Big data-based intelligent supply chain management optimization method and system

Abstract

The present application relates to the technical field of sales volume prediction, and in particular to a big data-based intelligent supply chain management optimization method and system. First, time series data of sales volume of all products is obtained to avoid single product bias; sudden factors such as promotion can cause the time series data of sales volume to fluctuate suddenly, so the time series data is divided into multiple sales volume data segments, and the demand burst index of each sales volume data segment is quantified to represent short-term fluctuations; further cross-product correlation analysis of sales volume and demand burst index is performed, and periodic changes and other characteristics are fused to obtain a fluctuation sensitivity index of each sales volume data segment in each product, which is used to represent the adaptability of the sales data segment to fluctuation changes; finally, based on the index dynamics, the prediction weight is adaptively adjusted and the future sales volume of each product is predicted, so that the prediction is more in line with market changes, the accuracy and reliability are improved, and precise replenishment control is supported.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sales volume prediction, in particular to a big data-based intelligent supply chain management optimization method and system. BACKGROUND

[0002] With the intensification of global market competition and the diversification of consumer demand, the complexity and uncertainty of supply chain management have significantly increased. With the popularity of the Internet of Things, ERP systems and online transaction platforms, a large amount of sales data is continuously accumulated, forming rich big data resources. By integrating data mining, time series prediction and optimization scheduling methods, dynamic perception, prediction and optimization configuration of supply chain demand can be achieved, enhancing supply chain flexibility and collaborative efficiency.

[0003] In the prior art, replenishment in supply chain inventory management usually relies on prediction of historical sales volume data to estimate future demand for various products and develop a replenishment plan. However, the weight is often static during prediction. However, due to the variety of products in the supply chain, there are mutual influences between different products, and the consumption characteristics of products also differ at different times. Static weights often ignore the mutual influence between different products and short-term demand fluctuations in the market, resulting in a significant decrease in the credibility of the final prediction accuracy. SUMMARY

[0004] To solve the technical problem that the weight is often static during prediction, but due to the variety of products in the supply chain, there are mutual influences between different products, and the consumption characteristics of products also differ at different times, static weights often ignore the mutual influence between different products and short-term demand fluctuations in the market, resulting in a significant decrease in the credibility of the final prediction accuracy. The purpose of the present application is to provide a big data-based intelligent supply chain management optimization method and system. The technical solution adopted is as follows:

[0005] A big data-based intelligent supply chain management optimization method, comprising:

[0006] Obtaining sales volume time series data of multiple products;

[0007] Segmenting each sales volume time series data based on its trend characteristics to obtain multiple sales volume data segments. In each sales volume data segment of each product, analyze the difference characteristics between sales volume and predicted sales volume, as well as the change amplitude difference of sales volume between sales volume data segments, to determine the demand burst index of each sales volume data segment of each product;

[0008] The correlation characteristics between the time series data of the sales volume of different products, the difference between the demand burst indicators of the sales volume data segments, the periodic variation characteristics of the sales volume of each product, and the demand burst indicators of the sales volume data segments of each product are analyzed, so as to quantify the fluctuation sensitive indicators of each product in each sales volume data segment;

[0009] The preset weight is adjusted based on the fluctuation sensitive indicators of each sales volume data segment of each product, so as to predict the sales volume in the future period.

[0010] Further, the demand burst indicator acquisition method comprises:

[0011] The sales volume prediction data segment corresponding to each sales volume data segment of each product is acquired based on the ARIMA model;

[0012] The absolute value of the difference between the sales volume mean in each sales volume data segment of each product and the predicted sales volume mean in the corresponding sales volume prediction data segment is calculated as a first burst fluctuation factor;

[0013] In the sales volume time series data of each product, the absolute value of the difference between the sales volume mean in each sales volume data segment and the sales volume mean in the two most recent sales volume data segments is taken as a second burst fluctuation factor;

[0014] The product of the first burst fluctuation factor and the second burst fluctuation factor of each sales volume data segment of each product is normalized to obtain the demand burst indicator of each sales volume data segment of each product.

[0015] Further, the fluctuation sensitive indicator acquisition method comprises:

[0016] An optional product is taken as a target product, the correlation characteristics between the sales volume time series data of the target product and other products, and the difference between the demand burst indicators of the sales volume data segments are analyzed, and the consumption correlation degree between the target product and other products in each sales volume data segment is determined;

[0017] In the sales volume time series data of each product, the periodic variation characteristics of the sales volume are analyzed, and the consumption regularity indicators of each product are determined in combination with the demand burst indicators of the sales volume data segments of each product;

[0018] In each sales volume data segment of the target product, other products with a consumption correlation degree greater than a preset correlation threshold value with the target product are taken as associated products of the target product, and the consumption regularity indicators of the associated products are weighted and fused based on the consumption correlation degree between the target product and the associated products in each sales volume data segment, and the obtained weighted result is taken as the associated stable characteristic value of the target product in each sales volume data segment.

