Method and device for determining product replenishment quantity, and computer storable medium

Abstract

The present disclosure relates to a method and apparatus for determining product replenishment quantities. The method for determining product replenishment quantities includes: obtaining historical data of multiple reference products related to a target product; determining attribute data of product attributes of each reference product based on its historical data, wherein the product attributes include demand attributes, and the attribute data of the demand attributes include the historical demand of the reference product; clustering the multiple reference products based on the attribute data of the product attributes of each reference product and target values ​​of the attribute weights to obtain multiple reference classes, wherein the attribute weights include demand attribute weights; screening reference classes similar to the target product from the multiple reference classes as similar classes; predicting the future demand of the target product in a future cycle based on the attribute data of the demand attributes of each reference product in the similar class; and determining the future replenishment quantity of the target product in the future cycle based on the future demand of the target product.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a method and device for determining product replenishment quantity, and a computer storable medium. Background Art

[0002] In related technologies, the future sales volume of the target product in the future period is predicted by using the historical sales volume data of reference products related to the target product, and the future replenishment quantity is determined based on the predicted future sales volume. Summary of the Invention

[0003] In related technologies, sales volume can reflect demand to a certain extent. However, in special circumstances such as inventory shortages, sales volume cannot be equated with demand, and replenishment volume is closely related to demand. Therefore, the replenishment volume determined by predicting sales volume will deviate to a certain extent from the actual situation, and thus the storage resources cannot be fully and reasonably utilized.

[0004] In response to the above technical problems, the present disclosure proposes a solution that can improve the accuracy of replenishment and increase the utilization rate of warehouse resources.

[0005] According to a first aspect of the present disclosure, a method for determining a replenishment quantity of a product is provided, comprising: acquiring historical data of a plurality of reference products related to a target product; determining, for each reference product, attribute data of a product attribute of the reference product based on its historical data, wherein the product attribute includes a demand attribute, and the attribute data of the demand attribute includes a historical demand quantity of the reference product; clustering the plurality of reference products based on the attribute data of the product attributes of the respective reference products and target values ​​of attribute weights to obtain a plurality of reference classes, wherein the attribute weights include demand attribute weights; screening, from the plurality of reference classes, reference classes that are similar to the target product as similar classes; predicting a future demand quantity of the target product in a future period based on the attribute data of the demand attributes of the respective reference products in the similar classes; and determining a future replenishment quantity of the target product in the future period based on the future demand quantity of the target product.

[0006] In some embodiments, the product attributes further include other attributes besides the demand attributes, the attribute weights further include other attribute weights besides the demand attribute weights, and the target value of the demand attribute weights is greater than the target value of the other attribute weights.

[0007] In some embodiments, the method for determining product replenishment quantity further includes: before clustering the multiple reference products, obtaining the target value of the demand attribute weight and the initial value of the other attribute weight; determining the target value of the other attribute weight based on the target value of the demand attribute weight and the initial value of the other attribute weight, the target value of the other attribute weight is negatively correlated with the target value of the demand attribute weight, and positively correlated with the initial value of the other attribute weight.

[0008] In some embodiments, determining the target value of the other attribute weight based on the target value of the requirement attribute weight and the initial value of the other attribute weight includes: determining the difference between the maximum value of the target value of the requirement attribute weight and the target value of the requirement attribute weight; and determining the target value of the other attribute weight based on the product of the difference and the initial value of the other attribute weight.

[0009] In some embodiments, the product attributes further include attributes other than demand attributes, and the attribute weights further include attribute weights other than demand attribute weights. Clustering the multiple reference products to obtain multiple reference classes includes: for every two reference products, determining the distance between the two reference products in terms of demand attributes based on the attribute data of the corresponding demand attributes, characterizing the degree of similarity between the two reference products in terms of demand attributes; determining the distance penalty item corresponding to the two reference products based on the attribute data of the other attributes of the two reference products, the distance penalty item characterizing the degree of influence of the other attributes on the distance between the two reference products in terms of demand attributes; for every two reference products, determining the composite distance between the two reference products in terms of the product attributes based on the distance between the two reference products in terms of demand attributes and the corresponding distance penalty item; clustering the multiple reference products based on the composite distance between every two reference products in the multiple reference products to obtain multiple reference classes.

[0010] In some embodiments, the product attributes of each reference product include quantitative attributes and qualitative attributes, the quantitative attributes include the demand attributes and other quantitative attributes, the distance penalty item includes a first distance penalty item and a second distance penalty item, the first distance penalty item represents the degree of influence of other quantitative attributes on the distance between each two reference products in terms of demand attributes, and the second distance penalty item represents the degree of influence of qualitative attributes on the distance between each two reference products in terms of demand attributes.

[0011] In some embodiments, determining the distance penalty item corresponding to each two reference products includes: determining the first distance penalty item corresponding to each two reference products based on the attribute data of other quantitative attributes of each two reference products and the target values ​​of the corresponding attribute weights, and the target values ​​of the corresponding demand attribute weights; and / or determining the second distance penalty item corresponding to each two reference products based on the attribute data of the qualitative attributes of each two reference products and the target values ​​of the corresponding attribute weights, and the target values ​​of the corresponding demand attribute weights.

[0012] In some embodiments, determining the distance between each two reference products in terms of demand attributes includes: determining the difference in historical demand for each two reference products at each historical moment as the demand difference at each historical moment; and determining the distance between each two reference products in terms of demand attributes based on the square root of the sum of the squares of the demand differences for each two reference products at each historical moment.

[0013] In some embodiments, determining the first distance penalty item corresponding to each two reference products includes: determining the distance between each two reference products in terms of each other quantitative attribute based on the attribute data of each other quantitative attribute of each two reference products; determining the attribute weight ratio of each other quantitative attribute relative to the demand attribute based on the attribute weight of each other quantitative attribute and the attribute weight ratio of the demand attribute; determining the first distance penalty item corresponding to each two reference products based on the distance between each two reference products in terms of each other quantitative attribute and the attribute weight ratio of each other quantitative attribute relative to the demand attribute, the first distance penalty item being positively correlated with the distance between each two reference products in terms of each other quantitative attribute and the attribute weight ratio of each other quantitative attribute relative to the demand attribute.

[0014] In some embodiments, determining the first distance penalty item corresponding to each two reference products based on the distance between each two reference products in each other quantitative attribute and the attribute weight ratio of each other quantitative attribute relative to the demand attribute includes: using the attribute weight ratio of each other quantitative attribute of each two reference products relative to the demand attribute, performing a weighted operation on the distance between each two reference products in each other quantitative attribute to obtain the first distance penalty item corresponding to each two reference products.

[0015] In some embodiments, the attribute value of the qualitative attribute does not change over time, and determining the second distance penalty item corresponding to each two reference products includes: when the attribute value of each qualitative attribute of each two reference products is different, determining the distance between each two reference products in terms of each qualitative attribute as the product of a first value and the distance between each two reference products in terms of the demand attribute; when the attribute value of each qualitative attribute of each two reference products is the same, determining the distance between each two reference products in terms of each qualitative attribute as the product of a second value and the distance between each two reference products in terms of the demand attribute, the second value being less than the first value; determining the attribute weight ratio of each qualitative attribute relative to the demand attribute based on the ratio of the attribute weight of each qualitative attribute to the attribute weight of the demand attribute; and determining the second distance penalty item corresponding to each two reference products based on the distance between each two reference products in terms of each qualitative attribute and the attribute weight ratio of each qualitative attribute relative to the demand attribute, the second distance penalty item being positively correlated with the distance between each two reference products in terms of each qualitative attribute and the attribute weight ratio of each qualitative attribute relative to the demand attribute.

[0016] In some embodiments, determining the second distance penalty item corresponding to each two reference products based on the distance between each two reference products in terms of each qualitative attribute and the attribute weight ratio of each qualitative attribute relative to the demand attribute includes: using the attribute weight ratio of each qualitative attribute of each two reference products relative to the demand attribute, performing a weighted operation on the distance between each two reference products in terms of each qualitative attribute, and obtaining the second distance penalty item corresponding to each two reference products.

[0017] In some embodiments, the product attributes of each reference product include dynamic attributes and static attributes, the dynamic attributes include the demand attributes, the attribute values ​​of the dynamic attributes change with time, and the attribute values ​​of the static attributes do not change with time. From the multiple reference classes, reference classes similar to the target product are screened as similar classes, including: obtaining attribute data of the static attributes of the target product; for each reference class, determining the classification distance between the target product and each reference class based on the attribute data of the static attributes of each reference product in each reference class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes; from the multiple reference classes, screening the reference class with the smallest classification distance as the similar class.

[0018] In some embodiments, determining the classification distance between the target product and each reference class includes: determining the distance between the target product and each reference product in each reference class in terms of static attributes based on the attribute data of the static attributes of each reference product in each reference class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes; determining the classification distance between the target product and each reference class based on the respective distances between the target product and each reference product in each reference class in terms of static attributes.

[0019] In some embodiments, there are multiple static attributes, and determining the distance between the target product and each reference product in each reference class in terms of the static attributes includes: for each static attribute, when the attribute values ​​of each static attribute of the target product and each reference product in each reference class are different, determining the distance between the target product and each reference product in each reference class in terms of each static attribute as a first value; for each static attribute, when the attribute values ​​of each static attribute of the target product and each reference product in each reference class are the same, determining the distance between the target product and each reference product in each reference class in terms of each static attribute as a second value, the second value being smaller than the first value; determining the attribute weight ratio of each static attribute relative to the sum of the attribute weights of each static attribute based on the ratio of the attribute weight of each static attribute to the sum of the attribute weights of each static attribute; determining the distance between the target product and each reference product in each reference class in terms of the static attribute based on the distance between the target product and each reference product in each reference class in terms of each static attribute and the attribute weight ratio of each static attribute relative to the sum of the attribute weights of each static attribute.

[0020] In some embodiments, determining the distance between the target product and each reference product in each reference class in terms of static attributes based on the distance between the target product and each reference product in each reference class in terms of each static attribute and the attribute weight ratio of each static attribute relative to the sum of the attribute weights of each static attribute includes: performing a weighted operation on the distance between the target product and each reference product in each reference class in terms of each static attribute using the attribute weight ratio of each static attribute relative to the sum of the attribute weights of each static attribute to obtain the distance between the target product and each reference product in each reference class in terms of static attributes.

[0021] In some embodiments, determining the classification distance between the target product and each reference class based on the respective distances between the target product and each reference product in each reference class in terms of static attributes includes: determining the sum of the respective distances between the target product and each reference product in each reference class in terms of static attributes as the composite distance between the target product and each reference product in each reference class in terms of static attributes; determining the classification distance between the target product and each reference class based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual number of products of each reference class, the classification distance being positively correlated with the composite distance between the target product and each reference product in each reference class in terms of static attributes and negatively correlated with the total actual number of products in each reference class.

[0022] In some embodiments, determining the classification distance between the target product and each reference product in each reference class based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual product quantity of each reference product in each reference class includes: obtaining a preset adjustment parameter corresponding to the target product, the preset adjustment parameter being used to adjust the degree to which the classification distance is affected by the actual product quantity, and the larger the parameter value of the preset adjustment parameter, the smaller the degree to which the classification distance is affected by the actual product quantity; determining a tendency value regarding product quantity based on the preset adjustment parameter and the actual product quantity, the tendency value regarding product quantity representing the business party's tendency towards the total quantity of reference products in the reference class, the tendency value regarding product quantity being negatively correlated with the preset adjustment parameter and positively correlated with the actual product quantity; determining the classification distance between the target product and each reference class based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the tendency value regarding product quantity, the classification distance being negatively correlated with the tendency value regarding product quantity and positively correlated with the sum of the composite distances.

