Object data prediction method, prediction device and electronic equipment

By obtaining and clustering historical sequence data and element composition information of object categories, the problem of low sales forecasting accuracy in the supply chain due to intergenerational confusion and weak hierarchical relationships is solved, and more accurate prediction results are achieved.

CN119090546APending Publication Date: 2024-12-06LENOVO (BEIJING) LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411215138.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

In supply chain management, there are often intergenerational confusion and weak hierarchical relationships in product sales, resulting in low sales forecasting accuracy and inaccurate prediction results.

Method used

By obtaining historical sequence data and element composition information of multiple object categories, clustering, obtaining clustering information, and making predictions based on this to improve prediction accuracy.

Benefits of technology

Through clustering processing, combining historical sequence data and element composition information of object categories, the prediction accuracy is improved and the prediction results are more accurate.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119090546A_ABST
    Figure CN119090546A_ABST
Patent Text Reader

Abstract

The embodiment of the invention provides an object data prediction method and device and electronic equipment. The method comprises the steps that multiple object types and historical sequence data and element composition information corresponding to each object type are obtained; according to the historical sequence data and the element composition information corresponding to the plurality of object categories, clustering the plurality of object categories to obtain at least one piece of clustering information; each piece of clustering information corresponds to at least one object category; and predicting one or more corresponding target data in the plurality of object categories according to the plurality of object categories, the historical sequence data corresponding to the plurality of object categories and the at least one piece of clustering information.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a data prediction method, a prediction device and an electronic device for an object. Background Art

[0002] In supply chain management, regional planners make product-level sales forecasts based on historical sales data. However, in product sales, there is often confusion between generations, and the intergenerational relationship between products sold at the same time is not strong, resulting in low prediction accuracy and inaccurate prediction results when making sales forecasts. Summary of the invention

[0003] Embodiments of the present disclosure provide a method, a device and an electronic device for predicting data of an object.

[0004] In a first aspect, an embodiment of the present disclosure provides a method for predicting data of an object, the method comprising:

[0005] Acquire multiple object categories and historical sequence data and element composition information corresponding to each of the object categories;

[0006] Clustering the multiple object categories according to the historical sequence data and the element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category;

[0007] One or more target data corresponding to the multiple object categories are predicted according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and the at least one clustering information.

[0008] In a second aspect, an embodiment of the present disclosure provides a prediction device, the prediction device comprising an acquisition unit, a clustering unit and a prediction unit;

[0009] The acquisition unit is configured to acquire multiple object categories and historical sequence data and element composition information corresponding to each of the object categories;

[0010] The clustering unit is configured to cluster the multiple object categories according to the historical sequence data and the element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category;

[0011] The prediction unit is configured to predict one or more corresponding target data in the multiple object categories according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and the at least one clustering information.

[0012] In a third aspect, an embodiment of the present disclosure provides an electronic device, the electronic device comprising a memory and a processor;

[0013] The memory is used to store a computer program that can be run on the processor;

[0014] The processor is configured to execute the method as described in any one of the first aspects when running the computer program.

[0015] In a fourth aspect, an embodiment of the present disclosure provides a computer storage medium, wherein the computer storage medium stores a computer program, and when the computer program is executed by at least one processor, the method as described in any one of the first aspects is implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of product category information provided in an embodiment of the present disclosure;

[0017] Figure 2 A schematic diagram of a flow chart of a method for predicting data of an object provided by an embodiment of the present disclosure;

[0018] Figure 3 A schematic diagram of initial category information provided by an embodiment of the present disclosure;

[0019] Figure 4 A partial schematic diagram of target category information provided by an embodiment of the present disclosure;

[0020] Figure 5 A schematic diagram of the structure of a prediction device provided by an embodiment of the present disclosure;

[0021] Figure 6 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure Figure 1 ;

[0022] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure Figure 2 . DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. It is understood that the specific embodiments described herein are only used to explain the relevant application, rather than to limit the application. It should also be noted that, for the convenience of description, only the parts related to the relevant application are shown in the drawings.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which the present disclosure belongs. The terms used herein are only for the purpose of describing the embodiments of the present disclosure and are not intended to limit the present disclosure.

[0025] In the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it will be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0026] It should be pointed out that the terms "first\second\third" involved in the embodiments of the present disclosure are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present disclosure described here can be implemented in an order other than that illustrated or described here.

[0027] In the supply chain planning stage, planners in each region need to use object category (Product Profile) information to construct hierarchical relationships and make product-level sales forecasts based on historical sales data, such as sales trends in the past one or two years.

[0028] For example, see Figure 1 , which shows a schematic diagram of product category information provided by an embodiment of the present disclosure. Figure 1 As shown in the figure, the product category information includes a tree relationship of four levels, namely the brand level, the series (or product series, Product Series) level, the central processing unit brand (CPU Brand) level and the family level. Among them, the brand level includes 1 object category (a1); the series level includes 2 object categories (b1, b2); the central processing unit brand level includes 3 object categories (c1, c2, c3); the family level includes 2 object categories (d1, d2). Here, the series level is the next level of the brand level, the central processing unit brand level is the next level of the series level, and the family level is the next level of the central processing unit brand level. c1 and c2 can be considered as the element composition of b1, c3 can be considered as the element composition of b2, and d1 and d2 can be considered as the element composition of c1.