[0019] The product of the correlation stability characteristic value of the target product under each sales volume data segment and the consumption regularity index of the target product is negatively correlated and normalized, and the value is taken as the fluctuation sensitivity index of the target product under each sales volume data segment.

[0020] Further, the consumption correlation degree acquisition method comprises:

[0021] The overall correlation characteristics between the sales volume time series data between the target product and each other product are analyzed to determine the supply correlation factor between the target product and each other product;

[0022] In all sales volume data segments of the target product, any one is taken as a target data segment, and in the sales volume data segments of each other product, the sales volume data segment with the closest time distance to the time median of the target data segment is taken as the comparative data segment of the target data segment;

[0023] The absolute value of the difference between the demand burst index between the target data segment of the target product and the comparative data segment of each other product is negatively correlated and normalized, and the value is taken as the demand correlation factor between the target product and each other product under the target data segment;

[0024] The demand correlation factor between the target product and each other product under the target data segment is multiplied by the supply correlation factor between the target product and each other product, and the obtained product is normalized, and the value is taken as the consumption correlation degree between the target product and each other product under the target data segment.

[0025] Further, the supply correlation factor acquisition method comprises:

[0026] For any product, the sales volume time series data of the product is linearly fitted based on the least square method, and the slope value of the fitted straight line is obtained;

[0027] Between the target product and each other product, the absolute value of the difference between the slope values of the fitted straight lines of the sales volume time series data is negatively correlated and normalized to obtain the trend consistency factor between the target product and each other product;

[0028] The Pearson correlation coefficient between the sales volume time series data of the target product and each other product is normalized, and the value is taken as the contact factor between the target product and each other product;

[0029] The product of the contact factor and the trend consistency factor between the target product and each other product is normalized, and the value is taken as the supply correlation factor between the target product and each other product.

[0030] Further, the consumption regularity index acquisition method comprises:

[0031] In the time series data of the sales volume of each product, the frequency with the maximum amplitude in the frequency spectrum obtained based on Fourier transform is taken as the dominant frequency, the period is calculated according to the dominant frequency, and the sales volume time series data is divided into periods to obtain all period data segments;

[0032] The similarity between the period data segments of each product is analyzed to determine the period stability index of each product;

[0033] The value of the mean of the demand burst index of all sales volume data segments in each product after negative correlation mapping and normalization is multiplied by the period stability index of each product, and the value of the product after normalization is taken as the consumption regularity index of each product.

[0034] Further, the period stability index acquisition method comprises:

[0035] In the time series, the DTW value between each adjacent two period data segments is calculated and negatively correlated and normalized to obtain a period stability factor, and the mean of the period stability factors between all adjacent period data segments is taken as the period stability index of each product.

[0036] Further, the fluctuation sensitivity index of each sales volume data segment based on each product is adjusted to a preset weight to predict the sales volume in the future period, comprising:

[0037] Each sales volume data segment of each product is predicted based on an ARIMA model, and the obtained weight is taken as the preset weight;

[0038] The preset weight of each sales volume data segment is multiplied by the corresponding fluctuation sensitivity index to obtain the adaptive prediction weight of each sales volume data segment;

[0039] Based on the weighted moving average method and the adaptive prediction weight of the sales volume data segment, the sales volume prediction time series data of each product in the future period is obtained.

[0040] Further, the sales volume data segment acquisition method comprises:

[0041] The sales volume time series data of each product is segmented based on the Python ruptures library to obtain all sales volume data segments corresponding to each product.

[0042] A big data-based intelligent supply chain management optimization system comprises a processor and a memory, and the memory stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to realize the steps of the big data-based intelligent supply chain management optimization method.

[0043] The present application has the following advantages:

[0044] Firstly, the sales time series data of various products are acquired to realize full-category coverage and avoid supply chain imbalance caused by prediction deviation of a single category; since the sales of products are often affected by sudden factors such as promotion activities, the sales time series data of each product are divided into multiple sales data segments based on trend characteristics, and the change characteristics of sales in each sales data segment are analyzed to quantify the demand burst index, which is used to represent the short-term fluctuation of product sales and is helpful for subsequent more accurate prediction and replenishment control. Further, since there are various products in the supply chain, there may be dependency relationships such as functional coupling or structural matching between different products, and the consumption stability characteristics exhibited by each product in the sales process can also be used as a reference for subsequent sales prediction and replenishment control, so the correlation analysis of sales and demand burst indexes across products is performed, and the periodic change characteristics of product sales and the numerical characteristics of demand burst indexes are fused to solve the precision loss caused by isolated prediction, obtain the fluctuation sensitivity index of each product in each sales data segment, and help accurately identify the true consumption characteristics of the product; finally, the adaptive allocation of prediction weights is realized based on the fluctuation sensitivity index, so as to predict the sales of each product in the future period, which can make the final prediction result more consistent with the real market changes and effectively improve the prediction accuracy and credibility. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0046] Figure 1 A method flowchart of a big data-based intelligent supply chain management optimization method provided by an embodiment of the present application;

