A commodity determination method, device, electronic device and storage medium

Abstract

An embodiment of the present application provides a method, apparatus, electronic device, and computer storage medium for commodity determination. The method includes: determining a value comparison result of the category to which each candidate commodity belongs; the value comparison result represents a comparison result of the value of the category to which each candidate commodity belongs with a reference value; obtaining a commodity fusing ratio of the category to which each candidate commodity belongs according to a comparison result between the value comparison result and a set threshold; using a trained fusion model to rank each candidate commodity to obtain a ranking result, and determining a first commodity to be fused from each candidate commodity based on the ranking result and the commodity fusing ratio; the fusion model is used to reflect the relationship between the self-characteristics of each candidate commodity and the order volume and / or click-through rate.

Description

Technical Field

[0001] The present application relates to the field of e-commerce operation technology, and in particular to a commodity determination method, device, electronic device and computer storage medium. Background Art

[0002] Due to the operational needs of the e-commerce platform's search business, it is necessary to allocate additional traffic to specific products and improve the display positions of these products in the search ranking results, so as to achieve the business goal of increasing product exposure or clicks. This intervention will disrupt the original search ranking results; however, for some of the promoted products with improved display positions, since they cannot handle the traffic brought by this position after the ranking is improved, their own efficiency is low, resulting in a decline in traffic distribution efficiency, causing a negative impact on the category or the market; in addition, even if some of the promoted products cannot handle the traffic brought by the ranking improvement for a period of time, these products will still continue to be given a high exposure, which will lead to a continuous decline in traffic distribution efficiency, and thus, the business goal of increasing product exposure or clicks cannot be achieved. Summary of the invention

[0003] The present application provides a commodity determination method, device, electronic device and computer storage medium.

[0004] The technical solution of this application is implemented as follows:

[0005] The present application provides a method for determining a product, the method comprising:

[0006] Determine a value comparison result of the category to which each of the selected commodities belongs; the value comparison result indicates a comparison result between the value of the category to which each of the selected commodities belongs and the reference value;

[0007] According to the comparison result between the value comparison result and the set threshold value, the product fuse ratio of the category to which each of the selected products belongs is obtained;

[0008] Using the trained fusion model, each of the candidate products is sorted to obtain a sorting result, and based on the sorting result and the product fuse ratio, the first product that needs to be blown is determined from each of the candidate products; the fusion model is used to reflect the relationship between the characteristics of each candidate product itself and the order volume and / or click-through rate.

[0009] In some embodiments, the method further comprises:

[0010] According to the comparison result between the value comparison result and the set threshold, a second commodity to be downgraded is obtained; the second commodity includes at least one commodity to be selected;

[0011] The second product that needs to be downgraded is downgraded according to historical data related to the click rate of the second product.

[0012] In some embodiments, the set threshold includes a first set threshold and a second set threshold, and obtaining the product fuse ratio of the category to which each selected product belongs according to the comparison result between the value comparison result and the set threshold includes:

[0013] When it is determined that the value comparison result is less than the first set threshold, the increment of the negative days is set to 1; the first set threshold is a value less than 0;

[0014] When it is determined that the value comparison result is greater than or equal to the second set threshold, the increment of the negative days is set to -1; the second set threshold is a value greater than 0;

[0015] When it is determined that the value comparison result is between the first set threshold and the second set threshold, setting the increment of negative days to 0;

[0016] Based on the increment of the negative days, the product fuse ratio of the category to which each of the to-be-selected products belongs is determined.

[0017] In some embodiments, determining the product fuse ratio of the category to which each of the selected products belongs based on the increment of the negative days includes:

[0018] The product of the increment of the negative days and the set value is determined as the product fuse ratio of the category to which each of the selected products belongs; the set value is a value between 0 and 1.

[0019] In some embodiments, the fusion model includes a first model and a second model, and the use of the trained fusion model to sort each of the selected products to obtain a sorting result includes:

[0020] Using the trained first model, scoring each of the selected products to obtain a first scoring result;

[0021] Using the trained second model, scoring each of the selected products to obtain a second scoring result;

[0022] An average score of the first scoring result and the second scoring result is calculated, and the average score is sorted to obtain a sorting result.

[0023] In some embodiments, the first model is used to reflect the relationship between the exposure and order volume of each to-be-selected product.

[0024] In some embodiments, the set threshold includes a third set threshold, and obtaining the second product that needs to be downgraded according to the comparison result between the value comparison result and the set threshold includes:

[0025] Determine the category that needs to be downgraded by comparing the value comparison result with the third set threshold value; the third set threshold value is a value less than 0;

[0026] According to a preset strategy, a second product that needs to be downgraded is selected from the categories that need to be downgraded.

[0027] In some embodiments, the step of demoting the second product to be demoted according to historical data related to the click rate of the second product includes:

[0028] Determining a downgrade ratio of the second product according to historical data related to the click rate of the second product;

[0029] The second commodity that needs to be downgraded is downgraded using the downgrade ratio.

[0030] The embodiment of the present application also proposes a commodity determination device, which includes a determination module, an acquisition module and a processing module, wherein:

[0031] A determination module, used to determine a value comparison result of the category to which each to-be-selected commodity belongs; the value comparison result indicates a comparison result between the value of the category to which each to-be-selected commodity belongs and a reference value;

[0032] An obtaining module, used for obtaining the product fuse ratio of the category to which each of the selected products belongs according to the comparison result of the value comparison result and the set threshold value;

[0033] A processing module is used to use the trained fusion model to sort each of the candidate products to obtain a sorting result, and based on the sorting result and the product fuse ratio, determine the first product that needs to be blown from each of the candidate products; the fusion model is used to reflect the relationship between the characteristics of each candidate product itself and the order volume and / or click-through rate.

[0034] An embodiment of the present application provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the commodity determination method provided by one or more of the aforementioned technical solutions when executing the program.

[0035] An embodiment of the present application provides a computer storage medium, which stores a computer program; after the computer program is executed, it can implement the commodity determination method provided by one or more of the above-mentioned technical solutions.

