Supply chain demand prediction and dynamic inventory optimization method based on big data

Abstract

The invention relates to a supply chain demand prediction and dynamic inventory optimization method based on big data, and relates to the technical field of inventory scheduling prediction, and the optimization steps are as follows: S1, collecting data, preprocessing the collected data, and uploading the preprocessed data to a data analysis module; s2, a data analysis module which performs grid division on different areas of the supply chain, analyzes supply demands of different grid division areas based on historical data, and judges factors which influence the demands of the supply chain of the divided areas; and S3, based on the obtained factors, obtaining the size degree of each influence factor existing in different divided areas in real time, and carrying out weighted comprehensive calculation to obtain an influence value. Complex multi-factor influences are presented in a quantitative form, the comprehensive effect of different factors on supply chain demands is scientifically measured, quantitative index support is provided for decision making, the warehouse stock amount is intelligently adjusted based on the influence value, and the stock level can be better matched with actual demands.

Description

Technical Field

[0001] This application relates to the field of inventory scheduling and forecasting technology, and in particular to a big data-based supply chain demand forecasting and dynamic inventory optimization method. Background Technology

[0002] In today's rapidly globalizing and digitalizing business environment, supply chain management is crucial for a company's operational efficiency and economic benefits. As market competition intensifies and customer demands become more diverse, personalized, and rapidly changing, supply chain demand forecasting and inventory management face significant challenges.

[0003] Traditional supply chain demand forecasting methods often rely on simple historical data trend analysis or experience-based judgment, making it difficult to accurately capture complex dynamic changes in the market. On the one hand, the market environment is influenced by numerous factors, such as differences in consumption habits across regions, varying levels of economic development, diverse promotional activities, and demand fluctuations during special periods like holidays. These factors are intertwined, and traditional methods cannot comprehensively and accurately consider them. On the other hand, with the booming development of e-commerce and the rise of new retail models, the supply chain structure has become increasingly complex, evolving from multi-level distribution to online-offline integration, drastically increasing the difficulty of inventory management. Excessive inventory leads to warehousing costs and capital tied up, while insufficient inventory may trigger stockout risks, impacting customer satisfaction and corporate reputation. To address this, we propose a big data-based supply chain demand forecasting and dynamic inventory optimization method. Summary of the Invention

[0004] The purpose of this application is to provide a method for supply chain demand forecasting and dynamic inventory optimization based on big data. The optimization steps are as follows: S1. Collect data, preprocess the collected data, and upload it to the data analysis module; S2, Data Analysis Module: Divides the supply chain into different regions into grids, analyzes the supply demand in different grid regions based on historical data, and determines the factors affecting the supply chain demand in each region. S3. Based on the obtained factors, the magnitude of each influencing factor in different regions is obtained in real time, and the influence value is calculated by weighted comprehensive calculation. Based on the obtained influence value, the warehouse inventory is intelligently adjusted.

[0005] Preferably, the data collected in step S1 includes order data, inventory data, logistics data, policy data, geographic data, and regional characteristics; Data preprocessing includes data cleaning, outlier detection, and normalization.

[0006] Preferably, the specific steps for grid division of different regions of the supply chain in step S2 are as follows: The specific objectives of grid division are clearly defined in advance, namely, to optimize inventory scheduling requirements; Define the geographical scope and customer density indicators, acquire geographical, business and population density data, and integrate them after cleaning and coordinate transformation to form a dataset; Based on clustering algorithms, grids are generated according to data density and business characteristics. Large grids are generated in areas with high demand from supply chain customers, and small grids are generated in areas with low demand from supply chain customers. Calculate the business density and data variance within each grid, evaluate the rationality of the grid division, and optimize the grid. The divided grids are coded and labeled with attribute information for management.

[0007] Preferably, the specific process of the clustering algorithm and the optimized grid is as follows: Parameters are set according to the business scenario, including the neighborhood radius and minimum number of samples, which are determined through grid search and empirical values; Traverse the dataset. If the number of samples in the neighborhood of a point is greater than or equal to the minimum number of samples, mark it as a core point and create a new cluster. Expand the cluster based on the core point by adding all points in its neighborhood to the cluster. Repeat this process until there are no more new points to add. Points that are not core points and are not in any cluster are marked as noise points. Based on the clustering results, a grid is generated on the basis of each cluster. In clustering areas with high customer demand and dense data, neighboring clusters are merged to generate a large grid; in clustering areas with low customer demand and sparse data, small grids are generated directly by merging individual clusters or small areas. Calculate the number of customers and total orders within each grid to assess the grid's rationality. If the business volume within a grid differs too much, readjust the neighborhood radius and minimum sample size parameters, and repeat the clustering and grid generation steps until the grid is reasonable.