[0047] Figure 2 A sales time series data schematic diagram of a certain product provided by an embodiment of the present application;

[0048] Figure 3A method flow chart of a fluctuation sensitive index acquisition method provided by an embodiment of the present application;

[0049] Figure 4 A system block diagram of a big data based smart supply chain management optimization system provided by an embodiment of the present application;

[0050] Figure 5 A system structure schematic diagram of a big data based smart supply chain management optimization system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0051] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined purposes, the following describes in detail the specific implementation, structure, features and effects of a big data based smart supply chain management optimization method and system according to the present application, with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0053] The following specifically describes the specific scheme of a big data based smart supply chain management optimization method and system provided by the present application, with reference to the accompanying drawings.

[0054] Please refer to Figure 1 which shows a method flow chart of a big data based smart supply chain management optimization method provided by an embodiment of the present application, and the method includes the following steps:

[0055] Step S1: acquiring time series data of sales volume of multiple products.

[0056] In modern manufacturing and circulation system, the supply chain link is highly complex, the demand fluctuation is frequent and the uncertainty is strong. In order to ensure that the supply meets the market demand, the future sales state within a period of time is usually predicted based on the past sales of products, that is, the future demand of various products is calculated, and then the product replenishment plan is made accordingly. Therefore, in the embodiment of the present application, the time series data of sales volume of multiple products on the supply chain can be collected based on the Internet of Things, ERP system and other online transaction platforms. Specifically, the daily sales volume of various products of an enterprise within the past month can be collected through the sales terminal system of the enterprise, so as to draw the time series data of sales volume of each product with time / day as the horizontal axis and sales volume as the vertical axis. Please refer to Figure 2Fig. 1 shows a schematic view of time series data of sales volume of a certain product in an embodiment of the present application; and product IDs of various products are also needed, which are used to distinguish product categories.

[0057] It should be noted that when drawing the time series data of sales volume of each product, the data can be preprocessed, such as missing value filling (using forward filling method), and the pre-processing process is a known technology and will not be repeated here.

[0058] Step S2: segmenting each time series data of sales volume based on its trend characteristics to obtain multiple sales volume data segments; in each sales volume data segment of each product, analyzing the difference characteristics between sales volume and predicted sales volume, and the change amplitude difference of sales volume between sales volume data segments, to determine the demand burst indicator of each sales volume data segment of each product.

[0059] In the process of predicting product demand based on product sales volume time series data to replenish inventory and realize dynamic perception of the whole supply chain, since traditional supply chain prediction often regards sales data as a continuous homogeneous sequence, it often ignores the inherent periodic fluctuations, such as promotion activities, seasonal changes, temporary project start or customer concentrated procurement, etc. For these products with short-term sudden fluctuations, the long-term demand trend of their sales volume time series data cannot reflect their real market demand, so in the embodiment of the present application, it is necessary to analyze the short-term sudden fluctuation of each product, which helps to avoid the misjudgment of demand trend by the final prediction model, so as to realize more accurate and stable sales prediction and replenishment control.

[0060] Firstly, the trend segmentation method can be used to segment and divide each product's sales volume time series data based on its trend characteristics to obtain multiple sales volume data segments. At this time, each sales volume data segment will have a more explicit trend characteristic, which will also be more convenient for subsequent analysis of short-term mutations.

[0061] Preferably, in an embodiment of the present application, the method for obtaining sales volume data segments comprises:

[0062] The Python ruptures library can realize automatic trend segmentation, which can automatically identify trend mutation points (such as slope change, volatility change, platform switching, etc.) in time series data, and does not need to preset the number of segments, but automatically determines the optimal segmentation point through optimization algorithm, so in this embodiment of the present application, the Python ruptures library is used to segment each product's sales volume time series data to obtain all sales volume data segments corresponding to each product.

[0063] It should be noted that the process of Python ruptures library realizing automatic segmentation is a known technology and will not be repeated here.

[0064] At this point, the sales volume time series data of each product can be segmented to obtain the sales volume data segment. Since the short-term sudden fluctuation of sales volume can be characterized by the change of sales volume, in the embodiment of the present application, the difference between each sales volume data segment of each product and the predicted sales volume is analyzed to characterize the deviation degree between the actual sales volume and the predicted state of the product, which is used to measure the mutation characteristics. At the same time, the change amplitude difference of the sales volume between each sales volume data segment of each product is analyzed to quantify the mutation degree of the actual sales volume in time sequence. The two are combined to obtain a comprehensive index that can measure the short-term sudden fluctuation of each sales volume data segment of each product, i.e., the demand burst index.