[0036] The embodiment of the present application proposes a commodity determination method, device, electronic device and computer storage medium, the method comprising: determining a value comparison result of the category to which each to-be-selected commodity belongs; the value comparison result represents a comparison result between the value of the category to which each to-be-selected commodity belongs and a reference value; obtaining a commodity fuse ratio of the category to which each to-be-selected commodity belongs based on a comparison result between the value comparison result and a set threshold; using a trained fusion model to sort each to-be-selected commodity to obtain a sorting result, and based on the sorting result and the commodity fuse ratio, determining a first commodity that needs to be blown from each to-be-selected commodity; the fusion model is used to reflect the relationship between the characteristics of each to-be-selected commodity itself and the order volume and / or click-through rate.

[0037] It can be seen that the embodiment of the present application utilizes the relationship between the value comparison result of the category to which each to-be-selected commodity belongs and the set threshold value to obtain the merchandise fuse ratio of the relevant category; further, the merchandise that needs to be fused in the relevant category is determined according to the merchandise fuse ratio; compared to the related technology which requires adjusting the fuse ratio of each to-be-selected commodity based on manual experience, the embodiment of the present application does not require manual labor to spend a lot of time analyzing data to obtain the merchandise fuse ratio, which can save a lot of human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is a flowchart of a commodity determination method in an embodiment of the present application;

[0039] Figure 2a A schematic diagram of a process for determining a commodity that needs to be fused in an embodiment of the present application;

[0040] Figure 2b This is a schematic diagram of the process of downgrading the selected products in the embodiment of the present application;

[0041] Figure 2c A curve diagram showing the comparison results of the unique visitor (UV) value of the product that needs to be blown before and after the blown in the embodiment of the present application;

[0042] Figure 3 A schematic diagram of the structure of a commodity determination device according to an embodiment of the present application;

[0043] Figure 4 A schematic diagram of the structure of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0044] The present application will be further described in detail below in conjunction with the accompanying drawings and Examples. It should be understood that the embodiments provided herein are only used to explain the present application and are not intended to limit the present application. In addition, the embodiments provided below are partial embodiments for implementing the present application, rather than providing all embodiments for implementing the present application. In the absence of conflict, the technical solutions recorded in the embodiments of the present application can be implemented in any combination.

[0045] It should be noted that, in the embodiments of the present application, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a method or device including a series of elements includes not only the elements explicitly recorded, but also includes other elements not explicitly listed, or also includes elements inherent to the implementation of the method or device. In the absence of further limitations, an element defined by the sentence "includes a ..." does not exclude the presence of other related elements (such as steps in a method or units in a device, for example, a unit may be a part of a circuit, a part of a processor, a part of a program or software, etc.) in the method or device including the element.

[0046] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, I and / or J can represent the three situations that I exists alone, I and J exist at the same time, and J exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of I, J, and R can represent including any one or more elements selected from the set consisting of I, J, and R.

[0047] For example, the commodity determination method provided in the embodiment of the present application includes a series of steps, but the commodity determination method provided in the embodiment of the present application is not limited to the recorded steps. Similarly, the commodity determination device provided in the embodiment of the present application includes a series of modules, but the commodity determination device provided in the embodiment of the present application is not limited to including the modules explicitly recorded, and may also include modules required to obtain relevant time series data or perform processing based on time series data.

[0048] The embodiments of the present application can be applied to a computer system composed of a server, and can operate with many other general or special computing system environments or configurations. Here, the server can be a distributed cloud computing technology environment including a small computer system, a large computer system, and so on.

[0049] Electronic devices such as server terminals can implement corresponding functions through the execution of program modules. Generally, program modules can include routines, programs, target programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in a distributed cloud computing environment, where tasks are performed by remote processing devices connected through a communication network. In a distributed cloud computing environment, program modules can be located on local or remote computing system storage media including storage devices.

[0050] In the related art, in order to identify inefficient products among the selected products, the products that have been downgraded and disconnected are first identified through experience, and then deleted from the product candidate set. However, this identification method is not only labor-intensive, but also requires operations for each business activity. Especially during the promotion period, there are as many as hundreds of business activities, which makes it difficult to realize automatic identification of inefficient products, reducing work efficiency.

[0051] In view of the above technical problems, the following embodiments are proposed.

[0052] In some embodiments of the present application, the commodity determination method can be implemented using a processor in a commodity determination device, and the above-mentioned processor can be at least one of an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor.

[0053] Figure 1 is a flowchart of a commodity determination method in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0054] Step 100: Determine the value comparison result of the category to which each to-be-selected commodity belongs; the value comparison result represents the comparison result between the value of the category to which each to-be-selected commodity belongs and the reference value.

[0055] In the embodiment of the present application, the selected product refers to the activity product whose display position in the search ranking result is improved by allocating additional non-natural traffic, for example, the promoted product; for example, assuming that the initial display position of a certain activity product in the search ranking result is ranked eighth, if the activity product is promoted, the current display position of the activity product may be ranked sixth, and the activity product that was originally ranked sixth is now ranked seventh. Here, the activity product can refer to any type of product traded by the e-commerce platform or seller through the Internet; for example, it can be clothing products, food products, etc.

[0056] Exemplarily, each commodity to be selected may be a commodity in the same business activity, or may be a commodity in multiple different business activities; for example, a business activity may be a new product activity, a sales promotion activity, and the like.

[0057] Exemplarily, each commodity to be selected may belong to the same category, or may belong to multiple different categories; for example, the categories may be clothing, food, etc.

[0058] In the embodiment of the present application, there is no limitation on the method for determining the value comparison result of the category to which each candidate product belongs. For example, the value comparison result of the category to which each candidate product belongs can be determined by an ab experiment, or by other methods. The following is an explanation using the ab experiment as an example.

[0059] For example, when the value comparison result of the category to which each candidate product belongs is determined through an ab experiment, the value comparison result may be a uv value comparison result, a UCVR comparison result, or other indicators that can determine the category to which each candidate product belongs under the ab experiment. The embodiment of the present application does not limit this.

[0060] Exemplarily, when each to-be-selected product belongs to the same category, the UV value of the category to which each to-be-selected product belongs is equal to the ratio of the sales volume under the category to the number of visitors.

[0061] Exemplarily, when each to-be-selected product belongs to multiple different categories, the UV value of each of the multiple categories can be calculated in the above manner, which will not be repeated here.

