Method and device for determining product attribute value, equipment and storage medium

Abstract

The embodiment of the invention provides a method and device for determining a product attribute value, equipment and a storage medium. The method described herein comprises: for a target product, based on an attribute value constraint condition of a target attribute of the target product, determining a plurality of candidate attribute values satisfying the attribute value constraint condition; utilizing a trained output quantity estimation model to determine estimated output quantities corresponding to the candidate attribute values when the attribute values of the target product are the candidate attribute values based on the candidate attribute values; and determining a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product, so that the total output measurement of the product obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target. Therefore, the output quantity of the product can be predicted by utilizing the trained output quantity prediction model based on the attribute value of the product, and then the attribute value of the product is continuously adjusted so as to improve the total output measurement of the product.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computers, and more particularly, to methods, apparatuses, devices, and computer-readable storage media for determining product attribute values. Background Art

[0002] Product attribute values directly influence consumer ordering intentions, and multiple products may compete with each other. A single product's attribute value not only affects its own output, but also the output of other similar products, ultimately impacting the product's overall output. Therefore, determining the optimal value of a product's attribute values warrants careful consideration. Summary of the Invention

[0003] In a first aspect of the present disclosure, a method for determining a product attribute value is provided. The method comprises: determining, for a target product, a plurality of candidate attribute values that satisfy the attribute value constraints based on the attribute value constraints of a target attribute of the target product; utilizing a trained output quantity estimation model to determine, based on the plurality of candidate attribute values, an estimated output quantity corresponding to each candidate attribute value when the attribute value of the target product is each candidate attribute value, wherein the output quantity indicates the number of external outputs of the target product within a period of time, and the output quantity estimation model is capable of representing the relationship between the attribute value and the output quantity; and determining a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product so that a total output metric of the product, obtained based on the attribute value and the estimated output quantity corresponding to the attribute value, satisfies a preset target.

[0004] In a second aspect of the present disclosure, a device for determining a product attribute value is provided. The device includes: a candidate attribute value determination module configured to determine, for a target product, a plurality of candidate attribute values that satisfy attribute value constraints based on the attribute value constraints of the target attribute of the target product; an output quantity estimation module configured to use a trained output quantity estimation model to determine, based on the plurality of candidate attribute values, an estimated output quantity corresponding to each candidate attribute value when the attribute value of the target product is each candidate attribute value, wherein the output quantity indicates the number of external outputs of the target product within a period of time, and the output quantity estimation model is capable of representing the relationship between the attribute value and the output quantity; and a product attribute value determination module configured to determine, from the plurality of candidate attribute values, a target candidate attribute value as the attribute value of the target product so that a total output metric of the product obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The electronic device includes at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions, when executed by the at least one processing unit, causing the electronic device to perform the method of the first aspect of the present disclosure.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium and can be executed by a processor to perform the method according to the first aspect of the present disclosure.

[0007] It should be understood that the content described in this section is not intended to limit the key features or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other features, advantages and aspects of the various embodiments of the present disclosure will become more apparent hereinafter with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0009] Figure 1 A schematic diagram illustrating an example environment in which embodiments of the present disclosure can be implemented;

[0010] Figure 2 A flowchart illustrating a process of determining a product attribute value according to some embodiments of the present disclosure is shown;

[0011] Figure 3 A block diagram showing an apparatus for determining a product attribute value according to some embodiments of the present disclosure; and

[0012] Figure 4 A block diagram of an electronic device is shown in which one or more embodiments of the present disclosure may be implemented. DETAILED DESCRIPTION

[0013] The following describes embodiments of the present disclosure in more detail with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments described herein. Instead, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0014] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, i.e., "including but not limited to." The term "based on" should be understood as "based at least in part on." The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment." The term "some embodiments" should be understood as "at least some embodiments." Other explicit and implicit definitions may be included below.

[0015] As used herein, the term "attribute" of a product may refer to a property that is unique to or possessed by a product under certain circumstances, such as, but not limited to, size, color, price, material, etc. The term "target attribute" refers to one or more attributes that are considered or of interest when determining output quantities. The term "attribute value" may refer to the numerical value corresponding to the attribute. For example, if the target attribute is price, the attribute value of the target attribute may be the numerical value corresponding to the price.

[0016] As used herein, the term "output" may refer to the number of products output to the outside in a period of time. For example, depending on different application scenarios, "output" may represent the sales volume, order volume, output volume, etc. of a product. The term "total output metric" may refer to a metric that indicates the output of multiple or multiple products to the outside in a period of time. For example, depending on different application scenarios, "total output metric" may represent the total sales volume, total sales (or total sales profit), total order volume, total order amount, total output, total revenue, etc. of multiple or multiple products in a period of time.

[0017] It should be noted that this article uses the example of price as the target attribute, the attribute value of the target attribute (also referred to as the attribute value) as the numerical value corresponding to the price, the output volume as sales volume, and the total output metric as revenue as an example for illustrative description. It is understood that the target attribute, the attribute value of the target attribute, the output volume, and the total output metric can also be any other appropriate content, and this disclosure does not limit this.

[0018] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0019] As used herein, the term "model" can learn the association between corresponding inputs and outputs from training data, so that after training is completed, corresponding outputs can be generated for given inputs. The generation of the model can be based on machine learning technology. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs by using multiple layers of processing units. In this article, "model" may also be referred to as "machine learning model", "machine learning network" or "network", and these terms are used interchangeably in this article.

