Virtual commodity supply chain scheduling system and method based on artificial intelligence

Abstract

The invention provides a virtual commodity supply chain scheduling system and method based on artificial intelligence, relates to the technical field of computers, is applied to a commodity supply system, and comprises a plurality of supply libraries; the scheduling system comprises a data acquisition module used for acquiring user behavior data, external supply data and inventory state data of each supply library; the intelligent decision module is used for obtaining user demand prediction data according to the user behavior data based on a preset prediction model; determining a corresponding risk level according to each piece of inventory state data, and obtaining a corresponding weight adjustment coefficient based on each risk level; and the decision execution module is used for carrying out inventory supply scheduling on each supply library according to the user demand prediction data and each weight adjustment coefficient. According to the invention, various data are widely collected through the data acquisition module, and deep analysis and processing are carried out through the intelligent decision-making module, so that accurate prediction of user demands and automatic intelligent assessment of inventory risks are realized, and the uncertainty of manual intervention is reduced.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an artificial intelligence-based virtual commodity supply chain scheduling system and method. Background Art

[0002] With the rapid development of digital technology, the scale and types of virtual goods transactions in the market have shown explosive growth, covering multiple fields such as digital content (such as e-books, video memberships), online services (such as cloud storage services), and virtual gifts (such as virtual props on social media).

[0003] However, existing inventory scheduling systems and technologies face a series of significant challenges in modern virtual goods supply chain management. First, user demand is highly uncertain and volatile. For example, demand for video memberships can fluctuate significantly due to the release of popular TV shows, competitor promotions, or unexpected events (such as the premature withdrawal of a popular TV series). Existing systems often rely solely on historical sales data for linear forecasting, resulting in poor demand forecasting accuracy. Second, traditional inventory decision support systems primarily rely on simple rules and statistical analysis, lacking the ability to mine complex data patterns. They are unable to provide decision makers with comprehensive, in-depth analysis and multi-dimensional decision-making recommendations, making them unable to cope with complex market environments. Consequently, existing systems often rely on manual or simple rules for inventory scheduling, resulting in low efficiency, slow response, insufficient risk assessment, and simplistic decision-making. These systems are unable to flexibly adapt to changes in market and user demand. Traditional methods adjust inventory based solely on historical sales, ignoring user behavior dynamics and external supply conditions, resulting in low supply and demand matching accuracy. Furthermore, the lack of intelligent interaction between modules further limits the system's flexibility and efficiency. Summary of the Invention

[0004] The problem solved by the present invention is one or more of the above-mentioned related technical problems.

[0005] To solve the above problems, the present invention provides a virtual commodity supply chain scheduling system and method based on artificial intelligence.

[0006] In a first aspect, the present invention provides an artificial intelligence-based virtual commodity supply chain scheduling system, which is applied to a commodity supply system, wherein the commodity supply system includes multiple supply warehouses; the virtual commodity supply chain scheduling system includes a data acquisition module, an intelligent decision module, and a decision execution module; The data collection module is used to collect user behavior data, external supply data and inventory status data of each supply warehouse; The intelligent decision-making module is configured to obtain user demand forecast data based on the user behavior data based on a preset forecast model; determine a corresponding risk level based on each inventory status data, and obtain a corresponding weight adjustment coefficient based on each risk level; The decision execution module is used to perform inventory supply scheduling for each supply depot based on the user demand forecast data and each weight adjustment coefficient; Among them, the preset prediction model is constructed based on multiple LSTM networks and attention mechanisms.

[0007] Optionally, the user behavior data includes category data and time data of user benefits received; the user demand prediction data obtained based on the user behavior data based on the preset prediction model includes: Integrating the category data and the time data to obtain a time series data set; The time series data set is input into the preset prediction model to obtain the user demand prediction data.

[0008] Optionally, the inventory status data includes inventory quantity and replenishment cycle; and determining the corresponding risk level according to each inventory status data includes: Construct a multivariate linear regression model based on historical inventory data and corresponding historical replenishment cycles; The inventory quantity and the replenishment cycle are calculated based on the multivariate linear regression model to obtain the corresponding risk level.

[0009] Optionally, the intelligent decision-making module is further used to: Based on a preset similarity model, the external supply data is matched with the data categories in the supply library to obtain a corresponding equivalent mapping table.

[0010] Optionally, the externally supplied data includes competitor category data, and the preset similarity model includes a BERTT model; the matching of the externally supplied data with the data categories in the supply library based on the preset similarity model to obtain a corresponding equivalence mapping table includes: Performing vector conversion on the competitor product category data to obtain first semantic vector data; Performing vector conversion on each of the data categories to obtain corresponding second semantic vector data, and performing similarity calculation on the first semantic vector data and each of the second semantic vector data based on a similarity calculation method to obtain corresponding similarity data; A temporary data category corresponding to the competitor category data is determined based on all the similarity data, and a corresponding equivalence mapping table is obtained based on all the temporary data categories.

[0011] Optionally, the inventory status data includes inventory data; and the decision execution module is specifically configured to: Obtaining initial weight data of each of the supply warehouses, and adjusting the corresponding initial weight data based on the weight adjustment coefficient to obtain corresponding target weight data; Obtaining demand data corresponding to each supply warehouse according to each target weight data and the user demand forecast data; The demand data is compared with the corresponding inventory data to obtain corresponding comparison results, and inventory scheduling is performed on each supply warehouse according to each comparison result.

[0012] Optionally, after comparing the demand data with the corresponding inventory data to obtain corresponding comparison results, and performing inventory scheduling for each supply warehouse according to each comparison result, the method further includes: When the inventory data of each category of data in any of the supply warehouses does not meet the demand, the supply warehouses are screened according to the equivalent mapping table to obtain corresponding substitute scheduling strategies.