[0062] Here, when the value comparison result is a UV value comparison result, the reference value can be a reference UV value pre-set according to actual business activities; it can also be the UV value of the category to which each candidate product belongs before the rights are raised. In this way, the UV value comparison result can represent the comparison result of the UV value of the category to which each candidate product belongs and the UV value of the category to which each candidate product belongs before the rights are raised; illustratively, the UV value of the category to which each candidate product belongs before the rights are raised can be calculated in the above manner, which will not be repeated here.

[0063] For example, suppose that category 1 includes products 1 to 10, and the search ranking results of these 10 products are products 1 to 10 in order. If product 8 is promoted, the display position of product 8 is promoted to the display position of product 6. At this time, product 8 is the selected product. Then the UV value of the category to which each selected product belongs represents the UV value of category 1 to which the selected product 8 belongs, that is, the UV value of category 1 after the promotion of product 8; the UV value of the category to which each selected product belongs before the promotion represents the UV value of category 1 before the promotion of product 8, that is, the UV value of category 1 before the promotion of product 8.

[0064] Exemplarily, the UV value of the category to which each selected product belongs represents the UV value corresponding to the experimental group b in the ab experiment; the reference UV value represents the UV value corresponding to the control group a in the ab experiment; at this time, the UV value comparison result = (UV value corresponding to the experimental group b - UV value corresponding to the control group a) / UV value corresponding to the control group a.

[0065] Exemplarily, when the value comparison result is a UCVR comparison result, the reference value may be a reference UCVR pre-set according to actual business activities; or it may be the UCVR of the category to which each candidate product belongs before being entitled. Here, the method for determining the UCVR comparison result is similar to the above-mentioned UV value comparison result and will not be repeated here.

[0066] Exemplarily, when each to-be-selected product belongs to the same category, the UCVR of the category to which each to-be-selected product belongs is equal to the ratio of the total order lines under the category to the number of visitors.

[0067] Step 101: According to the comparison result of the value comparison result and the set threshold value, the product fuse ratio of the category to which each selected product belongs is obtained.

[0068] In one embodiment, the set threshold may include a first set threshold and a second set threshold; wherein the first set threshold is a value less than 0, and the second set threshold is a value greater than 0; illustratively, the first set threshold may also be the inverse of the second set threshold; for example, when the first set threshold is -0.5%, the second set threshold may be 0.5%.

[0069] Exemplarily, for the implementation manner of obtaining the product fusing ratio of each product category to be selected according to the comparison result of the value comparison result and the set threshold, it may be: when it is determined that the value comparison result is less than the first set threshold, the increment of the negative days is set to 1; the first set threshold is a value less than 0; when it is determined that the value comparison result is greater than or equal to the second set threshold, the increment of the negative days is set to -1; the second set threshold is a value greater than 0; when it is determined that the value comparison result is between the first set threshold and the second set threshold, the increment of the negative days is set to 0; based on the increment of the negative days, the product fusing ratio of each product category to be selected is determined; hereinafter, the embodiments of the present application will be described by taking the uv value comparison result as an example.

[0070] Exemplarily, assume that the first set threshold is -0.5% and the second set threshold is 0.5%. When it is determined that the uv value comparison result is less than -0.5%, the increment of the negative days is set to 1; when it is determined that the uv value comparison result is greater than or equal to 0.5%, the increment of the negative days is set to -1; when it is determined that the uv value comparison result is between -0.5% and 0.5%, the increment of the negative days is set to 0.

[0071] Exemplarily, the character n is used to represent the negative days, and the initial value of the negative days n is 0; here, the negative days n are in days; if it is determined on the Tth day that the increment of the negative days is set to 1, then the negative days n on the Tth day is equal to the negative days n on the (T - 1)th day plus the set increment 1; if it is determined on the Tth day that the increment of the negative days is set to -1, then the negative days n on the Tth day is equal to the negative days n on the (T - 1)th day plus the set increment -1; if it is determined on the Tth day that the increment of the negative days is set to 0, then the negative days n on the Tth day is equal to the negative days n on the (T - 1)th day plus the set increment 0.

[0072] In the embodiments of the present application, according to the increment of the negative days determined every day, the negative days n of each day can be correspondingly obtained; thus, based on the negative days n of each day, the product fusing ratio of each product category to be selected can be correspondingly determined; it can be understood that if the negative days on the (T - 1)th day are different from the negative days on the Tth day, then the product fusing ratio of each product category determined on the (T - 1)th day is different from the product fusing ratio of each product category determined on the Tth day.

[0073] It can be seen that in the embodiments of the present application, through the increment of the negative days determined every day, the daily update of the product fusing ratio can be realized, so that the business activities corresponding to the products to be selected do not have a negative impact on the overall market, and the efficiency of traffic distribution is improved.

[0074] In some embodiments, determining the product fuse ratio of the category to which each to-be-selected product belongs based on the increment of negative days can include: multiplying the increment of negative days by a set value to determine the product fuse ratio of the category to which each to-be-selected product belongs; and setting the value to a value between 0 and 1.

[0075] Exemplarily, the set value may be a value between 0 and 1, for example, the set value may be 5%; here, the specific value of the set value may be determined based on actual business activities, and is not limited in the embodiments of the present application.

[0076] In an embodiment of the present application, after obtaining the increment of negative days according to the above steps, the current negative days can be determined based on the increment of negative days; then, the current negative days are multiplied by the set value, and the product of the two is used as the product fuse ratio of the category to which each selected product belongs.

[0077] For example, when the number of negative days on the T-1th day is 1, the increment of the number of negative days on the Tth day is 1, and the set value is 5%, it can be determined that the number of negative days on the Tth day is 2. At this time, the number of negative days on the Tth day, 2, is multiplied by the set value 5%, that is, 10% can be used as the product fuse ratio of the category to which each selected product belongs on the Tth day.

[0078] It can be seen that in the embodiment of the present application, the product circuit-breaking ratio can be automatically determined every day according to the above method, and then the subsequent product circuit-breaking ratio is performed according to the product circuit-breaking ratio determined every day; that is, there is no need to adjust the circuit-breaking ratio of the promotional products based on manual experience. In this way, while improving the circuit-breaking efficiency during the promotional activities, manpower is effectively saved.