[0065] Preferably, in an embodiment of the present application, the demand burst index acquisition method comprises:

[0066] The sales volume prediction data segment corresponding to each sales volume data segment of each product can be obtained based on the ARIMA model.

[0067] The absolute value of the difference between the sales volume mean value in each sales volume data segment of each product and the predicted sales volume mean value in the corresponding sales volume prediction data segment is calculated as the first burst fluctuation factor. The greater the first burst fluctuation factor, the greater the deviation degree between the average level of the actual sales volume and the average level of the predicted sales volume in the sales volume data segment of the product. The greater the deviation degree, the greater the change characteristics of the sales volume data in the actual sales scenario, which is considered as the greater the degree of short-term sudden fluctuation.

[0068] Then, the short-term sudden fluctuation can be analyzed from another aspect, i.e., the numerical value change difference of the actual sales volume in time sequence: in the sales volume time series data of each product, the absolute value of the difference between the sales volume mean value in each sales volume data segment and the sales volume mean value in the two nearest sales volume data segments in time sequence is taken as the second burst fluctuation factor. The greater the second burst fluctuation factor, the more prominent the numerical value change characteristics of the sales volume in the sales volume data segment compared to the adjacent nearest sales volume data segments, and thus the greater the degree of short-term sudden fluctuation of the sales volume data segment.

[0069] Finally, based on the foregoing analysis, the first burst fluctuation factor and the second burst fluctuation factor of each sales volume data segment are positively correlated with the degree of short-term burst fluctuation of the sales volume data segment, so in this case, the product demand burst index of each sales volume data segment of each product is the normalized value of the product of the first burst fluctuation factor and the second burst fluctuation factor of each sales volume data segment of each product. The greater the demand burst index, the greater the degree of burst change in the sales volume in the sales volume data segment. The normalization is a technique well known to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization. The specific normalization method is not limited here.

[0070] It should be noted that the process of obtaining predicted data by the ARIMA model is a known technique, and the specific process will not be described here.

[0071] Step S3: Analyze the correlation characteristics between the sales volume time series data of different products, the differences between the demand burst indexes of the sales volume data segments, the periodic variation characteristics of the sales volume of each product, and the demand burst indexes of the sales volume data segments of each product, thereby quantifying the fluctuation sensitive index of each product in each sales volume data segment.

[0072] There are many types of products in the supply chain, and the consumption rhythm, fluctuation amplitude, replenishment cycle, and demand driving factors of each product in actual operation are significantly different. If a unified prediction and replenishment strategy is used, it is easy to lead to insufficient inventory of high demand fluctuation products and inventory accumulation of low demand fluctuation products. Therefore, in the embodiment of the present application, it is considered necessary to deeply understand the consumption law of various products to realize differentiated prediction of products, so it is necessary to analyze the consumption law of the product and identify its consumption characteristics in the supply chain.

[0073] In actual sales scenarios, the consumption of some products often does not occur in isolation, but synchronously with other products that have functional coupling or structural matching relationship. By analyzing the consumption law of the matching products that have a use dependency or correlation with the product, it can be determined whether the demand burst fluctuation of the product is part of the systemic demand. At the same time, the consumption characteristics of each product can also be analyzed separately: when a certain product shows stable consumption and strong periodicity characteristics in the long-term sales process, it means that it has a clear and predictable consumption law. The demand change of such product is less affected by external interference, has consistent rhythm, and has high regularity and controllability. Therefore, even if there is a burst demand fluctuation, the impact on the inventory replenishment of the product will be small.

[0074] Therefore, in the step, the sales volume between products is jointly consumed and analyzed, the difference characteristics between the demand burst indexes of the sales volume data segments are considered, and the periodic change characteristics of the product sales volume and the numerical characteristics of the demand burst indexes of the sales volume data segments are combined, so as to quantify the fluctuation sensitive indexes of each product in each sales volume data segment, which is helpful to accurately identify the real consumption characteristics of the product.

[0075] Preferably, in an embodiment of the present application, the acquisition method of the fluctuation sensitive index comprises:

[0076] Referring to Figure 3 , a method flowchart of the acquisition method of the fluctuation sensitive index in an embodiment of the present application is shown, and the method comprises the following steps:

[0077] Step S301: Optionally, one product is selected as a target product, the correlation characteristics between the sales volume time series data of the target product and other products and the differences between the demand burst indexes of the sales volume data segments are analyzed, and the consumption correlation degree between the target product and other products in each sales volume data segment is determined.

[0078] For the convenience of explanation and description, one product is selected as a target product in all products.