[0029] It should be noted that the hierarchical relationship is effective for all industries of physical products, such as electronic products, automobiles, beauty products, etc. At the same time, the hierarchical relationship can also be applied to some virtual products. The hierarchical relationship can include product levels (such as Figure 1It can also include geographical levels (such as country, province, city), sales channel levels (such as e-commerce platforms, dealers, major customers), etc. Figure 1 In the embodiment, the hierarchical relationship is described by taking electronic products as an example, and the hierarchical relationship specifically refers to the product hierarchical relationship. In addition, this embodiment does not specifically limit the number of hierarchies and the number of object categories included in each hierarchical level.

[0030] In product sales, there may be product line confusion and generational confusion, such as the simultaneous sales of two generations of products. In addition, due to the unknowable external market environment and / or the unstatistical internal promotion rhythm, the intergenerational relationship between products sold at the same time is not strong. As a result, when making sales forecasts, the variance of the data input into machine learning (product category information, historical sales data, etc.) becomes larger, the prediction accuracy is significantly reduced, and the prediction results are inaccurate.

[0031] Based on this, the disclosed embodiment provides a method for predicting data of an object, the method comprising: obtaining multiple object categories and historical sequence data and element composition information corresponding to each object category; clustering the multiple object categories according to the historical sequence data and element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category; predicting one or more corresponding target data in the multiple object categories according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and the at least one clustering information. In this way, the multiple object categories are clustered according to the historical sequence data and element composition information to obtain clustering information, and then when making predictions, not only the influence of the historical sequence data of the object category itself is taken into account, but also the clustering information is combined. Object categories with similar historical sequence data and / or similar element composition information correspond to the same clustering information, which will have an impact on each other, thereby improving the prediction accuracy and making the predicted target data more accurate.

[0032] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings.

[0033] In one embodiment of the present disclosure, see Figure 2 , which shows a flow chart of a method for predicting data of an object provided by an embodiment of the present disclosure. Figure 2 As shown, the method may include:

[0034] S101, obtaining multiple object categories and historical sequence data and element composition information corresponding to each object category.

[0035] It should be noted that the embodiments of the present disclosure provide a data prediction method for an object (or simply referred to as a prediction method), which can be applied to a data prediction device for an object (or simply referred to as a prediction device), or an electronic device integrated with the device. Here, the electronic device can be, for example, a computer, a smart phone, a tablet computer, a laptop computer, a server, etc., and the embodiments of the present disclosure do not specifically limit this.

[0036] In addition, the prediction method can be generalized to all prediction scenarios with intergenerational relationships, such as 3C electronic products, automobiles, beauty products, etc., without specific limitation. For example, the object in the embodiment of the present disclosure is an electronic product, specifically a laptop computer, and the hierarchical relationship specifically refers to the product hierarchical relationship.

[0037] It should also be noted that Figure 1 As shown in the example, the product category information includes multiple levels, each level includes multiple object categories, and the prediction method provided by the embodiment of the present disclosure can perform prediction on any one or more object categories in any level.

[0038] Taking the series level as an example, the multiple object categories contained in the series level may refer to a series of laptop computers. While obtaining the multiple object categories, the historical sequence data and element composition information of each object category are also obtained.

[0039] Among them, the historical sequence data is a sequence composed of at least one historical data, for example, the historical data is the sales data of the series of products (i.e., the object category) in different months / quarters / years, etc. In this case, the historical sequence data uses time as a division node, that is, the historical sequence data includes at least one historical data arranged in chronological order, so it can be called time series data; or, the historical data can also be the sales data of the series of products in different cities. In this case, the historical sequence data uses cities as division nodes, that is, the historical sequence data can refer to time series data, or it can refer to historical data arranged in order of other characteristics, and there is no specific limitation on this. This embodiment takes time series data as an example for explanation, and the time sequence can be the listing time or the sales time, and there is no specific limitation on this. Specifically, the time series data can be the sales data at different time points within the target historical duration. The target historical duration can be N (N is an integer greater than or equal to 1) years before the current time, or M (M is an integer greater than or equal to 1) months, or other specified durations, etc. Exemplarily, the time series data refers to the sales data of laptops from January to June in the past year.

[0040] Element composition information (or simply element composition) represents the component composition information of the object category. For example, for a laptop computer, the element composition information may include elements such as the processor, screen, casing, etc., and may also include information such as the model, manufacturer, and price of these elements.

[0041] Here, the component composition information (corresponding to the element composition information) of the object category is the endogenous property of the object, and the sales data (corresponding to the historical sequence data) is the exogenous property of the object. Endogenous properties refer to the characteristics of the object itself, and exogenous properties refer to the appearance.

[0042] S102: Cluster the multiple object categories according to the historical sequence data and element composition information corresponding to the multiple object categories to obtain at least one clustering information.

[0043] Each piece of clustering information corresponds to at least one object category.

[0044] It should be noted that the process of dividing a collection of physical or abstract objects into multiple classes consisting of similar objects is called clustering. The cluster generated by clustering is a collection of a group of data objects that are similar to objects in the same cluster and different from objects in other clusters. For example, Table 1 shows an example of the object category, historical sequence data, and element composition information obtained in step S101, which are collectively referred to as original data, and Table 2 shows the data after clustering (referred to as target data).