[0079] In some embodiments, based on the comparison result of the value comparison result with the set threshold, a second product that needs to be downgraded can also be obtained, wherein the second product includes at least one product to be selected; specifically, the comparison result of the value comparison result with the third set threshold can be used to determine the category that needs to be downgraded; according to the preset strategy, the second product that needs to be downgraded is selected from the category that needs to be downgraded.

[0080] Exemplarily, the above-mentioned set threshold also includes a third set threshold, and the third set threshold is a value less than 0, for example, the third set threshold can be -0.5%; wherein the third set threshold can be the same as the value of the above-mentioned first set threshold.

[0081] In an embodiment of the present application, after obtaining the UV value comparison result, the UV value comparison result is compared with the third set threshold to obtain a comparison result; if it is determined according to the comparison result that the UV value comparison result is less than the third set threshold, then it means that the category to which each to-be-selected commodity belongs is a category that needs to be downgraded; conversely, if it is determined according to the comparison result that the UV value comparison result is greater than or equal to the third set threshold, then it means that the category to which each to-be-selected commodity belongs is not a category that needs to be downgraded.

[0082] In the embodiment of the present application, after determining that the category to which each to-be-selected commodity belongs is a category that needs to be downgraded, a second commodity that needs to be downgraded can be selected from the category that needs to be downgraded according to a preset strategy.

[0083] Here, the preset strategy may be the top 70% of the exposure and the bottom 30% of the click rate of the products in the category that needs to be downgraded. It is understandable that the preset strategy can be adjusted according to the actual activity scenario, and the embodiment of the present application does not limit this.

[0084] For example, assuming that the category that needs to be downgraded is Category 2, and Category 2 includes 100 items to be selected, according to the above preset strategy, 70 (100*70%) items to be selected with the highest exposure can be selected from Category 2, and then 21 items to be selected with the lowest click-through rate can be selected from these 70 items to be selected. In this way, the 21 items to be selected that are finally determined are the second items that need to be downgraded.

[0085] For example, if the UV value comparison result is less than the third set threshold, it means that the traffic distribution efficiency of the category to which the current selected product belongs is decreasing; that is, some of the selected products under this category cannot bear the traffic brought by the ranking improvement; it can be seen that the second product is some selected products that will cause traffic loss; and the embodiment of the present application reduces the weight of these selected products and their exposure, so that the exposure is allocated to the products that bring positive changes in efficiency, thereby improving traffic efficiency. That is, it can solve the problem in the related art that even if the traffic distribution efficiency of some selected products decreases, these products will still be given high exposure.

[0086] Step 102: Use the trained fusion model to sort each candidate product to obtain a sorting result, and based on the sorting result and the product fuse ratio, determine the first product that needs to be fuse-cut from each candidate product.

[0087] Exemplarily, the fusion model is used to reflect the relationship between the characteristics of each to-be-selected product and the order volume and / or click-through rate; the fusion model may include a first model and a second model; wherein the first model is used to reflect the relationship between the characteristics of each to-be-selected product and the order volume, where the characteristics themselves represent characteristics related to the exposure of the product, that is, the first model is used to reflect the relationship between the exposure of each to-be-selected product and the order volume; here, the first model may be referred to as a high-exposure and low-turnover model; the second model may be a click-through rate (CTR) model, which is used to reflect the relationship between the characteristics of each to-be-selected product and the click-through rate.

[0088] It can be seen that the embodiment of the present application can quickly determine the selected products that need to be fused based on the sorting results and the product fused ratio, thereby ensuring the traffic efficiency during the promotion period.

[0089] In some embodiments, using the trained fusion model to sort each product to be selected to obtain a sorting result may include: using the trained first model to score each product to be selected to obtain a first scoring result; using the trained second model to score each product to be selected to obtain a second scoring result; calculating the average score of the first scoring result and the second scoring result, sorting the average score, and obtaining a sorting result.

[0090] Exemplarily, the label data label of the high-exposure and low-speed model (first model) may be defined as follows, wherein the positive sample is defined as shown in expression (1):

[0091]

[0092] The definition of negative samples is shown in expression (2):

[0093]

[0094] Among them, pv test Indicates the exposure of the test bucket, pv base Indicates the exposure of the base bucket; ctr cid3 Indicates the average click rate under the three-level category; orderlines test Indicates the number of orders in the test bucket, orderlines base Indicates the number of orders in the base bucket; here, the test bucket corresponds to the experimental group b in the ab experiment, and the base bucket corresponds to the experimental group a in the ab experiment.

[0095] Exemplarily, when training the above-mentioned high-exposure and low-rotation model, it is first necessary to obtain the corresponding training data set; here, the training data set consists of a large number of paired feature vectors features and corresponding label data label (0-1 labels); in the training data set, if the category to which each to-be-selected product belongs meets the above-mentioned positive sample definition, the label data label is 1; if the category to which each to-be-selected product belongs meets the above-mentioned negative sample definition, the label data label is 0.

[0096] Exemplarily, the feature vectors in the training data set are generated based on the original features; here, the original features can be continuous features, for example, the CTR, conversion rate (Click Value Rate, CVR), IPV (number of times entering the product details page), revenue per thousand impressions (Revenue Per Mille, RPM), gross merchandise volume (Gross Merchandise Volume, GMV), and addcart times (addcart), etc. of each selected product in the third-level category to which it belongs on the same day, in the past 3 days, in the past 7 days, and in the past 15 days; the original features can also be discrete features, such as whether to place advertisements, whether there is site-wide dynamic sales, etc. It should be noted that the embodiment of the present application does not limit the original feature type of the above-mentioned high-exposure and low-conversion model. For example, the above-mentioned continuous features can also include features such as store points and brand points.

[0097] Here, the above original features are screened and cleaned, and then encoded and converted (including feature crossover) to generate feature vectors. The feature vectors and their corresponding 0-1 labels are input into the high-exposure and low-speed model for training to obtain the trained high-exposure and low-speed model.

[0098] Exemplarily, when making predictions, the feature vector associated with each product to be selected is input into the trained high-exposure, low-turnover model, and the model outputs a number between 0 and 1, which is used to rank each product to be selected.