[0023] In some embodiments, the product attributes of each reference product include dynamic attributes and static attributes, the dynamic attributes include the demand attributes, and predicting the future demand of the target product in the future period includes: determining the distance between the target product and each reference product in the similarity class in terms of static attributes based on the attribute data of the static attributes of each reference product in the similarity class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes; determining the similarity weight of each reference product in the similarity class based on the distance between the target product and each reference product in the similarity class in terms of static attributes, the similarity weight of each reference product in the similarity class is positively correlated with the corresponding distance in terms of static attributes; performing a weighted operation on the attribute data of the demand attributes of each reference product in the similarity class according to each similarity weight to obtain the future demand of the target product in the future period.

[0024] In some embodiments, determining the similarity weights of each reference product in the similarity class includes: determining the sum of the distances between the target product and each reference product in the similarity class in terms of static attributes as the sum of similarity distances; and determining the similarity weight of the reference product based on the ratio of the distance between the target product and each reference product in the similarity class in terms of static attributes to the sum of similarity distances.

[0025] In some embodiments, the future period includes multiple future moments, the future demand for the target product in the future period includes the future demand at each future moment, and predicting the future demand for the target product in the future period also includes: obtaining future impact factors that are expected to affect the future demand at each future moment in the future period, the future impact factors reflecting the impact of events expected to occur at each future moment in the future period on the future demand at the future moment; and adjusting the future demand for the target product at each future moment according to the future impact factors at each future moment.

[0026] In some embodiments, predicting the future demand for the target product in the future period further includes: if the target product is not a new product, obtaining historical replenishment data for the target product; determining, based on the historical replenishment data for the target product, a historical influencing factor that affects the future demand for the future period, the historical influencing factor reflecting the impact of an event that has occurred within the historical time corresponding to the historical data on the future demand for the future period; and adjusting the future demand for the target product in the future period based on the historical influencing factor.

[0027] In some embodiments, the historical influencing factors include at least one of a replenishment fulfillment rate influencing factor, a replenishment authenticity rate influencing factor, and a replenishment duration influencing factor. The replenishment fulfillment rate influencing factor reflects the impact of historical replenishment fulfillment rates on future demand, the replenishment authenticity rate influencing factor reflects the impact of historical replenishment authenticity rates on future demand, and the replenishment duration influencing factor reflects the impact of historical replenishment durations on future demand.

[0028] In some embodiments, for each reference product, based on its historical data, determining the attribute data of the product attributes of each reference product includes: for the demand attributes of each reference product, inputting the sales data in the historical data of each reference product and the relevant data of the influencing factors affecting the historical demand of each reference product into the regression model to obtain the historical demand of each reference product.

[0029] In some embodiments, determining the attribute data of the product attributes of each reference product further includes: smoothing the historical demand of each reference product to obtain the smoothed historical demand for clustering the multiple reference products.

[0030] In some embodiments, for each reference product, the relevant data of demand-sensitive influencing factors affecting the historical demand of each reference product includes historical traffic data, historical product value data, and historical product activity data within the historical time period to which the historical data belongs.

[0031] In some embodiments, the historical data is data of each historical moment within a historical period, the historical period to which the historical data of the reference product belongs is located within a reference historical time interval with the current moment as the end moment, the future period includes multiple future moments, the historical period and the future period have the same cycle length and the time interval between adjacent historical moments is the same as the time interval between adjacent future moments.

[0032] In some embodiments, predicting the future demand for the target product in the future period includes: predicting the initial future demand for the target product in the future period based on the attribute data of the demand attributes of each reference product in the similar class; in the case where the target product is an old product, predicting the future sales of the target product in the future period based on the attribute data of the sales attributes in the historical data of the target product; and determining the future demand for the target product in the future period based on the predicted future sales and the initial future demand.

[0033] In some embodiments, based on the predicted future sales and the initial future demand, predicting the future demand of the target product in the future period includes: obtaining the future sales weight and the future demand weight corresponding to the future sales and the initial future demand respectively, when the duration between the first sale time of the target product and the current time is greater than the duration threshold, the future sales weight is greater than the future demand weight, when the duration between the first sale time of the target product and the current time is less than or equal to the duration threshold, the future sales weight is less than the future demand weight, and the sum of the future sales weight and the future demand weight is 1; using the future sales weight and the future demand weight, perform a weighted operation on the future sales and the initial future demand to obtain the future demand of the target product in the future period.

[0034] In some embodiments, the future sales weight is positively correlated with the time between the first sale time of the target product and the current time, and the future demand weight is negatively correlated with the time between the first sale time of the target product and the current time.

[0035] In some embodiments, determining the future replenishment quantity of the target product in the future period includes: obtaining parameter values ​​of one or more replenishment parameters related to replenishment; and determining the future replenishment quantity of the target product in the future period using a replenishment strategy based on the future demand quantity of the target product, current inventory status information, and the parameter values ​​of the one or more replenishment parameters.

[0036] In some embodiments, at least one of the type of product attribute, the target value of the attribute weight of the product attribute, the clustering processing method, the number of reference products in the similar class, the similarity weight, the weighting operation method, the future impact factor, and the historical impact factor can be adjusted by the business party.

[0037] In some embodiments, the preset adjustment parameters can be adjusted by the business party.

[0038] In some embodiments, the input data and smoothing method of the regression model can be adjusted by the business party.

[0039] According to a second aspect of the present disclosure, a device for determining a replenishment quantity of a product is provided, comprising: an acquisition module configured to acquire historical data of a plurality of reference products related to a target product; a first determination module configured to determine, for each reference product, attribute data of a product attribute of the reference product based on its historical data, wherein the product attribute includes a demand attribute, and the attribute data of the demand attribute includes a historical demand quantity of the reference product; a clustering module configured to cluster the plurality of reference products according to the attribute data of the product attributes of the respective reference products and a target value of an attribute weight, to obtain a plurality of reference classes, wherein the attribute weight includes a demand attribute weight; a screening module configured to screen, from the plurality of reference classes, a reference class similar to the target product as a similar class; a prediction module configured to predict a future demand quantity of the target product in a future period based on the attribute data of the demand attributes of the respective reference products in the similar class; and a second determination module configured to determine a future replenishment quantity of the target product in the future period based on the future demand quantity of the target product.

[0040] According to a third aspect of the present disclosure, a device for determining a product replenishment quantity is provided, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute the method for determining a product replenishment quantity described in any of the above embodiments based on instructions stored in the memory.

[0041] According to a fourth aspect of the present disclosure, a computer storable medium is provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the method for determining the product replenishment quantity described in any of the above embodiments is implemented.

[0042] In the above embodiment, the accuracy of replenishment can be improved and the utilization rate of storage resources can be increased. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0044] The present disclosure can be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0045] Figure 1 is a flow chart illustrating a method for determining a product replenishment quantity according to some embodiments of the present disclosure;

[0046] Figure 2 is a schematic diagram showing the temporal relationship between a target product k and a reference product j according to some embodiments of the present disclosure.

[0047] Figure 3 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to some embodiments of the present disclosure;

[0048] Figure 4 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to other embodiments of the present disclosure;

[0049] Figure 5A is a block diagram illustrating a data reading and processing module according to some embodiments of the present disclosure;

[0050] Figure 5B is a block diagram illustrating a product clustering and classification module according to some embodiments of the present disclosure;

[0051] Figure 5C is a block diagram illustrating a target product demand forecasting module according to some embodiments of the present disclosure;

[0052] Figure 5D is a block diagram illustrating a target product dynamic replenishment module according to some embodiments of the present disclosure;

[0053] Figure 6 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to further embodiments of the present disclosure;

[0054] Figure 7 is a block diagram illustrating a computer system for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0055] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure.

[0056] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0057] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0058] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0059] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0060] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0061] Figure 1 is a flow chart illustrating a method for determining a product replenishment quantity according to some embodiments of the present disclosure.

[0062] like Figure 1 As shown, the method for determining product replenishment includes: step S1, obtaining historical data of multiple reference products related to the target product; step S2, for each reference product, determining the attribute data of the product attributes of each reference product based on its historical data; step S3, clustering the multiple reference products based on the attribute data of the product attributes of each reference product and the target value of the attribute weight to obtain multiple reference classes; step S4, screening reference classes similar to the target product from the multiple reference classes as similar classes; step S5, predicting the future demand of the target product in the future cycle based on the attribute data of the demand attributes of each reference product in the similar class; and step S6, determining the future replenishment of the target product in the future cycle based on the future demand of the target product. In some embodiments, the method for determining product replenishment is performed by a device for determining product replenishment. For example, the target product can be a new product or an old product. Old products include sub-new products and other old products other than sub-new products. Sub-new products are old products that are second only to new products.

[0063] In the above embodiment, the demand attributes are taken into account during clustering, and the corresponding attribute weights are assigned to the product attributes, so that the demand attributes play a semi-supervisory role in the clustering process, which can obtain more accurate clustering results, thereby obtaining more accurate similarity classes, and determining the future replenishment quantity through the future demand predicted by the similarity classes, fully considering the close relationship between the replenishment quantity and the demand quantity, and to a certain extent reducing the deviation between the replenishment quantity and the actual existence caused by the sales volume failing to truly reflect the demand volume in special circumstances such as inventory shortages, which can improve the accuracy of replenishment and thus improve the utilization rate of warehousing resources. For example, in special circumstances such as inventory shortages, sales volume is lower than demand volume, and the replenishment volume predicted based on sales volume is less than the actual replenishment volume, which will lead to the inability to fully and reasonably utilize warehousing resources. The replenishment volume predicted by the present disclosure through demand guidance is more accurately close to the actual market demand, thereby being able to fully and reasonably utilize warehousing resources and improve the utilization rate of warehousing resources.

[0064] In step S1, historical data of multiple reference products related to the target product is obtained. In some embodiments, the historical data is data for each historical moment within a historical period, the historical period to which the historical data of the reference product belongs is within a reference historical time interval ending at the current moment, the future period includes multiple future moments, the historical period and the future period have the same period length, and the time interval between adjacent historical moments is the same as the time interval between adjacent future moments.

[0065] In some embodiments, the historical data for multiple reference products includes both system data and manual data. For example, the data recorded by the e-commerce platform system, including procurement, warehousing, shipment, and sales, covering the entire product production, supply, and sales chain, is referred to as system data. Manual data refers to data related to new product sales and inventory management that is generated by business parties or operators based on their own experience or shared among multiple departments or entities, and that is not yet entered into the system, including but not limited to various instructions and other related information.

[0066] For example, the first sales time of target product k and old product p is represented by t. The target product and old products are distinguished by adding subscripts. The subscript k represents the target product, and the subscript p refers to multiple old products. Define t+δ as the current prediction time node, and δ is the time difference between the first sales time of the target product and the old product and the current prediction time node. a Indicates the future demand of the target product from the start time of the future cycle predicted by the current forecast time node, that is, the duration of the future cycle. b Indicates how long before the current forecast time node the selected old products were on sale, that is, the earliest starting time of the historical data of each reference product is limited. i It represents the forecast lead time of the target product, that is, the difference between the current forecast time node and the starting point of the time period corresponding to the demand to be forecasted (future cycle).

[0067] Among these old products, we need to select those products that correspond to the target product in time, that is, the old products that meet the "t k +δ-T b ≤t j +δ+lt k ” and “t j +δ+lt k +T a ≤ t k +δ” condition, as reference products. These reference products have complete T a For example, the complete T a The starting time of the historical data may be different. You can align their starting times to complete the T a Each historical moment of the historical period to which the historical data of the duration belongs is normalized to [1,T a ] interval, so as to unify the processing and improve the convenience. In some embodiments, the start time can also be set to any natural number according to the actual usage. The above-mentioned normalized time interval can be of any granularity that meets the actual needs. The relationship between each time point can be referred to Figure 2 Schematic diagram of the timing relationship between target product k and reference product j.