[0020] As explained above, determining the proper value of product attributes is a critical issue. Some approaches currently employ fitting attribute value elasticity and linear optimization to determine the attribute values of individual products. In scenarios with multiple products, cross-attribute value elasticity is often used to determine whether two products are substitutable or complementary, thereby enabling better pricing decisions.

[0021] The current method for determining product attribute values lacks quantification of errors during the calculation and fitting process, and the resulting attribute values lack robustness, which in turn affects the overall output measurement of the product.

[0022] In view of this, the present disclosure provides a method for determining product attribute values. In this method, for a target product, based on the attribute value constraints of the target attribute of the target product, multiple candidate attribute values that satisfy the attribute value constraints are determined; a trained output quantity estimation model is used to determine, based on the multiple candidate attribute values, the estimated output quantities corresponding to each candidate attribute value when the target product's attribute value is each candidate attribute value, where the output quantity indicates the number of external outputs of the target product within a period of time, and the output quantity estimation model can represent the relationship between the attribute value and the output quantity; and a target candidate attribute value is determined from the multiple candidate attribute values as the attribute value of the target product so that a total product output metric obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

[0023] In this way, the output of a product can be predicted based on the attribute value of the product using the trained output estimation model, and the attribute value of the product can be continuously adjusted to improve the overall output metric of the product.

[0024] Example environment and basic working principles

[0025] Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown. The environment 100 involves a model application device 120, which is configured to estimate the output of a product based on the attribute value of the target attribute of the product (for example, estimating the sales volume of a product based on the numerical value of the price). In this document, unless otherwise specified, the attribute value is also the attribute value of the target attribute. Here, for the sake of ease of discussion, the product for which the output needs to be determined is referred to as the target product. In the embodiment of the present disclosure, the target product can be any product. Specifically, the target product can be a logistics product in the logistics industry. Logistics products can be, for example, express heavy cargo, express less-than-truckload and other products. A trained output estimation model 130 is deployed in the model application device 120. In the embodiment of the present disclosure, the estimated output of the target product is determined with the help of the trained output estimation model 130.

[0026] The output estimation model 130 can be configured as any suitable type of model. In some embodiments, the output estimation model 130 can be a regression model such as XGBRegressor, GBDT, LightGBM, etc. Of course, any other type of model is also applicable.

[0027] In the disclosed embodiment, for different types of products, the output volume estimation model 130 corresponding to the product can be trained based on the product's sample data. Alternatively, for the same product, if the product's pricing is affected by the application scenario, the output volume estimation model 130 for the product in different application scenarios can be trained based on sample data from different application scenarios.

[0028] Model application device 120 can use its deployed output quantity estimation model 130 to estimate the output quantity corresponding to each candidate attribute value for a product. For example, multiple candidate attribute values for a product can be used as input to output quantity estimation model 130, which then processes and outputs the estimated output quantity corresponding to each candidate attribute value.

[0029] In the disclosed embodiment, based on the output volume estimated by the output volume estimation model 130, a total output metric for the product can be further derived based on each candidate attribute value and the corresponding estimated output volume. Furthermore, based on the total output metric for the product, a candidate attribute value whose total output metric satisfies a condition is determined from among the multiple candidate attribute values to serve as the attribute value for the product, i.e., to set the price. For example, the condition for satisfying the total output metric can be maximizing the total output metric. For example, the total output metric can be profit and / or revenue.

[0030] In the environment 100, the model training device 125 is configured to train the output estimation model 130. The training of the output estimation model 130 can be implemented by any appropriate computing system or device, such as a server, a cloud computing device, an edge computing node, etc. In some embodiments, the training of the output estimation model 130 can be completed at the model training device 125, and the trained output estimation model 130 can be provided to the model application device 120 for use. In some embodiments, the training of the output estimation model 130 can be partially or entirely implemented at the model application device 120.

[0031] In environment 100, the model application device 120 or the model training device 125 can be any type of device with computing capabilities, including a terminal device or a server device. The terminal device can be any type of mobile terminal, fixed terminal or portable terminal, including a mobile phone, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, a media computer, a multimedia tablet, a personal communication system (PCS) device, a personal navigation device, a personal digital assistant (PDA), an audio / video player, a digital camera / camcorder, a positioning device, a television receiver, a radio broadcast receiver, an e-book device, a gaming device or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. The server device can, for example, include a computing system / server, such as a mainframe, an edge computing node, an electronic device in a cloud environment, and the like. Although shown as separate devices, the model application device 120 and the model training device 125 can be the same device, or included in the same system.

[0032] It should be understood that the structure and functionality of environment 100 are described for exemplary purposes only and do not imply any limitation on the scope of the present disclosure.

[0033] Example embodiment for determining product attribute values

[0034] As mentioned above, in the embodiment of the present disclosure, the output volume estimation model 130 is used to estimate the output volume of a product based on its attribute values. The following describes the training process of the output volume estimation model 130 corresponding to the logistics product mentioned above as an example.

[0035] In an embodiment of the present disclosure, sample data of a target product may be obtained. The sample data may include, for example, product profile data of the target product. For example, the product profile data may include the product's services, attribute values, weight, volume, and so on.

[0036] Alternatively or additionally, the sample data for the target product may further include at least one historical attribute value of the target product and historical output data corresponding to the historical attribute value. The historical attribute value and the corresponding historical output data may be obtained from order logs. For example, order logs may include historical output information for each product on a daily basis, as well as information such as order type, attribute value, and discount.