[0099] For example, assuming that after a feature vector related to the selected product 1 is input into the high-exposure and low-rotation model, the output number of the model is 0.8, 0.2 (1-0.8=0.2) is used as the score of the high-exposure and low-rotation model for the selected product 1.

[0100] Exemplarily, after scoring each product to be selected according to the high-exposure-low-rotation model, the score of each product to be selected can be obtained, that is, the first scoring result.

[0101] Exemplarily, the label data label of the CTR model (second model) may be defined as follows, wherein a positive sample is defined as: there is a click in the last 7 days; a negative sample is defined as: there is no click in the last 7 days.

[0102] Similarly, when training the above CTR model, we first need to obtain the corresponding training data set; here, the training data set consists of a large number of paired feature vectors features and corresponding label data label (0-1 label); in the training data set, if the category to which each candidate product belongs meets the above positive sample definition, the label data label is 1; if the category to which each candidate product belongs meets the above negative sample definition, the label data label is 0.

[0103] Exemplarily, the feature vectors in the training data set are generated based on the original features; here, the original features can be continuous features, which can include product features, for example, the normalized values ​​of the total site sales, the number of good reviews under search, the number of search attention, CVR, CTR, etc. of each selected product on the same day, in the past 3 days, in the past 7 days, and in the past 15 days; it can also include user features, such as the user's age, gender, etc.; it can also include product brand features, such as brand clicks, brand points, etc.; it can also include store features, such as store points, store good review rate, store bad review rate, store weather vane, etc. The original features can also be discrete features, which can include product features, such as whether it is a self-operated product; it can also include user features, such as users of different age segments; it can also include product brand features, such as whether it is a set brand; it can also include store features, such as store stratification. It should be noted that the embodiment of the present application does not limit the original feature type of the above-mentioned CTR model.

[0104] Here, the original features of the above CTR model are screened and cleaned, and then encoded (including feature crossover) to generate feature vectors. The feature vectors and their corresponding 0-1 labels are input into the CTR model for training to obtain a trained CTR model.

[0105] Exemplarily, when making a prediction, a feature vector associated with each candidate product is input into a trained CTR model, and the model outputs a number between 0 and 1, which is used to rank each candidate product.

[0106] For example, assuming that after a feature vector related to the selected product 1 is input into the CTR model, the output number of the model is 0.4, then 0.4 is used as the score of the CTR model for the selected product 1.

[0107] Exemplarily, after scoring each product to be selected according to the CTR model, a score of each product to be selected can be obtained, that is, a second scoring result.

[0108] For example, the score of each selected product can be obtained according to the above-mentioned high exposure and low rotation model and CTR model, that is, the first scoring result and the second scoring result; here, the average score of each selected product in the first scoring result and the second scoring result can be calculated by expression (3); and the average score values ​​are sorted in descending order to obtain the final sorting result.

[0109]

[0110] Here, score represents the above fusion model; score 1 represents the first model mentioned above, that is, the high exposure and low rotation model; score 2 Represents the second model mentioned above, namely the CTR model.

[0111] Exemplarily, assuming that the category that needs to be downgraded is Category 3, and Category 3 includes three selected products, namely, Product 1 to Product 3, if the scoring results of Product 1 to Product 3 through the high exposure and low conversion model are 0.2, 0.5, and 0.3 respectively; the scoring results of Product 1 to Product 3 through the CTR model are 0.4, 0.5, and 0.5 respectively; the average score of Product 1 to Product 3 in the first scoring result and the second scoring result is calculated, and the average score of Product 1 to Product 3 is 0.3, 0.5, and 0.4 respectively; after sorting the above average score values ​​in order from high to low, it can be determined that the final sorting results of Product 1 to Product 3 in Category 3 are 0.3, 0.4, and 0.5.

[0112] In the embodiment of the present application, after the ranking result of each to-be-selected commodity is obtained, the first commodity that needs to be fused is determined from each to-be-selected commodity according to the ranking result and the commodity fused ratio.

[0113] For example, assume that a certain category includes 10 items to be selected, and the final sorting results of these 10 items to be selected are item 1 to item 10 in descending order; if the item fuse ratio of the category is determined to be 10% according to step 101, then the number of the first items that need to be fused can be determined as 1 (10*10%=1); then, the item 10 at the end can be fused.

[0114] It can be seen that the embodiment of the present application scores each candidate product through the first model and the second model respectively, and jointly determines the first product that needs to be blown according to the scoring results of the two, so that the accuracy of product blowing can be improved.

[0115] In some embodiments, the second product that needs to be demoted can also be demoted based on historical data related to the click-through rate of the second product. Specifically, it can include: determining the demotion ratio of the second product based on the historical data related to the click-through rate of the second product; and using the demotion ratio to demote the second product that needs to be demoted.

[0116] For example, firstly, historical data related to the click rate of the second product is obtained. After the historical data related to the click rate of the second product is obtained, the real-time downgrade parameter γ of each candidate product in the second product can be determined according to expression (4). i,t :

[0117]

[0118] Among them, α and β are adjustable coefficients; ctri,t represents the click rate of the i-th selected item at time t, ctri,t represents the click rate of the i-th selected item at time t, and ctr c,t-1 It represents the average click rate of the category to which the i-th selected product belongs at time t-1.

[0119] Here, the above real-time weight reduction parameter γ can be expressed as i,t Make corrections to obtain the corrected real-time weight reduction parameters

[0120]

[0121] According to expression (5), in real-time weight reduction parameter γ i,t When it is greater than or equal to 0.5, it is considered that the corresponding selected product will not cause traffic loss. At this time, the selected product will not be downgraded. On the contrary, when the real-time downgrade parameter γ i,t When it is less than 0.5, the corresponding selected product is considered to be an inefficient product that causes traffic loss; at this time, the real-time downgrade parameter γ it *2 value The downgrade ratio is used as the downgrade ratio for downgrading the candidate product, and the candidate product is downgraded by utilizing the downgrade ratio.

[0122] Exemplarily, historical data related to the click-through rate of the second product can be obtained once every hour, and the downgrade ratio of each selected product can be determined according to the above formula; in this way, the embodiment of the present application can downgrade the products that cause real-time efficiency losses in the market and categories to a certain extent by statistically analyzing the hourly feedback historical data of the selected products.

