer service level that represents the percentage of demand that is capable of being met from available inventory, and the stockout penalty (SP), represents a penalty cost associated with stockouts, which is based on the lost sales or the impact on customer satisfaction.

The proposed SCN-BT system aims to increase the precision of demand forecasting and optimize the SCN to improve efficiency, reduce costs, inventory demand planning, and effectively meet customer demand by combining the capabilities of LSTM for demand prediction and the GWO algorithm for supply chain optimization. Algorithm 1 outlines the procedures for using a hybridized LSTM-GWO to estimate demand in a supply chain network.

Algorithm 1: Hybrid LSTM-GWO Algorithm
Input:Historical Time Series Data of Demand in the supply chain network.
Output:Predicted Demand Values
Step 1:Data Preprocessing: – Load and normalize the historical demand data.
Step 2:Split the preprocessed data into training and testing sets.
Step 3:Initialize the LSTM model with the desired architecture and hyperparameters.
Step 4:Initialize the parameters of GWO.
Step 5:Train the LSTM model using the training set.
Step 6:Initialize the best solution – Set the initial best solution as $\alpha$ position.
Step 7:Calculate the best fitness value using the objective function as in Eq. (6).
Step 8:Repeat until the maximum number of iterations is reached.
Step 9:Update the positions of the grey wolves and check the positions.
Step 10:Calculate the fitness value for each updated position using the objective function.
Step 11:Update the alpha, beta, and delta positions based on the fitness values.
Step 12:Update the best solution if a new global best is found.
Step 13:Repeat Steps 8–12 until the maximum number of iterations is reached.
Step 14:Apply the best solution (alpha position) to the LSTM forecast values.
Step 15:Display or export the final demand forecasts from the hybrid LSTM and GWO algorithms.
Step 16:End

The proposed system SCN-BT aims to optimize supply chain networks by providing a comprehensive approach to demand forecasting. It combines the LSTM model, GWO algorithm, and blockchain technology. LSTM models are employed for demand prediction, as they are known for their ability to handle complex time series data and capture temporal dependencies. GWO optimizes demand forecasting, enhancing accuracy, efficiency, and transparency throughout the supply chain. Blockchain technology ensures transparency and trust among supply chain stakeholders, allowing them to collaborate more effectively and make informed decisions. By implementing blockchain technology, the SCN-BT system enables secure and efficient communication, reduces costs, and advances supply chain management methods. By setting this concept into reality, supply chain stakeholders may collaborate better, make better decisions, and cut costs, thereby advancing supply chain management methods.

4. Results and discussion

This experimental investigation aims to evaluate the performance and efficacy of the suggested SCN-BT system structure model to increase decision-making effectiveness, boost interaction among supply chain actors, and improve supply chain efficiency. Through extensive experiments and performance evaluations, The effects of blockchain technology, smart contracts, demand forecasting, optimization algorithms, and inventory planning are examined by supply chain metrics like predictive accuracy, RMSE, sMAPE, demand prediction, and operational efficiency.

4.1 Dataset

The historical time series dataset of sales [26] is utilized in the proposed SCN-BT system for demand forecasting in a supply chain network. The dataset includes monthly reports on annual sales from different retail points, with an average of 20000 monthly records. All the datasets are gathered, the attributes are standardized, and they are split into two distinct training and test datasets for the predictive analysis by the SCN-BT framework.

4.2 Experimental analysis of SCN-BT framework

The SCN-BT system’s performance is examined with month-wise and average predictive accuracy based on actual sales as shown in Fig. 3. The performance of the SCN-BT methodology is compared with existing predictive models ARIMA, AR-MDN, KNN, and SVM models, about RMSE, Mean Absolute Error (MAE), and sMAPE and the results are tabulated in Table 1. The average demand prediction of the SCN-BT system is comparatively studied with other predictive models as displayed in Fig. 4.

Table 1
Evaluation metrics of SCN-BT and other predictive models

Methodology	RMSE	MAE	sMAPE
ARIMA	2568.32	357.89	0.1613
AR-MDN	3526.89	305.26	0.1813
KNN	3218.15	364.12	0.1416
SVM	2165.42	295.16	0.1347
SCN-BT	1841.25	260.74	0.1002

Figure 3.

Month-wise and average predictive accuracy by SCN-BT system.

Figure 4.

Average demand prediction by SCN-BT and other predictive models.

Based on the abovementioned findings, SCN-BT can be used as a demand prediction model. Four optimization techniques are then used in various combinations, and the model is thoroughly assessed using average values. The information is annual sales data to analyze multiple combinations. Regarding average demand forecast and optimization efficiency, the results of various hybrid models and the proposed framework are shown in Figs 5 and 6, respectively.

Figure 5.

Evaluation of SCN-BT system with other hybrid models in terms of accuracy rate.