[0079] Between the target product and each of the other products, the overall correlation characteristics between the sales volume time series data can be analyzed first: for any product, the sales volume time series data of the product are linearly fitted based on the least square method, and the slope value of the fitted straight line is obtained, which can represent the long-term sales trend of the sales volume of the product. Between the target product and each of the other products, the difference absolute value between the slope values of the fitted straight lines of the sales volume time series data is calculated. The smaller the difference absolute value is, the more consistent the long-term trends of the sales volume between the target product and each of the other products are. Therefore, the difference absolute value is negatively correlated and normalized to correct the logical relationship, and a trend consistency factor between the target product and each of the other products is obtained. The greater the trend consistency factor is, the higher the correlation degree of the sales volume between the two products is, and the more likely the consumption process is to be linked in the actual scene. The negative correlation mapping and normalization processing here can adopt the formula , wherein, represents an exponential function with the natural constant e as the base, and x represents the independent variable.

[0080] Then the Pearson correlation coefficient between the sales time series data of the target product and each of the other products is calculated. The closer the Pearson correlation coefficient is to 1, the greater the correlation between the sales of the two products, and the greater the correlation. Therefore, the normalized value of the Pearson correlation coefficient is used as the correlation factor between the target product and each of the other products. The greater the correlation factor, the higher the supply correlation between the two products. Since the Pearson correlation coefficient can be positive or negative, the normalization here can use the abs function.

[0081] The product of the correlation factor between the target product and each of the other products and the trend consistency factor is normalized to obtain the supply correlation factor between the target product and each of the other products. Based on the foregoing analysis, the greater the supply correlation factor, the higher the supply correlation between the two products.

[0082] When the supply correlation between two products is high and the degree of mutation in sales during demand surges is consistent, it is more likely that the two products are used simultaneously in actual applications, and their consumption is not independent but interdependent. Therefore, the difference between the demand surge indicators of the sales data segments of the products can be analyzed in detail to quantify the consumption dependence between the two products in the local sales data segments.

[0083] Since the segmentation of the sales time series data of different products may not be consistent, the sales data segments between the products need to be aligned first. In the sales data segments of the target product, any one is selected as the target data segment. In the sales data segments of each of the other products, the sales data segment with the closest time distance to the time median of the target data segment is selected as the comparison data segment of the target data segment. For example, if a sales data segment corresponds to a time series of 10-20, the time median is 15. By comparing the time medians of the sales data segments, the alignment can be achieved, which can make the sales data segments overlap as much as possible, thereby more accurately capturing the data features.

[0084] ​Then, an absolute value of a difference between the demand burst index of the target data segment of the target product and the comparative data segment under each of the other products is calculated. The smaller the absolute value of the difference is, the more consistent the sales fluctuation characteristics of the two products are under the condition of a burst demand, and thus the higher the consumption dependency between the two products is. Therefore, the absolute value of the difference is negatively correlated and normalized to correct the logical relationship, so as to obtain a demand correlation factor between the target product and each of the other products under the target data segment. The greater the demand correlation factor is, the higher the consumption dependency between the target product and each of the other products under the target data segment is, and the stronger the correlation is. The negatively correlated and normalized processing can be performed by using the formula wherein, represents an exponential function with a natural constant e as a base, and x represents an independent variable.

[0085] Finally, the demand correlation factor between the target product and each of the other products under the target data segment is multiplied by the supply correlation factor between the target product and each of the other products, and the normalized value of the obtained product is taken as a consumption correlation degree between the target product and each of the other products under the target data segment. Based on the foregoing logic and analysis, the greater the consumption correlation degree is, the higher the consumption dependency between the two products is, that is, the higher the correlation degree between the sales is, and there is a certain mutual influence relationship between the two products. The normalization is a technology known to those skilled in the art, and the normalization function can be linear normalization or standard normalization. The specific normalization method is not limited herein.

[0086] Thus, the consumption correlation degree between each product and the remaining products under each sales data segment can be obtained.

[0087] Step S302: In the sales time series data of each product, the periodic change characteristics of the sales are analyzed, and the demand burst index of each sales data segment of each product is combined to determine the consumption rule index of each product.

[0088] In the sales time series data of each product, the maximum amplitude frequency in the frequency spectrum is obtained as a main frequency based on Fourier transform, and the period (T) is calculated according to the main frequency. wherein, represents the period, represents the frequency.

[0089] The similarity between the period data segments of each product is analyzed to determine the period stability index of each product: in time sequence, the DTW value between each adjacent two period data segments is calculated, the smaller the DTW value, the higher the similarity between the sales data of the adjacent two period data segments, that is, the change of sales in the two adjacent period data segments tends to be consistent, and the demand period is relatively stable, so the DTW value is negatively correlated and normalized to correct the logical relationship, and the period stability factor is obtained, and the average of the period stability factors between all adjacent period data segments is taken as the period stability index of each product. The greater the period stability index, the better the overall change stability of the sales time sequence data of the product, and the more likely the product has a more regular consumption characteristic.