[0123] The embodiment of the present application proposes a commodity determination method, device, electronic device and computer storage medium, the method comprising: determining a value comparison result of the category to which each to-be-selected commodity belongs; the value comparison result represents a comparison result between the value of the category to which each to-be-selected commodity belongs and a reference value; obtaining a commodity fuse ratio of the category to which each to-be-selected commodity belongs based on a comparison result between the value comparison result and a set threshold; using a trained fusion model to sort each to-be-selected commodity to obtain a sorting result, and based on the sorting result and the commodity fuse ratio, determining a first commodity that needs to be blown from each to-be-selected commodity; the fusion model is used to reflect the relationship between the characteristics of each to-be-selected commodity itself and the order volume and / or click-through rate. It can be seen that the embodiment of the present application utilizes the relationship between the value comparison result of the category to which each candidate product belongs and the set threshold value to obtain the product fuse ratio of the relevant category; then, the products that need to be fused in the relevant category are determined according to the product fuse ratio; compared with the related technology that requires adjusting the fuse ratio of each candidate product based on manual experience, the embodiment of the present application does not require manual labor to spend a lot of time analyzing data to obtain the product fuse ratio, which can save a lot of human resources; in addition, according to the above embodiment, since the second products that need to be downgraded are some candidate products that will cause traffic loss, the embodiment of the present application can improve the traffic distribution efficiency by downgrading these candidate products.

[0124] In order to better reflect the purpose of this application, based on the above-mentioned embodiments of this application, further explanation is given by taking the right-claimed product as an example. Figure 2a Schematic diagram of a process for determining a commodity that needs to be fused in an embodiment of the present application; Figure 2a As shown, the process includes the following steps:

[0125] Step A1: Turn on the fuse.

[0126] For example, before determining the products that need to be fused, the operator sends a fuse-enabling instruction to the e-commerce platform, and the e-commerce platform activates the fuse after receiving the instruction.

[0127] Step A2: Obtain the trained first model and second model.

[0128] Exemplarily, the historical data of each commodity with rights in the business activities can be used to train the first model and the second model respectively to obtain the trained first model and the second model.

[0129] Step A3: Rank the commodities with the rights being raised by integrating the scores.

[0130] Exemplarily, after obtaining the trained first model and the second model, each of the mentioned weighted products is scored using the two models, and the average score corresponding to each mentioned weighted product is calculated, and the average score corresponding to each mentioned weighted product is sorted in descending order to obtain the scoring sorting result.

[0131] Step A4: Determine the UV value comparison result of the category to which the product to be promoted belongs.

[0132] Exemplarily, the UV value comparison result of each category of each commodity in the business activity under the ab experiment is determined; the UV value comparison result represents the comparison result of the UV value of the category to which each commodity with raised rights belongs and the UV value of the category to which each commodity without raised rights belongs.

[0133] Step A5: Determine whether the uv value comparison result is less than -0.5%.

[0134] Exemplarily, -0.5% represents the first set threshold. If the judgment result is yes, step A6 is executed; otherwise, step A7 is executed.

[0135] Step A6: The increment of negative days is set to 1.

[0136] Exemplarily, the initial value of the negative days n is 0, and when it is determined that the uv value comparison result is less than -0.5%, the increment of the negative days is set to 1.

[0137] Step A7: Determine whether the uv value comparison result is greater than or equal to 0.5%.

[0138] Exemplarily, 0.5% represents the second set threshold. If the judgment result is yes, step A8 is executed; otherwise, step A9 is executed.

[0139] Step A8: The increment of negative days is set to -1.

[0140] Exemplarily, when it is determined that the uv value comparison result is greater than or equal to 0.5%, the increment of the negative days is set to -1.

[0141] Step A9: The increment of negative days is set to 0.

[0142] Exemplarily, when it is determined that the UV value comparison result is less than 0.5%, the increment of the negative days is set to 1.

[0143] Step A10: Determine the product fuse ratio.

[0144] Exemplarily, after obtaining the final number of negative days n according to the above steps, 0.5%*n is used as the product circuit breaker ratio, and the tail of the above scoring and sorting results is fused according to the product circuit breaker ratio, and then, the products that need to be fused are determined; here, the product circuit breaker ratio can be determined once a day based on the UV value data fed back.

[0145] It can be seen that the embodiment of the present application can automatically adjust the product circuit breaker ratio in a timely manner through the daily UV value feedback data of the category to which each right-claimed product belongs. The goal is to ultimately achieve non-negative efficiency of the overall market, improve the efficiency of manual circuit breakers during big promotions and on weekdays, and effectively save manpower.

[0146] For example, the general weight adjustment problem is defined as shown in expression (6):

[0147]

[0148] Here, f(x ij ) represents the business goal; the conversion decrease of the market is represented by the function g(x ij ) indicates that the business goal is maximized when the market decline does not exceed △; the flow obtained by the weighted commodity is S, S∈C, C is a constant; x i Indicates that the exposure of the i-th item being adjusted is greater than 0; x ij It indicates that for the j-th user, the exposure of the i-th adjusted product will be increased.

[0149] Based on the definition shown in the above expression (6), the product fuse target, that is, the product fuse target, is constructed; here, it can be determined that the product fuse target should be consistent with the target of the privilege escalation activity, maximizing the number of clicks under the constraint of no loss in the market, corresponding to expression (7):

[0150]

[0151] Here, Pctr i *x ij It represents the ideal number of clicks for the i-th weighted product for the j-th user.

[0152] According to the product melting target shown in expression (7), we construct the hypothetical conditions. Here, we assume that the traffic distribution of the privilege escalation activity is reasonable and there is no need to fuse the product. However, in reality, the traffic distribution will not reach the optimal solution, resulting in a difference between the actual number of clicks and the ideal number of clicks. The product melting target is to optimize the difference between the two, so there is a sub-target, as shown in expression (8):

[0153]

[0154] here, It represents the actual number of clicks on the i-th weighted product for the j-th user.

[0155] In the embodiment of the present application, the real click count shown in expression (8) can be optimized by disconnecting the first product that is determined to be disconnected; or by downgrading the second product that needs to be downgraded. and ideal click count Pctr i *x ij The difference between the two can achieve the business goal of increasing product exposure or clicks.