[0013] Optionally, the artificial intelligence-based virtual commodity supply chain scheduling system further includes an abnormality feedback module; The abnormal feedback module is used to trigger an alternative supply process when the artificial intelligence-based virtual commodity supply chain scheduling system fails to send a replenishment order to the supply warehouse; The abnormal feedback module is also used to automatically check the weight adjustment coefficient and verify the feedback result of the instruction issued by the decision execution module.

[0014] Optionally, the artificial intelligence-based virtual commodity supply chain scheduling system further includes a collaborative network module, and the collaborative network module is used to enable the artificial intelligence-based virtual commodity supply chain scheduling system to interact collaboratively with an external entity system.

[0015] In a second aspect, the present invention provides a virtual commodity supply chain scheduling method, which is applied to the aforementioned artificial intelligence-based virtual commodity supply chain scheduling system. The scheduling method includes: Collecting user behavior data, external supply data, and inventory status data of each of the supply warehouses; Based on a preset prediction model, obtain user demand prediction data according to the user behavior data; Determining a corresponding risk level according to each of the inventory status data, and obtaining a corresponding weight adjustment coefficient based on each of the risk levels; Inventory supply scheduling is performed on each supply warehouse according to the user demand forecast data and each weight adjustment coefficient.

[0016] The beneficial effects of the artificial intelligence-based virtual commodity supply chain scheduling system and method of the present invention are: This AI-based virtual commodity supply chain scheduling system is applied to commodity supply systems involving multiple supply bases. The entire system consists of three key modules: a data acquisition module, an intelligent decision-making module, and a decision execution module. These three modules work together to achieve efficient scheduling of the virtual commodity supply chain.

[0017] The data collection module is responsible for collecting three key types of data: user behavior data, external supply data, and inventory status data for each supply depot. User behavior data covers the specific operations users perform during the benefit redemption process, such as the type of benefit received (e.g., video membership, coupons, etc.) and the time of redemption. This data can be used to understand user preferences and demand patterns. External supply data primarily includes data on competing product categories, which is used to understand the types of benefits offered by other suppliers in the market and provide an external reference for supply chain scheduling. Inventory status data includes information such as the inventory quantity and replenishment cycle of benefits in each supply depot, allowing for real-time understanding of each depot's inventory status and supply capacity.

[0018] As the decision-making core of the system, the intelligent decision-making module mainly performs the following tasks: Based on the preset prediction model and the collected user behavior data, it predicts the user's future needs and generates user demand prediction data. This prediction model can comprehensively consider multi-dimensional information such as the user's historical collection behavior and time factors to provide a demand basis for subsequent inventory scheduling. According to the inventory status data of each supply warehouse, the risk level corresponding to each supply warehouse is determined. The risk level assessment comprehensively considers factors such as whether the inventory quantity is sufficient and whether the replenishment cycle is timely. Based on the assessed risk level, the corresponding weight adjustment coefficient is further calculated to adjust the weight distribution of each supply warehouse in inventory scheduling, so as to make the inventory distribution more reasonable and reduce the risk of insufficient or excessive inventory.

[0019] The decision execution module is responsible for scheduling inventory supply for each supply depot based on the user demand forecast data and weight adjustment coefficients generated by the intelligent decision module. Specifically, it allocates predicted user demand to each supply depot according to the adjusted weights, ensuring that each depot's inventory can effectively meet user demand while avoiding inventory backlogs or stockouts, thereby achieving a dynamic balance in the supply chain.

[0020] In summary, the present invention widely collects various types of data through the data acquisition module, and uses the intelligent decision-making module to conduct in-depth analysis and processing, thereby achieving accurate prediction of user needs and intelligent assessment of inventory risks, making supply chain decisions more scientific and intelligent, and reducing the uncertainty of manual intervention and experience-based decisions. At the same time, the weight of each supply warehouse is dynamically adjusted according to the risk level to optimize inventory allocation and avoid inventory backlogs and shortages. Accurate demand forecasting and effective inventory scheduling reduce inventory management costs, improve operational efficiency, and optimize the supply chain cost structure. In addition, accurate demand forecasting ensures that user needs are met in a timely manner, and flexible scheduling strategies reduce delays in rights collection due to inventory problems, thereby improving user satisfaction and loyalty. In short, this artificial intelligence-based virtual commodity supply chain scheduling system fully utilizes the value of data through close collaboration of various modules, realizes intelligent, flexible and efficient management of the supply chain, and is of great significance to improving the operational efficiency and market competitiveness of virtual commodity supply companies. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a structural diagram of a virtual commodity supply chain scheduling system based on artificial intelligence according to an embodiment of the present invention; Figure 2 This is a second structural diagram of an artificial intelligence-based virtual commodity supply chain scheduling system according to an embodiment of the present invention; Figure 3 The figure is a flow chart of a virtual commodity supply chain scheduling method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0022] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, specific embodiments of the present invention are described in detail below with reference to the accompanying drawings. Although certain embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as being limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0023] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0024] The term "including" and its variations used in this document are open inclusions, that is, "including but not limited to"; the term "based on" means "based at least in part on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one other embodiment"; the term "some embodiments" means "at least some embodiments"; the term "optionally" means "optional embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc. mentioned in the present invention are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0025] It should be noted that the modifications of "one" and "multiple" mentioned in the present invention are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".

[0026] The names of the messages or information exchanged between multiple devices in the embodiments of the present invention are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0027] In response to the problems existing in the above-mentioned related technologies, an embodiment of the present invention provides a virtual commodity supply chain scheduling system and method based on artificial intelligence.