[0037] Alternatively or additionally, the sample data of the target product may also include attribute value feature data of other products whose similarity to the target product meets preset conditions, for example, attribute value feature data constructed based on the pricing of similar products by competing merchants.

[0038] Alternatively or additionally, the sample data of the target product may further include product output data associated with the target product. For example, the product output data may include monthly provincial-level express delivery volume, consumer price index, average monthly oil price, and other product output data.

[0039] It is understood that corresponding sample data can be obtained according to the needs of model training. The embodiment of the present disclosure does not limit the type or content of the sample data, etc.

[0040] Furthermore, the acquired sample data of the target product may be preprocessed to obtain feature data of the target product represented by a matrix.

[0041] In the embodiment of the present disclosure, the preprocessing of sample data may include data denoising. Data denoising is the filtering and screening of data containing outliers in the data set. Any appropriate method can be used to achieve denoising. For example, the method that can be adopted is the 3sigma denoising method. Alternatively or additionally, the preprocessing of sample data may also include preprocessing of continuous numerical and categorical discrete features. Specifically, for continuous numerical features, the main methods adopted include constructing a time window and counting the numerical change information within a time period, including the maximum value, minimum value, average value, variance, ratio of the total sum, etc. For some features with large values, normalization is performed to prevent overfitting in the training of subsequent models. For categorical discrete features, the main methods adopted include one-hot coding. That is, a vector containing only 0 or 1 values is constructed, and the vector length can be the total number of categories. For some features with more categorical values, embedding features are constructed by training the model to realize the conversion of feature values to low-dimensional numerical vectors. After all data are subjected to feature preprocessing, the data set will become a feature matrix for subsequent training of the output estimation model 130 .

[0042] Furthermore, the feature data may be used to train the output quantity estimation model 130 , and the trained output quantity estimation model 130 may represent the corresponding relationship between the attribute value and the output quantity of the target product.

[0043] In the embodiment of the present disclosure, as mentioned above, for the case where the target product is a logistics product, the logistics product usually adopts a segmented pricing method. For example, for a certain logistics product, the attribute value of this product is priced in stages based on the weight range of the product to be transported. Specifically, for the weight of the items to be transported that fall within the weight range of 0 to 20 kg, its attribute value is a fixed value. For the weight of the items to be transported that fall within the weight range of 20 to 60 kg, its attribute value is based on the fixed value corresponding to the range of 0 to 20 kg, and the unit price corresponding to the weight gain per unit weight gain is the first weight gain value. Taking this into account, corresponding output volume estimation models can be constructed for different weight ranges.

[0044] In the embodiment of the present disclosure, an output quantity prediction model can be constructed based on a machine learning method, and the relationship between all product attribute values and output quantities can be constructed through a fusion pricing method. As an example, the present disclosure uses the XGBRegressor model to construct an output quantity prediction model, which can learn the correspondence between the attribute values and output quantities of products. In other words, the model can learn the output quantity changes brought about by the changes in the attribute values of products. Therefore, the current output quantity features are used as labels, and the historical output quantity features can be used as features used for training the model. XGBRegressor is an application of XGBoost to regression problems. XGBoost is an additive model. For a given sample x i (In the embodiment of the present disclosure, sample x i It can be the feature data mentioned above), and use XGBoost to its label value (In this embodiment, the label value Predictions can be made for the output quantities mentioned above).

[0045] For example, if F represents the hypothesis space, that is, the set of all decision trees (i.e., F = {f(x) = ω q (x)}(q:R m →T,ω∈R T ), there are K decision trees, f k Represents the kth decision tree (also known as f k ∈F), CART tree is used in XGBoost, and the present disclosure can start from 1 to calculate the f corresponding to K decision trees k (x i ) and determine the accumulated result as the label value The present disclosure can be based on the loss function and the regularization term ∑ k Ω(f kThe loss function of the XGBoost model is determined by the sum of γT and ω. If T represents the number of leaf nodes, ω represents the score of leaf nodes, and γ and λ are manually set hyperparameters, the present disclosure can also be based on γT and To determine the Ω(f) in the regularization term.

[0046] The above describes the training process of the output volume estimation model 130. The following describes the process of pricing products based on the output volume estimation model 130.

[0047] In the embodiment of the present disclosure, attribute value constraints of the target product can be set for the target product. The attribute value constraints of the target product can be set, for example, using the following rules: set upper and lower limits for attribute value changes; set attribute value constraints for multiple products based on the positioning of the target product to complete the segmentation of product attribute values; finally, it is also necessary to consider the attribute value constraints of different products themselves and the constraints on the attribute values of competing products. Take the intersection of all the attribute value constraints obtained to obtain the optimal limiting conditions as the attribute value constraints of the target product. In the following text, the attribute value of the target product is adjusted within the range defined by the attribute value constraints.

[0048] Furthermore, based on the attribute value constraints of the target product, multiple candidate attribute values that satisfy the attribute value constraints can be determined. For example, after determining the optimization constraints, since the attribute value itself has strong constraints (not an infinite decimal), all feasible solutions can be found in a deterministic space as multiple candidate attribute values for the target product.

[0049] Furthermore, the trained output quantity estimation model 130 may be used to determine, based on a plurality of candidate attribute values, the estimated output quantities corresponding to the respective candidate attribute values when the attribute value of the target product is the respective candidate attribute value.