[0156] Figure 2b The flowchart of downgrading the rights of the goods whose rights are raised in the embodiment of the present application is as follows: Figure 2b As shown, the flowchart includes five modules, namely, real-time data bus (JDQ) module, Doris module, Mysql module, real-time demotion module and monitoring dashboard module; wherein, the JDQ module is used to perform data tracking on the user's behavior data on the promoted product at the bottom layer; the Doris module is used to obtain the tracking data from the JDQ module and perform real-time data processing, and store the data processing results; the Mysql module is used to obtain the corresponding data processing results from the JDQ module for subsequent data processing, and the data can be data related to the UV value; the real-time demotion module is used to determine the demotion ratio of each promoted product every hour; the monitoring dashboard module is used to view the performance of the second product after being downgraded.

[0157] Exemplarily, the above-mentioned monitoring dashboard module is also used to monitor the product circuit breaking situation in real time, and deposit the result data of each dimension on the data platform, so as to provide a more comprehensive explanation for the product circuit breaking. Here, the monitoring dashboard module can be used to monitor whether the product circuit breaker is effective, which can be determined based on the exposure improvement of the ab experiment, the exposure ratio of the circuit-broken products, and the distribution of the number of downgraded degrees; the monitoring dashboard module can be used to determine whether the efficiency of the product circuit breaker meets the requirements, which can be determined based on the overall efficiency, the efficiency of the category, the UV value under the triggering search term (query), the efficiency of the circuit-broken products and the non-circuited products, the efficiency comparison of the circuit-broken products before and after the circuit breaker, and the comparison of the traffic efficiency triggered by the circuit-broken products before and after the circuit breaker; the monitoring dashboard module can also be used to evaluate the fusion model, which can be determined in the following ways: The recall rate of the circuit breaker = the products that are correctly predicted to be fused / the products that actually need to be fused; The precision rate of the circuit breaker = the products that are correctly predicted to be fused / the products that are predicted to be fused; The missed recall rate of the circuit breaker = the products that need to be fused but are not predicted / the products that actually need to be fused; The products that need to be fused with negative UV values; The products that are correctly predicted to be fused; Here, after the inefficient products that correctly need to be fused are fused, the UV value comparison result of the next day turns from negative to positive, refer to Figure 2c .

[0158] Figure 2c The following is a curve diagram showing the comparison results of the UV values ​​of the product that needs to be fused before and after fusion in the embodiment of the present application, as shown in FIG. Figure 2c As shown in the figure, starting from time t, if the products that need to be fused are not fused, the UV value comparison result under the ab experiment will decrease, that is, the rights-raising products that have not been fused will cause traffic loss to the market; if the products that need to be fused are fused, the UV value comparison result under the ab experiment will increase, that is, after the products are fused, there will be no traffic loss to the market, thus improving the traffic distribution efficiency.

[0159] Figure 3 Schematic diagram of the composition structure of the commodity determination device in the embodiment of the present application. Figure 3 As shown, the device includes: a determination module 300, an acquisition module 301 and a processing module 302, wherein:

[0160] The determination module 300 is used to determine the value comparison result of the category to which each to-be-selected commodity belongs; the value comparison result represents the comparison result between the value of the category to which each to-be-selected commodity belongs and the reference value;

[0161] Obtaining module 301, for obtaining the product fuse ratio of the category to which each of the selected products belongs according to the comparison result of the value comparison result and the set threshold value;

[0162] The processing module 302 is used to use the trained fusion model to sort each of the candidate products to obtain a sorting result, and based on the sorting result and the product fuse ratio, determine the first product that needs to be blown from each of the candidate products; the fusion model is used to reflect the relationship between the characteristics of each candidate product itself and the order volume and / or click-through rate.

[0163] In some embodiments, the processing module 302 is further configured to:

[0164] According to the comparison result between the value comparison result and the set threshold, a second commodity to be downgraded is obtained; the second commodity includes at least one commodity to be selected;

[0165] The second product that needs to be downgraded is downgraded according to historical data related to the click rate of the second product.

[0166] In some embodiments, the set threshold includes a first set threshold and a second set threshold, and the obtaining module 301 is used to obtain the product fuse ratio of the category to which each of the selected products belongs according to the comparison result of the value comparison result and the set threshold, including:

[0167] When it is determined that the value comparison result is less than the first set threshold, the increment of the negative days is set to 1; the first set threshold is a value less than 0;

[0168] When it is determined that the value comparison result is greater than or equal to the second set threshold, the increment of the negative days is set to -1; the second set threshold is a value greater than 0;

[0169] When it is determined that the value comparison result is between the first set threshold and the second set threshold, setting the increment of negative days to 0;

[0170] Based on the increment of the negative days, the product fuse ratio of the category to which each of the to-be-selected products belongs is determined.

[0171] In some embodiments, the obtaining module 301 is used to determine the product fuse ratio of the category to which each of the selected products belongs based on the increment of the negative days, including:

[0172] The product of the increment of the negative days and the set value is determined as the product fuse ratio of the category to which each of the selected products belongs; the set value is a value between 0 and 1.

[0173] In some embodiments, the fusion model includes a first model and a second model, and the processing module 302 is used to sort each of the selected products using the trained fusion model to obtain a sorting result, including:

[0174] Using the trained first model, scoring each of the selected products to obtain a first scoring result;

[0175] Using the trained second model, scoring each of the selected products to obtain a second scoring result;

[0176] An average score of the first scoring result and the second scoring result is calculated, and the average score is sorted to obtain a sorting result.

[0177] In some embodiments, the first model is used to reflect the relationship between the exposure and order volume of each to-be-selected product.

[0178] In some embodiments, the set threshold includes a third set threshold, and the obtaining module 301 is used to obtain the second commodity that needs to be downgraded according to the comparison result between the value comparison result and the set threshold, including:

[0179] Determine the category that needs to be downgraded by comparing the value comparison result with the third set threshold value; the third set threshold value is a value less than 0;

[0180] According to a preset strategy, a second product that needs to be downgraded is selected from the categories that need to be downgraded.

[0181] In some embodiments, the processing module 302 is used to downgrade the second product that needs to be downgraded according to historical data related to the click rate of the second product, including:

[0182] Determining a downgrade ratio of the second product according to historical data related to the click rate of the second product;

[0183] The second commodity that needs to be downgraded is downgraded using the downgrade ratio.