[0090] Based on the analysis in step S2, the greater the demand burst index of the sales data segment of the product, the greater the degree of sudden change of sales, and then the regularity of consumption should be reduced. Therefore, the average of the demand burst index of all sales data segments of each product is negatively correlated and normalized, and the value is multiplied by the period stability index of each product, and the normalized value of the product is taken as the consumption regularity index of each product. At this time, the greater the consumption regularity index, the smaller the degree of sudden demand fluctuation of the product in the long-term sales process, and the stronger the regularity of consumption, so the external interference is smaller, and the influence on the subsequent prediction result is also smaller. The negative correlation mapping and normalization processing can use the formula wherein, represents an exponential function with natural constant e as the base, and x represents the independent variable. The selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited here.

[0091] It should be noted that the calculation process of Fourier transform and DTW value is a known technology, and the specific process is not repeated here.

[0092] Step S303: Based on the numerical characteristics of the consumption correlation degree between the target product and other products in each sales data segment, the associated products are screened out, and the associated stability characteristic value of the target product in each sales data segment is determined based on the consumption regularity index of the associated products and the consumption correlation degree between the associated products and the target product.

[0093] Based on the calculation in step S301, the consumption correlation degree between the target product and each product in each sales data segment of the target product is obtained, and then in each sales data segment of the target product, the other products with a consumption correlation degree greater than a preset correlation threshold value are taken as the associated products of the target product.

[0094] When the consumption regularity of the associated product of the target product is stronger, it can be explained from the side that the consumption regularity of the target product will be stronger, so here, the consumption regularity index of the associated product is weighted and fused by using the consumption correlation degree between the target product and the associated product under each sales data segment, that is, the consumption correlation degree between the target product and each associated product under each sales data segment is multiplied by the consumption regularity index of each associated product, and finally the sum of the products obtained between the target product and all associated products under each sales data segment is taken as the associated stability characteristic value of the target product under each sales data segment. The greater the associated stability characteristic value, the higher the regularity degree of the consumption characteristics of the target product under the sales data segment.

[0095] It should be noted that the associated stability characteristic value should be a normalized value, and the specific normalization method is a technical means familiar to those skilled in the art, which is not limited and elaborated here; the preset association threshold is 0.7, and the specific value can be adjusted according to the implementation scene, which is not limited here.

[0096] Step S304: The associated stability characteristic value of the target product under each sales data segment and the consumption regularity index of the target product are fused to determine the fluctuation sensitive index of the target product under each sales data segment.

[0097] If the consumption regularity of a certain product itself is stronger, and the consumption stability represented by the associated product is higher, it can be considered that the short-term sudden demand fluctuation of the product has less influence on the prediction of future sales, so the associated stability characteristic value of the target product under each sales data segment is multiplied by the consumption regularity index of the target product. The greater the product, the stronger the sales regularity of the target product, which has strong long-term trend characteristics and can better adapt to short-term sudden fluctuations, so the product is negatively correlated and normalized to be used as the fluctuation sensitive index of the target product under each sales data segment. The greater the fluctuation sensitive index, the greater the influence of the sudden fluctuation in the sales data segment of the target product on the future prediction. The negative correlation mapping and normalization processing here can adopt the formula wherein, represents an exponential function with natural constant e as the base, and x represents the independent variable. The selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited here.

[0098] At this point, the fluctuation sensitive index of each sales data segment of each product can be obtained.

[0099] Step S4: Adjust the preset weight based on the fluctuation sensitive index of each sales data segment of each product to predict the sales in the future period.

[0100] The fluctuation sensitive index of each sales volume data segment of each product represents the sensitivity of the sales of the sales volume data segment to fluctuation. Since the traditional prediction model usually relies on the average characteristics of the data and cannot capture the short-term sudden fluctuation in the sales data, in this step, the preset weight can be adjusted based on the fluctuation sensitive index of each sales volume data segment of each product, so as to predict the sales volume in the future period, and ensure that the prediction model can pay more attention to the real sales state of the market.

[0101] Preferably, in an embodiment of the present application, the preset weight is adjusted based on the fluctuation sensitive index of each sales volume data segment of each product, so as to predict the sales volume in the future period, comprising:

[0102] First, the sales volume data segment of each product is predicted based on the traditional ARIMA model, and the obtained weight is taken as the preset weight.

[0103] Since the greater the fluctuation sensitive index, the greater the impact of the sudden fluctuation in the sales volume data segment of the target product on the future prediction, the preset weight of each sales volume data segment is multiplied by the corresponding fluctuation sensitive index to obtain the adaptive prediction weight of each sales volume data segment, so as to improve the weight of the data segment with high fluctuation sensitivity and reduce the weight of the data segment with low fluctuation sensitivity in the adaptive adjustment process.

[0104] Finally, under each product, the sales volume prediction time series data of each product in the future period (the acquisition of the period is recorded in step S3) is obtained based on the weighted moving average method and the adaptive prediction weight of the sales volume data segment.

[0105] The sales volume prediction time series data of each product in the future period obtained at this time can more accurately reflect the demand trend of the current product, so the replenishment quantity can be calculated based on the comparison between the sales volume prediction time series data and the current actual inventory, combined with the preset safety stock, which can ensure that the inventory of the product can meet the sales demand in the future period while effectively avoiding excessive accumulation.