[0008] The preferred method for determining the factors influencing supply chain demand in different regions is as follows: Collect historical data for each grid area, and draw histograms and heatmaps to visualize the temporal trends and spatial distribution differences in demand for each grid area. By calculating the mean, standard deviation, and correlation coefficient statistics, the relationships between variables are analyzed, and potential factors that may be related to changes in demand are screened out. We selected a multiple linear regression model to analyze the linear relationship between demand and various influencing factors, and calculated the degree of influence of each factor on changes in demand. Based on the calculation results output by the model, the results are sorted, and the three groups of factors with the highest impact on supply chain demand are selected; the influencing factors identified in the analysis are then applied to actual demand forecasting and inventory management. The factors identified included promotional policies, geographical location, and holidays.

[0009] The preferred algorithm steps for determining the factors influencing regional supply chain demand are as follows: The statistical analysis of the relationships between variables is performed to calculate the mean of demand data for each grid area, representing the average level of demand in that area. The standard deviation is used to measure the dispersion of demand data around the mean, reflecting the magnitude of demand fluctuations. The correlation coefficients between different variables are calculated to determine the strength and direction of the correlation, and potential factors that may be related to changes in demand are screened out. The multiple linear regression model is based on the principle of least squares, and the coefficients corresponding to each factor in the model are continuously adjusted to minimize the sum of squared errors between the model's predicted values ​​and the actual demand values, thus obtaining the degree of influence of each factor on changes in demand.

[0010] Preferably, the factors obtained in step S3 include promotional policy factors, geographical development factors, and holiday factors; among which, promotional policy factors are obtained through local websites and supply chain merchants; geographical development factors are obtained through digital maps to obtain the transportation and building development of grid cities; and holiday factors are obtained based on whether the day is a holiday. Real-time access to policies offered by different merchants in different regions, the level of development of the regions, and the impact of holidays on supply chain inventory; The values ​​of each factor are quantified, and the weight values ​​of different factors are obtained based on the analytic hierarchy process. The influence value is obtained through weighted calculation. When the impact is significant, the warehouse inventory should be replenished promptly. When the impact value is small, the amount of replenishment to the warehouse inventory is reduced, and the warehouse inventory is dynamically adjusted.

[0011] Preferably, the steps for quantifying the values ​​of each factor are as follows: Obtain promotional information and convert discount percentages into numerical values; Collect commercial density and traffic density data for each city grid, and calculate a weighted comprehensive score for geographical development. By integrating information from national statutory holidays and e-commerce shopping festivals, a holiday impact index is constructed. Holidays are divided into long holidays and short holidays, and different basic impact coefficients are assigned to each. The demand order growth rate of each holiday type is calculated to form a holiday impact index.

[0012] Preferably, the formula for converting the discount percentage into a numerical value is as follows: Based on the formula, it can be concluded that the greater the discount and the closer the value is to 1, the stronger the pull effect on inventory demand, that is, the greater the demand on the supply chain. The steps for calculating the comprehensive score of geographical development level are as follows: Commercial density and traffic density are standardized using the following formulas: The regional commercial density is calculated as the number of shopping malls / shops in the region divided by the area of ​​the region; the regional traffic density is calculated as the length of the main road plus the number of subway stations × 3 plus the number of express delivery outlets × 2. The weighted composite score is calculated using the following formula: Geographical development score = Standardized commercial density value × w1 + Standardized traffic density value × w2 Where w1 and w2 are weight values, which are determined by the AHP method; The calculation steps for the holiday impact index are as follows: Long holidays are defined as statutory holidays of ≥3 days or major e-commerce promotions, while short holidays are defined as ordinary weekends. The formula for calculating the growth rate of demand orders is: The formula for calculating the order growth rate of the holiday impact index is: α is the basic impact coefficient, which is a pre-set coefficient based on the type of holiday. It is used to reflect the basic impact of different types of holidays on inventory demand and other aspects. The α value is different for different types of holidays. The basic impact coefficient is relatively high for long holidays and relatively low for short holidays. The holiday impact index is obtained by calculation.