[0045] Table 1

[0046]

[0047] Table 2

[0048]

[0049] For example, the historical sequence data specifically refers to the sales volume of each month from January to June in the past year, and the element composition information specifically refers to the component composition information. As shown in Table 1, four object categories L1, L2, L3 and L4 are known. The historical sequence data corresponding to L1 is C1 and the corresponding element composition information is E1. The historical sequence data corresponding to L2 is C2 and the corresponding element composition information is E2. The historical sequence data corresponding to L3 is C3 and the corresponding element composition information is E3. The historical sequence data corresponding to L4 is C4 and the corresponding element composition information is E4.

[0050] As shown in Table 1, the historical sequence data C1 corresponding to the object category L1 is C1 = (c11, c12, c13, c14, c15, c16), where c11 to c16 represent the sales volume of the object category L1 from January to June, and the rest are similar and will not be elaborated here.

[0051] As shown in Table 2, after clustering the four object categories L1, L2, L3, and L4, they are divided into two categories, and two cluster information A and B are obtained, where cluster information A corresponds to object categories L1 and L2, and cluster information B corresponds to object categories L3 and L4. In other words, the similarity between object categories L1 and L2 is high, and the similarity between object categories L3 and L4 is high, but the difference between object categories L1 and L2 and object categories L3 and L4 is high.

[0052] In this way, after clustering, several object categories with high similarity are clustered into one category, corresponding to the same cluster information. If the similarity between an object category and the other object categories is very low and does not meet the standard of being clustered into one category, then the object category will become a category on its own and correspond to a separate cluster information.

[0053] That is to say, for each clustering information, at least one object category corresponding to the clustering information satisfies the similarity condition; the similarity condition includes: the similarity between an object category corresponding to the clustering information and any other object category does not reach a threshold, or the similarity between multiple object categories corresponding to the clustering information is greater than a threshold.

[0054] It should be noted that the threshold represents the critical value of whether two object categories are similar. If the similarity between two object categories reaches the threshold, the two object categories are considered similar and can be grouped into one category. Otherwise, they are considered dissimilar and are not grouped into one category. The specific value of the threshold is not limited in the embodiments of the present disclosure.

[0055] In some embodiments, obtaining multiple object categories and historical sequence data and element composition information corresponding to each object category may include:

[0056] Acquire multiple object categories of at least one level in the initial category information and historical sequence data and element composition information corresponding to each object category; the element corresponding to the element composition information is a level below at least one level;

[0057] Correspondingly, the method may further include:

[0058] According to at least one clustering information, the category information corresponding to multiple object categories of at least one level is updated, and the category information corresponding to the elements of the next level of at least one level is updated to obtain the target category information.

[0059] It should be noted that the prediction method provided in the embodiment of the present disclosure may specifically be a multi-level supply chain prediction method. The tree structure composed of multiple levels is called initial category information. For example, Figure 1The product category information shown is an initial category information. That is, the initial category information includes one or more levels, each level includes one or more object categories, and there is a superior-subordinate relationship between adjacent levels, for example Figure 1 In the diagram, the series level is the level above the CPU brand level, and the CPU brand level is the level below the series level.

[0060] Here, Figure 3 As shown in the example, the initial category information includes a five-level tree relationship, and the five levels are region, brand, series, CPU brand and family. Specifically, the regional level includes 1 object category (e1); the brand level includes 2 object categories (f1, f2); the series level includes 4 object categories (g1, g2, g3, g4), among which g1 and g2 are the element composition information of f1 at the next level, and g3 and g4 are the element composition information of f2 at the next level; the CPU brand level includes 6 object categories (h1, h2, h3, h4, h5, h6), among which h1 and h2 are the element composition information of g1 at the next level, h3 is the element composition information of g2 at the next level, h4 and h5 are the element composition information of g3 at the next level, and h6 is the element composition information of g4 at the next level; the family level includes 4 object categories (j1, j2, j3, j4), among which j1 and j2 are the element composition information of h1 at the next level, and j3 and j4 are the element composition information of h4 at the next level.

[0061] That is, in the embodiment of the present disclosure, for any object category in any level in the initial category information, the object category in the next level can be used as the element composition information of the object category in the level. For example, the object categories h1 and h2 in the CPU brand level are used as two elements in the element composition information of the object category g1 in the series level.

[0062] It should also be noted that the elements of the next level as element composition information is only an example. The element composition information can also be obtained by other means, or the union of the elements of the next level and the elements obtained by other means can be used as the element composition information of a certain object category. There is no specific limitation on this.

[0063] It should also be noted that in the embodiments of the present disclosure, clustering can be performed on only one level or on multiple levels separately. After re-clustering multiple object categories, the category information corresponding to the multiple object categories in the level can be updated according to the clustering information, and the category information corresponding to the elements of the next level can be updated to update the initial level information to the target category information.

[0064] For example, after clustering the four object categories L1, L2, L3 and L4, cluster information C corresponds to object category L1, and cluster information D corresponds to object categories L2, L3 and L4. That is, the category information corresponding to object category L1 has not been updated, while the category information corresponding to object categories L2, L3 and L4 has been updated.