[0184] In practical applications, the above-mentioned determination module 300, obtaining module 301 and processing module 302 can all be implemented by a processor located in an electronic device, and the processor can be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor.

[0185] In addition, each functional module in this embodiment can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or software functional modules.

[0186] 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 relevant technology 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 of this embodiment. The aforementioned storage medium includes: a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, and other media that can store program codes.

[0187] Specifically, the computer program instructions corresponding to a commodity determination method in this embodiment can be stored on a storage medium such as a CD, a hard disk, or a USB flash drive. When the computer program instructions corresponding to a commodity determination method in the storage medium are read or executed by an electronic device, any commodity determination method in the aforementioned embodiments is implemented.

[0188] Based on the same technical concept as the above embodiments, see Figure 4 , which shows an electronic device 400 provided by the present application, which may include: a memory 401 and a processor 402; wherein,

[0189] Memory 401, used for storing computer programs and data;

[0190] The processor 402 is configured to execute a computer program stored in the memory to implement any one of the commodity determination methods in the foregoing embodiments.

[0191] In practical applications, the memory 401 may be a volatile memory, such as RAM; or a non-volatile memory, such as ROM, flash memory, hard disk drive (HDD) or solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 402.

[0192] The processor 402 may be at least one of ASIC, DSP, DSPD, PLD, FPGA, CPU, controller, microcontroller, and microprocessor. It is understandable that for different object attribute determination devices, the electronic device used to implement the processor function may also be other, which is not specifically limited in the embodiments of the present application.

[0193] In some embodiments, the functions or modules included in the device provided in the embodiments of the present application can be used to execute the method described in the above method embodiments. The specific implementation can refer to the description of the above method embodiments. For the sake of brevity, it will not be repeated here.

[0194] The above description of various embodiments tends to emphasize the differences between the various embodiments. The same or similar aspects can be referenced to each other, and for the sake of brevity, they will not be repeated herein.

[0195] The methods disclosed in the various method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0196] The features disclosed in the various product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0197] The features disclosed in the various method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0198] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of hardware embodiments, software embodiments, or embodiments in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage and optical storage, etc.) that contain computer-usable program code.

[0199] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0200] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0201] The above are only preferred embodiments of the present application and are not intended to limit the protection scope of the present application.

Claims

1. A method for determining a product, It is characterized in that The method comprises: Determine the value comparison result of the category to which each candidate product belongs; the value comparison result represents the comparison result between the value of the category to which each candidate product belongs and the reference value; the candidate product represents the active product whose display position in the search ranking result is improved by allocating additional non-natural traffic; According to the comparison result between the value comparison result and the set threshold, the increment of the negative days corresponding to each of the selected products is obtained, and based on the product of the increment of the negative days and the set value, the product fuse ratio of the category to which each of the selected products belongs is determined; Using the trained fusion model, each of the candidate products is sorted to obtain a sorting result, and based on the sorting result and the product fuse ratio, the first product that needs to be blown is determined from each of the candidate products; wherein the fusion model includes a first model and a second model, the first model is used to reflect the relationship between the exposure and order volume of each candidate product, and the second model is used to reflect the relationship between the characteristics of each candidate product itself and the click-through rate.

2. The method according to claim 1, It is characterized in that The method further comprises: According to the comparison result between the value comparison result and the set threshold, a second commodity to be downgraded is obtained; the second commodity includes at least one commodity to be selected; The second product that needs to be downgraded is downgraded according to historical data related to the click rate of the second product.

3. The method according to claim 1, It is characterized in that The set threshold includes a first set threshold and a second set threshold, and the increment of the number of negative days obtained according to the comparison result between the value comparison result and the set threshold includes: When it is determined that the value comparison result is less than the first set threshold, the increment of the negative days is set to 1; the first set threshold is a value less than 0; When it is determined that the value comparison result is greater than or equal to the second set threshold, the increment of the negative days is set to -1; the second set threshold is a value greater than 0; When it is determined that the value comparison result is between the first set threshold and the second set threshold, the increment of the negative days is set to 0.

4. The method according to claim 1, It is characterized in that The method of using the trained fusion model to sort each of the selected products to obtain a sorting result includes: Using the trained first model, scoring each of the selected products to obtain a first scoring result; Using the trained second model, scoring each of the selected products to obtain a second scoring result; An average score of the first scoring result and the second scoring result is calculated, and the average score is sorted to obtain a sorting result.

5. The method according to claim 2, It is characterized in that The set threshold includes a third set threshold, and obtaining the second commodity that needs to be downgraded according to the comparison result between the value comparison result and the set threshold includes: Determine the category that needs to be downgraded by comparing the value comparison result with the third set threshold value; the third set threshold value is a value less than 0; According to a preset strategy, a second product that needs to be downgraded is selected from the categories that need to be downgraded.

6. The method according to claim 2, It is characterized in that The step of downgrading the second product to be downgraded according to the historical data related to the click rate of the second product includes: Determining a downgrade ratio of the second product according to historical data related to the click rate of the second product; The second commodity that needs to be downgraded is downgraded using the downgrade ratio.

7. A commodity determination device, It is characterized in that The device comprises: A determination module, used to determine the value comparison result of the category to which each candidate commodity belongs; the value comparison result represents the comparison result between the value of the category to which each candidate commodity belongs and the reference value; the candidate commodity represents the active commodity whose display position in the search ranking result is improved by allocating additional non-natural traffic; A obtaining module, used for obtaining the increment of the negative days corresponding to each of the selected commodities according to the comparison result of the value comparison result and the set threshold value, and determining the commodity fuse ratio of the category to which each of the selected commodities belongs based on the product of the increment of the negative days and the set value; A processing module is used to use the trained fusion model to sort each of the candidate products to obtain a sorting result, and based on the sorting result and the product fuse ratio, determine the first product that needs to be blown from each of the candidate products; wherein the fusion model includes a first model and a second model, the first model is used to reflect the relationship between the exposure and order volume of each candidate product, and the second model is used to reflect the relationship between the characteristics of each candidate product itself and the click-through rate.

8. An electronic device, It is characterized in that The device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of claims 1 to 6 when executing the program.

9. A computer storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images