[0028] like Figure 1 As shown, an embodiment of the present invention provides an artificial intelligence-based virtual commodity supply chain scheduling system, which is applied to a commodity supply system, and the commodity supply system includes multiple supply warehouses; the virtual commodity supply chain scheduling system includes a data acquisition module, an intelligent decision-making module and a decision execution module.

[0029] The data collection module is used to collect user behavior data, external supply data and inventory status data of each supply warehouse.

[0030] Specifically, user behavior data collection may include: Benefit claiming behavior: recording the type of benefits claimed by the user (e.g., video membership, coupons), claiming timestamp, claiming quantity, and claiming channel (app / PC / third-party platform). User profile data: preference tags generated based on historical claiming behavior (e.g., "frequent video membership claimers," "groups sensitive to lifestyle services").

[0031] Externally supplied data collection may include: Competitive product category data: such as the names, prices, display locations, and promotional activities (e.g., "iQiyi Student Annual Pass 50% Off") of the top 10 benefits on competing platforms. Market trend data: search volume for popular benefit categories and holiday promotion trends (e.g., surges in benefit sales during "618" and "Double 11").

[0032] Inventory status data collection may include: Real-time inventory balance: the current inventory quantity of each supplier's equity (e.g., "Tencent Video Monthly Pass: 800 units"). Replenishment cycle data: the supplier's most recent replenishment completion time and the average replenishment cycle (e.g., 48 hours).

[0033] The data collection module ensures comprehensive and real-time data, avoiding decision-making biases associated with a single data source. Secondly, it provides high-quality data input for subsequent modules, directly supporting the accuracy and flexibility of supply chain scheduling. Finally, by monitoring inventory status in real time, the supply chain's risk response capabilities are enhanced.

[0034] The intelligent decision-making module is configured to obtain user demand forecast data based on the user behavior data based on a preset forecast model; determine a corresponding risk level based on each inventory status data, and obtain a corresponding weight adjustment coefficient based on each risk level. The preset forecast model is constructed based on multiple LSTM networks and an attention mechanism.

[0035] Specifically, the user demand forecasting process takes user behavior data (e.g., redemption time, benefit type, quantity, and preference tags) as input. This behavior data can be converted into time series features (e.g., periodic patterns and user profiles). A machine learning model is trained on historical user behavior data to generate forecasts of virtual product demand for different future periods (e.g., hourly, daily, and weekly). This is known as user demand forecast data (e.g., demand for 3,000 "Tencent Video Memberships" in the next 24 hours).

[0036] The inventory risk assessment and weighting process uses inventory status data (inventory balance, replenishment cycle) to calculate the risk level (such as high / medium / low risk). A weighting adjustment factor is then generated based on the risk level (high risk reduces the weight, low risk maintains or increases the weight).

[0037] This AI-based virtual commodity supply chain scheduling system directly guides inventory allocation through forecast data, uses weight coefficients to quantify inventory risks in real time, and achieves optimal resource allocation, thereby ensuring the timeliness and stability of equity supply, and ultimately significantly enhancing user satisfaction.

[0038] The intelligent decision-making module offers several benefits: First, by deeply integrating multi-source data, it enables accurate forecasting of user demand and intelligent assessment of inventory risk, enhancing the scientific and intelligent nature of supply chain decision-making. Second, a dynamic weight adjustment mechanism ensures rational inventory allocation, reduces the uncertainty of manual intervention and empirical decision-making, and optimizes the supply chain's cost structure.

[0039] The decision execution module is used to perform inventory supply scheduling for each supply warehouse based on the user demand forecast data and each weight adjustment coefficient.

[0040] Specifically, the decision execution module is responsible for converting the user demand forecast data and weight adjustment coefficients generated by the intelligent decision module into specific inventory scheduling operations. Its workflow is as follows: The decision execution module is responsible for translating the output of the intelligent decision module into actual inventory scheduling operations. Based on user demand forecast data and the weight adjustment coefficients for each supplier, inventory supply scheduling is performed for each supplier. Specifically, the module first obtains the initial weight data for each supplier and calculates the target weight data based on the weight adjustment coefficients. Subsequently, based on the user demand forecast data, the module calculates the corresponding demand allocation for each supplier and compares these allocations with the current inventory data. Based on the comparison results, the module generates corresponding inventory scheduling instructions, such as replenishment or reduction plans, and sends them to each supplier through the system interface for execution.

[0041] The decision-making execution module optimizes inventory management processes and reduces inventory costs through precise demand allocation and inventory comparison. Secondly, the automated scheduling mechanism improves operational efficiency, accelerates supply chain responsiveness, and ensures timely and stable equity supply. Finally, this module effectively improves customer satisfaction and strengthens the company's competitive advantage in the market.

[0042] In this embodiment, this AI-based virtual commodity supply chain scheduling system is applied to a commodity supply system that includes multiple supply depots. The entire system consists of three key modules: a data acquisition module, an intelligent decision-making module, and a decision execution module. These three modules work together to achieve efficient scheduling of the virtual commodity supply chain.

[0043] The data collection module is responsible for collecting three key types of data: user behavior data, external supply data, and inventory status data for each supply depot. User behavior data covers the specific operations users perform during the benefit redemption process, such as the type of benefit received (e.g., video membership, coupons, etc.) and the time of redemption. This data can be used to understand user preferences and demand patterns. External supply data primarily includes data on competing product categories, which is used to understand the types of benefits offered by other suppliers in the market and provide an external reference for supply chain scheduling. Inventory status data includes information such as the inventory quantity and replenishment cycle of benefits in each supply depot, allowing for real-time understanding of each depot's inventory status and supply capacity.