[0065] For example, Figure 3 Update the series level and CPU brand level in Figure 4 This is a partial schematic diagram of the updated target category information. After clustering, object category g1 and object category g3 are clustered into one category, and these two object categories can be mapped to a new category information g5. Similarly, object category g2 and object category g4 are clustered into one category, and these two object categories can be mapped to a new category information g6. For the next level (i.e., the CPU brand level), elements h1, h2, h4, and h5 serve as the element composition information of the new category information g5. At the same time, in the CPU brand level, elements h1 and h4 are clustered into one category, and these two object categories can be mapped to a new category information h7. Similarly, elements h2 and h5 are clustered into one category, and these two object categories can be mapped to a new category information h8.

[0066] For another example, in the initial data, one category information corresponds to X sub-series products (X is an integer greater than 1). After clustering, in the target data, Y (Y is an integer greater than 0 and less than X) of the X sub-series products are clustered into one category, and the other XY are clustered into another category. Then the Y sub-series products correspond to a new category information, and the XY sub-series products correspond to another new category information. Thus, the category information is updated.

[0067] Further, for step S102, in an example provided in the present disclosure, clustering the multiple object categories according to the historical sequence data and element composition information corresponding to the multiple object categories to obtain at least one clustering information may include:

[0068] Determine similarities between the multiple object categories based on historical sequence data and element composition information corresponding to the multiple object categories;

[0069] Clustering the multiple object categories according to the similarities between the multiple object categories to obtain at least one clustering information;

[0070] The similarity between multiple object categories includes at least one of the following:

[0071] a first similarity between historical sequence data corresponding to a plurality of object categories;

[0072] a second similarity between sequence trends of historical sequence data corresponding to a plurality of object categories;

[0073] The third similarity between the element composition information corresponding to the plurality of object categories.

[0074] It should be noted that the first similarity represents the similarity between the historical sequence data of two object categories; the sequence trend represents the direction, trend, smoothness, etc. of the historical data in the historical sequence data, and the second similarity represents the similarity of the sequence trend; the third similarity represents the similarity of the element composition information of the two object categories. For example, the object categories are different series of laptop computer products, and their element composition information may include the brand, model, specifications, etc. of the CPU, memory, display, etc. The third similarity refers to the overall similarity of these components.

[0075] In some embodiments, the historical sequence data includes at least one historical data, and the step of determining the first similarity may include:

[0076] For each object category, determining an average value of at least one historical data, and obtaining a plurality of average values ​​corresponding to a plurality of object categories;

[0077] The fourth similarity between the multiple average values ​​is determined as the first similarity between the historical sequence data corresponding to the multiple object categories.

[0078] It should be noted that, here, the similarity between the historical data in the historical sequence data can be used to represent the similarity between the historical sequence data. In other embodiments, the similarity can also be calculated by using the median, peak value, valley value, or weighted average value after weighting based on actual factors, and there is no specific limitation on this.

[0079] Taking the average value as an example, as shown in Table 1 above, at least one historical data refers to the sales volume of each month from January to June, and the average value C1_mean of the time series data corresponding to the object category L1 (that is, the average value of c11, c12, c13, c14, c15, and c16), the average value C2_mean of the time series data corresponding to the object category L2 (that is, the average value of c21, c22, c23, c24, c25, and c26), the average value C3_mean of the time series data corresponding to the object category L3 (that is, the average value of c31, c32, c33, c34, c35, and c36), and the average value C4_mean of the time series data corresponding to the object category L4 (that is, the average value of c41, c42, c43, c44, c45, and c46) are calculated. Then, the similarity between each of the four average values ​​C1_mean, C2_mean, C3_mean, and C4_mean is calculated (recorded as the fourth similarity), and the fourth similarity is used as the first similarity, and finally 6 first similarities are obtained. For example, if the fourth similarity between the average value C1_mean and the average value C3_mean is similarity 1, then the first similarity between the object category L1 and the object category L3 is similarity 1.

[0080] In the disclosed embodiment, the fourth similarity may be calculated between two of the four mean values ​​C1_mean, C2_mean, C3_mean, and C4_mean according to the Mean Absolute Error Loss (MAE) algorithm. It is to be understood that MAE is only an example and is not limited to this algorithm.

[0081] In some embodiments, the step of determining the second similarity may include:

[0082] For each object category, determining a sequence trend of the historical sequence data according to the historical sequence data corresponding to the object category;

[0083] A second similarity between sequence trends of the plurality of historical sequence data is determined.

[0084] Exemplarily, the sequence trend is a smooth trend. As shown in Table 1, the smooth trend C1_lifecycle of the time series data corresponding to the object category L1, the smooth trend C2_lifecycle of the time series data corresponding to the object category L2, the smooth trend C3_lifecycle of the time series data corresponding to the object category L3, and the smooth trend C4_lifecycle of the time series data corresponding to the object category L4 are calculated. Then, the second similarities are calculated between the four smooth trends C1_lifecycle, C2_lifecycle, C3_lifecycle, and C4_lifecycle, respectively, to obtain six second similarities.