[0013] In summary, this application includes at least one of the following beneficial technical effects: 1. Collecting and preprocessing data removes noise and fills in missing values, resulting in higher data quality and providing an accurate and reliable data foundation for subsequent analysis. This makes the analysis results more credible. Uploading the processed data to the data analysis module facilitates centralized storage and retrieval of data, allowing different analysts or analysis tasks to quickly obtain the required data and improving overall work efficiency. 2. Dividing the supply chain into grids allows for more refined analysis granularity, revealing unique characteristics and differences in supply and demand across different regions. This avoids general analysis masking regional characteristics and helps identify factors influencing supply chain demand based on historical data. It also helps identify key drivers such as regional economic level, population density, and consumption habits, providing a basis for subsequent targeted strategy development. 3. Real-time acquisition of the magnitude of each factor and weighted comprehensive calculation of the impact value, presenting the complex multi-factor impact in a quantitative form, scientifically measuring the comprehensive effect of different factors on supply chain demand, providing quantitative indicator support for decision-making, and intelligently adjusting warehouse inventory based on the impact value, so that the inventory level can better match the actual demand, avoid inventory backlog tying up capital, or inventory shortage leading to stockout losses, and improve inventory management efficiency and overall supply chain benefits. Attached Figure Description

[0014] Figure 1 This is a flowchart of the optimized steps of the present invention; Figure 2 This is a flowchart of the steps for dividing the supply chain into grids in this invention. Detailed Implementation

[0015] The following is in conjunction with the appendix Figure 1 - Appendix Figure 2 This application will be described in further detail.

[0016] A big data-based supply chain demand forecasting and dynamic inventory optimization method, comprising the following optimization steps: S1. Collect data, preprocess the collected data, and upload it to the data analysis module; S2, Data Analysis Module: Divides the supply chain into different regions into grids, analyzes the supply demand in different grid regions based on historical data, and determines the factors affecting the supply chain demand in each region. S3. Based on the obtained factors, the magnitude of each influencing factor in different regions is obtained in real time, and the influence value is calculated by weighted comprehensive calculation. Based on the obtained influence value, the warehouse inventory is intelligently adjusted.

[0017] In the data collection and preprocessing upload phase of this application, a standardized data collection process was implemented to remove data impurities, ensuring data quality and laying a solid foundation for subsequent analysis. Data was centrally uploaded to the data analysis module for unified management and retrieval, improving data flow and utilization efficiency. In the regional grid division and demand factor analysis phase, detailed regional segmentation enabled precise insights into supply and demand differences, preventing regional characteristics from being overlooked in general analyses. In-depth mining of influencing factors based on historical data provided a clear direction for supply chain management strategies. During factor quantification and intelligent inventory adjustment, scientific quantitative methods were used to comprehensively consider the impact of multiple factors, using precise numerical values ​​to support decision-making. Dynamically adjusting inventory based on impact values ​​effectively balances inventory levels with actual demand, avoiding the risks of inventory backlog or shortages, finding the optimal solution between capital occupation and supply security, improving the overall operational efficiency and economic benefits of the supply chain, and enhancing its flexibility and responsiveness.

[0018] The data collected in step S1 includes order data, inventory data, logistics data, policy data, geographic data, and regional characteristics; Data preprocessing includes data cleaning, outlier detection, and normalization.

[0019] The rich data sources in this application provide ample information for subsequent analysis, enabling companies to accurately assess business performance, identify potential problems and opportunities, and make more realistic production, procurement, and distribution decisions. Data cleaning removes duplicate, erroneous, and incomplete data, making it cleaner and more accurate; outlier detection identifies data that deviates from the normal range, preventing it from misleading the analysis results; normalization unifies the data scale, eliminates the influence of different units, makes the data more comparable, and enhances the stability and accuracy of the analysis model.