[0044] As the decision-making core of the system, the intelligent decision-making module mainly performs the following tasks: Based on the preset prediction model and the collected user behavior data, it predicts the user's future needs and generates user demand prediction data. This prediction model can comprehensively consider multi-dimensional information such as the user's historical collection behavior and time factors to provide a demand basis for subsequent inventory scheduling. According to the inventory status data of each supply warehouse, the risk level corresponding to each supply warehouse is determined. The risk level assessment comprehensively considers factors such as whether the inventory quantity is sufficient and whether the replenishment cycle is timely. Based on the assessed risk level, the corresponding weight adjustment coefficient is further calculated to adjust the weight distribution of each supply warehouse in inventory scheduling, so as to make the inventory distribution more reasonable and reduce the risk of insufficient or excessive inventory.

[0045] The decision execution module is responsible for scheduling inventory supply for each supply depot based on the user demand forecast data and weight adjustment coefficients generated by the intelligent decision module. Specifically, it allocates predicted user demand to each supply depot according to the adjusted weights, ensuring that each depot's inventory can effectively meet user demand while avoiding inventory backlogs or stockouts, thereby achieving a dynamic balance in the supply chain.

[0046] In summary, this embodiment widely collects various types of data through the data acquisition module, and uses the intelligent decision-making module to conduct in-depth analysis and processing, thereby achieving accurate prediction of user demand and intelligent assessment of inventory risks, making supply chain decisions more scientific and intelligent, and reducing the uncertainty of manual intervention and experience-based decisions. At the same time, the weight of each supply warehouse is dynamically adjusted according to the risk level to optimize inventory allocation and avoid inventory backlogs and shortages. Accurate demand forecasting and effective inventory scheduling reduce inventory management costs, improve operational efficiency, and optimize the supply chain cost structure. In addition, accurate demand forecasting ensures that user needs are met in a timely manner, and flexible scheduling strategies reduce delays in rights collection due to inventory problems, thereby improving user satisfaction and loyalty. In short, this artificial intelligence-based virtual commodity supply chain scheduling system fully utilizes the value of data through close collaboration of various modules, realizes intelligent, flexible and efficient management of the supply chain, and is of great significance to improving the operational efficiency and market competitiveness of virtual commodity supply companies.

[0047] Optionally, the user behavior data includes category data and time data of user benefits received; the user demand prediction data obtained based on the user behavior data based on the preset prediction model includes: Integrating the category data and the time data to obtain a time series data set; The time series data set is input into the preset prediction model to obtain the user demand prediction data.

[0048] Specifically, the process of integrating categorical data and time data into a time series dataset is as follows: Combine the categorical data and time data of user benefit redemption to form a time series dataset. For example, if user A redeems a "Video Membership Monthly Card" at 10:00 AM on July 1, 2025, the record will be (2025-07-01 10:00, Video Membership Monthly Card).

[0049] The integrated data is cleaned and normalized to remove noise and outliers, making the data suitable for model input.

[0050] Extract key features from time series, such as the periodicity of pickup times (hours, days, weeks) and category frequency. Build a prediction model using multiple LSTM networks and an attention mechanism. The LSTM network captures long-term dependencies in the time series, while the attention mechanism focuses on key time nodes.

[0051] The model training process uses historical data to train the model, optimizing the algorithm to adjust model parameters and minimize the error between the predicted and actual values. Time series data is fed into the trained model, which outputs a forecast of the number of benefits redeemed for each future time period. For example, it is predicted that the number of "Video Membership Monthly Cards" redeemed between 20:00 and 22:00 on July 2, 2025 will be 500.

[0052] The structure of the preset prediction model involves building a neural network consisting of multiple LSTM layers. Each LSTM layer contains multiple neurons, responsible for capturing the dependencies between different time steps in the sequence. For example, an LSTM unit can record the claim trends of a certain benefit over the past week. Stacking multiple LSTM layers together increases the depth and complexity of the model, enabling it to learn more abstract and deeper time series features. For example, the first LSTM layer extracts short-term trends, and the second LSTM layer builds on this to extract longer-term trends. Sharing some parameters between different LSTM layers reduces the risk of overfitting and improves training efficiency.

[0053] The output of the LSTM network is linearly transformed through a fully connected layer to obtain a dimension that matches the attention mechanism. The attention score calculation process uses the scaled dot product attention mechanism to calculate the relevance score between each time step. Assuming there are three matrices: query, key, and value, the attention score matrix is obtained by taking the dot product of the query and key. The score matrix is then scaled by dividing it by the square root of the key vector dimension to stabilize the gradient.

[0054] Finally, the attention scores are normalized using the softmax function to obtain the weights for each time step. These weights are then multiplied by the corresponding elements of the value matrix and summed to obtain the output of the attention mechanism.

[0055] The mean squared error (MSE) loss function can be used during model training to measure the difference between the predicted value and the true value. The Adam optimization algorithm is used to automatically adjust the learning rate and accelerate model convergence. During training, the predicted value is calculated through forward propagation, then compared with the true value to calculate the loss. Backpropagation is then used to update the model parameters, and this process is repeated until the model converges.

[0056] The model evaluation and optimization process often includes: Evaluation indicators: Use indicators such as mean absolute error (MAE) and root mean square error (RMSE) to evaluate the predictive performance of the model.

[0057] Cross-validation: The time series cross-validation method is used to ensure the stability and generalization ability of the model in different time periods.

[0058] Hyperparameter adjustment: Perform grid search or random search on the number of LSTM network layers, number of neurons, parameters of the attention mechanism, etc. to find the optimal hyperparameter combination.