[0085] It should also be noted that the smoothing trend of data refers to an analysis method that reveals the long-term trend of data by removing noise and random fluctuations in the data. The smoothing trend of data can be a simple smoothing trend, a weighted smoothing trend, or an exponential smoothing trend, which is not specifically limited in this embodiment.

[0086] In the disclosed embodiment, the second similarity may be calculated between each of the four smooth trends C1_lifecycle, C2_lifecycle, C3_lifecycle, and C4_lifecycle according to a dynamic time warping (DTW) algorithm. It is to be understood that DTW is only an example and is not limited to this algorithm.

[0087] In some embodiments, as shown in Table 1, the third similarities are calculated for each of the four element composition information E1, E2, E3, and E4, respectively, to obtain six third similarities. In the disclosed embodiment, the element composition information of the object category includes multiple elements, and the multiple elements form a vector E.

[0088] In the disclosed embodiment, multiple elements in the element composition information can be represented by vectors according to an embedding algorithm, so a vector composed of multiple elements can also be called an embedding vector. Here, the third similarity can also be calculated between two of the four component composition information E1, E2, E3, and E4 according to the mean absolute error loss algorithm MAE. It can be understood that MAE is only an example and is not limited to this algorithm.

[0089] It should be noted that, in the embodiment of the present disclosure, the similarity between two object categories may include a first similarity, a second similarity, and a third similarity, or may include only one or two of the three.

[0090] Alternatively, a target similarity may be calculated based on two or three of the three factors to characterize the similarity between the two object categories. Therefore, in some embodiments, the method may further include:

[0091] For any two object categories, the target similarity between the two object categories is determined according to the corresponding first similarity and / or second similarity and / or third similarity; the similarities between the multiple object categories include each target similarity.

[0092] It should be noted that, any two similarities (the first similarity and the second similarity, or the first similarity and the third similarity, or the second similarity and the third similarity) or all three similarities can be calculated through weighted operation to obtain the target similarity between the two object categories; the target similarity between the two object categories can also be obtained by inputting any two similarities or all three similarities into the calculation model, and there is no specific limitation on this.

[0093] Taking the similarity between object categories including the first similarity, the second similarity and the third similarity as an example, in some embodiments, all the first similarities form a first similarity matrix, all the second similarities form a second similarity matrix, and all the third similarities form a third similarity matrix; clustering the multiple object categories according to the similarities between the multiple object categories to obtain at least one clustering information may include:

[0094] Normalizing the first similarity matrix, the second similarity matrix, and the third similarity matrix;

[0095] Clustering is performed on multiple object categories according to the normalized first similarity matrix, the second similarity matrix, and the third similarity matrix to obtain at least one clustering information.

[0096] It should be noted that, taking 4 object categories as an example, the 4 average values ​​calculated constitute an average value matrix, the 4 sequence trends calculated constitute a sequence trend matrix, and the 4 element composition information calculated constitute an element composition matrix; the 6 first similarities calculated according to the 4 average values ​​constitute a first similarity matrix, and the 6 second similarities calculated according to the 4 sequence trends constitute a second similarity matrix; the 6 third similarities calculated according to the 4 element composition information constitute a third similarity matrix.

[0097] Here, each average value in the average value matrix can be processed by the first algorithm in pairs to obtain a first similarity matrix; each sequence trend in the sequence trend matrix can be processed by the second algorithm in pairs to obtain a second similarity matrix; each element composition information in the element composition matrix can be processed by the first algorithm in pairs to obtain a third similarity matrix.

[0098] In some embodiments, the first algorithm is a mean absolute error loss algorithm MAE, and the second algorithm is a dynamic time warping algorithm DTW.

[0099] Furthermore, the disclosed embodiment can also normalize the first similarity matrix, the second similarity matrix, and the third similarity matrix to unify the scale according to the standard deviation normalization method (also known as the Z-Score normalization method). Here, the Z-Score converts the original data into a standardized Z-Score score by calculating the standard deviation and mean of the data, so that data of different magnitudes can be converted into Z-Score scores of a unified scale for comparison, thereby improving data comparability.

[0100] It should also be noted that the embodiment of the present disclosure can cluster multiple object categories by calculating loss values, so the first similarity matrix can be called a mean loss matrix, the second similarity matrix can be called a life cycle loss matrix, and the third similarity matrix can be called a mean loss matrix.

[0101] In addition, the normalized first similarity matrix, the second similarity matrix, and the third similarity matrix can form a loss space. Clustering multiple object categories according to the normalized first similarity matrix, the second similarity matrix, and the third similarity matrix can include:

[0102] The loss values ​​of different clustering schemes are solved according to the normalized first similarity matrix, the second similarity matrix and the third similarity matrix to obtain a clustering scheme with a loss value less than a loss threshold or a minimum loss value.

[0103] It should be noted that, taking three object categories (L1, L2, L3) as an example, there are the following four clustering schemes:

[0104] Clustering scheme 1: L1, L2 and L3 are one class;

[0105] Clustering scheme 2: L1 is one class, L2 and L3 are one class;

[0106] Clustering scheme 3: L2 is one class, L1 and L3 are one class;

[0107] Clustering scheme 4: L3 is one class, and L1 and L2 are one class.