[0050] For example, multiple candidate attribute values can be used as inputs to the output quantity estimation model 130, which outputs an estimated output quantity corresponding to each candidate attribute value. In other words, the output quantity estimation model 130 can be used to output the estimated output quantity corresponding to each candidate attribute value at once. Furthermore, a correspondence between candidate attribute values and corresponding estimated output quantities can be established, that is, a mapping table from attribute values to output quantities can be constructed.

[0051] In this way, the constructed mapping table can be used for subsequent product pricing. When determining the target candidate attribute value from multiple candidate attribute values as the product pricing, the mapping table can be directly read. This can reduce the number of calls to the output estimation model 130 (as mentioned above, the output estimation model 130 only needs to be called once), improve optimization efficiency, and obtain a deterministic optimal solution within a limited range of feasible solutions. In other words, the target candidate attribute value is determined from multiple candidate attribute values as the product pricing.

[0052] Furthermore, a target candidate attribute value may be determined from a plurality of candidate attribute values as the attribute value of the target product. Furthermore, the determined target candidate attribute value must satisfy the following conditions: a total product output metric obtained based on the target candidate attribute value and the estimated output corresponding to the target candidate attribute value satisfies a preset target.

[0053] In some embodiments of the present disclosure, an initial attribute value can be determined from multiple candidate attribute values; if the total output metric of the product obtained based on the initial attribute value and the estimated output quantity corresponding to the initial attribute value does not meet the preset target, a preset iterative algorithm is used to determine the target candidate attribute value from the multiple candidate attribute values based on the initial attribute value as the attribute value of the target product, that is, the pricing of the target product.

[0054] For example, an iterative algorithm, such as a heuristic nonlinear optimization algorithm, can be used to obtain the optimal solution within the optimization constraints. Specifically, a target candidate attribute value can be determined from multiple candidate attribute values for the product, serving as the target product's pricing. In the disclosed embodiments, a simulated annealing optimization algorithm can be used. Specifically, two optimization objectives are: optimizing gross merchandise volume (GMV) and optimizing the balance between GMV and profit.

[0055] In some scenarios, it may involve pricing of multiple different target products with related relationships. For this scenario, when determining the pricing of these multiple target products, if p i Denotes the attribute value of the i-th product after price adjustment, and uses Q(p i ) represents the output of the i-th product, then the p corresponding to N products can be i Q(p i ) to determine the GMV optimization target. i0 Denote the initial attribute value of the i-th product, and use Q(p i0 ) represents the benchmark output under the initial attribute value of the i-th product, c represents the cost of the i-th product, and β represents the optimized proportion of profit. Then, based on the corresponding N products The maximum value of the accumulated results The sum of GMV and profit is used to determine the optimal coordination.

[0056] It is understood that in addition to the simulated annealing algorithm, embodiments of the present disclosure may also employ heuristic optimization algorithms such as evolutionary algorithms, particle swarm optimization algorithms, and ant colony optimization algorithms to determine optimal attribute values. Furthermore, in addition to heuristic optimization algorithms, other optimization algorithms may also be employed, and the present disclosure does not limit the specific implementation of the algorithms.

[0057] In some embodiments of the present disclosure, different value ranges of a preset dimension of a target product correspond to different attribute value constraints, and the trained output quantity estimation model includes multiple output quantity estimation models corresponding to different value ranges. As mentioned above, for logistics products, the weight of the object to be transported needs to be priced separately based on the value range that the weight falls into. For example, for the weight of the object to be transported of 0-20kg, it is necessary to determine the basic attribute value x; for 20kg-60kg, it is necessary to determine the attribute value y per unit weight increase within this range; for above 60kg, it is necessary to determine the attribute value z per unit weight increase within this range. It can be understood that the attribute value per unit weight increase refers to the attribute value per unit weight based on the basic attribute value. For example, if the basic attribute value for the range of 0-20kg is x, and the attribute value for the unit weight increase for the range of 20kg-60kg is y, then for any weight m in the range of 20kg-60kg, its attribute value p = x + y × (m-20).

[0058] In some embodiments of the present disclosure, the different value ranges of the preset dimension of the target product include a first value range and a second value range. There is no intersection between the second value range and the first value range. For example, as mentioned above, 0 to 20 kg is the first value range, and 20 kg to 60 kg is the second value range. Accordingly, the first value range corresponds to the first attribute value constraint and the first output quantity estimation model, and the second value range corresponds to the second attribute value constraint and the second output quantity estimation model. It is understandable that it may further include a third value range, a fourth value range and the corresponding output quantity estimation model, and so on. The present disclosure is not limited to this.

[0059] For the target product, Table 1 shows the model indicator effects of the regression model built using XGBRegressor under segmented modeling.

[0060] Table 1

[0061] Model indicator statistics r2_score mae mape wmape 0-20kg 0.8537 12.9459 0.3351 0.3254 20-60kg 0.8470 11.8790 0.2979 0.2245 60kg and above 0.9469 10.7312 0.2975 0.2060

[0062] Among them, r2_score represents the degree of explanation of the independent variable on the dependent variable in the regression analysis. The closer it is to 1, the better the model fit is; mae represents the mean absolute error; mape represents the mean absolute error percentage; and wmape represents the weighted mean absolute error percentage.

[0063] The following uses segmented modeling to obtain two output estimation models as an example to explain how to determine the attribute values of the target product at each stage (that is, the value range mentioned above).