[0059] Predictive models can accurately predict user demand, proactively understand changes in equity demand, and facilitate inventory optimization. Inventory can be adjusted based on these forecasts to reduce overstocks and shortages, lowering costs. This ensures timely and stable equity supply, enhancing user satisfaction. Furthermore, quantitative evidence is provided to support operational strategies and improve scientific decision-making.

[0060] Optionally, the inventory status data includes inventory quantity and replenishment cycle; and determining the corresponding risk level according to each inventory status data includes: Construct a multivariate linear regression model based on historical inventory data and corresponding historical replenishment cycles; The inventory quantity and the replenishment cycle are calculated based on the multivariate linear regression model to obtain the corresponding risk level.

[0061] Specifically, historical inventory data, including inventory quantity and replenishment cycle, was collected and cleaned to remove outliers and missing values. A dataset was constructed with inventory quantity and replenishment cycle as independent variables and inventory risk level as the dependent variable. A multivariate linear regression model was trained using this historical data, and the regression coefficients were determined. The resulting model form is: Risk level = β0 + β1 × inventory quantity + β2 × replenishment cycle + ε, where β0 is the constant term, β1 and β2 are the relevant regression coefficients, and ε is the error term, representing the portion of risk level variation that the model fails to explain. This term reflects the impact of factors other than inventory quantity and replenishment cycle on risk level that are not included in the model.

[0062] By correlating inventory quantities and replenishment cycles with risk levels, we provide a quantitative assessment method, enhancing the objectivity of risk assessment. This approach provides essential data support for inventory management, enabling decision-makers to plan ahead and effectively reduce inventory disruptions and backlogs. Furthermore, it enables rapid risk assessment and timely identification of inventory issues requiring attention, thereby improving inventory management efficiency. By optimizing inventory strategies, we can significantly reduce inventory costs.

[0063] Optionally, the intelligent decision-making module is further used to: Based on a preset similarity model, the external supply data is matched with the data categories in the supply library to obtain a corresponding equivalent mapping table.

[0064] Optionally, the externally supplied data includes competitor category data, and the preset similarity model includes a BERTT model; the matching of the externally supplied data with the data categories in the supply library based on the preset similarity model to obtain a corresponding equivalence mapping table includes: Performing vector conversion on the competitor product category data to obtain first semantic vector data; Performing vector conversion on each of the data categories to obtain corresponding second semantic vector data, and performing similarity calculation on the first semantic vector data and each of the second semantic vector data based on a similarity calculation method to obtain corresponding similarity data; A temporary data category corresponding to the competitor category data is determined based on all the similarity data, and a corresponding equivalence mapping table is obtained based on all the temporary data categories.

[0065] Specifically, the collected competitor product category data is pre-processed by cleaning, tokenizing, and removing stop words. For example, the competitor product benefit name "Super Value Video Membership Seasonal Card + Exclusive Coupon Combination" is cleaned to "Video Membership Seasonal Card + Coupon Combination" and tokenized to obtain "Video Membership Seasonal Card Coupon Combination."

[0066] The same preprocessing operation is performed on the data categories in the supply library. For example, the benefit category "video membership monthly card" in the supply library is cleaned and segmented into "video membership monthly card".

[0067] The pre-trained BERT model was used to convert the processed competitor category data into semantic vectors. The BERT model captures the deep semantic information of the text and represents each competitor category data as a fixed-dimensional vector. For example, the competitor benefit "Video Season Membership + Coupon Combo" was converted into a vector [0.2, 0.5, 0.7, …, 0.1].

[0068] The BERT model is also used to convert data categories in the supply library into semantic vectors. For example, the benefit category "Video Membership Monthly Card" in the supply library is converted into a vector [0.3, 0.4, 0.6, …, 0.2].

[0069] Cosine similarity is used to calculate the similarity between the semantic vectors of the competitor category data and the supplier category data. Cosine similarity measures the degree of directional similarity between two vectors, with values ranging from -1 to 1. Values closer to 1 indicate greater similarity.

[0070] We calculated the cosine similarity between the vectors for each competitor category and all the vectors for the supply library. For example, we calculated the cosine similarity between the vector [0.2, 0.5, 0.7, …, 0.1] for the competitor benefit "Video Seasonal Membership Card + Coupon Combo" and the vector [0.3, 0.4, 0.6, …, 0.2] for "Video Monthly Membership Card" in the supply library, resulting in a similarity of 0.85.

[0071] Based on the similarity data, we match competitor category data with a similarity above a certain threshold (e.g., 0.8) with the data categories in the supply library to determine the temporary data category corresponding to the competitor category data. For example, the similarity between the competitor benefit "Video Membership Quarterly Pass + Coupon Combo" and the "Video Membership Monthly Pass" in the supply library is 0.85, which is higher than the threshold of 0.8. Therefore, "Video Membership Monthly Pass" is selected as the temporary data category.

[0072] Based on all temporary data categories, an equivalence mapping table is constructed. This table records the equivalence relationship between competitor product category data and data categories in the supply library, and can also be annotated with equivalence coefficients (i.e., similarity data). For example, the equivalence coefficient between "Competitor Benefit: Video Membership Quarterly Pass + Coupon Combo" and "Supply Library Benefit: Video Membership Monthly Pass" recorded in the equivalence mapping table is 0.85.

[0073] Using an equivalence mapping table, the system can quickly identify equivalent competing products as replacements when stock is low, effectively improving resource utilization and reducing the risk of losing user demand due to stockouts. Furthermore, real-time tracking of competing product rights and market dynamics enables the platform to flexibly adjust its rights strategy and enhance its market competitiveness. Furthermore, by introducing competing product equivalent rights, the platform reduces its reliance on a single supplier and mitigates inventory risks caused by supplier issues or poor inventory management, thereby improving the stability and flexibility of the supply chain.