[0108] At this time, the loss value can be calculated in the loss space according to the three similarity matrices to obtain the loss value of each clustering scheme. The clustering scheme with the smallest loss value is taken as the final clustering scheme, and the clustering information is set.

[0109] It should also be noted that if the number of object categories is huge, calculating the loss values ​​of all clustering schemes may take too long or consume a lot of computing power. At this time, you can also set a loss threshold. When the loss value of a clustering scheme is less than the loss threshold, it will be used as the final clustering scheme. The loss threshold can be set based on actual conditions.

[0110] It should also be noted that multiple object categories can be clustered according to the K-means clustering algorithm to obtain the category grouping with the smallest loss value. Among them, the K-means clustering algorithm is an iterative clustering analysis algorithm, and its steps are: first divide the data into K groups, then randomly select K objects as the initial cluster centers, and then calculate the distance between each object and each seed cluster center, and assign each object to the cluster center closest to it. The cluster center and the objects assigned to them represent a cluster. Each time a sample is assigned, the cluster center of the cluster will be recalculated based on the existing objects in the cluster. This process will be repeated until a certain termination condition is met. The termination condition can be that no (or the minimum number) of objects are reassigned to different clusters, no (or the minimum number) of cluster centers change again, and the error sum of squares is locally minimized. Multiple object categories can also be clustered by calculating the Euclidean distance corresponding to each point in the loss space, which is not specifically limited.

[0111] S103: predict one or more corresponding target data in the multiple object categories according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and at least one clustering information.

[0112] It should be noted that multiple object categories, historical sequence data corresponding to multiple object categories, and at least one clustering information are input into the prediction model, and the prediction model performs prediction and outputs the prediction result. Here, according to actual needs, a certain object category, several object categories, or all object categories can be predicted to obtain the corresponding prediction result. Among them, the prediction result can be the target data, and the target data can be the same type of data as the historical data in the historical sequence data. For example, the historical sequence data refers to sales data, and the target data is the future sales data or sales data sequence.

[0113] Alternatively, the target data may also be element data, in which case the composition information of each element may also be input into the prediction model, and the prediction model performs predictions to select the elements that are most conducive to increasing sales.

[0114] In the disclosed embodiment, the prediction model is a pre-trained model with prediction function, such as a neural network model and the like.

[0115] It should also be noted that when making predictions, it is also possible to make predictions on the components in each level of the aforementioned updated target category information. For example, two object categories in the initial category information are merged into one in the target category information, and sales forecasts can be made for the entire category information obtained after the merger.

[0116] As shown in Table 2, the target data adds clustering information on the basis of the original data. After obtaining the clustering information, the four object categories L1, L2, L3 and L4 and their corresponding time series data, and the two clustering information A and B are input into the prediction model for prediction. The clustering information can be considered as the label of the corresponding object category. Since the clustering information of the object category is added, when predicting one or more of the multiple object categories, the related data of similar object categories will also have an impact, making the prediction result more accurate.

[0117] In summary, the disclosed embodiment provides a prediction method, and the method flow is briefly described as follows:

[0118] For any level in the initial category information, if the current level is known (for example, the current level is Figure 3 The historical sequence data can be time series data (also called category time series), and the historical sequence data of all object categories in the series level are uniformly recorded as the historical sequence data matrix {C_m} (C represents the object category in the series level, m is the number of object categories, for example Figure 3 The element composition information is represented by an embedding vector, and the embedding vectors of the element composition signals of all object categories of the series level are uniformly recorded as the element composition matrix {E_m}.

[0119] Perform the following steps on the current level:

[0120] 1. Calculate the average value C_mean of each historical sequence data and the sequence trend C_lifecycle (here is the smooth trend) of each historical sequence data; m average values ​​C_mean constitute the average value matrix {C_m_mean}, and m sequence trends constitute the sequence trend matrix {C_m_lifecycle};

[0121] 2. Perform MAE processing on the m average values ​​in the average value matrix {C_m_mean} in pairs, and the obtained fourth similarities form the first similarity matrix (mean loss matrix) L_mean;

[0122] 3. Perform DTW processing on the m sequence trends in the sequence trend matrix {C_m_lifecycle} in pairs to obtain the second similarity matrix (lifecycle loss matrix) L_lifecycle;

[0123] 4. Perform MAE processing on the m embedding vectors in the element composition matrix {E_m} in pairs to obtain the third similarity matrix (mean loss matrix) L_emb;

[0124] 5. Normalize the three similarity matrices (i.e., three loss matrices) to unify the scale;

[0125] 6. Solve the loss value in the loss space and obtain the clustering solution with a loss value less than the loss threshold or the minimum loss value. The clustering solution with the minimum loss value is used for prediction.

[0126] The disclosed embodiment provides a prediction method for data-driven reconstruction of object hierarchical relationships, and prediction is performed after obtaining target category information from the endogenous and exogenous properties of objects at each level. Within each level, the similarity between historical sequence data and element composition information is comprehensively considered through the consideration of the average value of historical sequence data corresponding to multiple object categories, the sequence trend of historical sequence data corresponding to multiple object categories, and the element composition information corresponding to multiple object categories, correcting the data offset problem in intergenerational relationships or hierarchical predictions, improving the prediction accuracy, and feeding back the obtained clustering information to the supply chain and product team to optimize the category information.