[0064] For the case where the value of the preset dimension of the target product is within the first value range, a plurality of first candidate attribute values that meet the first attribute value constraint condition can be determined based on the first attribute value constraint condition corresponding to the first value range. Furthermore, a trained first output quantity estimation model corresponding to the first value range can be used to determine the estimated output quantities corresponding to the respective first candidate attribute values of the target product based on these multiple first candidate attribute values. Furthermore, the target first candidate attribute value can also be determined from the multiple first candidate attribute values as the first attribute value corresponding to the value of the preset dimension of the target product within the first preset value range (for example, the basic attribute value x mentioned above) so that the total output metric of the product obtained based on the first attribute value and the estimated output quantity corresponding to the first attribute value meets the first preset target.

[0065] That is, the first output quantity estimation model corresponding to the first value range can be used to obtain the first attribute value of the target product corresponding to the first value range.

[0066] In some scenarios, after determining the first attribute value corresponding to the first value range, the second attribute value corresponding to the second value range (for example, the attribute value y per unit weight gain mentioned above) is further determined. In other words, the attribute value per unit weight gain for the next value range needs to be determined in combination with the already determined basic attribute value x.

[0067] In some embodiments, when the value of a preset dimension of a target product falls within a second value range, multiple second candidate attribute values that satisfy the second attribute value constraint are determined based on a second attribute value constraint corresponding to the second value range. Furthermore, multiple corresponding third candidate attribute values can be obtained based on the sum of the first attribute value and the attribute values of each second candidate attribute value. Furthermore, a trained second output quantity estimation model corresponding to the second value range can be used to determine, based on the multiple third candidate attribute values, the estimated output quantities corresponding to each third candidate attribute value of the target product when the attribute value falls within the second value range. Furthermore, a target third candidate attribute value can be determined from the multiple third candidate attribute values. This target third candidate attribute value satisfies the following condition: a total product output metric derived based on the target third candidate attribute value and the estimated output quantity corresponding to the target third candidate attribute value satisfies a second preset target. Furthermore, a target second candidate attribute value corresponding to the target third candidate attribute value can be determined as the second attribute value of the target product when the value of the preset dimension falls within the second preset value range.

[0068] In some embodiments, for the value of the preset dimension of the target product that is within the second value range, the first attribute value and each second candidate attribute value can be directly used as the input of the trained third output quantity estimation model to determine the corresponding estimated output quantity. In this case, the training of the third output quantity estimation model may be different from the second output quantity estimation model mentioned above. Therefore, it is not necessary to determine the sum of the first attribute value and each second candidate attribute value as the third candidate attribute value. Furthermore, the target second candidate attribute value can be determined from multiple second candidate attribute values as the second attribute value of the unit value when the second value range is used. The total output measurement target also needs to be considered when determining the target second candidate attribute value.

[0069] It is understood that if a third value range, a fourth value range, and so on are included, the attribute value for the next value range needs to be determined based on the attribute value determined for the previously determined value range. For example, if the attribute value (z) per unit weight gain corresponding to the third value range needs to be determined, z needs to be determined based on the already determined base attribute value x and attribute value y per unit weight gain. Similarly, attribute values for multiple value ranges (or stages) can be determined.

[0070] In some embodiments of the present disclosure, the determined attribute values can also be adjusted. Specifically, special cases where no optimal solution is found or where there are separate constraints are handled separately. The attribute values of multiple products are processed by business-based rule-making to meet the rationalization of all attribute values, thereby completing the subsequent attribute value launch. The specific methods adopted are:

[0071] 1. For products whose quality is better than that of their peers but whose attribute values are lower, appropriate price increases should be made.

[0072] 2. For products that have better quality than their peers but have excessively high attribute values, appropriate price reductions should be made.

[0073] 3. For products whose quality is inferior to that of their peers but whose attribute values are high, appropriate price reductions should be made.

[0074] 4. For products whose quality is inferior to that of their peers but whose attribute values are too low, appropriate price increases should be made.

[0075] 5. For different products with large differences in attribute values after differentiation, the attribute values should be readjusted.

[0076] In this way, the embodiments of the present disclosure can predict the output of a product based on the attribute value of the product using a trained output estimation model, and then continuously adjust the attribute value of the product to improve the overall output metric of the product.

[0077] Example Process

[0078] Figure 2 FIG. 2 is a flow chart illustrating a process 200 for determining a product attribute value according to some embodiments of the present disclosure. The process 200 may be implemented at the model application device 120 .

[0079] In block 210 , for a target product, the model application device 120 determines a plurality of candidate attribute values that satisfy the attribute value constraint conditions based on the attribute value constraint conditions of the target attribute of the target product.

[0080] In box 220, the model application device 120 uses the trained output quantity estimation model to determine the estimated output quantity corresponding to each candidate attribute value when the attribute value of the target product is each candidate attribute value based on multiple candidate attribute values. The output quantity indicates the number of external outputs of the target product within a period of time. The output quantity estimation model can represent the relationship between attribute value and output quantity.

[0081] In block 230 , the model application device 120 determines a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product so that a total output metric of the product obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

[0082] In some embodiments, different value ranges of a preset dimension of a target product correspond to different attribute value constraints, and the trained output quantity estimation model includes a plurality of output quantity estimation models corresponding to different value ranges, respectively.

[0083] In some embodiments, the different value ranges of the preset dimension of the target product include a first value range and a second value range, and there is no intersection between the second value range and the first value range; and determining multiple candidate attribute values that satisfy the attribute value constraint includes: for the value of the preset dimension of the target product within the first value range, based on the first attribute value constraint corresponding to the first value range, determining multiple first candidate attribute values that satisfy the first attribute value constraint; and for the value of the preset dimension of the target product within the second value range, based on the second attribute value constraint corresponding to the second value range, determining multiple second candidate attribute values that satisfy the second attribute value constraint.