[0074] Optionally, the inventory status data includes inventory data; and the decision execution module is specifically configured to: Obtaining initial weight data of each of the supply warehouses, and adjusting the corresponding initial weight data based on the weight adjustment coefficient to obtain corresponding target weight data; Obtaining demand data corresponding to each supply warehouse according to each target weight data and the user demand forecast data; The demand data is compared with the corresponding inventory data to obtain corresponding comparison results, and inventory scheduling is performed on each supply warehouse according to each comparison result.

[0075] Optionally, after comparing the demand data with the corresponding inventory data to obtain corresponding comparison results, and performing inventory scheduling for each supply warehouse according to each comparison result, the method further includes: When the inventory data of each category of data in each supply warehouse does not meet the demand, each supply warehouse is screened according to the equivalent mapping table to obtain a corresponding substitute scheduling strategy.

[0076] Specifically, the decision execution module reads the initial weight data for each supply warehouse from the system database. This initial weight data reflects the inventory allocation ratio of each supply warehouse under normal circumstances. For example, the initial weights of supply warehouses A, B, and C are 0.4, 0.3, and 0.3, respectively.

[0077] The initial weights of each supply warehouse are adjusted based on the weight adjustment coefficients provided by the intelligent decision-making module to calculate the target weight data. For example, if the weight adjustment coefficients of supply warehouse A are -0.1, supply warehouse B is +0.05, and supply warehouse C is 0, then the adjusted weights are A: 0.3, B: 0.35, and C: 0.3.

[0078] Combine the target weight data with the user demand forecast data to calculate the demand data for each supply warehouse. For example, if the total demand forecast is 3,000 units, then under the adjusted weights, the demand data for Supply Warehouse A is 3,000 x 0.3 = 900 units, Supply Warehouse B is 3,000 x 0.35 = 1,050 units, and Supply Warehouse C is 3,000 x 0.3 = 900 units.

[0079] The current inventory data of each supply warehouse is compared with the calculated demand data. If the inventory data of a supply warehouse does not meet the demand data, the backup scheduling mechanism is triggered. For example, it is found that the current inventory data of supply warehouse B is 800 pieces, which is less than the demand data of 1050 pieces.

[0080] Based on the equivalence mapping table, we screen out competing product benefits with equivalent relationships as alternative options. For example, the equivalence mapping table shows that the equivalence coefficient between the competing platform's "Quarterly Video Membership Card" and our platform's "Monthly Video Membership Card" is 0.85, and there is sufficient inventory.

[0081] A replacement scheduling strategy is formed by shifting some of the demand from a supply base with insufficient inventory to a competing supply base with equivalent equity. For example, 250 units of demand could be shifted from supply base B to competing supply base D to meet the total demand.

[0082] This process optimizes inventory allocation, reduces overstocks and out-of-stocks, and lowers management costs through dynamic weight adjustment and precise demand calculation. Automated scheduling reduces manual operations, improving supply chain responsiveness and operational efficiency, enabling faster adaptation to market and demand changes. This ensures users receive the benefits they need in a timely manner, enhancing user experience and satisfaction, and strengthening platform loyalty. Using an equivalence mapping table to screen for alternative benefits reduces the risk of insufficient inventory or supplier issues, ensuring a stable and reliable supply chain.

[0083] For example, the system's supply library contains a benefit category called "Monthly Video Membership Card," and the predicted demand for the next 24 hours is 3,000 units. The initial weights are for supply libraries A (0.4), B (0.3), and C (0.3). After adjusting the weight adjustment coefficient, the target weights are A (0.3), B (0.35), and C (0.3). The calculated demand for each supply library is 900 units for A, 1,050 units for B, and 900 units for C. If supply library B currently has only 800 units in stock, it cannot meet the demand. Based on the equivalence mapping table, it is found that the equivalence coefficient between the competing platform's "Quarterly Video Membership Card" and the platform's "Monthly Video Membership Card" is 0.85, and competing supply library D has sufficient inventory. Therefore, the 250 units of demand that cannot be met by supply library B are transferred to competing supply library D, forming a backup scheduling strategy to ensure that the total demand is met.

[0084] It should be noted that in addition to adjusting through the equivalent mapping table, there are several other ways to handle the situation where the supply warehouse inventory data does not meet the requirements: 1. Allocate inventory from other supply warehouses: If another supply warehouse has sufficient inventory, the excess inventory can be allocated to the supply warehouse with insufficient inventory. For example, if supply warehouse A has sufficient inventory, some inventory can be temporarily allocated to supply warehouse B to meet demand.

[0085] 2. Emergency Replenishment: Place an expedited replenishment order with the supplier to speed up the replenishment process. This is suitable for situations where the supplier has a short replenishment cycle and a quick response.

[0086] 3. Delayed fulfillment of some demands: Divide demands into urgent and non-urgent ones. Urgent ones are met first, while non-urgent ones are delivered later. This is applicable to situations where demands have varying degrees of urgency.

[0087] 4. Adjust weights and redistribute demand: Based on real-time inventory conditions, adjust the weights of each supply warehouse again and redistribute demand to other supply warehouses.

[0088] Example: Supply Base B has 800 units in stock, demand is 1,050 units, and there is a shortage of 250 units. Reduce the weight of Supply Base B from 0.35 to 0.25 and distribute the excess demand to other supply bases.

[0089] These methods can be used individually or in combination to flexibly deal with the problem of insufficient inventory in the supply warehouse, ensuring the stable operation of the supply chain and the timely satisfaction of user needs.