[0127] In another embodiment of the present disclosure, see Figure 5 , which shows a schematic diagram of the composition structure of a prediction device provided by an embodiment of the present disclosure. Figure 5 As shown, the prediction device 30 includes an acquisition unit 301, a clustering unit 302 and a prediction unit 303;

[0128] An acquisition unit 301 is configured to acquire multiple object categories and historical sequence data and element composition information corresponding to each object category;

[0129] The clustering unit 302 is configured to cluster the multiple object categories according to the historical sequence data and element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category;

[0130] The prediction unit 303 is configured to predict one or more corresponding target data in the multiple object categories according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and at least one clustering information.

[0131] In some embodiments, the acquisition unit 301 is further configured to acquire multiple object categories of at least one level in the initial category information and historical sequence data and element composition information corresponding to each object category.

[0132] The element corresponding to the element composition information is at least one level below.

[0133] In some embodiments, Figure 5As shown, the prediction device 30 may also include an updating unit 304, which is configured to update the category information corresponding to multiple object categories of at least one level and update the category information corresponding to the elements of the next level of at least one level based on at least one clustering information to obtain target category information.

[0134] In some embodiments, the clustering unit 302 is further configured to determine the similarity between multiple object categories based on the historical sequence data and element composition information corresponding to the multiple object categories; cluster the multiple object categories based on the similarity between the multiple object categories to obtain at least one clustering information.

[0135] Among them, the similarity between multiple object categories includes at least one of the following: a first similarity between historical sequence data corresponding to multiple object categories; a second similarity between sequence trends of historical sequence data corresponding to multiple object categories; a third similarity between element composition information corresponding to multiple object categories.

[0136] In some embodiments, the historical sequence data includes at least one historical data, such as Figure 5 As shown, the prediction device 30 may also include a determination unit 305, configured to determine, for each object category, an average value of at least one historical data, and obtain multiple average values ​​corresponding to multiple object categories; and determine a fourth similarity between the multiple average values ​​as a first similarity between the historical sequence data corresponding to the multiple object categories.

[0137] In some embodiments, the determination unit 305 is further configured to determine, for each object category, a sequence trend of the historical sequence data according to the historical sequence data corresponding to the object category; and determine a second similarity between the sequence trends of multiple historical sequence data.

[0138] In some embodiments, the determination unit 305 is further configured to determine, for any two object categories, a target similarity between the two object categories according to the corresponding first similarity and / or second similarity and / or third similarity.

[0139] The similarities between the multiple object categories include each target similarity.

[0140] In some embodiments, all first similarities form a first similarity matrix, all second similarities form a second similarity matrix, and all third similarities form a third similarity matrix. The clustering unit 302 is further configured to normalize the first similarity matrix, the second similarity matrix, and the third similarity matrix; cluster multiple object categories according to the normalized first similarity matrix, the second similarity matrix, and the third similarity matrix to obtain at least one clustering information.

[0141] In some embodiments, for each clustering information, at least one object category corresponding to the clustering information satisfies a similarity condition; the similarity condition includes: the similarity between an object category corresponding to the clustering information and any other object category does not reach a threshold, or the similarity between multiple object categories corresponding to the clustering information is greater than a threshold.

[0142] It can be understood that in this embodiment, a "unit" can be a part of a circuit, a part of a processor, a part of a program or software, etc., and of course it can also be a module, or it can be non-modular. Moreover, the components in this embodiment can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of a software functional module.

[0143] If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0144] Therefore, this embodiment provides a computer storage medium, which stores a computer program. When the computer program is executed by at least one processor, the steps of the data prediction method for the object described in any one of the aforementioned embodiments are implemented.

[0145] Based on the above-mentioned composition of a prediction device 30 and a computer storage medium, see Figure 6 , which shows a schematic diagram of the composition structure of an electronic device provided by an embodiment of the present disclosure Figure 1 .like Figure 6As shown, the electronic device 40 may include: a communication interface 401, a memory 402 and a processor 403; each component is coupled together via a bus system 404. It is understood that the bus system 404 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 404 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 6 Various buses are labeled as bus system 404. Among them, the communication interface 401 is used for receiving and sending signals in the process of sending and receiving information between other external network elements;

[0146] Memory 402, used to store computer programs that can be run on processor 403;

[0147] The processor 403 is configured to execute, when running the computer program:

[0148] Acquire multiple object categories and historical sequence data and element composition information corresponding to each object category;

[0149] Clustering the multiple object categories according to the historical sequence data and element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category;

[0150] One or more corresponding target data of the multiple object categories are predicted according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and at least one clustering information.

[0151] It can be understood that the memory 402 in the embodiment of the present disclosure can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memory 402 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0152] The processor 403 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 403. The above processor 403 may be a general processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present disclosure can be implemented or executed. The general processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present disclosure can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor are combined to perform. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 402, and the processor 403 reads the information in the memory 402 and completes the steps of the above method in combination with its hardware.

[0153] It is understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present disclosure, or a combination thereof.

[0154] For software implementation, the techniques described herein can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0155] Optionally, as another embodiment, the processor 403 is further configured to execute the method described in any one of the aforementioned embodiments when running the computer program.