[0068] In step S2, for each reference product, attribute data of the product attributes of each reference product is determined based on its historical data. The product attributes include demand attributes. The attribute data of the demand attributes include the historical demand volume of the reference product.

[0069] In some embodiments, for each reference product's demand attributes, sales data from each reference product's historical data and relevant data on factors influencing each reference product's historical demand are input into a regression model to obtain each reference product's historical demand. For example, for each reference product, relevant data on demand-sensitive factors influencing each reference product's historical demand may include historical traffic data, historical product value data (price), and historical product activity data (such as promotions) within the historical period to which the historical data pertains.

[0070] In the above embodiment, the historical demand data is restored by using the historical sales data and the relevant data of the factors affecting the historical demand of each reference product. Not only the impact of sales on demand is taken into account, but also the relevant data of other influencing factors are taken into account. This allows the historical demand of the reference product to be restored more accurately, thereby further improving the accuracy of clustering and similarity determination, further improving the accuracy of demand forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0071] In some embodiments, the historical demand for each reference product may be smoothed to obtain smoothed historical demand for clustering the multiple reference products. For example, the smoothing may be performed using a mean value method, an interpolation method, a weighted method, a sampling method, or the like.

[0072] In the above embodiment, the restored historical demand for each reference product may contain extreme demand data, such as extremely high or extremely low historical demand. These extreme phenomena may be caused by accidental economic, weather, environmental factors, system or operator errors, or sudden retail events (such as promotions, price reductions, discounts, etc.). To address these issues, smoothing can reduce noise, thereby further improving the accuracy of clustering and similarity determination, further improving the accuracy of demand forecasts, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0073] In step S3, multiple reference products are clustered based on the attribute data of each reference product and the target attribute weights to obtain multiple reference clusters. The attribute weights include the demand attribute weights. For example, the target attribute weights are greater than 0 and less than or equal to 1.

[0074] In some embodiments, product attributes further include attributes other than demand attributes. Attribute weights further include attribute weights other than demand attribute weights. The target value of the demand attribute weight is greater than the target values ​​of the other attribute weights. In the above embodiment, the target value of the demand attribute weight is greater than the target values ​​of the other attribute weights, which can fully reflect the importance of the demand attribute in the clustering process, thereby making the semi-supervisory role of the demand attribute in the clustering process stronger than other attributes, further improving the accuracy of demand forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0075] In some embodiments, before clustering multiple reference products, a target value for the demand attribute weight and initial values ​​for the other attribute weights are obtained; and a target value for the other attribute weights is determined based on the target value for the demand attribute weight and the initial values ​​for the other attribute weights. The target value for the other attribute weights is negatively correlated with the target value for the demand attribute weight and positively correlated with the initial values ​​for the other attribute weights. For example, the initial values ​​for the other attribute weights are greater than 0 and less than or equal to 1.

[0076] In the above embodiment, the target values ​​of other attribute weights are determined by referring to the target value of the demand attribute weight, which implicitly constructs the correlation relationship between the demand attribute and other attributes, and reflects the comprehensiveness between the attribute weights of various product attributes, thereby further strengthening the importance of demand attributes in the clustering process and their semi-supervisory role in the clustering process, further improving the accuracy of demand forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0077] In some embodiments, determining the target values ​​of other attribute weights based on the target value of the requirement attribute weight and the initial values ​​of the other attribute weights can be achieved in the following manner.

[0078] First, determine the difference between the maximum target value of the requirement attribute weight and the target value of the requirement attribute weight. For example, if the target value of the requirement attribute weight is greater than 0 and less than or equal to 1, the maximum target value of the requirement attribute weight is 1. Then, determine the target values ​​of the other attribute weights based on the product of this difference and the initial values ​​of the other attribute weights.

[0079] The target value of the required attribute weight is represented as α0, and the initial values ​​of the other attribute weights of n other attributes are represented as the vector [α1,α2,…,α n ] as an example, the target values ​​of the demand attribute weight and n other attribute weights are expressed as the vector [α0, (1-α0)α1, (1-α0)α2, …, (1- α0)α n]. For example, the target value of the requirement attribute weight reflects the feature importance of the requirement attribute, and the above vector can be called the attribute importance vector.

[0080] In some embodiments, the target value of the demand attribute weight can be set by the business party, and the initial values ​​of other attribute weights can be obtained based on the rough set clustering method.

[0081] In some embodiments, taking the example that product attributes also include other attributes in addition to demand attributes, and the attribute weights also include other attribute weights in addition to the demand attribute weights, multiple reference products can be clustered in the following manner to obtain multiple reference classes.

[0082] First, for each two reference products, the distance between the two reference products in terms of the demand attributes is determined based on the attribute data of the corresponding demand attributes. The distance between the two reference products in terms of the demand attributes represents the degree of similarity between the two reference products in terms of the demand attributes.

[0083] In some embodiments, the difference in historical demand between two reference products at each historical moment is determined as the demand difference at each historical moment. The distance between the two reference products in terms of demand attributes is determined based on the square root of the sum of the squares of the demand differences between the two reference products at each historical moment. This means that the Euclidean distance between the two reference products in terms of demand attributes is calculated. In other embodiments, Hamming distance, Minkowski distance, etc. may also be used.

[0084] The historical data of each reference product is the data of each historical moment in the historical period. The historical period to which the historical data of the reference product belongs is located in the reference historical time interval with the current moment as the end moment. The future period includes multiple future moments. The period length of the historical period and the future period is the same, and the time interval between adjacent historical moments is the same as the time interval between adjacent future moments. The quantitative attributes in the product attributes are expressed as l1∈(0,L), 0 is the demand attribute. For example, the Euclidean distance between each two reference products in terms of demand attributes is Indicates that the 0th product attribute (i.e., demand attribute) of the i-th reference product is a The attribute value at time t within the cycle (i.e., historical demand). Indicates that the 0th product attribute (i.e., demand attribute) of the jth reference product is a The attribute value at the tth moment in the cycle (i.e., historical demand). t∈(1,T a ) means aligning the different moments of the historical cycles of different reference products to (1, T a ) time interval. The duration of the historical cycle and the future cycle is T aThe explanation of this formula is applicable to other similar formulas in the present disclosure.

[0085] Secondly, based on the attribute data of other attributes of each two reference products, a distance penalty term corresponding to each two reference products is determined. The distance penalty term represents the degree of influence of other attributes on the distance between each two reference products in terms of demand attributes.

[0086] In some embodiments, the product attributes of each reference product include quantitative attributes and qualitative attributes. Quantitative attributes include demand attributes and other quantitative attributes. A distance penalty term includes a first distance penalty term and a second distance penalty term. The first distance penalty term represents the degree of influence of other quantitative attributes on the distance between each two reference products with respect to the demand attributes. The second distance penalty term represents the degree of influence of qualitative attributes on the distance between each two reference products with respect to the demand attributes. Quantitative attributes are quantifiable product attributes, while qualitative attributes are non-quantifiable product attributes.

[0087] In the above embodiment, the penalty of the distance of other attributes to the demand attributes is considered from the two aspects of quantitative attributes and quantitative attributes, and the clustering distance calculation is realized more comprehensively and accurately, thereby further improving the accuracy of clustering, and further improving the accuracy of similar classes, thereby further improving the accuracy of demand forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0088] In some embodiments, the first distance penalty item corresponding to each two reference products may be determined based on attribute data of other quantitative attributes of each two reference products and target values ​​of corresponding attribute weights and target values ​​of corresponding demand attribute weights.

[0089] For example, determining the first distance penalty item corresponding to every two reference products may be achieved through the following steps 1) to 3).

[0090] 1) Determining, based on the attribute data of each other quantitative attribute of each two reference products, the distance between each two reference products in terms of each other quantitative attribute. In some embodiments, the distance between each two reference products in terms of each other quantitative attribute includes, but is not limited to, the Euclidean distance, which reflects the degree of similarity between each two reference products in terms of each other quantitative attribute.

[0091] 2) Determine the attribute weight ratio of each other quantitative attribute relative to the requirement attribute based on the attribute weight of each other quantitative attribute and the attribute weight of the requirement attribute.

[0092] 3) Determine a first distance penalty term corresponding to each pair of reference products based on the distance between each pair of reference products in each other quantitative attribute and the weight ratio of each other quantitative attribute relative to the required attribute. The first distance penalty term is positively correlated with the distance between each pair of reference products in each other quantitative attribute and the weight ratio of each other quantitative attribute relative to the required attribute.

[0093] In some embodiments, a weighted operation is performed on the distance between each two reference products in terms of each other quantitative attribute using the attribute weight ratio of each other quantitative attribute of each two reference products relative to the demand attribute to obtain a first distance penalty item corresponding to each two reference products.

[0094] For example, other quantitative attributes in product attributes are expressed as l1∈(1,L), and the first distance penalty term corresponding to reference product i and reference product j is expressed as

[0095] In the above embodiment, the attribute weight ratio of each other quantitative attribute relative to the demand attribute reflects the degree of influence of each other quantitative attribute on the demand attribute. The greater the degree of influence, the greater the first distance penalty term. The greater the distance between two reference products in each other quantitative attribute, the greater the first distance penalty term. In other words, the present disclosure comprehensively considers the penalties of other quantitative attributes on the distance of demand attributes in terms of similarity and influence, which is more objective and close to reality, thereby realizing cluster distance calculation more comprehensively and accurately, thereby further improving the accuracy of clustering, further improving the accuracy of similarity classes, and further improving the accuracy of demand forecasts, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0096] In some embodiments, a second distance penalty term corresponding to each of the two reference products may be determined based on the attribute data of the qualitative attributes of each of the two reference products, the target values ​​of the corresponding attribute weights, and the target values ​​of the corresponding demand attribute weights. The attribute values ​​of the qualitative attributes do not change over time.

[0097] For example, determining the second distance penalty item corresponding to every two reference products may be achieved through the following steps i) to iii).

[0098] i) When the attribute values ​​of each qualitative attribute of each two reference products are different, determining the distance between each two reference products in each qualitative attribute as: the product of the first value and the distance between each two reference products in the demand attribute.

[0099] When the attribute value of each qualitative attribute of each two reference products is the same, the distance between the two reference products in each qualitative attribute is determined as: the product of the second value and the distance between the two reference products in the demand attribute. The second value is less than the first value. For example, the second value is 0 and the first value is 1.

[0100] ii) Determine the attribute weight ratio of each qualitative attribute relative to the requirement attribute based on the ratio of the attribute weight of each qualitative attribute to the attribute weight of the requirement attribute.

[0101] iii) determining a second distance penalty item corresponding to each of the two reference products based on the distance between each of the two reference products in terms of each qualitative attribute and the attribute weight ratio of each qualitative attribute to the demand attribute.

[0102] In some embodiments, a weighted operation is performed on the distance between each two reference products in terms of each qualitative attribute using the attribute weight ratio of each qualitative attribute of each two reference products relative to the demand attribute to obtain a second distance penalty item corresponding to each two reference products.

[0103] For example, the qualitative attribute in the product attributes is expressed as l2∈(L+1,n), and the second distance penalty term corresponding to each two reference products is expressed as Represent the attribute values ​​of the qualitative attribute l2 of reference products i and j respectively. is an indicative function, which takes a first value when the condition in the brackets is true and a second value when the condition in the brackets is false. In some embodiments, the attribute value of the qualitative attribute l2 of the reference products i and j can be a numerical attribute value.