[0084] In some embodiments, determining the estimated output quantities corresponding to the attribute values of the target product when each candidate attribute value is determined includes: for the value of the preset dimension of the target product within the first value range, using a trained first output quantity estimation model corresponding to the first value range, based on multiple first candidate attribute values, determining the estimated output quantities corresponding to the attribute values of the target product when each first candidate attribute value is determined.

[0085] In some embodiments, determining a target candidate attribute value from multiple candidate attribute values as the attribute value of a target product includes: for the value of a preset dimension of the target product is within a first value range, determining a target first candidate attribute value from multiple first candidate attribute values as the first attribute value corresponding to the value of the preset dimension of the target product within the first preset value range so that the total output metric of the product obtained based on the first attribute value and the estimated output quantity corresponding to the first attribute value meets the first preset target.

[0086] In some embodiments, determining the estimated output quantities corresponding to the attribute values of the target product when they are each candidate attribute value includes: for the value of the preset dimension of the target product within the second value range, obtaining the corresponding multiple third candidate attribute values based on the sum of the first attribute value and the attribute values of each second candidate attribute value; and using the trained second output quantity estimation model corresponding to the second value range, determining the estimated output quantities corresponding to the attribute values of the target product when they are each third candidate attribute value based on the multiple third candidate attribute values.

[0087] In some embodiments, determining a target candidate attribute value from multiple candidate attribute values as the attribute value of the target product includes: for the value of the preset dimension of the target product is within the second value range, determining a target third candidate attribute value from multiple third candidate attribute values, and the target third candidate attribute value satisfies the following conditions: the total output metric of the product obtained based on the target third candidate attribute value and the estimated output quantity corresponding to the target third candidate attribute value meets the second preset target; and determining the target second candidate attribute value corresponding to the target third candidate attribute value as the second attribute value of the unit value when the value of the preset dimension of the target product is within the second preset value range.

[0088] In some embodiments, determining a target candidate attribute value from a plurality of candidate attribute values as the attribute value of a target product includes: determining an initial attribute value from a plurality of candidate attribute values; if a total output metric of the product obtained based on the initial attribute value and the estimated output quantity corresponding to the initial attribute value does not meet a preset target, then determining a target candidate attribute value from a plurality of candidate attribute values using a preset iterative algorithm based on the initial attribute value as the attribute value of the target product.

[0089] In some embodiments, the output quantity prediction model is trained in the following manner: obtaining sample data of the target product, the sample data including one or more of the following: product portrait data of the target product, at least one historical attribute value of the target product and historical output quantity data corresponding to the historical attribute value, attribute value feature data of other products whose similarity to the target product meets preset conditions, and product output data associated with the target product; preprocessing is performed based on the obtained sample data of the target product to obtain feature data of the target product represented by a matrix; and using the feature data to train the output quantity prediction model, the trained output quantity prediction model represents the correspondence between the attribute value and output quantity of the target product.

[0090] Example device

[0091] Figure 3 A block diagram of an apparatus 300 for determining a product attribute value according to some embodiments of the present disclosure is shown. The apparatus 300 may be implemented in the model application device 120. Each module / component in the apparatus 300 may be implemented by hardware, software, firmware, or any combination thereof.

[0092] The device 300 includes a candidate attribute value determination module 310, which is configured to determine, for a target product, a plurality of candidate attribute values that satisfy the attribute value constraint conditions based on the attribute value constraint conditions of the target attribute of the target product. The device 300 also includes an output quantity estimation module 320, which is configured to use a trained output quantity estimation model to determine, based on a plurality of candidate attribute values, the estimated output quantities corresponding to the respective candidate attribute values when the attribute value of the target product is each candidate attribute value. The output quantity indicates the number of external outputs of the target product within a period of time. The output quantity estimation model can represent the relationship between the attribute value and the output quantity. The device 300 also includes a product attribute value determination module 330, which is configured to determine a target candidate attribute value from a plurality of candidate attribute values as the attribute value of the target product so that the total output metric of the product obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

[0093] In some embodiments, different value ranges of a preset dimension of a target product correspond to different attribute value constraints, and the trained output quantity estimation model includes a plurality of output quantity estimation models corresponding to different value ranges, respectively.

[0094] In some embodiments, the different value ranges of the preset dimension of the target product include a first value range and a second value range, and there is no intersection between the second value range and the first value range; and determining multiple candidate attribute values that satisfy the attribute value constraint includes: for the value of the preset dimension of the target product within the first value range, based on the first attribute value constraint corresponding to the first value range, determining multiple first candidate attribute values that satisfy the first attribute value constraint; and for the value of the preset dimension of the target product within the second value range, based on the second attribute value constraint corresponding to the second value range, determining multiple second candidate attribute values that satisfy the second attribute value constraint.

[0095] In some embodiments, determining the estimated output quantities corresponding to the attribute values of the target product when each candidate attribute value is determined includes: for the value of the preset dimension of the target product within the first value range, using a trained first output quantity estimation model corresponding to the first value range, based on multiple first candidate attribute values, determining the estimated output quantities corresponding to the attribute values of the target product when each first candidate attribute value is determined.