[0156] In yet another embodiment of the present disclosure, see Figure 7 , which shows a schematic diagram of the composition structure of an electronic device provided by an embodiment of the present disclosure Figure 2 .like Figure 7 As shown, the electronic device 40 at least includes the prediction device 30 described in any one of the aforementioned embodiments.

[0157] In the disclosed embodiment, for the electronic device 40, historical sequence data and element composition information corresponding to multiple object categories are comprehensively considered to obtain clustering information, thereby improving prediction accuracy and making the predicted target data more accurate.

[0158] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the protection scope of the present disclosure.

[0159] It should be noted that in the present disclosure, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0160] The serial numbers of the above-mentioned embodiments of the present disclosure are only for description and do not represent the advantages or disadvantages of the embodiments.

[0161] The methods disclosed in several method embodiments provided in the present disclosure can be arbitrarily combined without conflict to obtain new method embodiments.

[0162] The features disclosed in several product embodiments provided in the present disclosure can be arbitrarily combined without conflict to obtain new product embodiments.

[0163] The features disclosed in several method or device embodiments provided in the present disclosure may be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0164] The above is only a specific embodiment of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present disclosure, which should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A method for predicting data of an object, the method comprising: Acquire multiple object categories and historical sequence data and element composition information corresponding to each of the object categories; Clustering the multiple object categories according to the historical sequence data and the element composition information corresponding to the multiple object categories to obtain at least one clustering information; Each piece of clustering information corresponds to at least one of the object categories; One or more target data corresponding to the multiple object categories are predicted according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and the at least one clustering information.

2. The method according to claim 1, wherein: The acquiring of multiple object categories and historical sequence data and element composition information corresponding to each of the object categories includes: Acquire the plurality of object categories of at least one level in the initial category information and the historical sequence data and the element composition information corresponding to each of the object categories; the element corresponding to the element composition information is a level below the at least one level; The method further comprises: According to the at least one clustering information, the category information corresponding to the multiple object categories of the at least one level is updated, and the category information corresponding to the elements of the next level of the at least one level is updated to obtain the target category information.

3. The method according to claim 1, wherein: The clustering of the multiple object categories according to the historical sequence data and the element composition information corresponding to the multiple object categories to obtain at least one clustering information includes: Determining similarities between the multiple object categories based on the historical sequence data and the element composition information corresponding to the multiple object categories; Clustering the multiple object categories according to the similarities between the multiple object categories to obtain the at least one clustering information; The similarities between the multiple object categories include at least one of the following: a first similarity between the historical sequence data corresponding to the multiple object categories; a second similarity between the sequence trends of the historical sequence data corresponding to the multiple object categories; A third similarity between the element composition information corresponding to the multiple object categories.

4. The method according to claim 3, wherein: The historical sequence data includes at least one historical data, and the step of determining the first similarity includes: For each of the object categories, determining an average value of the at least one historical data, and obtaining a plurality of average values ​​corresponding to the plurality of object categories; The fourth similarity between the multiple average values ​​is determined as the first similarity between the historical sequence data corresponding to the multiple object categories.

5. The method according to claim 3, wherein: The step of determining the second similarity comprises: For each of the object categories, determining a sequence trend of the historical sequence data according to the historical sequence data corresponding to the object category; Determine a second similarity between sequence trends of a plurality of the historical sequence data.

6. The method according to claim 3, wherein: The method further comprises: For any two of the object categories, the target similarity between the two object categories is determined according to the corresponding first similarity and / or the second similarity and / or the third similarity; the similarities between the multiple object categories include each of the target similarities.

7. The method according to claim 3, wherein: All the first similarities form a first similarity matrix, all the second similarities form a second similarity matrix, and all the third similarities form a third similarity matrix; The clustering of the multiple object categories according to the similarities between the multiple object categories to obtain the at least one clustering information includes: Normalizing the first similarity matrix, the second similarity matrix, and the third similarity matrix; The plurality of object categories are clustered according to the normalized first similarity matrix, the second similarity matrix, and the third similarity matrix to obtain the at least one clustering information.

8. The method according to any one of claims 1 to 7, wherein: For each of the clustering information, at least one of the object categories corresponding to the clustering information satisfies a similarity condition; the similarity condition includes: the similarity between one of the object categories corresponding to the clustering information and any other object categories does not reach a threshold, or the similarity between multiple object categories corresponding to the clustering information is greater than the threshold.

9. A prediction device, comprising an acquisition unit, a clustering unit and a prediction unit; The acquisition unit is configured to acquire a plurality of object categories and historical sequence data and element composition information corresponding to each of the object categories; The clustering unit is configured to cluster the multiple object categories according to the historical sequence data and the element composition information corresponding to the multiple object categories to obtain at least one clustering information; each clustering information corresponds to at least one object category; The prediction unit is configured to predict one or more corresponding target data in the multiple object categories according to the multiple object categories, the historical sequence data corresponding to the multiple object categories, and the at least one clustering information.

10. An electronic device, comprising a memory and a processor; The memory is used to store a computer program that can be run on the processor; The processor is configured to execute the method according to any one of claims 1 to 8 when running the computer program.