[0104] For example, the qualitative attributes of clothing color can have values ​​such as red, green, and yellow. These can be numerically converted to values ​​such as 1, 2, and 3, making them easier for computers to process. The values ​​of 1, 2, and 3 here are meaningless. Another example is the qualitative attributes of product value, which can have values ​​such as high, medium, and low. These values ​​can be numerically converted to values ​​such as 1, 2, and 3. For example, for time series data on prices, the mean, mode, or median can be extracted to represent the magnitude. By performing equidistant classification on multiple products with a distance of Δ, we can also obtain a similar classification of 1, 2, 3, etc., except that these numerical values ​​can represent the price range from low to high, but the numbers themselves are meaningless.

[0105] In the above embodiment, the attribute weight ratio of each qualitative attribute relative to the demand attribute reflects the degree of influence of each qualitative attribute on the demand attribute. The greater the degree of influence, the greater the first distance penalty term. The greater the distance between each two reference products in each qualitative attribute, the greater the first distance penalty term. In other words, the present disclosure comprehensively considers the penalty of the distance between the qualitative attributes in terms of similarity and influence on the demand attributes, which is more objective and close to reality, thereby realizing the clustering distance calculation more comprehensively and accurately, thereby further improving the accuracy of clustering, and further improving the accuracy of similarity classes, thereby further improving the accuracy of demand forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0106] Then, for each two reference products, the composite distance between the two reference products in terms of product attributes is determined according to the distance between the two reference products in terms of demand attributes and the corresponding distance penalty items.

[0107] In some embodiments, the composite distance between reference product i and reference product j in terms of product attributes can be expressed as follows:

[0108] Finally, the multiple reference products are clustered based on the composite distance between each two reference products to obtain multiple reference clusters. For example, clustering methods include but are not limited to hierarchical clustering, k-means clustering, mean-shift clustering, etc.

[0109] In the above embodiment, by introducing a penalty term that characterizes the degree of influence of other attributes on the distance between each two reference products in terms of demand attributes, the distance calculation in the clustering process not only considers the demand attributes closely related to the replenishment quantity, but also considers the indirect influence of other attributes, thereby realizing the clustering distance calculation more comprehensively and accurately, thereby further improving the accuracy of clustering, and further improving the accuracy of similar classes, thereby further improving the accuracy of demand quantity forecasting, further improving the accuracy of replenishment, further improving the utilization rate of warehouse resources, and further reducing the waste of warehouse resources.

[0110] In step S4, a reference class similar to the target product is screened from the plurality of reference classes as a similar class.

[0111] In some embodiments, the product attributes of each reference product include dynamic attributes and static attributes. Dynamic attributes include demand attributes. Dynamic attributes have attribute values ​​that change over time. Static attributes have attribute values ​​that do not change over time.

[0112] For example, the following steps (1) to (3) can be used to screen reference classes similar to the target product from multiple reference classes as similar classes.

[0113] (1) Obtain attribute data of static attributes of the target product.

[0114] (2) For each reference class, the classification distance between the target product and each reference class is determined based on the attribute data of the static attributes of each reference product in each reference class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes. The classification distance reflects the overall similarity between the target product and each reference product in each reference class.

[0115] (3) From multiple reference classes, select the reference class with the smallest classification distance as the similar class.

[0116] In the above embodiment, the target product and the reference class's static attribute attribute data are used to determine the overall similarity between the target product and each reference class, thereby realizing the target product classification process (i.e., screening similar classes). This is applicable not only to old products, but also to new products. Thus, the target product classification can be realized without relying on historical data (attribute data of dynamic attributes), thereby improving the versatility of the method for determining product replenishment quantities.

[0117] Based on (2) in screening reference classes similar to the target product, in some embodiments, the distance between the target product and each reference product in each reference class in terms of static attributes is determined according to the attribute data of the static attributes of each reference product in each reference class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes; and the classification distance between the target product and each reference class is determined according to the respective distances between the target product and each reference product in each reference class in terms of static attributes.

[0118] In some embodiments, taking the case where there are multiple static attributes as an example, the distance between the target product and each reference product in each reference class in terms of the static attribute may be determined in the following manner.

[0119] First, for each static attribute, if the attribute value of each static attribute of the target product and each reference product in each reference class is different, the distance between the target product and each reference product in each reference class with respect to each static attribute is determined as a first value, for example, 1.

[0120] For each static attribute, if the target product and each reference product in each reference class have the same attribute value for each static attribute, the distance between the target product and each reference product in each reference class with respect to each static attribute is determined to be a second value. The second value is less than the first value. For example, the second value is 0.

[0121] Then, according to the ratio of the attribute weight of each static attribute to the sum of the attribute weights of all static attributes, the attribute weight proportion of each static attribute relative to the sum of the attribute weights of all static attributes is determined.

[0122] Finally, the distance between the target product and each reference product in each reference class in terms of static attributes is determined based on the distance between the target product and each reference product in each reference class in terms of each static attribute and the proportion of the attribute weight of each static attribute relative to the sum of the attribute weights of each static attribute.

[0123] In some embodiments, a weighted operation is performed on the distance between the target product and each reference product in each reference class in terms of each static attribute by utilizing the ratio of the attribute weight of each static attribute relative to the sum of the attribute weights of each static attribute, thereby obtaining the distance between the target product and each reference product in each reference class in terms of the static attribute.

[0124] For example, the target product and each reference product in the reference class have m static attributes in total, where the static attributes l3∈[1,m]. The distance between the target product k and each reference product j in each reference class in terms of the m static attributes is expressed as They represent the attribute values ​​of the static attribute l3 of the target product k and the reference product j respectively. It is an indicator function. It takes the first value when the condition in the brackets is true and the second value when the condition in the brackets is false. is the sum of the attribute weights of each static attribute. So far, the distance between the target product and each reference product in each reference class in terms of static attributes can be calculated.

[0125] In some embodiments, the classification distance between the target product and each reference class may be determined in the following manner.

[0126] First, the sum of the distances between the target product and each reference product in each reference class in terms of static attributes is determined as the composite distance between the target product and each reference product in each reference class in terms of static attributes.

[0127] Then, the classification distance between the target product and each reference class is determined based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual number of products in each reference class. The classification distance is positively correlated with the composite distance between the target product and each reference product in each reference class in terms of static attributes and negatively correlated with the actual number of products in each reference class.

[0128] In some embodiments, the classification distance between the target product and each reference class can be determined based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual product quantity of each reference product in each reference class through the following steps I)-III).

[0129] 1) Obtaining a preset adjustment parameter corresponding to the target product. The preset adjustment parameter is used to adjust the degree to which the classification distance is affected by the actual number of products. The larger the value of the preset adjustment parameter, the less the classification distance is affected by the actual number of products.

[0130] II) Determine a product quantity preference value based on the preset adjustment parameter and the actual product quantity. The product quantity preference value indicates the business party's preference for the total quantity of reference products in the reference category. The stronger the preference, the larger the preference value. The product quantity preference value is negatively correlated with the preset adjustment parameter and positively correlated with the actual product quantity.

[0131] III) Determine the categorization distance between the target product and each reference product in each reference category based on the composite distances between the target product and each reference product in each reference category in terms of static attributes and the propensity score for product quantity. The categorization distance is negatively correlated with the propensity score for product quantity and positively correlated with the sum of the composite distances.

[0132] For example, clustering results in C reference classes, where there are num c reference products, the preset adjustment parameter is expressed as b, and the classification distance between the reference category c of the target product k is expressed as This is the tendency value regarding the number of products. For example, b∈[1,+∞).

[0133] In the above embodiment, by introducing preset adjustment parameters, the business party's purpose tendency for the total number of reference products in the reference category (i.e., the number of samples) can be introduced into the classification process to adjust the degree of influence of the actual product quantity on the classification distance and improve the flexibility of determining the product replenishment quantity.

[0134] In step S5, the future demand for the target product in the future period is predicted based on the attribute data of the demand attributes of each reference product in the similar class.

[0135] Taking the example that the product attributes of each reference product include dynamic attributes and static attributes, and the dynamic attributes include demand attributes, the above step S5 can be implemented in the following manner.

[0136] First, the distance between the target product and each reference product in the similar class in terms of the static attribute is determined based on the attribute data of the static attribute of each reference product in the similar class, the attribute data of the static attribute of the target product, and the target value of the attribute weight of the static attribute.

[0137] Then, based on the distance between the target product and each reference product in the similarity class in terms of static attributes, the similarity weight of each reference product in the similarity class is determined. The similarity weight of each reference product in the similarity class is positively correlated with the corresponding distance in terms of static attributes.

[0138] In some embodiments, the sum of the distances between the target product and each reference product in the similarity class in terms of static attributes is determined as the sum of similarity distances; the similarity weight of the reference product is determined based on the ratio of the distance between the target product and each reference product in the similarity class in terms of static attributes to the sum of similarity distances. For example, taking target product k as an example, the similarity weight of reference product j in the similarity class is

[0139] Finally, according to each similarity weight, a weighted operation is performed on the attribute data of the demand attributes of each reference product in the similarity class to obtain the future demand of the target product in the future period.

[0140] The similarity class c includes num4 reference products, and the attribute data of the demand attributes of each reference product includes 1 to T a This T a As an example, the historical demand at a certain moment, the future demand of the target product in the next cycle The subscript 0 in represents the demand attribute, the subscript 1 represents the historical moment 1 in the historical cycle of the reference product, and the superscript 1 represents the reference product 1 in the similar class. It represents the historical demand of reference product 1 in the similar category at historical moment 1. Similarly, it can be known that the historical demand of other reference products in the similar category and other historical moments are represented.

[0141] Taking the example that a future period includes multiple future moments, and the future demand for the target product in the future period includes the future demand at each future moment, predicting the future demand for the target product in the future period also includes: obtaining a future impact factor that is expected to affect the future demand at each future moment in the future period, the future impact factor reflecting the impact of an event expected to occur at each future moment in the future period on the future demand at that future moment; and adjusting the future demand for the target product at each future moment based on the future impact factor at each future moment. In some embodiments, for each future moment, the product of the future impact factor and the corresponding future demand is used as the adjusted future demand. For example, for each future moment, if there is no event expected to occur at that future moment, the future impact factor is 1.

[0142] In some embodiments, there may be demand-sensitive information such as promotions, discounts, price reductions, economic fluctuations, policy factors, and weather and environmental factors in the future period. These demand-sensitive information will affect the future demand at the corresponding future moment. The business party can generate the future impact factor of the corresponding moment based on this predicted demand-sensitive information. For example, the future impact factor of the future moment t is The future demand of the adjusted target product at each future moment is expressed as represents the future demand before adjustment at the future time t, and v represents the type of demand-sensitive information.

[0143] In the above embodiment, by introducing future influencing factors and adjusting future demand, the accuracy of demand forecasting can be further improved (making future demand more in line with actual conditions), the accuracy of replenishment can be further improved, the utilization rate of warehouse resources can be further improved, and the waste of warehouse resources can be further reduced.

[0144] In some embodiments, predicting the future demand for the target product in the future period further includes the following steps.

[0145] First, if the target product is not a new product, obtain the historical replenishment data of the target product. Figure 2 As shown, the time period of historical replenishment data is [t k ,t k + δ], called the replenishment time interval, δ ≥ 0. In particular, when the value is 0, it means the first forecast and there is no corresponding historical replenishment data.

[0146] Then, based on the historical replenishment data of the target product, the historical impact factors that affect the future demand in the future cycle are determined. The historical impact factors reflect the impact of events that have occurred within the historical time corresponding to the historical data on the future demand in the future cycle.