Demand forecasting is a crucial responsibility in supply chain management, and numerous models have been created to address this issue. For demand forecasting, one well-liked method is to employ LSTM-GWO in the SCN-BT system, which helps anticipate future demand because it can capture sequential trends and correlations in time-series data. As seen in Fig. 3, SCN-BT has demonstrated promising results of 128.93 in average demand forecast of all months, compared to other approaches like ARIMA, AR-MDN, KNN, and SVM, regression-based approach, due to its capacity for handling long-term dependencies and nonlinear patterns. It offers insights into the suggested system’s effectiveness and can capture complicated correlations between previous and future demand measurements. The performance of the SCN-BT model is evaluated using the RMSE, MAE, and sMAPE metrics, as shown in Fig. 4, and its results are contrasted with those of other prediction models with 1841.25, 260.74, and 0.1002, respectively, the demand prediction rate of the SCN-BT beats other RMSE, MAE, and sMAPE models. Regardless of the direction of the errors, the MAE metric analyzes the average absolute difference between the expected and actual demand levels to determine the average forecasting error. The RMSE metric measures the standard deviation of the forecasting mistakes. Regarding the proposed and other comparison models, the sMAPE statistic evaluates the average percentage difference symmetrically, considering over- and under-forecasting errors. Predictive accuracy, which quantifies the percentage of accurately anticipated demand values within a specific tolerance level or error margin, is a metric that evaluates the accuracy of the demand forecasts produced by the LSTM-GWO algorithm in the SCN-BT system.

Figure 6.

Evaluation of SCN-BT system with other hybrid models in terms of efficiency.

The SCN-BT system achieved the maximum demand forecast accuracy of 92.18. It proved the efficiency of optimization as 94.25 in the supply chain network, according to the data in Figs 5 and 6. The high demand prediction accuracy is a big plus for the SCN-BT system since precise demand forecasts are essential for improving inventory control, production scheduling, and all other supply chain activities. The proposed system’s capacity to produce accurate and consistent demand projections helps to reduce inventory costs, prevent stockouts or instances where there are too many supplies, and raise customer satisfaction.

The smart contract incorporated in the SCN-BT system also increases the efficiency of the optimization process. Blockchain’s decentralized structure eliminates the need for intermediaries, lowers transaction costs, and permits real-time collaboration among network users. The LSTM-GWO algorithm’s enhanced execution speed enables quicker decision-making and reaction to changing market conditions. The automated execution and implementation of contractual obligations are advantages of integrating smart contract technology. Smart contracts can allow smooth interactions and transactions based on the anticipated demand values. For instance, they can begin manufacturing orders, start inventory replenishment, or change pricing plans depending on expected demand, enhancing the effectiveness and responsiveness of the supply chain. As shown in Figs 5 and 6, the SCN-BT framework shows that smart contracts are helping the supply chain network perform better by enhancing efficiency and predicting demand more precisely. The supply chain management and demand forecasting processes could be revolutionized by this combination, opening the door to more dependable and efficient operations.

The proposed system SCN-BT optimized the supply chain efficiently and provided the most accurate demand forecasts. This is significant since accurate demand projections enhance supply chain activities like inventory management, manufacturing scheduling, etc. By removing intermediaries, decreasing transaction costs, and enabling cooperation among network users, the LSTM-GWO and smart contract built into the system improve the effectiveness of the optimization process. By facilitating interactions and transactions predicated on estimated demand values, the SCN-BT improves supply chain responsiveness and efficiency.

5. Conclusion

The SCN-BT system in this study was primarily concerned with demand forecasting, which predicts future demand using past time series data from sales. The suggested system uses the LSTM-GWO to produce the best results in the supply chain network. The system uses smart contracts to automate and enforce the implementation of agreements among supply chain actors. Combining the LSTM-GWO algorithm and blockchain’s smart contract technology has produced outstanding outcomes for demand forecasting and supply chain network optimization. By using smart contracts, the algorithm’s execution is guaranteed to be accurate and correct, fostering more stakeholder accountability. Integrating smart contract technology has also increased the effectiveness of the optimization process. Blockchain’s decentralized structure eliminates intermediaries and lowers transaction costs, enabling real-time cooperation and quicker decision-making. Timely responses to changing market conditions are made possible by this increased efficiency. The method strives to reduce carrying costs, avoid stockouts, and maintain an appropriate ratio between supply and demand by matching the available inventory with the projected demand. In this study, a supply chain network is used to assess the efficiency and performance of the LSTM-GWO algorithm with smart contracts. Additional testing and tweaking may be necessary to ensure the system’s applicability to various sectors, supply chain arrangements, and market dynamics. Furthermore, Smart contracts within the SCN-BT system could be further investigated to improve their capabilities and use. This may require researching more complex automation and data transmission options, such as conditional execution, real-time tracking, or connection to IoT devices.

References

Ulas

. Digital transformation process and SMEs. Procedia Computer Science. 2019; 158: 662-671.

Abbasimehr

Shabani

. A new framework for predicting customer behaviour in terms of RFM by considering the temporal aspect based on time series techniques. Journal of Ambient Intelligence and Humanized Computing. 2021; 12(1): 515-531.