[0020] The specific steps for grid-dividing different regions of the supply chain in step S2 are as follows: In supply chain management, the specific goal of pre-defining grid division is to optimize inventory scheduling needs. By accurately dividing regional grids, we can gain a more detailed understanding of the demand characteristics of different regions, thereby achieving efficient allocation of inventory resources. Set the geographic scope and customer density indicators, acquire geographic data, business data and population density data, clean the data to remove duplicate, erroneous and incomplete data records to ensure the accuracy and consistency of the data. At the same time, since the coordinate systems of different data sources are different, coordinate transformation is required to unify all data under the same geographic coordinate framework. The cleaned and coordinate-transformed data are integrated to form a comprehensive dataset. In this dataset, each record is associated with a specific geographical location, business information, and population-related information, providing comprehensive data support for subsequent grid division. Clustering algorithms group data based on the similarity between data points. In supply chain regional grid division, based on the data density and business characteristics of the integrated dataset, clustering algorithms are used to divide the geographic space into different grids. In areas of high customer demand in the supply chain, large grids are generated due to the dense data points and obvious business characteristics, in order to cover more areas of concentrated demand; while in areas of low customer demand, data points are sparse, resulting in small grids. This division method can better meet actual needs and improve the targeting of inventory scheduling. Calculate the service density within each grid, which is the ratio of service volume to grid area, to measure the activity level of services within the grid. Calculate the data variance to reflect the dispersion of data within the grid and assess the stability and consistency of the data. The rationality of the grid division is evaluated based on the calculated business density and data variance. If it is found that the business density of some grids is unreasonable (such as the business volume of large grids being too low or the business volume of small grids being too high), or the data variance is too large (indicating abnormal data fluctuations within the grid), the grid is adjusted and optimized. Possible optimization methods include re-dividing grid boundaries, merging or splitting grids to ensure that each grid can accurately reflect its internal business characteristics and needs. Each grid is uniquely coded to facilitate identification and location within the supply chain management system. The grid's attribute information, such as geographical location, main business type, customer group characteristics, and demand fluctuation patterns, is labeled. This attribute information will provide detailed reference for subsequent business decisions such as inventory scheduling and logistics distribution, enabling effective management of different grid areas in the supply chain.

[0021] The specific process of clustering algorithm and mesh optimization is as follows: Formulas for clustering algorithms and optimized mesh processes The settings parameters include: The neighborhood radius is ∈, the minimum number of samples is MinPts, and the dataset is D={x1,x2,...,x n}; The core point determination is for any point x. i ∈D, its neighborhood is defined as: N ∈ (x i )={x j ∈D|dist(x i ,x j )≤∈} Where N ∈ (x i ) represents x i The neighborhood of ∈ includes the neighborhood of point x. i The set of all points whose distance is in the range ∈; x i ,x j Represented as a point in the dataset, dist(x) i ,x j ) represents the calculation point x i With x j The function of the distance between them is the Euclidean distance function, used to determine x. j Is it in x? i ∈ within the neighborhood; If the number of samples in the neighborhood satisfies: |N ∈ (x i )|≥MinPts Then x i Marked as the core point, denoted as x i∈CorePoints; For each core point x i ∈CorePoints, create a new cluster C k And through neighborhood expansion, and: Where C k The table shows the set of data points aggregated by the rules in the k-th cluster. Represents all distances C with index less than k. l Perform a union operation, ensuring that x j It will not be repeatedly assigned to an already generated cluster; Points not included in any cluster are marked as noise: Where Noise represents the set of noise points. This indicates that for all clusters C k Perform a union operation; Number of customers Total number of orders Demand(C) k = f(number of customers, total number of orders) Grid merging rule: If Demand(C k )>θ high And Distance(C k C l If ) < δ, then merge C k and C l .

[0022] The steps to determine the factors influencing supply chain demand in different regions are as follows: Collect historical data for each grid area, and draw histograms and heatmaps to visualize the temporal trends and spatial distribution differences in demand for each grid area. By calculating the mean, standard deviation, and correlation coefficient statistics, the relationships between variables are analyzed, and potential factors that may be related to changes in demand are screened out. We selected a multiple linear regression model to analyze the linear relationship between demand and various influencing factors, and calculated the degree of influence of each factor on changes in demand. Based on the calculation results output by the model, the results are sorted, and the three groups of factors with the highest impact on supply chain demand are selected; the influencing factors identified in the analysis are then applied to actual demand forecasting and inventory management. The factors identified included promotional policies, geographical location, and holidays.