[0147] In some embodiments, the historical influencing factors include at least one of a replenishment fulfillment rate influencing factor, a replenishment authenticity rate influencing factor, and a replenishment duration influencing factor. The replenishment fulfillment rate influencing factor reflects the impact of the historical replenishment fulfillment rate on future demand, the replenishment authenticity rate influencing factor reflects the impact of the historical replenishment authenticity rate on future demand, and the replenishment duration influencing factor reflects the impact of the historical replenishment duration on future demand (e.g., the impact of delays in the historical replenishment process on future demand). The historical influencing factors are also referred to as replenishment performance influencing factors, and different historical influencing factors correspond to different replenishment performances. For example, the replenishment fulfillment rate influencing factor, the replenishment authenticity rate influencing factor, and the replenishment duration influencing factor correspond to replenishment performance u1, u2, and u3, respectively.

[0148] In some embodiments, the replenishment fulfillment rate influencing factor may be determined as follows.

[0149] First, based on the historical replenishment data for the target product, the average ratio of the actual replenishment quantity to the expected replenishment quantity during at least one replenishment process in the historical replenishment data is determined as the historical replenishment fulfillment rate. The actual replenishment quantity is also called the actual delivery quantity, and the expected replenishment quantity is also called the expected delivery quantity. Then, based on the historical replenishment fulfillment rate, the replenishment fulfillment rate influencing factor is determined. The replenishment fulfillment rate influencing factor is negatively correlated with the historical replenishment fulfillment rate.

[0150] For example, taking the historical replenishment times δ as an example, the replenishment satisfaction rate influencing factor is

[0151] For example, taking the historical replenishment times δ as an example, the replenishment authenticity rate influencing factor is This is the historical replenishment authenticity rate.

[0152] For example, still taking the historical replenishment times δ as an example, the replenishment time influence factor Reflects delays in historical replenishment processes.

[0153] Finally, the future demand of the target product in the future period is adjusted based on the historical impact factor. For example, at each future moment, the product of the historical impact factor and the corresponding future demand is used as the adjusted future demand.

[0154] The historical impact factor includes r u1 And this adjustment is in TD 2 As an example, the future demand after this adjustment can be expressed as follows: For example, if there are multiple historical impact factors, the multiple historical impact factors are multiplied together and then multiplied by the future demand before adjustment. For another example, the adjustment based on historical impact factors can also be based on TD 1 is carried out on the basis of.

[0155] In the above embodiment, for target products that are not brand-new products, historical influencing factors for adjusting future demand are determined through historical replenishment data, which can further improve the accuracy of demand forecasts (making future demand more in line with actual conditions), further improve the accuracy of replenishment, further improve the utilization rate of warehouse resources, and further reduce the waste of warehouse resources.

[0156] In some embodiments, the future demand for the target product in the future period can also be predicted in the following manner.

[0157] First, the initial future demand for the target product in the future period is predicted based on the attribute data of the demand attributes of each reference product in the similar class.

[0158] Then, in the case that the target product is an old product, the future sales volume of the target product in the future period is predicted based on the attribute data of the sales volume attribute in the historical data of the target product.

[0159] Finally, based on the predicted future sales and initial future demand, determine the future demand for the target product in the future cycle.

[0160] In the above embodiment, by introducing the historical sales data of the target product to predict the future sales of the target product, the impact of the historical sales of the target product itself on the future demand is reflected, which can further improve the accuracy of demand forecasting, further improve the accuracy of replenishment, further improve the utilization rate of warehouse resources, and further reduce the waste of warehouse resources.

[0161] In some embodiments, a future sales weight and a future demand weight corresponding to the future sales volume and the initial future demand volume are obtained, and a weighted operation is performed on the future sales volume and the initial future demand volume using the future sales weight and the future demand weight to obtain the future demand volume of the target product in the future period. When the time between the first sale time of the target product and the current time is greater than a time threshold, the future sales weight is greater than the future demand weight. When the time between the first sale time of the target product and the current time is less than or equal to the time threshold, the future sales weight is less than the future demand weight. The sum of the future sales weight and the future demand weight is 1.

[0162] The current time in the above embodiment refers to the current forecast time. The future sales weight reflects the degree of influence of the historical sales data of the target product on the future demand forecast, and the future demand weight reflects the degree of influence of the historical demand data of the reference product in the similar category on the future demand forecast.

[0163] In the above embodiment, since the target product has been sold for too long, the impact of similar products on its future demand will be reduced. The present disclosure assigns corresponding weights to future sales and initial future demand. When the target product has been sold for a long time, the focus is on considering the impact of the target product's own historical sales on future demand. When the target product has been sold for a short time, the focus is on considering the impact of demand data of reference products (also called similar products) in similar categories of the target product on future demand. This can further improve the accuracy of demand forecasting, further improve the accuracy of replenishment, further improve the utilization rate of warehouse resources, and further reduce the waste of warehouse resources.

[0164] In some embodiments, the future sales weight is positively correlated with the time between the first sale time of the target product and the current time, and the future demand weight is negatively correlated with the time between the first sale time of the target product and the current time.

[0165] In the above embodiment, the future sales weight is positively correlated with the time between the first sale time of the target product and the current time, and the future demand weight is negatively correlated with the time between the first sale time of the target product and the current time. This more accurately reflects the impact of the data that is focused on on future demand due to different sales times, which can further improve the accuracy of demand forecasts, further improve the accuracy of replenishment, further improve the utilization rate of warehouse resources, and further reduce the waste of warehouse resources.

[0166] Take the initial future demand as TD 3 , future sales volume is S, future demand weight is ω1, and future sales weight is ω2. For example, the future demand of the target product in the next cycle is TD=ω1×TD 3 +ω2×S. The initial future demand can also be TD 1 or TD 2 .

[0167] In some embodiments, after determining the future demand for the target product at each future moment in the future cycle, the future demand at each future moment is summed up to determine the future replenishment quantity. For example, Similarly, TD 1 or TD 2 Or TD=ω1×TD 3 +ω2×S performs a similar summation operation.

[0168] In step S6, the future replenishment quantity of the target product in the future period is determined based on the future demand quantity of the target product.

[0169] In some embodiments, parameter values ​​of one or more replenishment parameters related to replenishment are obtained; and a replenishment strategy is used to determine a future replenishment quantity of the target product in a future period based on the future demand for the target product, the current inventory status information, and the parameter values ​​of the one or more replenishment parameters. For example, the current inventory status information includes the current inventory quantity of the target product.

[0170] In some embodiments, the replenishment strategy includes a continuous replenishment strategy, an (s, S) strategy, an (r, Q) strategy, and the like.

[0171] In some embodiments, after determining the future replenishment quantity of the target product in the future cycle, the determined future replenishment quantity and relevant data of the current replenishment forecast process may also be stored for use in determining the replenishment quantity in subsequent other cycles.

[0172] In the above embodiment, at least one of the following can be adjusted by the business party: the type of product attribute, the target value of the attribute weight of the product attribute, the clustering method, the number of reference products in the similarity category, the similarity weight, the weighting method, the future impact factor, and the historical impact factor. Preset adjustment parameters can also be adjusted by the business party. The input data of the regression model and the smoothing method can also be adjusted by the business party.

[0173] Through human-computer interaction, business parties or operators can intervene in any stage of system operation in real time according to actual conditions, thereby improving replenishment flexibility.

[0174] Figure 3 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to some embodiments of the present disclosure.

[0175] like Figure 3 As shown, the device 3 for determining product replenishment quantity includes an acquisition module 31 , a first determination module 32 , a clustering module 33 , a screening module 34 , a prediction module 35 and a second determination module 36 .

[0176] The acquisition module 31 is configured to acquire historical data of multiple reference products related to the target product, for example, by performing the following steps: Figure 1 Step S1 shown.

[0177] The first determination module 32 is configured to determine the attribute data of the product attributes of each reference product based on its historical data, the product attributes including demand attributes, and the attribute data of the demand attributes including the historical demand of the reference product, for example, performing the following steps: Figure 1 Step S2 is shown.

[0178] The clustering module 33 is configured to perform clustering processing on multiple reference products based on the attribute data of the product attributes of each reference product and the target value of the attribute weight to obtain multiple reference classes. The attribute weight includes the demand attribute weight, for example, Figure 1 Step S3 shown.

[0179] The screening module 34 is configured to screen a reference class similar to the target product from a plurality of reference classes as a similar class, for example, performing the following steps: Figure 1 Step S4 is shown.

[0180] The prediction module 35 is configured to predict the future demand of the target product in the future period based on the attribute data of the demand attributes of each reference product in the similar class, for example, by executing step S5 as shown in FIG1 .

[0181] The second determining module 26 is configured to determine the future replenishment quantity of the target product in the future period according to the future demand quantity of the target product, for example, Figure 1Step S6 is shown.

[0182] Figure 4 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to some other embodiments of the present disclosure.

[0183] like Figure 4 As shown, the device 4 for determining product replenishment quantity includes a data storage module 40, a data reading and processing module 41, a product clustering and classification module 42, a target product demand prediction module 43, and a target product dynamic replenishment module 44.

[0184] The data storage module 40 is configured to store system data and manual data of all products.

[0185] The data reading and processing module 41 is configured to read the system data and manual data of all products from the data storage module 40 , and obtain historical data of multiple reference products related to the target product.

[0186] In some embodiments, the data reading and processing module 41 includes the following Figure 5A The data reading, writing and integration submodule 411, the data time related operation submodule 412, the sales data demand restoration submodule 413 and the demand data smoothing processing submodule 414 are shown.

[0187] The data read / write and integration submodule 411 is configured to read relevant system data and / or manual data from the data storage module 40 based on a storage medium or an application interface. The data read by the data read / write and integration submodule 411 may be all data that is readable as of the time of reading, or the most recent data that is readable as of the time of reading. The data read by the data read / write and integration submodule 411 may be stored in the currently running memory or on a specific other disk to facilitate access to the data.

[0188] Regarding writing, the data read / write and integration submodule 411 is configured to integrate the data from the data reading and processing module 41 and write it to the data storage module 40, allowing the business side to monitor the processing of the corresponding data as needed. When the relevant data is written to the data storage module 40 again, the corresponding memory or space is released. Each module may also have the function of reading the corresponding data from the data storage module as appropriate.

[0189] The data reading, writing, and integration submodule 411 is further configured to write the processed information to the data storage module 40 after any submodule completes the corresponding processing, according to the business party's settings. The data reading, writing, and integration submodule 411 is also configured to return comprehensive information of all submodules after each submodule completes execution. For example, data can be written using various high-level programming languages ​​or database technologies. In some embodiments, the data structure can be in the form of a linked list, tree, graph, or other forms. Optionally, the data storage file can be a variety of file types such as xls, xlsx, mdb, and mdl.

[0190] The data time related operation submodule 412 is configured to obtain historical data of multiple reference products related to the target product based on the data of the old product obtained by the data reading, writing and integration submodule 411 from the data storage module 40. For example, the data time related operation submodule 412 can be based on Figure 1 The historical data of multiple reference products related to the target product are obtained in the manner described in the above embodiment.

[0191] In some embodiments, the historical data of multiple reference products related to the target product obtained by the data time related operation submodule 412 can be transmitted back to the data storage module 40 through the data reading, writing and integration submodule 411 and then presented to the business side. The business side can make adjustments according to the situation, including but not limited to shortening the T when there are more reference products. b Or the reference product is less and T is increased b The above process can be repeated until the business party is satisfied.

[0192] The sales data demand restoration submodule 413 is configured to input the sales data in the historical data of each reference product and the relevant data of the factors affecting the historical demand of each reference product into the regression model for the demand attribute of each reference product, and obtain the historical demand of each reference product. In some embodiments, the regression model includes a linear regression model or a multivariate linear regression model. For example, the specific implementation process can refer to the method based on Figure 1 A description of the relevant steps and effects.