[0090] Alternatively, as Figure 2 As shown, the artificial intelligence-based virtual commodity supply chain scheduling system also includes an abnormality feedback module; The abnormal feedback module is used to trigger an alternative supply process when the artificial intelligence-based virtual commodity supply chain scheduling system fails to send a replenishment order to the supply warehouse; The abnormal feedback module is also used to automatically check the weight adjustment coefficient and verify the feedback result of the instruction issued by the decision execution module.

[0091] Specifically, one of the functions of the exception feedback module is to trigger alternative supply processes: When the AI-based virtual goods supply chain scheduling system fails to send a replenishment order to the supply warehouse, the exception feedback module detects this anomaly in real time and immediately triggers the backup supply process. It selects a suitable backup supplier from a pre-set list based on established priority and quickly sends a replenishment order to it, ensuring supply continuity.

[0092] For example, the system sends a replenishment order to Supply Store A, but the order fails due to a network failure. Upon detecting the failure, the exception feedback module immediately selects Supply Store D from the list of potential suppliers and sends a replenishment order to Supply Store D, ensuring timely replenishment of the equity product without affecting subsequent supply.

[0093] Function 2: Automatic verification and feedback verification: The Abnormal Feedback Module regularly and automatically verifies the rationality of the weight adjustment coefficient. It evaluates the weight adjustment coefficient based on real-time inventory status, historical sales data, and market dynamics, applying preset verification rules (such as indicators such as inventory fill rate and demand matching after weight adjustment). If the weight adjustment coefficient is found to be unreasonable, such as causing inventory allocation to deviate significantly from actual demand, the module automatically triggers the weight recalculation process and reports the issue to the Intelligent Decision-Making Module for further analysis and optimization.

[0094] After the decision-making and execution module issues inventory scheduling instructions, the exception feedback module tracks and verifies the execution results in real time. By interacting with real-time data from the supply warehouse system, logistics system, and other systems, the module compares actual execution with the instruction requirements, such as replenishment quantity and time. If any deviation is detected, such as the supply warehouse failing to complete replenishment on time or replenishing insufficient quantities, the exception feedback module promptly reports the issue to the decision-making and execution module and initiates appropriate remedial measures based on pre-set rules, such as reissuing replenishment instructions or adjusting subsequent scheduling plans.

[0095] The exception feedback module effectively addresses issues such as replenishment failures and unreasonable weighting by triggering alternative supply processes and automatically verifying weight adjustment coefficients. This reduces the risk of system interruptions due to abnormal situations and ensures stable supply chain operations. Furthermore, it verifies the feedback results of decision-making execution instructions, promptly identifying and correcting deviations, ensuring the accuracy of information such as inventory data and scheduling instructions, and providing reliable data support for precise supply chain decision-making. This automated verification and feedback mechanism enables rapid identification of issues and the implementation of appropriate measures, reducing manual intervention time and improving supply chain responsiveness and overall scheduling efficiency.

[0096] For example, when the system sends a replenishment order to Supply Depot B, the order fails due to a system failure. The exception feedback module immediately detects this anomaly, triggering the alternative supply process and selecting Supply Depot C from the list of alternative suppliers to send the replenishment order. Simultaneously, the exception feedback module automatically verifies the weight adjustment coefficients and discovers that the weight adjustment by Supply Depot B has resulted in an unreasonable allocation of some demand. This module promptly provides feedback to the intelligent decision-making module for optimization and adjustment. After the decision-making execution module issues a new dispatch instruction, the exception feedback module verifies the execution results to ensure that Supply Depot C completes the replenishment on time and then provides feedback to the system.

[0097] Alternatively, as Figure 2 As shown, the artificial intelligence-based virtual commodity supply chain scheduling system also includes a collaborative network module, which is used to enable the artificial intelligence-based virtual commodity supply chain scheduling system to interact collaboratively with an external entity system.

[0098] Specifically, the collaborative network module's primary function is to enable system-wide collaborative interaction. Through standard interface protocols (such as RESTful APIs and Web Services), it seamlessly connects the AI-based virtual goods supply chain scheduling system with external physical systems (such as suppliers' inventory management systems, logistics distribution systems, and e-commerce platform systems). This enables real-time sharing and synchronization of data, such as inventory information, order status, and logistics progress, between these systems, enabling efficient information flow and collaborative work.

[0099] This module supports cross-system business process orchestration and management, allowing for customizable collaborative business processes. For example, in the replenishment process, the collaborative network module can coordinate suppliers' production plans, logistics and distribution arrangements, and platform inventory updates based on pre-set business rules, ensuring efficient collaboration and smooth execution throughout the replenishment process.

[0100] The collaborative network module enables real-time data exchange and collaboration between the AI-based virtual commodity supply chain scheduling system and external physical systems, breaking down information silos and improving collaboration efficiency and information sharing across all supply chain links. Furthermore, through cross-system collaborative business process management, it can better integrate supply chain resources, optimize resource allocation, improve resource utilization, and reduce operating costs. Furthermore, in the face of market fluctuations, demand changes, or emergencies, the collaborative network module can quickly coordinate resources from all parties, flexibly adjust supply chain strategies, and enhance supply chain resilience and adaptability.

[0101] For example, the collaborative network module connects in real time with the supplier's inventory management system through an interface. When the system predicts that a particular equity commodity is running low and needs to be restocked, the collaborative network module transmits the replenishment request to the supplier's inventory management system in real time. The supplier then schedules production and updates inventory information accordingly. Simultaneously, the collaborative network module works in conjunction with the logistics and distribution system to track the progress of shipments in real time after the supplier ships, and feeds this information back to the platform. Users can check the replenishment progress of equity commodities on the platform, enhancing the user experience. Furthermore, the platform can share sales data with the e-commerce platform system through the collaborative network module, allowing it to proactively adjust replenishment plans based on sales trends and optimize inventory management.