Giusti

Manerba

Bruno

Tadei

. Synchromodal logistics: An overview of critical success factors, enabling technologies, and open research issues. Transportation Research Part E: Logistics and Transportation Review. 2019; 129: 92-110.

Elhusseiny

Crispim

. SMEs, Barriers and Opportunities on adopting Industry 4.0: A review. Procedia Computer Science. 2022; 196: 864-871.

Feizabadi

. Machine learning demand forecasting and supply chain performance. International Journal of Logistics Research and Applications. 2022; 25(2): 119-142.

Khan

Saqib

Alyas

Rehman

Saeed

Zeb

Mohamed

. Effective demand forecasting model using business intelligence empowered with machine learning. IEEE Access. 2020; 8: 116013-116023.

Seyedan

Mafakheri

. Predictive big data analytics for supply chain demand forecasting: Methods, applications, and research opportunities. Journal of Big Data. 2020; 7(1): 1-22.

Idrees

Alam

Agarwal

. A prediction approach for stock market volatility based on time series data. IEEE Access. 2019; 7: 17287-17298.

Punia

Singh

Madaan

. From predictive to prescriptive analytics: A data-driven multi-item newsvendor model. Decision Support Systems. 2020; 136: 113340.

10.

Hasan

Kamal

Daowd

Eldabi

Koliousis

Papadopoulos

. Critical analysis of the impact of big data analytics on supply chain operations. Production Planning & Control. 2022; 1-25.

11.

Wan

Huang

Holtskog

. Blockchain-enabled information sharing within a supply chain: A systematic literature review. IEEE Access. 2020; 8: 49645-49656.

12.

Omar

Jayaraman

Debe

Hasan

Salah

Omar

. Supply chain inventory sharing using ethereum blockchain and smart contracts. IEEE Access. 2021; 10: 2345-2356.

13.

Yuan

Qiu

Chang

Lam

. Analysis of coordination mechanism of supply chain management information system from the perspective of block chain. Information Systems and e-Business Management. 2020; 18: 681-703.

14.

Falatouri

Darbanian

Brandtner

Udokwu

. Predictive analytics for demand forecasting – a comparison of sarima and lstm in retail scm. Procedia Computer Science. 2022; 200: 993-1003.

15.

Mei

Zeng

. A novel fractional time delayed grey model with Grey Wolf Optimizer and its applications in forecasting the natural gas and coal consumption in Chongqing China. Energy. 2019; 178: 487-507.

16.

Zhu

Kouhizadeh

. Blockchain technology, supply chain information, and strategic product deletion management. IEEE Engineering Management Review. 2019; 47(1): 36-44.

17.

Kopyto

Lechler

von der Gracht

Hartmann

. Potentials of blockchain technology in supply chain management: Long-term judgments of an international expert panel. Technological Forecasting and Social Change. 2020; 161: 120330.

18.

Abbasimehr

Shabani

Yousefi

. An optimized model using LSTM network for demand forecasting. Computers & Industrial Engineering. 2020; 143: 106435.

19.

Sun

Liu

Wang

Huang

. Monthly runoff forecasting using variational mode decomposition coupled with gray wolf optimizer-based long short-term memory neural networks. Water Resources Management. 2022; 36(6): 2095-2115.

20.

Chandriah

Naraganahalli

. RNN/LSTM with modified Adam optimizer in deep learning approach for automobile spare parts demand forecasting. Multimedia Tools and Applications. 2021; 80(17): 26145-26159.

21.

Zio

Zhang

. A hybrid hourly natural gas demand forecasting method based on the integration of wavelet transform and enhanced Deep-RNN model. Energy. 2019; 178: 585-597.

22.

Pereira

Frazzon

. A data-driven approach to adaptive synchronization of demand and supply in omni-channel retail supply chains. International Journal of Information Management. 2021; 57: 102165.

23.

Punia

Nikolopoulos

Singh

Madaan

Litsiou

. Deep learning with long short-term memory networks and random forests for demand forecasting in multi-channel retail. International Journal of Production Research. 2020; 58(16): 4964-4979.

24.

Gao

. The analysis of commodity demand predication in supply chain network based on particle swarm optimization algorithm. Journal of Computational and Applied Mathematics. 2022; 400: 113760.

25.

Zhang

. Application of improved BP neural network based on e-commerce supply chain network data in the forecast of aquatic product export volume. Cognitive Systems Research. 2019; 57: 228-235.

26.

https://www.kaggle.com/datasets/knightbearr/sales-product-data.

Enterprise supply chain network optimization algorithm based on blockchain-distributed technology under the background of digital economy

Abstract

Keywords

1. Introduction

3. System model

3.3 Prediction and optimization in SCN-BT system

A. Demand Prediction

Hyperparameters Setting

B. System Optimization

Parameters Setting

4.1 Dataset

4.2 Experimental analysis of SCN-BT framework

Table 1 Evaluation metrics of SCN-BT and other predictive models

References

Table 1
Evaluation metrics of SCN-BT and other predictive models