[0096] In some embodiments, determining a target candidate attribute value from multiple candidate attribute values as the attribute value of a target product includes: for the value of a preset dimension of the target product is within a first value range, determining a target first candidate attribute value from multiple first candidate attribute values as the first attribute value corresponding to the value of the preset dimension of the target product within the first preset value range so that the total output metric of the product obtained based on the first attribute value and the estimated output quantity corresponding to the first attribute value meets the first preset target.

[0097] In some embodiments, determining the estimated output quantities corresponding to the attribute values of the target product when they are each candidate attribute value includes: for the value of the preset dimension of the target product within the second value range, obtaining the corresponding multiple third candidate attribute values based on the sum of the first attribute value and the attribute values of each second candidate attribute value; and using the trained second output quantity estimation model corresponding to the second value range, determining the estimated output quantities corresponding to the attribute values of the target product when they are each third candidate attribute value based on the multiple third candidate attribute values.

[0098] In some embodiments, determining a target candidate attribute value from multiple candidate attribute values as the attribute value of the target product includes: for the value of the preset dimension of the target product is within the second value range, determining a target third candidate attribute value from multiple third candidate attribute values, and the target third candidate attribute value satisfies the following conditions: the total output metric of the product obtained based on the target third candidate attribute value and the estimated output quantity corresponding to the target third candidate attribute value meets the second preset target; and determining the target second candidate attribute value corresponding to the target third candidate attribute value as the second attribute value of the unit value when the value of the preset dimension of the target product is within the second preset value range.

[0099] In some embodiments, determining a target candidate attribute value from a plurality of candidate attribute values as the attribute value of a target product includes: determining an initial attribute value from a plurality of candidate attribute values; if a total product output metric obtained based on the initial attribute value and the estimated output quantity corresponding to the initial attribute value does not meet a preset target, then determining a target candidate attribute value from a plurality of candidate attribute values based on the initial attribute value using a preset iterative algorithm as the attribute value of the target product.

[0100] In some embodiments, the output quantity prediction model is trained in the following manner: obtaining sample data of the target product, the sample data including one or more of the following: product portrait data of the target product, at least one historical attribute value of the target product and historical output quantity data corresponding to the historical attribute value, attribute value feature data of other products whose similarity to the target product meets preset conditions, and product output data associated with the target product; preprocessing is performed based on the obtained sample data of the target product to obtain feature data of the target product represented by a matrix; and using the feature data to train the output quantity prediction model, the trained output quantity prediction model represents the correspondence between the attribute value and output quantity of the target product.

[0101] The units included in the device 300 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units can be implemented using software and / or firmware, such as machine executable instructions stored on a storage medium. In addition to or as an alternative to machine executable instructions, some or all of the units in the device 300 can be implemented at least in part by one or more hardware logic components. By way of example and not limitation, exemplary types of hardware logic components that can be used include field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0102] Figure 4 4 shows a block diagram of an electronic device 400 in which one or more embodiments of the present disclosure may be implemented. Figure 4The illustrated electronic device 400 is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein. Figure 4 The electronic device 400 shown can be used to implement Figure 1 model application device 120 or model training device 125.

[0103] like Figure 4 As shown, electronic device 400 is in the form of a general electronic device. Components of electronic device 400 may include, but are not limited to, one or more processors or processing units 410, memory 420, storage device 430, one or more communication units 440, one or more input devices 450, and one or more output devices 460. Processing unit 410 may be a real or virtual processor and is capable of performing various processes according to programs stored in memory 420. In a multi-processor system, multiple processing units execute computer-executable instructions in parallel to increase the parallel processing capabilities of electronic device 400.

[0104] The electronic device 400 typically includes a plurality of computer storage media. Such media can be any available media accessible to the electronic device 400, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 420 can be a volatile memory (e.g., registers, cache, random access memory (RAM)), a non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. The storage device 430 can be a removable or non-removable medium and can include a machine-readable medium, such as a flash drive, a disk, or any other medium that can be used to store information and / or data (e.g., training data for training) and can be accessed within the electronic device 400.

[0105] The electronic device 400 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 4 As shown in FIG, a magnetic disk drive for reading from or writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to a bus (not shown) by one or more data media interfaces. Memory 420 may include a computer program product 425 having one or more program modules configured to perform various methods or actions of various embodiments of the present disclosure.

[0106] The communication unit 440 enables communication with other electronic devices via a communication medium. Additionally, the functions of the components of the electronic device 400 can be implemented in a single computing cluster or multiple computing machines that can communicate via a communication connection. Thus, the electronic device 400 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0107] Input device 450 may be one or more input devices, such as a mouse, keyboard, or trackball. Output device 460 may be one or more output devices, such as a display, a speaker, or a printer. Electronic device 400 may also communicate with one or more external devices (not shown) via communication unit 440 as needed, such as a storage device, a display device, or the like, with one or more devices that allow a user to interact with electronic device 400, or with any device that allows electronic device 400 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0108] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which one or more computer instructions are stored, wherein the one or more computer instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0109] Various aspects of the present disclosure are described herein with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products implemented according to the present disclosure. It should be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer-readable program instructions.

[0110] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, such that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device is generated that implements the functions / actions specified in one or more blocks in the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, where these instructions cause the computer, programmable data processing device, and / or other device to operate in a specific manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing various aspects of the functions / actions specified in one or more blocks in the flowchart and / or block diagram.

[0111] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device, so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0112] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple implementations of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a part for a module, program segment or instruction, and a part for a module, program segment or instruction comprises one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0113] While various implementations of the present disclosure have been described above, the foregoing description is intended to be illustrative, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is selected to best explain the principles of the implementations, their practical applications, or improvements to existing technologies, or to enable others skilled in the art to understand the implementations disclosed herein.