[0102] “It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of relevant countries and regions, and corresponding operation portals are provided for users to choose to authorize or refuse.” like Figure 3 As shown, an embodiment of the present invention provides a virtual commodity supply chain scheduling method, which is applied to the artificial intelligence-based virtual commodity supply chain scheduling system. The scheduling method includes: Collecting user behavior data, external supply data, and inventory status data of each of the supply warehouses; Based on the preset prediction model, user demand prediction data is obtained according to user behavior data; Determine the corresponding risk level based on each inventory status data, and obtain the corresponding weight adjustment coefficient based on each risk level; The inventory supply of each supply warehouse is scheduled according to the user demand forecast data and the weight adjustment coefficients.

[0103] Although the present invention is disclosed as above, the protection scope of the present invention is not limited thereto. Those skilled in the art may make various changes and modifications without departing from the spirit and scope of the present invention, and these changes and modifications will fall within the protection scope of the present invention.

Claims

1. A virtual commodity supply chain scheduling system based on artificial intelligence, characterized by: Applied to a commodity supply system, the commodity supply system includes multiple supply warehouses; the virtual commodity supply chain scheduling system includes a data acquisition module, an intelligent decision module and a decision execution module; The data collection module is used to collect user behavior data, external supply data and inventory status data of each supply warehouse; The intelligent decision-making module is configured to obtain user demand forecast data based on the user behavior data based on a preset forecast model; determine a corresponding risk level based on each inventory status data, and obtain a corresponding weight adjustment coefficient based on each risk level; The decision execution module is used to perform inventory supply scheduling for each supply depot based on the user demand forecast data and each weight adjustment coefficient; Among them, the preset prediction model is constructed based on multiple LSTM networks and attention mechanisms.

2. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 1 is characterized in that: The user behavior data includes category data and time data of the user's rights and interests; the user demand prediction data obtained based on the user behavior data based on the preset prediction model includes: Integrating the category data and the time data to obtain a time series data set; The time series data set is input into the preset prediction model to obtain the user demand prediction data.

3. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 1 is characterized in that: The inventory status data includes inventory quantity and replenishment cycle; Determining the corresponding risk level according to each inventory status data includes: Construct a multivariate linear regression model based on historical inventory data and corresponding historical replenishment cycles; The inventory quantity and the replenishment cycle are calculated based on the multiple linear regression model to obtain the corresponding risk level.

4. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 1, characterized in that: The intelligent decision-making module is also used to: Based on a preset similarity model, the external supply data is matched with the data categories in the supply library to obtain a corresponding equivalence mapping table.

5. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 4 is characterized in that: The externally supplied data includes competitor category data, and the preset similarity model includes a BERTT model; the preset similarity model is used to match the externally supplied data with the data categories in the supply library to obtain a corresponding equivalence mapping table, including: Performing vector conversion on the competitor product category data to obtain first semantic vector data; Performing vector conversion on each of the data categories to obtain corresponding second semantic vector data, and performing similarity calculation on the first semantic vector data and each of the second semantic vector data based on a similarity calculation method to obtain corresponding similarity data; A temporary data category corresponding to the competitor category data is determined based on all the similarity data, and a corresponding equivalence mapping table is obtained based on all the temporary data categories.

6. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 4 is characterized in that: The inventory status data includes inventory data; the decision execution module is specifically used to: Obtaining initial weight data of each of the supply warehouses, and adjusting the corresponding initial weight data based on the weight adjustment coefficient to obtain corresponding target weight data; Obtaining demand data corresponding to each supply warehouse according to each target weight data and the user demand forecast data; The demand data is compared with the corresponding inventory data to obtain corresponding comparison results, and inventory scheduling is performed on each supply warehouse according to each comparison result.

7. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 6, characterized in that: After comparing the demand data with the corresponding inventory data to obtain corresponding comparison results, and performing inventory scheduling for each supply warehouse according to each comparison result, the method further includes: When the inventory data of the category data in any of the supply warehouses does not meet the demand, the supply warehouses are screened according to the equivalent mapping table to obtain a corresponding substitute scheduling strategy.

8. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 1 is characterized in that: The artificial intelligence-based virtual commodity supply chain scheduling system also includes an abnormality feedback module; The abnormal feedback module is used to trigger an alternative supply process when the artificial intelligence-based virtual commodity supply chain scheduling system fails to send a replenishment order to the supply warehouse; The abnormal feedback module is also used to automatically check the weight adjustment coefficient and verify the feedback result of the instruction issued by the decision execution module.

9. The artificial intelligence-based virtual commodity supply chain scheduling system according to claim 1, characterized in that: The artificial intelligence-based virtual commodity supply chain scheduling system further includes a collaborative network module, which is used to enable the artificial intelligence-based virtual commodity supply chain scheduling system to interact collaboratively with an external entity system.

10. A virtual commodity supply chain scheduling method, characterized in that: Applied to the artificial intelligence-based virtual commodity supply chain scheduling system according to any one of claims 1 to 9, the scheduling method includes: Collecting user behavior data, external supply data, and inventory status data of each of the supply warehouses; Based on a preset prediction model, obtain user demand prediction data according to the user behavior data; Determining a corresponding risk level according to each of the inventory status data, and obtaining a corresponding weight adjustment coefficient based on each of the risk levels; Inventory supply scheduling is performed on each supply warehouse according to the user demand forecast data and each weight adjustment coefficient.

Images