[0106] It should be noted that the ARIMA model and the weighted moving average method are both known technologies, and the specific process is not described here. The preset safety stock of each product can be calculated based on the known safety stock formula, which is not described here.

[0107] In order to facilitate calculation, all index data involved in the calculation in the embodiment of the present application are subjected to data preprocessing, and then the dimensional influence is cancelled. The means for removing the dimensional influence is a technology known to those skilled in the art, which is not limited here.

[0108] In summary, firstly, the sales time series data of multiple products are acquired to realize full-category coverage and avoid supply chain imbalance caused by prediction deviation of a single category; since the sales of products are often affected by sudden factors such as promotion activities, the sales time series data of each product is divided into multiple sales data segments based on trend characteristics, and the change characteristics of sales in each sales data segment are analyzed to quantify the demand burst index, which is used to represent the short-term fluctuation of product sales and is helpful for subsequent more accurate prediction and replenishment control. Further, since there are various products in the supply chain, there may be dependency relationships such as functional coupling or structural matching between different products, and the consumption stability characteristics exhibited by each product during the sales process can also be used as a reference for subsequent sales prediction and replenishment control, so the correlation analysis of sales and demand burst indexes across products is performed, and the periodic variation characteristics of product sales and the numerical characteristics of demand burst indexes are fused to solve the precision loss caused by isolated prediction, obtain the fluctuation sensitive index of each product in each sales data segment, and help accurately identify the true consumption characteristics of the product; finally, the adaptive allocation of prediction weights is realized based on the fluctuation sensitive index to predict the sales of each product in the future period, which can make the final prediction result more consistent with the real market changes and effectively improve the prediction accuracy and reliability.

[0109] The embodiment of the present application also provides a big data-based intelligent supply chain management optimization system, please refer to Figure 4 which shows a system block diagram, including a data acquisition module 401 for realizing step S1 in the above method embodiment; a sudden demand analysis module 402 for realizing step S2 in the above method embodiment; a fluctuation analysis module 403 for realizing step S3 in the above method embodiment; and a sales prediction module 404 for realizing step S4 in the above method embodiment.

[0110] It should be noted that the system provided in the above embodiments is only exemplified by the division of the above functional modules, and in actual application, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the computer device is divided into different functional modules to complete all or part of the above described functions. In addition, the big data-based intelligent supply chain management optimization system and the big data-based intelligent supply chain management optimization method embodiment provided in the above embodiments belong to the same concept, and the specific implementation process is described in detail in the method embodiment, which will not be repeated here.

[0111] Please refer to Figure 5Fig. 1 shows a system structure schematic diagram of a big data-based intelligent supply chain management optimization system according to an embodiment of the present application, which comprises a processor 500, a memory 501, a bus 502 and a communication interface 503, the processor 500, the communication interface 503 and the memory 501 being connected through the bus 502; wherein the memory 501 can contain a high-speed random access memory, the bus 502 can be an ISA bus, a PCI bus or an EISA bus, etc., the processor 500 can be an integrated circuit chip with signal processing capability; the memory 501 stores at least one instruction, at least one program, a code set or an instruction set, which are loaded and executed by the processor to implement the steps of a big data-based intelligent supply chain management optimization method.

[0112] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0113] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. A smart supply chain management optimization method based on big data, characterized in that, The method includes: Obtain time-series sales data for multiple products; Based on the trend characteristics of each sales volume time series data, the data is segmented to obtain multiple sales volume data segments; within each product's sales volume data segment, the differences between the sales volume and the predicted sales volume, as well as the differences in the magnitude of sales volume changes between sales volume data segments, are analyzed to determine the demand surge index for each sales volume data segment of each product. By analyzing the correlation characteristics between time-series sales volume data of different products, the differences between demand surge indicators in sales volume data segments, the periodic change characteristics of sales volume of each product, and the demand surge indicators in sales volume data segments of each product, the volatility sensitivity indicators of each product in each sales volume data segment are quantified. The preset weights are adjusted based on the volatility sensitivity indicators of each sales volume data segment for each product, thereby predicting the sales volume in future periods. The methods for obtaining the demand surge indicator include: Based on the ARIMA model, obtain the sales volume prediction data segment corresponding to each sales volume data segment for each product; Calculate the absolute value of the difference between the average sales volume in each sales volume data segment for each product and the average predicted sales volume in the corresponding sales volume prediction data segment, and use this as the first sudden fluctuation factor; In the time series data of sales volume for each product, the absolute value of the difference between the average sales volume in each sales volume data segment and the average sales volume of the two most recent sales volume data segments is used as the second sudden fluctuation factor. The normalized value of the product of the first sudden fluctuation factor and the second sudden fluctuation factor for each sales volume data segment of each product is used as the demand suddenness indicator for each sales volume data segment of each product. The method for obtaining the volatility-sensitive indicator includes: Choose one product as the target product, analyze the correlation characteristics between the sales volume time series data of the target product and other products, as well as the differences between the demand surge indicators of the sales volume data segments, and determine the consumption correlation between the target product and other products in each sales volume data segment. In the time series data of sales volume for each product, analyze the periodic variation characteristics of sales volume, and combine it with the demand surge indicators of each product's sales volume data segment to determine the consumption pattern indicators of each product. In each sales volume data segment of the target product, other products with a consumption correlation greater than a preset correlation threshold are considered as related products of the target product. The consumption correlation between the target product and related products in each sales volume data segment is used to weight and fuse the consumption pattern indicators of related products. The weighted result is used as the correlation stable feature value of the target product in each sales volume data segment. The product of the stable correlation characteristic value of the target product under each sales volume data segment and the consumption pattern indicator of the target product is negatively correlated and normalized, and then used as the fluctuation-sensitive indicator of the target product under each sales volume data segment.