[0193] In some embodiments, the historical demand for each reference product obtained by the sales data demand restoration submodule 413 can be transmitted back to the data storage module 40 via the data reading, writing, and integration submodule 411 and presented to the business side. The business side can make adjustments based on circumstances, including but not limited to expanding the time interval to obtain more regression samples if the relationship is unclear due to insufficient regression samples, or providing new attribute-demand correspondences based on the characteristics of a particular product. The above process can be repeated until the business side is satisfied.

[0194] The demand data smoothing submodule 414 is configured to smooth the historical demand for each reference product obtained through regression, obtaining smoothed historical demand for clustering the multiple reference products. For example, the specific implementation process can refer to the description of the relevant steps and effects based on FIG1 .

[0195] In some embodiments, the demand data smoothing submodule 414 can transmit the data back to the data storage module 40 via the data reading, writing, and integration submodule 411, and then present it to the business side. The business side can make adjustments as needed, including but not limited to adjusting the number of data points used in the mean smoothing method or the specific spline interpolation calculation method. Repeat the above process until the business side is satisfied.

[0196] The product clustering and classification module 42 is configured to Figure 1 For example, the specific implementation process can refer to the steps S3 to S5 based on Figure 1 A description of the relevant steps and effects.

[0197] In some embodiments, the product clustering and classification module 42 includes the following: Figure 5B The data reading, writing and integration submodule 421 , the feature extraction submodule 422 , the feature data processing and importance determination submodule 423 , the custom distance calculation submodule 424 , the clustering submodule 425 , and the classification submodule 426 are shown.

[0198] In some embodiments, the data reading, writing and integration submodule 421 has the same or similar functions as the data reading, writing and integration submodule 411 .

[0199] The feature extraction submodule 422 is configured to obtain attribute data of product attributes of multiple reference products from the data reading and processing module 41. The feature extraction submodule 422 is also configured to read relevant system data or manual data from the data storage module 40 through the data reading, writing and integration submodule 421.

[0200] The extraction of product features by the feature extraction submodule 422 can be divided into two parts: static attributes and dynamic attributes.

[0201] On the one hand, these product attributes (i.e., product features) include static attributes such as weight, volume, and packaging (i.e., static features). For the target product, these static attributes have corresponding fixed values ​​before and after the first sale. For legacy products, these static attributes also remain consistent over the corresponding time dimension. The corresponding attribute values ​​of static attributes can be extracted from the product attribute information table maintained by the platform or system.

[0202] On the other hand, these product attributes also include dynamic attributes. For example, dynamic attributes include product value (price), flow, demand, etc. that change over time. For the target product, the attribute values ​​corresponding to these dynamic attributes in the expected forecast period (i.e., future period) are unknown and recorded as null. For the reference product, through the processing of the data time-related operation submodule 412 in the data reading and processing module 41, the corresponding data length of the reference product is T a Through the feature extraction submodule 422, the static features (ie, fields) of the target product and the static features of the reference product and the dynamic features (ie, fields) of a specific time period can be obtained, as well as the static data and time series data corresponding to the above fields.

[0203] The feature extraction submodule 422 can also extract the qualitative and quantitative attributes of the reference product. For example, the specific implementation method can refer to the Figure 1 A description of the relevant steps and effects.

[0204] In some embodiments, the feature extraction submodule 422 can be transmitted back to the data storage module 40 through the data reading, writing and integration submodule 421 and then presented to the business side. The business side can make adjustments according to the situation, including but not limited to adding or reducing certain attribute fields based on its own experience.

[0205] The feature data processing and importance determination submodule 423 is configured to obtain the target value of the attribute weight of each reference product for clustering. Figure 1 A description of the relevant steps and effects.

[0206] In some embodiments, the feature data processing and importance determination submodule 423 can be transmitted back to the data storage module 40 via the data reading, writing, and integration submodule 421 and then presented to the business side. The business side can make adjustments based on the circumstances, including but not limited to modifying the values ​​of Δ and α0 based on its own experience. For example, the business side can dynamically set a weight threshold for each attribute weight, i.e., an importance threshold, and remove attribute features with attribute weights less than this threshold. These removed attributes will not participate in the subsequent clustering process. This operation can achieve a certain degree of dimensionality reduction.

[0207] The custom distance calculation submodule 424 is configured to determine the composite distance between each two reference products in terms of product attributes. Figure 1 A description of the relevant steps and effects.

[0208] In some embodiments, the custom distance calculation submodule 424 can be transmitted back to the data storage module 40 through the data reading, writing, and integration submodule 421 and then presented to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the attribute weights involved in the distance calculation process or adjusting the normalization method of the time series data, or even deciding whether to perform normalization on the time series data.

[0209] The clustering submodule 425 is configured to perform clustering processing on the multiple reference products according to the composite distance between each two reference products in the multiple reference products to obtain multiple reference clusters. For example, the specific implementation process can refer to the method based on Figure 1 A description of the relevant steps and effects.

[0210] In some embodiments, the clustering submodule 425 can transmit the data to the data storage module 40 through the data reading, writing and integration submodule 421 and then present it to the business side. The business side can make adjustments according to the situation, including but not limited to adjusting the selected clustering method and adjusting specific parameters in the clustering method.

[0211] The classification submodule 426 is configured to select a reference class similar to the target product from a plurality of reference classes as a similar class. Figure 1 The classification submodule 426 can also obtain the classification distance vector of each classification distance between the target product and each reference product in the similarity class c. The classification distance vector is [Dis2(k,1),…,Dis2(k,j),…,Dis2(k,num c )].

[0212] In some embodiments, the classification submodule 426 can transmit the data back to the data storage module 40 through the data reading, writing and integration submodule 421 and then present it to the business side. The business side can make adjustments according to the situation, including but not limited to adjusting the calculation method of the classification distance or the value of b.

[0213] The target product demand forecasting module 43 is configured to forecast the future demand of the target product in the future period based on the attribute data of the demand attributes of each reference product in the similar category. Figure 1 A description of the relevant steps and effects.

[0214] In some embodiments, the target product demand forecasting module 43 includes the following: Figure 5C The shown submodules include a data reading, writing and integration submodule 431 , a similar product data processing submodule 432 , a demand sensitive information processing submodule 433 , a replenishment data processing and impact characterization submodule 434 , and a target product demand calculation submodule 435 .

[0215] The data reading, writing and integration submodule 431 has the same or similar functions as the data reading, writing and integration submodules 411 and 421 .

[0216] The similar product data processing submodule 432 is configured to determine the future demand TD of the target product in the future period. 1 For example, the specific implementation process can refer to Figure 1 A description of the relevant steps and effects.

[0217] In some embodiments, the similar product data processing submodule 432 can transmit the data back to the data storage module 40 through the data reading, writing, and integration submodule 431 and then present it to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the weighted calculation format or dynamically adding or removing reference products in the similar category based on demand.

[0218] The demand sensitive information processing submodule 433 is configured to obtain the future impact factor and adjust the future demand quantity TD determined by the similar product data processing submodule 432 based on the future impact factor. 1 , get the future demand TD 2 For example, the specific implementation process can refer to Figure 1 A description of the relevant steps and effects of the future impact factor. The future impact factor is also known as the demand-sensitive information factor. The advantage of this submodule is that it can incorporate newly acquired relevant information into the demand forecast during the forecast lead time before finalizing the forecast and replenishment, demonstrating its sensitivity.

[0219] In some embodiments, the demand-sensitive information processing submodule 433 can transmit the information back to the data storage module 40 via the data reading, writing, and integration submodule 431, and then present it to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the size of the future impact factor and increasing or decreasing the corresponding demand-sensitive information types, i.e., the types of future impact factors.

[0220] The replenishment data processing and impact characterization submodule 434 is configured to determine the historical impact factor based on the historical replenishment data of the target product, and adjust the future demand quantity TD determined by the demand sensitive information processing submodule 433 according to the historical impact factor. 2 , get the future demand TD 3 For example, the specific implementation process can refer to Figure 1 A description of the steps involved and their effects. The historical impact factor is also known as the replenishment performance impact factor.

[0221] In some embodiments, replenishment data processing and impact characterization submodule 434 can transmit data back to data storage module 40 via data reading, writing, and integration submodule 431 for presentation to the business side. The business side can make adjustments based on circumstances, including but not limited to adjusting the calculation method of historical impact factors and increasing or decreasing the types of corresponding historical impact factors.

[0222] The target product demand calculation submodule 435 is configured to, after determining the future demand of the target product at each future moment in the future cycle, perform a summation operation on the future demand at each future moment. For example, the final future demand of the target product is

[0223] In some embodiments, the target product demand calculation submodule 435 may also be configured to perform the following operations.

[0224] First, the initial future demand for the target product in the future period is predicted based on the attribute data of the demand attributes of each reference product in the similar class.

[0225] Then, in the case that the target product is an old product, the future sales volume of the target product in the future period is predicted based on the attribute data of the sales volume attribute in the historical data of the target product.

[0226] Finally, based on the predicted future sales volume and initial future demand volume, determine the future demand volume of the target product in the next cycle. For example, the specific implementation process can refer to the Figure 1 A description of the relevant steps and effects.

[0227] In some embodiments, the target product demand calculation submodule 435 can transmit data back to the data storage module 40 via the data reading, writing, and integration submodule 431 and then present it to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the value of the final predicted future demand and the calculation method of the final demand forecast value.

[0228] The target product dynamic replenishment module 44 is configured to determine the future replenishment quantity of the target product in the future period according to the future demand of the target product. Figure 1 A description of the relevant steps and effects.

[0229] In some embodiments, the target product dynamic replenishment module 44 includes the following: Figure 5D The data reading, writing and integration submodule 441, the replenishment parameter setting submodule 442, the replenishment quantity calculation submodule 443, and the replenishment performance data integration submodule 444 are shown.

[0230] The data reading, writing and integration submodule 441 has the same or similar functions as the data reading, writing and integration submodules 411 , 421 , and 431 .

[0231] The replenishment parameter setting submodule 442 is configured to obtain the parameter values ​​of one or more replenishment parameters related to replenishment. Figure 1 A description of the relevant steps and effects.

[0232] The replenishment quantity calculation submodule 443 is configured to determine the future replenishment quantity of the target product in the future period using a replenishment strategy based on the future demand quantity of the target product, the current inventory information and the parameter values ​​of one or more replenishment parameters. Figure 1 A description of the relevant steps and effects.

[0233] In some embodiments, the replenishment quantity calculation submodule 443 can transmit data back to the data storage module 40 via the data reading, writing, and integration submodule 441 and then present it to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the replenishment quantity based on the procurement budget or changing the replenishment quantity calculation strategy.

[0234] Replenishment performance data integration submodule 444 is configured to store the determined future replenishment quantities and related data from the current replenishment forecast process in data storage module 40 via data reading, writing, and integration submodule 441, for use in determining replenishment quantities in subsequent cycles. For example, the current order quantity can be recorded and, when entering the next cycle, the replenishment fulfillment rate can be calculated based on this and the actual delivery quantity. Another example is recording the current order placement time and the order arrival time at the supplier. When entering the next cycle, this can be used in conjunction with the actual delivery date to determine the supplier lead time.

[0235] In some embodiments, replenishment performance data integration submodule 444 can transmit data back to data storage module 40 via data reading, writing, and integration submodule 441, and then present it to the business side. The business side can make adjustments based on the situation, including but not limited to adjusting the size of the stored relevant data, adding or deleting relevant data types, etc.