[0023] This application collects historical data to create histograms and heatmaps. Through visualization, it allows for a quick and intuitive understanding of how demand fluctuates over time and is spatially distributed across different grid areas. For example, heatmaps clearly show which areas have strong demand and which have weak demand, providing a clear direction for subsequent analysis and reducing the cost of understanding the complexity of the data. It also calculates statistical measures such as mean, standard deviation, and correlation coefficient. The mean reflects the central tendency of the data, the standard deviation reflects the degree of dispersion, and the correlation coefficient measures the degree of association between variables. Using these statistical measures, potential factors that may be related to changes in demand can be accurately screened from numerous variables, avoiding blind analysis and improving the efficiency and accuracy of factor selection. Furthermore, it employs a multiple linear regression model to quantitatively analyze the linear relationship between demand and various influencing factors, calculating the degree of influence of each factor on changes in demand. This quantitative approach allows enterprises to clearly understand the specific impact of different factors on demand, providing precise data support for resource allocation and strategy formulation.

[0024] The algorithm steps for determining the factors influencing regional supply chain demand are as follows: The statistical analysis of the relationships between variables is performed to calculate the mean of demand data for each grid area, representing the average level of demand in that area. The standard deviation is used to measure the dispersion of demand data around the mean, reflecting the magnitude of demand fluctuations. The correlation coefficients between different variables are calculated to determine the strength and direction of the correlation, and potential factors that may be related to changes in demand are screened out. The multiple linear regression model is based on the principle of least squares, and the coefficients corresponding to each factor in the model are continuously adjusted to minimize the sum of squared errors between the model's predicted values ​​and the actual demand values, thus obtaining the degree of influence of each factor on changes in demand.

[0025] This application calculates the average demand data for each grid area, which can intuitively present the average level of demand in that area. This provides enterprises with a basic reference value, making it easier to understand the scale of demand in different areas. By comparing the average demand of different city grid areas, it is possible to quickly locate areas with large and small demand scales, providing key basis for resource allocation and market strategy formulation, such as increasing inventory reserves or allocating more marketing resources in areas with high average demand.

[0026] The factors obtained in step S3 include promotional policy factors, geographical development factors, and holiday factors. Promotional policy factors are obtained through local websites and supply chain merchants; geographical development factors are obtained through digital maps to assess the transportation and building development of grid cities; and holiday factors are determined based on whether the day is a holiday. Real-time access to policies offered by different merchants in different regions, the level of development of the regions, and the impact of holidays on supply chain inventory; The values ​​of each factor are quantified, and the weight values ​​of different factors are obtained based on the analytic hierarchy process. The influence value is obtained through weighted calculation. When the impact is significant, the warehouse inventory should be replenished promptly. When the impact value is small, the amount of replenishment to the warehouse inventory is reduced, and the warehouse inventory is dynamically adjusted.

[0027] This application obtains promotional policy factors through local websites and supply chain merchants. Local websites provide public and extensive promotional information, covering various online and offline merchant activities; supply chain merchants can directly provide feedback on internal promotional strategies. Obtaining information through multiple channels ensures comprehensiveness, avoids missing important promotional activities, and allows enterprises to fully grasp market promotion dynamics and adjust inventory strategies in a timely manner. Digital maps are used to obtain information on the development of transportation and buildings in grid cities. Digital maps present geographical information in an intuitive and visual way, which can accurately reflect the region's transportation convenience and the distribution of commercial buildings.

[0028] The steps for quantifying each factor value are as follows: Obtain promotional information and convert discount percentages into numerical values; Collect commercial density and traffic density data for each city grid, and calculate a weighted comprehensive score for geographical development. By integrating information from national statutory holidays and e-commerce shopping festivals, a holiday impact index is constructed. Holidays are divided into long holidays and short holidays, and different basic impact coefficients are assigned to each. The demand order growth rate of each holiday type is calculated to form a holiday impact index.