2. The method for optimizing intelligent supply chain management based on big data according to claim 1, characterized in that, The method for obtaining the consumption correlation degree includes: Analyze the overall correlation characteristics between the target product and each other product's sales volume time-series data to determine the supply correlation factors between the target product and each other product; Select one of the sales volume data segments for the target product as the target data segment. Among the sales volume data segments for each other product, select the sales volume data segment whose median time is closest to the median time of the target data segment as the comparison data segment for the target data segment. The absolute value of the difference between the demand surge index of the target product and the comparative data segment under each other product is negatively correlated and normalized, and then used as the demand correlation factor between the target product and each other product under the target data segment. Multiply the demand correlation factor between the target product and each other product in the target data segment by the supply correlation factor between the target product and each other product, and then normalize the product to obtain the consumption correlation degree between the target product and each other product in the target data segment.

3. The method for optimizing intelligent supply chain management based on big data according to claim 2, characterized in that, The method for obtaining the supply-related factors includes: For any product, the sales volume time series data of the product is fitted with a straight line based on the least squares method, and the slope value of the fitted line is obtained. Between the target product and each other product, the absolute value of the difference between the slope values ​​of the fitted straight lines of the sales volume time series data is negatively correlated and normalized to obtain the trend consistency factor between the target product and each other product. The normalized value of the Pearson correlation coefficient between the target product and the time series data of the sales volume of each other product is used as the correlation factor between the target product and each other product. The normalized value of the product of the correlation factor between the target product and each other product and the trend consistency factor is used as the supply correlation factor between the target product and each other product.

4. The intelligent supply chain management optimization method based on big data according to claim 1, characterized in that, The methods for obtaining the consumption pattern indicators include: In the time series data of sales volume for each product, the frequency with the largest amplitude in the spectrum is obtained based on Fourier transform as the main frequency. The period is calculated based on the main frequency and the time series data of sales volume is divided into periods to obtain all periodic data segments. Analyze the similarities between the cyclical data segments for each product to determine the cyclical stability index for each product; The mean of the demand surge index for all sales volume data segments of each product is negatively correlated and normalized. This normalized value is then multiplied by the cyclical stability index for each product, and the resulting product is normalized again. This normalized product is then used as the consumption pattern index for each product.

5. The intelligent supply chain management optimization method based on big data according to claim 4, characterized in that, The method for obtaining the periodic stability index includes: In terms of time series, the DTW value between every two adjacent periodic data segments is calculated and negative correlation mapping and normalization are performed to obtain the periodic stability factor. The mean of the periodic stability factor between all adjacent periodic data segments is used as the periodic stability index for each product.

6. The intelligent supply chain management optimization method based on big data according to claim 4, characterized in that, The method of adjusting preset weights based on fluctuation-sensitive indicators for each sales volume data segment under each product, thereby predicting sales volume in future periods, includes: Based on the ARIMA model, the sales volume data segment for each product is predicted, and the resulting weights are used as preset weights. The preset weight of each sales volume data segment is multiplied by the corresponding volatility-sensitive indicator to obtain the adaptive prediction weight of each sales volume data segment. For each product, based on the weighted moving average method and the adaptive prediction weight of the sales volume data segment, the time series data of the sales volume forecast for the next period of each product is obtained.

7. The method for optimizing intelligent supply chain management based on big data according to claim 1, characterized in that, The method for obtaining the sales volume data segment includes: The Python-based ruptures library segments the time-series sales data for each product, thus obtaining all sales data segments corresponding to each product.

8. A smart supply chain management optimization system based on big data, characterized in that, It includes a processor and a memory, the memory storing at least one instruction, at least one program, code set or instruction set, and when the at least one instruction, at least one program, code set or instruction set is loaded and executed by the processor, it implements the steps of the smart supply chain management optimization method based on big data as described in any one of claims 1-7.

Images