[0236] In the above embodiment, the data storage module 40, the data reading and processing module 41, the product clustering and classification module 42, the target product demand forecasting module 43, and the target product dynamic replenishment module 44 can constitute a product demand forecasting and replenishment subsystem.

[0237] In some embodiments, the device 4 for determining product replenishment quantities further includes a manual data input module 45 and a visual system monitoring module 46. The business side uses the manual data input module 45 to input manual data before, during, and after the operation for use in the replenishment quantity determination process. It should be noted that manual data entered after the operation is completed and after the process enters the product clustering and classification module will not be used in the current replenishment quantity determination process. The business side can also use the manual data input module 45 to provide feedback on adjustments to various data during the human-computer interaction process. The visual system monitoring module 46 can view, in a visual format such as a table or image, data stored in the data storage module 40 that needs to be fed back to the business side for adjustments during the replenishment quantity determination process.

[0238] For example, the manual data input module 45 and the visual system monitoring module 46 constitute the interactive interface subsystem of the device 4 for determining the product replenishment quantity.

[0239] It should be noted that the dynamic nature of dynamic replenishment is reflected in its multi-stage replenishment decision-making. In other words, in the long term, multiple complete modules 41, 42, 43, and 44 can be executed until the final replenishment quantity is reached. However, the target product dynamic replenishment module 44 or its submodules can still be executed multiple times without fully executing the above modules 41, 42, 43, and 44. This operation is intended to facilitate the business side to flexibly improve the current conclusion based on their own experience or newly acquired information, highlighting the advantages of the human-computer interaction system, and does not represent the dynamic meaning of dynamic replenishment.

[0240] Figure 6 is a block diagram illustrating an apparatus for determining a product replenishment quantity according to further embodiments of the present disclosure.

[0241] like Figure 6 As shown, the apparatus 6 for determining product replenishment quantities includes a memory 61 and a processor 62 coupled to the memory 61. The memory 61 is configured to store instructions for executing the method for determining product replenishment quantities according to an embodiment. The processor 62 is configured to execute the method for determining product replenishment quantities according to any of the embodiments of the present disclosure based on the instructions stored in the memory 61.

[0242] Figure 7 is a block diagram illustrating a computer system for implementing some embodiments of the present disclosure.

[0243] like Figure 7 As shown, computer system 70 may be embodied as a general-purpose computing device. Computer system 70 includes memory 710, processor 720, and bus 700 that connects various system components.

[0244] Memory 710 may include, for example, system memory, non-volatile storage media, and the like. System memory, for example, stores an operating system, application programs, a boot loader, and other programs. System memory may include volatile storage media, such as random access memory (RAM) and / or cache memory. Non-volatile storage media, for example, stores instructions for executing at least one embodiment of the method for determining product replenishment quantity. Non-volatile storage media include, but are not limited to, disk storage, optical storage, and flash memory.

[0245] The processor 720 can be implemented as a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, or as discrete hardware components such as discrete gates or transistors. Accordingly, each module, such as the judgment module and the determination module, can be implemented by a central processing unit (CPU) executing instructions in a memory that execute corresponding steps, or by a dedicated circuit that executes the corresponding steps.

[0246] The bus 700 may use any of a variety of bus architectures, including, but not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, and a Peripheral Component Interconnect (PCI) bus.

[0247] The computer system 70 may also include an input / output interface 730, a network interface 740, a storage interface 750, and the like. These interfaces 730, 740, and 750, as well as the memory 710 and the processor 720, may be connected via a bus 700. The input / output interface 730 may provide a connection interface for input / output devices such as a display, mouse, and keyboard. The network interface 740 may provide a connection interface for various networked devices. The storage interface 750 may provide a connection interface for external storage devices such as floppy disks, USB flash drives, and SD cards.

[0248] Here, various aspects of the present disclosure are described with reference to flowcharts and / or block diagrams of methods, devices, and computer program products according to embodiments of the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks, can be implemented by computer-readable program instructions.

[0249] These computer-readable program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable device to produce a machine, so that the processor executes the instructions to produce means for implementing the functions specified in one or more blocks in the flowcharts and / or block diagrams.

[0250] These computer-readable program instructions may also be stored in a computer-readable memory, which cause the computer to operate in a specific manner to produce an article of manufacture, including instructions for implementing the functions specified in one or more blocks in the flowcharts and / or block diagrams.

[0251] The present disclosure can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects.

[0252] By using the method and device for determining product replenishment quantity and the computer storable medium in the above-mentioned embodiments, the accuracy of replenishment can be improved and the utilization rate of storage resources can be increased.

[0253] The method and apparatus for determining product replenishment quantities, as well as the computer-storable medium, according to the present disclosure have been described in detail. To avoid obscuring the underlying principles of the present disclosure, some details known in the art have been omitted. Based on the above description, those skilled in the art will readily understand how to implement the disclosed technical solutions.

Claims

1. A method for determining a product replenishment quantity, comprising: Obtain historical data of multiple reference products related to the target product; For each reference product, determining attribute data of a product attribute of each reference product based on its historical data, wherein the product attribute includes a demand attribute, and the attribute data of the demand attribute includes a historical demand amount of the reference product, wherein the product attribute also includes other attributes in addition to the demand attribute, and the attribute weight also includes other attribute weights in addition to the demand attribute weight, and a target value of the demand attribute weight is greater than the target values ​​of the other attribute weights; performing clustering processing on the plurality of reference products based on attribute data of product attributes and target values ​​of attribute weights of the respective reference products to obtain a plurality of reference classes, wherein the attribute weights include demand attribute weights; screening, from the plurality of reference classes, reference classes similar to the target product as similar classes; Predicting the future demand for the target product in a future period based on the attribute data of the demand attributes of each reference product in the similar class; Determine the future replenishment quantity of the target product in the future period according to the future demand quantity of the target product.

2. The method for determining product replenishment quantity according to claim 1, further comprising: Before clustering the plurality of reference products, obtaining a target value of the demand attribute weight and an initial value of the other attribute weights; The target value of the other attribute weight is determined according to the target value of the requirement attribute weight and the initial value of the other attribute weight. The target value of the other attribute weight is negatively correlated with the target value of the requirement attribute weight and positively correlated with the initial value of the other attribute weight.

3. The method for determining product replenishment quantity according to claim 2, wherein: Determining the target value of the other attribute weight according to the target value of the requirement attribute weight and the initial value of the other attribute weight includes: Determine the difference between the maximum value of the target value of the demand attribute weight and the target value of the demand attribute weight; The target value of the other attribute weight is determined according to the product of the difference and the initial value of the other attribute weight.

4. The method for determining product replenishment quantity according to any one of claims 1 to 2, wherein: The product attributes further include other attributes in addition to the demand attributes, and the attribute weights further include other attribute weights in addition to the demand attribute weights. Clustering is performed on the multiple reference products to obtain multiple reference classes including: For each two reference products, determining the distance between the two reference products in terms of the demand attributes based on the corresponding attribute data of the demand attributes, and indicating the degree of similarity between the two reference products in terms of the demand attributes; Determining, based on attribute data of other attributes of each of the two reference products, a distance penalty item corresponding to each of the two reference products, wherein the distance penalty item represents the degree of influence of the other attributes on the distance between each of the two reference products in terms of the demand attributes; For each two reference products, determining a composite distance between the two reference products in terms of the product attributes according to the distance between the two reference products in terms of the demand attributes and the corresponding distance penalty item; The multiple reference products are clustered according to the composite distance between every two reference products to obtain multiple reference classes.

5. The method for determining product replenishment quantity according to claim 4, wherein: The product attributes of each reference product include quantitative attributes and qualitative attributes, the quantitative attributes include the demand attributes and other quantitative attributes, the distance penalty item includes a first distance penalty item and a second distance penalty item, the first distance penalty item represents the degree of influence of other quantitative attributes on the distance between each two reference products in terms of demand attributes, and the second distance penalty item represents the degree of influence of qualitative attributes on the distance between each two reference products in terms of demand attributes.

6. The method for determining product replenishment quantity according to claim 5, wherein: Determining the distance penalty item corresponding to each two reference products includes: Determining a first distance penalty item corresponding to each of the two reference products based on attribute data of other quantitative attributes of each of the two reference products and target values ​​of corresponding attribute weights and target values ​​of corresponding demand attribute weights; and / or A second distance penalty item corresponding to each of the two reference products is determined based on the attribute data of the qualitative attributes of each of the two reference products and the target values ​​of the corresponding attribute weights and the target values ​​of the corresponding demand attribute weights.

7. The method for determining product replenishment quantity according to claim 4, wherein: Determining the distance between each two reference products in terms of the demand attributes includes: Determine the difference in historical demand of each of the two reference products at each historical moment as the demand difference at each historical moment; The distance between the two reference products in terms of demand attributes is determined according to the square root of the sum of the squares of the differences in demand quantities of the two reference products at various historical moments.

8. The method for determining product replenishment quantity according to claim 6, wherein: Determining a first distance penalty item corresponding to each of the two reference products includes: determining, based on the attribute data of each other quantitative attribute of each of the two reference products, a distance between the two reference products in terms of each other quantitative attribute; Determining the attribute weight ratio of each other quantitative attribute relative to the requirement attribute based on the attribute weight of each other quantitative attribute and the attribute weight of the requirement attribute; Based on the distance between each two reference products in each other quantitative attribute and the attribute weight ratio of each other quantitative attribute relative to the demand attribute, a first distance penalty item corresponding to each two reference products is determined, and the first distance penalty item is positively correlated with the distance between each two reference products in each other quantitative attribute and the attribute weight ratio of each other quantitative attribute relative to the demand attribute.

9. The method for determining product replenishment quantity according to claim 8, wherein: Determining, based on the distance between each two reference products in terms of each other quantitative attribute and the attribute weight ratio of each other quantitative attribute to the demand attribute, a first distance penalty item corresponding to each two reference products includes: By utilizing the attribute weight ratios of each other quantitative attribute of each of the two reference products relative to the demand attribute, a weighted operation is performed on the distance between each of the two reference products in terms of each other quantitative attribute to obtain a first distance penalty item corresponding to each of the two reference products.

10. The method for determining product replenishment quantity according to claim 6, wherein: The attribute value of the qualitative attribute does not change over time, and determining the second distance penalty item corresponding to each of the two reference products includes: In a case where the attribute values ​​of each qualitative attribute of each of the two reference products are different, determining the distance between each of the two reference products in terms of each qualitative attribute as the product of the first value and the distance between each of the two reference products in terms of the demand attribute; When the attribute value of each qualitative attribute of each of the two reference products is the same, determining the distance between each of the two reference products in terms of each qualitative attribute as the product of a second value and the distance between each of the two reference products in terms of the demand attribute, the second value being smaller than the first value; Determining the attribute weight ratio of each qualitative attribute relative to the requirement attribute based on the ratio of the attribute weight of each qualitative attribute to the attribute weight of the requirement attribute; Based on the distance between each two reference products in each qualitative attribute and the attribute weight ratio of each qualitative attribute relative to the demand attribute, a second distance penalty item corresponding to each two reference products is determined, and the second distance penalty item is positively correlated with the distance between each two reference products in each qualitative attribute and the attribute weight ratio of each qualitative attribute relative to the demand attribute.

11. The method for determining product replenishment quantity according to claim 10, wherein: Determining the second distance penalty item corresponding to each of the two reference products according to the distance between each of the two reference products in terms of each qualitative attribute and the attribute weight ratio of each qualitative attribute to the demand attribute includes: Using the attribute weight ratio of each qualitative attribute of each two reference products relative to the demand attribute, a weighted operation is performed on the distance between each two reference products in terms of each qualitative attribute to obtain a second distance penalty item corresponding to each two reference products.