[0029] The formula for converting the discount percentage into a numerical value is as follows: Based on the formula, it can be concluded that the greater the discount and the closer the value is to 1, the stronger the pull effect on inventory demand, that is, the greater the demand on the supply chain. The steps for calculating the comprehensive score of geographical development level are as follows: Commercial density and traffic density are standardized using the following formulas: The regional commercial density is calculated as the number of shopping malls / shops in the region divided by the area of ​​the region; the regional traffic density is calculated as the length of the main road plus the number of subway stations × 3 plus the number of express delivery outlets × 2. The weighted composite score is calculated using the following formula: Geographical development score = Standardized commercial density value × w1 + Standardized traffic density value × w2 Where w1 and w2 are weight values, which are determined by the AHP method; The calculation steps for the holiday impact index are as follows: Long holidays are defined as statutory holidays of ≥3 days or major e-commerce promotions, while short holidays are defined as ordinary weekends. The formula for calculating the growth rate of demand orders is: The formula for calculating the order growth rate of the holiday impact index is: α is the basic impact coefficient, which is a pre-set coefficient based on the type of holiday. It is used to reflect the basic impact of different types of holidays on inventory demand and other aspects. The α value is different for different types of holidays. The basic impact coefficient is relatively high for long holidays and relatively low for short holidays. The holiday impact index is obtained by calculation.

[0030] This application uses historical data to create visual charts and combines statistical measures such as mean, standard deviation, and correlation coefficient to identify three key influencing factors: promotional policies, geographical development, and holidays. Promotional policies are obtained from local websites and supply chain merchants. Geographical development is calculated using digital maps to determine commercial and traffic density, which is then standardized and weighted. Holidays are categorized into short-term and long-term based on the number of days and e-commerce promotions. Each factor is then quantified, and the analytic hierarchy process (AHP) is used to determine weights and calculate the impact value. When the impact value is large, warehouse inventory is replenished promptly; when the impact value is small, replenishment is reduced, achieving dynamic inventory adjustment. The application also provides calculation formulas for discount conversion values, comprehensive scores for geographical development, holiday impact index, and demand order growth rate. Through scientific quantification and dynamic management, this approach helps enterprises accurately respond to market changes, optimize supply chain inventory strategies, and improve operational efficiency and competitiveness.

[0031] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Identical components are represented by the same reference numerals. Therefore, all equivalent changes made to the structure, shape, and principle of this application should be covered within the scope of protection of this application.

Claims

1. A method for supply chain demand forecasting and dynamic inventory optimization based on big data, characterized in that, The optimization steps are as follows: S1. Collect data, preprocess the collected data, and upload it to the data analysis module; S2, Data Analysis Module: Divides the supply chain into different regions into grids, analyzes the supply demand in different grid regions based on historical data, and determines the factors affecting the supply chain demand in each region. S3. Based on the obtained factors, the magnitude of each influencing factor in different regions is obtained in real time, and the influence value is calculated by weighted comprehensive calculation. Based on the obtained influence value, the warehouse inventory is intelligently adjusted.

2. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 1, characterized in that, The data collected in step S1 includes order data, inventory data, logistics data, policy data, geographic data, and regional characteristics; Data preprocessing includes data cleaning, outlier detection, and normalization.

3. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 1, characterized in that, The specific steps for grid-dividing different regions of the supply chain in step S2 are as follows: The specific objectives of grid division are clearly defined in advance, namely, to optimize inventory scheduling requirements; Define the geographical scope and customer density indicators, acquire geographical, business and population density data, and integrate them after cleaning and coordinate transformation to form a dataset; Based on clustering algorithms, grids are generated according to data density and business characteristics. Large grids are generated in areas with high demand from supply chain customers, and small grids are generated in areas with low demand from supply chain customers. Calculate the business density and data variance within each grid, evaluate the rationality of the grid division, and optimize the grid. The divided grids are coded and labeled with attribute information for management.

4. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 3, characterized in that, The specific process of clustering algorithm and mesh optimization is as follows: Parameters are set according to the business scenario, including the neighborhood radius and minimum number of samples, which are determined through grid search and empirical values; Traverse the dataset. If the number of samples in the neighborhood of a point is greater than or equal to the minimum number of samples, mark it as a core point and create a new cluster. Expand the clustering based on the core point, add all points in its neighborhood to the cluster, and repeat this process until there are no new points to add. Points that are not core points and are not in any cluster are marked as noise points. Based on the clustering results, a grid is generated on the basis of each cluster. In clustering areas with high customer demand and dense data, neighboring clusters are merged to generate a large grid. In clustering areas with low customer demand and sparse data, small grids can be generated directly by merging individual clusters or small areas. Calculate the number of customers and total orders within each grid to assess the grid's rationality. If the business volume within a grid differs too much, readjust the neighborhood radius and minimum sample size parameters, and repeat the clustering and grid generation steps until the grid is reasonable.

5. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 1, characterized in that, The steps to determine the factors influencing supply chain demand in different regions are as follows: Collect historical data for each grid area, and draw histograms and heatmaps to visualize the temporal trends and spatial distribution differences in demand for each grid area. By calculating the mean, standard deviation, and correlation coefficient statistics, the relationships between variables are analyzed, and potential factors that may be related to changes in demand are screened out. We selected a multiple linear regression model to analyze the linear relationship between demand and various influencing factors, and calculated the degree of influence of each factor on changes in demand. Based on the calculation results output by the model, the results are sorted and the three groups of factors that have the highest impact on supply chain demand are selected. The influencing factors identified in the analysis will be applied to actual demand forecasting and inventory management. The factors identified included promotional policies, geographical location, and holidays.

6. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 5, characterized in that, The algorithm steps for determining the factors influencing regional supply chain demand are as follows: Calculate statistical measures to analyze the relationship between variables, calculate the mean of demand data for each grid area, representing the average level of demand in that area; measure the dispersion of demand data around the mean using the standard deviation, reflecting the magnitude of demand fluctuations; calculate the correlation coefficient between different variables to determine the strength and direction of the correlation, and screen out potential factors that may be related to changes in demand. The multiple linear regression model is based on the principle of least squares. It continuously adjusts the coefficients of each factor in the model to minimize the sum of squared errors between the model's predicted values ​​and the actual demand values, thereby obtaining the degree of influence of each factor on changes in demand.

7. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 1, characterized in that, The factors obtained in step S3 include promotional policy factors, geographical development factors, and holiday factors. Promotional policy factors are obtained through local websites and supply chain merchants; geographical development factors are obtained through digital maps to assess the transportation and building development of grid cities; and holiday factors are determined based on whether the day is a holiday. Real-time access to policies offered by different merchants in different regions, the level of development of the regions, and the impact of holidays on supply chain inventory; The values ​​of each factor are quantified, and the weight values ​​of different factors are obtained based on the analytic hierarchy process. The influence value is obtained through weighted calculation. When the impact is significant, the warehouse inventory should be replenished promptly. When the impact value is small, the amount of replenishment to the warehouse inventory is reduced, and the warehouse inventory is dynamically adjusted.

8. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 7, characterized in that, The steps for quantifying each factor value are as follows: Obtain promotional information and convert discount percentages into numerical values; Collect commercial density and traffic density data for each city grid, and calculate a weighted comprehensive score for geographical development. By integrating information from national statutory holidays and e-commerce shopping festivals, a holiday impact index is constructed. Holidays are divided into long holidays and short holidays, and different basic impact coefficients are assigned to each. The demand order growth rate of each holiday type is calculated to form a holiday impact index.

9. The supply chain demand forecasting and dynamic inventory optimization method based on big data according to claim 8, characterized in that, The formula for converting the discount percentage into a numerical value is as follows: Based on the formula, it can be concluded that the greater the discount and the closer the value is to 1, the stronger the pull effect on inventory demand, that is, the greater the demand on the supply chain. The steps for calculating the comprehensive score of geographical development level are as follows: Commercial density and traffic density are standardized using the following formulas: The regional commercial density is calculated as the number of shopping malls / shops in the region divided by the area of ​​the region; the regional traffic density is calculated as the length of the main road plus the number of subway stations × 3 plus the number of express delivery outlets × 2. The weighted composite score is calculated using the following formula: Geographical development score = Standardized commercial density value × w1 + Standardized traffic density value × w2 Where w1 and w2 are weight values, which are determined by the AHP method; The calculation steps for the holiday impact index are as follows: Long holidays are defined as statutory holidays of ≥3 days or major e-commerce promotions, while short holidays are defined as ordinary weekends. The formula for calculating the growth rate of demand orders is: The formula for calculating the order growth rate of the holiday impact index is: α is the basic impact coefficient, which is a pre-set coefficient based on the type of holiday. It is used to reflect the basic impact of different types of holidays on inventory demand and other aspects. The α value is different for different types of holidays. The basic impact coefficient is relatively high for long holidays and relatively low for short holidays. The holiday impact index is obtained by calculation.

Images