Claims

1. A method for determining a product attribute value, comprising: For a target product, based on an attribute value constraint condition of a target attribute of the target product, determining a plurality of candidate attribute values that satisfy the attribute value constraint condition; Determining, using a trained output quantity estimation model, based on the multiple candidate attribute values, an estimated output quantity corresponding to each candidate attribute value when the attribute value of the target product is the target product, wherein the output quantity indicates the number of outputs of the target product to the outside within a period of time, and the output quantity estimation model is capable of representing the relationship between the attribute value and the output quantity; as well as A target candidate attribute value is determined from the plurality of candidate attribute values as the attribute value of the target product so that a total output metric of the product obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

2. According to the method according to claim 1, different value ranges of the preset dimensions of the target product correspond to different attribute value constraints, and the trained output quantity estimation model includes multiple output quantity estimation models corresponding to the different value ranges.

3. The method according to claim 2, wherein the different value ranges of the preset dimension of the target product include a first value range and a second value range, and the second value range has no intersection with the first value range; And the determining of multiple candidate attribute values that satisfy the attribute value constraint condition includes: The value of the preset dimension of the target product is within the first value range, Determining, based on a first attribute value constraint corresponding to the first value range, a plurality of first candidate attribute values that satisfy the first attribute value constraint; as well as The value of the preset dimension of the target product is within the second value range, Based on a second attribute value constraint corresponding to the second value range, a plurality of second candidate attribute values that satisfy the second attribute value constraint are determined.

4. The method according to claim 3, wherein determining the attribute value of the target product as the estimated output quantity corresponding to each candidate attribute value comprises: The value of the preset dimension of the target product is within the first value range, The trained first output quantity estimation model corresponding to the first value range is used to determine the estimated output quantities corresponding to the respective first candidate attribute values when the attribute value of the target product is the first candidate attribute value based on the multiple first candidate attribute values.

5. The method according to claim 4, wherein determining a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product comprises: The value of the preset dimension of the target product is within the first value range, Determine a target first candidate attribute value from the multiple first candidate attribute values as the first attribute value corresponding to the preset dimension of the target product when the value is within the first preset value range so that the total output measurement of the product obtained based on the first attribute value and the estimated output quantity corresponding to the first attribute value meets the first preset target.

6. The method according to claim 5, wherein determining the attribute value of the target product as the estimated output corresponding to each candidate attribute value comprises: The value of the preset dimension of the target product is within the second value range, Obtaining corresponding multiple third candidate attribute values based on the sum of the first attribute value and the attribute values of each of the second candidate attribute values; as well as The trained second output quantity estimation model corresponding to the second value range is used to determine, based on the multiple third candidate attribute values, the estimated output quantities corresponding to the respective third candidate attribute values of the target product.

7. The method according to claim 6, wherein determining a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product comprises: For the target product, if the value of the preset dimension is within the second value range, a target third candidate attribute value is determined from the plurality of third candidate attribute values, and the target third candidate attribute value satisfies the following condition: a total output metric of the product obtained based on the target third candidate attribute value and an estimated output quantity corresponding to the target third candidate attribute value satisfies a second preset target; as well as The target second candidate attribute value corresponding to the target third candidate attribute value is determined as the second attribute value of the unit value when the value of the preset dimension of the target product is within the second preset value range.

8. The method according to claim 1, wherein determining a target candidate attribute value from the plurality of candidate attribute values as the attribute value of the target product comprises: determining an initial attribute value from the plurality of candidate attribute values; If the total output metric of the product obtained based on the initial attribute value and the estimated output quantity corresponding to the initial attribute value does not meet the preset target, the target candidate attribute value is determined from the multiple candidate attribute values using a preset iterative algorithm based on the initial attribute value as the attribute value of the target product.

9. The method according to any one of claims 1 to 8, wherein the output estimation model is trained by: Acquiring sample data of the target product, the sample data including one or more of the following: product portrait data of the target product, at least one historical attribute value of the target product and historical output data corresponding to the historical attribute value, attribute value feature data of other products whose similarity to the target product meets preset conditions, and product output data associated with the target product; Preprocessing is performed based on the acquired sample data of the target product to obtain feature data of the target product represented by a matrix; as well as The output quantity estimation model is trained using the feature data, and the trained output quantity estimation model represents the corresponding relationship between the attribute value and the output quantity of the target product.

10. A device for determining a product attribute value, comprising: a candidate attribute value determination module configured to determine, for a target product, a plurality of candidate attribute values that satisfy the attribute value constraint conditions based on the attribute value constraint conditions of the target attribute of the target product; an output quantity estimation module configured to use a trained output quantity estimation model to determine, based on the plurality of candidate attribute values, an estimated output quantity corresponding to each candidate attribute value when the attribute value of the target product is the same as that of the candidate attribute value, wherein the output quantity indicates the number of external outputs of the target product within a period of time, and the output quantity estimation model is capable of representing the relationship between the attribute value and the output quantity; as well as The product attribute value determination module is configured to determine a target candidate attribute value from the multiple candidate attribute values as the attribute value of the target product so that the product total output metric obtained based on the attribute value and the estimated output quantity corresponding to the attribute value meets a preset target.

11. An electronic device comprising: at least one processing unit; as well as At least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 9 when executed by the at least one processing unit.

12. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 9.

Images