12. The method for determining product replenishment quantity according to any one of claims 1-2, wherein: The product attributes of each reference product include dynamic attributes and static attributes. The dynamic attributes include the demand attributes. The attribute values ​​of the dynamic attributes change over time, and the attribute values ​​of the static attributes do not change over time. From the multiple reference classes, reference classes similar to the target product are screened as similar classes, including: Acquiring attribute data of static attributes of the target product; For each reference class, determining a classification distance between the target product and each reference class based on the attribute data of the static attributes of each reference product in the reference class, the attribute data of the static attributes of the target product, and the target value of the attribute weight of the static attributes; From the multiple reference classes, a reference class with the smallest classification distance is selected as a similar class.

13. The method for determining product replenishment quantity according to claim 12, wherein: Determining the classification distance between the target product and each reference class includes: Determining the distance between the target product and each reference product in each reference class in terms of the static attribute based on the attribute data of the static attribute of each reference product in each reference class, the attribute data of the static attribute of the target product, and the target value of the attribute weight of the static attribute; The classification distance between the target product and each reference class is determined according to each distance between the target product and each reference product in each reference class in terms of static attributes.

14. The method for determining product replenishment quantity according to claim 13, wherein: There are multiple static attributes, and determining the distance between the target product and each reference product in each reference class in terms of the static attribute includes: For each static attribute, when the attribute value of each static attribute of the target product and each reference product in each reference class is different, determining the distance between the target product and each reference product in each reference class with respect to each static attribute as a first value; For each static attribute, when the target product and each reference product in each reference class have the same attribute value for each static attribute, determining a distance between the target product and each reference product in each reference class with respect to each static attribute as a second value, the second value being smaller than the first value; Determining the attribute weight proportion of each static attribute relative to the sum of the attribute weights of each static attribute based on the ratio of the attribute weight of each static attribute to the sum of the attribute weights of each static attribute; The distance between the target product and each reference product in each reference class in terms of static attributes is determined based on the distance between the target product and each reference product in each reference class in terms of each static attribute and the attribute weight ratio of each static attribute to the sum of the attribute weights of each static attribute.

15. The method for determining product replenishment quantity according to claim 14, wherein: Determining the distance between the target product and each reference product in each reference class in terms of each static attribute, based on the distance between the target product and each reference product in each reference class in terms of each static attribute and the attribute weight ratio of each static attribute to the sum of the attribute weights of each static attribute, includes: By using the attribute weight ratio of each static attribute relative to the sum of the attribute weights of each static attribute, a weighted operation is performed on the distance between the target product and each reference product in each reference class in terms of each static attribute to obtain the distance between the target product and each reference product in each reference class in terms of static attributes.

16. The method for determining product replenishment quantity according to claim 13, wherein: Determining the classification distance between the target product and each reference product in each reference class according to the distances between the target product and each reference product in each reference class in terms of static attributes includes: Determining the sum of the distances between the target product and the reference products in each reference class in terms of static attributes as a composite distance between the target product and the reference products in each reference class in terms of static attributes; Based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual number of products of each reference product in each reference class, the classification distance between the target product and each reference class is determined. The classification distance is positively correlated with the composite distance between the target product and each reference product in each reference class in terms of static attributes and negatively correlated with the total actual number of products in each reference class.

17. The method for determining product replenishment quantity according to claim 16, wherein: Determining the classification distance between the target product and each reference product in each reference class according to the composite distance between the target product and each reference product in each reference class in terms of static attributes and the actual number of each reference product in each reference class includes: Obtaining a preset adjustment parameter corresponding to the target product, wherein the preset adjustment parameter is used to adjust the degree to which the classification distance is affected by the actual number of products, wherein the larger the parameter value of the preset adjustment parameter, the smaller the degree to which the classification distance is affected by the actual number of products; Determining a tendency value for product quantity based on the preset adjustment parameter and the actual product quantity, the tendency value for product quantity representing the business party's tendency for the total quantity of reference products in the reference category, the tendency value for product quantity being negatively correlated with the preset adjustment parameter and positively correlated with the actual product quantity; Based on the composite distance between the target product and each reference product in each reference class in terms of static attributes and the tendency value regarding the number of products, the classification distance between the target product and each reference class is determined, and the classification distance is negatively correlated with the tendency value regarding the number of products and positively correlated with the sum of the composite distances.

18. The method for determining product replenishment quantity according to claim 1, wherein: The product attributes of each reference product include dynamic attributes and static attributes, wherein the dynamic attributes include the demand attributes. Predicting the future demand of the target product in the future period includes: Determining the distance between the target product and each reference product in the similarity class in terms of the static attribute based on the attribute data of the static attribute of each reference product in the similarity class, the attribute data of the static attribute of the target product, and the target value of the attribute weight of the static attribute; Determining a similarity weight of each reference product in the similarity class based on a distance between the target product and each reference product in the similarity class in terms of static attributes, wherein the similarity weight of each reference product in the similarity class is positively correlated with the corresponding distance in terms of static attributes; According to each similarity weight, a weighted operation is performed on the attribute data of the demand attribute of each reference product in the similarity class to obtain the future demand of the target product in the future period.

19. The method for determining product replenishment quantity according to claim 18, wherein: Determining the similarity weight of each reference product in the similarity class includes: Determine the sum of the distances between the target product and the reference products in the similarity class in terms of static attributes as the sum of similarity distances; The similarity weight of the reference product is determined according to the ratio of the distance between the target product and each reference product in the similarity class in terms of static attributes to the sum of the similarity distances.

20. The method for determining product replenishment quantity according to claim 18, wherein: The future period includes multiple future moments, the future demand for the target product in the future period includes the future demand at each future moment, and predicting the future demand for the target product in the future period further includes: Obtaining a future impact factor that is expected to affect the future demand at each future moment in the future period, wherein the future impact factor reflects the impact of an event that is expected to occur at each future moment in the future period on the future demand at that future moment; According to the future impact factor at each future moment, the future demand quantity of the target product at each future moment is adjusted.

21. The method for determining product replenishment quantity according to claim 20, wherein: Forecasting the future demand for the target product in the next period also includes: If the target product is not a new product, obtaining historical replenishment data of the target product; Determining, based on historical replenishment data of the target product, historical impact factors that affect future demand for the future period, wherein the historical impact factors reflect the impact of events that have occurred during the historical period corresponding to the historical data on future demand for the future period; According to the historical impact factor, the future demand for the target product in the future period is adjusted.

22. The method for determining product replenishment quantity according to claim 21, wherein: The historical influencing factors include at least one of a replenishment fulfillment rate influencing factor, a replenishment authenticity rate influencing factor, and a replenishment duration influencing factor. The replenishment fulfillment rate influencing factor reflects the impact of the historical replenishment fulfillment rate on future demand, the replenishment authenticity rate influencing factor reflects the impact of the historical replenishment authenticity rate on future demand, and the replenishment duration influencing factor reflects the impact of the historical replenishment duration on future demand.

23. The method for determining product replenishment quantity according to claim 1, wherein: For each reference product, based on its historical data, the attribute data of the product attributes of each reference product is determined to include: For each reference product's demand attributes, the sales data in the historical data of each reference product and the relevant data of the factors affecting the historical demand of each reference product are input into the regression model to obtain the historical demand of each reference product.

24. The method for determining product replenishment quantity according to claim 23, wherein: The attribute data for determining the product attributes of each reference product further includes: The historical demand of each reference product is smoothed to obtain the smoothed historical demand, which is used to perform clustering on the multiple reference products.

25. The method for determining product replenishment quantity according to claim 23, wherein: For each reference product, the relevant data of demand-sensitive influencing factors affecting the historical demand of each reference product includes historical traffic data, historical product value data, and historical product activity data within the historical time period to which the historical data belongs.

26. The method for determining product replenishment quantity according to claim 1, wherein: The historical data is the data of each historical moment within the historical period. The historical period to which the historical data of the reference product belongs is located in the reference historical time interval with the current moment as the end moment. The future period includes multiple future moments. The historical period and the future period have the same cycle length and the time interval between adjacent historical moments is the same as the time interval between adjacent future moments.

27. The method for determining product replenishment quantity according to claim 1, wherein: Forecasting the future demand for the target product in the next period includes: Predicting the initial future demand for the target product in a future period based on the attribute data of the demand attributes of each reference product in the similar class; If the target product is an old product, predicting future sales of the target product in the future period based on attribute data of sales attributes in historical data of the target product; The future demand for the target product in the future period is predicted based on the predicted future sales volume and the initial future demand.

28. The method for determining product replenishment quantity according to claim 27, wherein: Predicting the future demand for the target product in the next period based on the predicted future sales volume and the initial future demand includes: Obtaining a future sales weight and a future demand weight corresponding to the future sales volume and the initial future demand, respectively, where if the duration between the first sales time of the target product and the current time is greater than a duration threshold, the future sales weight is greater than the future demand weight; and if the duration between the first sales time of the target product and the current time is less than or equal to the duration threshold, the future sales weight is less than the future demand weight, and the sum of the future sales weight and the future demand weight is 1; The future sales volume weight and the future demand volume weight are used to perform a weighted operation on the future sales volume and the initial future demand volume to obtain the future demand volume of the target product in the future period.

29. The method for determining product replenishment quantity according to claim 28, wherein: The future sales weight is positively correlated with the time between the first sale time of the target product and the current time, and the future demand weight is negatively correlated with the time between the first sale time of the target product and the current time.

30. The method for determining product replenishment quantity according to claim 1, wherein: Determining the future replenishment quantity of the target product in the future period includes: Get parameter values ​​of one or more replenishment parameters related to replenishment; A future replenishment quantity of the target product in the future period is determined using a replenishment strategy according to the future demand quantity of the target product, current inventory status information, and parameter values ​​of the one or more replenishment parameters.

31. The method for determining product replenishment quantity according to claim 21, wherein: At least one of the type of product attribute, the target value of the attribute weight of the product attribute, the clustering processing method, the number of reference products in the similarity class, the similarity weight, the weighting operation method, the future impact factor, and the historical impact factor can be adjusted by the business party.

32. The method for determining product replenishment quantity according to claim 17, wherein: The preset adjustment parameters can be adjusted by the business party.

33. The method for determining product replenishment quantity according to claim 24, wherein: The input data and smoothing method of the regression model can be adjusted by the business party.

34. A device for determining product replenishment quantity, comprising: an acquisition module configured to acquire historical data of a plurality of reference products related to a target product; A first determination module is configured to determine, for each reference product, attribute data of a product attribute of the reference product based on its historical data, wherein the product attribute includes a demand attribute, and the attribute data of the demand attribute includes a historical demand for the reference product, wherein the product attribute further includes other attributes in addition to the demand attribute, and the attribute weight further includes other attribute weights in addition to the demand attribute weight, and a target value of the demand attribute weight is greater than a target value of the other attribute weights; a clustering module configured to cluster the plurality of reference products according to attribute data of product attributes of each reference product and target values ​​of attribute weights to obtain a plurality of reference clusters, wherein the attribute weights include demand attribute weights; a screening module configured to screen, from the plurality of reference classes, a reference class similar to the target product as a similar class; a prediction module configured to predict the future demand of the target product in a future period based on the attribute data of the demand attributes of each reference product in the similar class; The second determining module is configured to determine the future replenishment quantity of the target product in the future period according to the future demand quantity of the target product.

35. A device for determining product replenishment quantity, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the method for determining a product replenishment quantity according to any one of claims 1 to 33 based on instructions stored in the memory.

36. A computer storable medium having computer program instructions stored thereon, which, when executed by a processor, implements the method for determining product replenishment quantity according to any one of claims 1 to 33.

Images