A method and system for fuzzy consolidation processing for a trading platform

Abstract

The application provides a fuzzy merging processing method and system for a transaction platform, and belongs to the technical field of data analysis, and specifically comprises the following steps: based on identifying a risk type of deviation and to-be-listed data of the game accessories of the merchant in different games, determining the association of the to-be-listed game accessories of the target risk type with other merchants, and combining the to-be-listed game accessory data of the target risk type, determining the merging processing merchant in the merchant, and based on the identification result of the batch listing target of the merging processing merchant in different game accessories, determining the identification deviation risk of different game accessories, and combining the identification deviation risk of the game accessories of the merchant with the game accessories, determining the automatic listing control strategy of the game accessories, and realizing the automatic listing processing of the game accessories.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, and in particular relates to a fuzzy merging processing method and system for trading platforms. Background Technology

[0002] In today's booming digital entertainment economy, game item trading platforms have evolved from simple trading venues into comprehensive player ecosystems. Within these communities, experienced players and merchants hold large inventories of fragmented in-game items, but the lack of efficient bulk management tools severely restricts market liquidity.

[0003] To address the aforementioned technical problems, invention patent application CN202310675015.7, "Image Batch Processing and Product Listing Method, Apparatus, Device and Storage Medium," utilizes an image batch processing script to call an image processing module to batch process locally stored images. It can also extract product information from the processed images and then batch publish the images and product information to target online stores, thereby achieving fast, efficient, and customizable product listing and providing a better user experience. However, the above technical solution has the following drawbacks: When batch listing game items, the price factors affecting different game items vary, and the reliability of image recognition models in recognizing these price factors also differs. Therefore, determining an automatic listing control strategy for game items based on the overall recognition deviation risk, and maximizing the overall listing efficiency while reducing the risks of automatic listing, has become an urgent technical problem to be solved.

[0004] Therefore, there is an urgent need for a fuzzy merging processing method and system for trading platforms. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted: Specifically, this application provides a fuzzy merging processing method for a trading platform, which includes: S1 uses the factors influencing the additional price of game items as a basis, and combines the distribution data of unlisted game items in different merchants to determine the identification strategy for batch listing targets of the game items. Based on the identification strategy, batch listing targets are determined. According to the fuzzy merging processing method, the degree of deviation between the image recognition result of the batch listing target and the secondary processing result of the merchant is determined. Using the degree of deviation, the identification deviation risk type of the batch listing target under different additional price influencing factors is determined. S2, based on the identified bias risk type and the merchant's pending listing data for different game items, determines the association between the pending listing game items of the target risk type and other merchants, and, in conjunction with the pending listing game item data of the target risk type, determines the merchants to be merged among the merchants; Based on the identification results of the batch listing targets of the merged processing merchants in different game items, S3 determines the identification deviation risk of different game items, and in combination with the identification deviation risk of the game items of the merchants that have the game items, determines the automatic listing control strategy of the game items.

[0006] The beneficial effects of this invention are as follows: By analyzing the factors influencing the added price of in-game items and the distribution data of unlisted in-game items across different merchants, a strategy for identifying bulk listing targets for in-game items is determined. During the bulk listing process, the current bulk listing demand for in-game items is dynamically determined based on their varying complexity and distribution across different merchants' inventories. This bulk listing demand is then used to determine which identification strategy to employ to efficiently and accurately identify the bulk listing targets, i.e., the difficulty and requirements of the current bulk listing identification process. Based on this difficulty and requirements, a preset identification strategy or a second identification strategy is adopted, i.e., a bulk listing target identification and control scheme. This ensures that the bulk listing targets used for verifying the reliability of the model's identification are linked to the actual listing demand of in-game items, achieving effective control over risk and verification reliability.

[0007] Based on the identification deviation risk of different game items and the identification deviation risk of game items from merchants, an automatic listing control strategy for game items is determined. During the batch listing of game items, the actual listing results of merchants identified as being processed for consolidation are used to statistically analyze the secondary modifications of each item at different merchants, calculate the identification deviation risk coefficient of each item, and, based on this coefficient and the risk distribution of the item in the overall market, determine whether the item is allowed to be automatically listed and the batch size for automatic listing, thus achieving a balance between listing efficiency and risk control.

[0008] Furthermore, the additional price-influencing factors include price-influencing factors other than wear and tear, including gradients and the type of print.

[0009] Furthermore, the distribution data of unlisted game items in different merchants is determined based on the number of game items in the unlisted game items of different merchants.

[0010] Furthermore, the method for determining the identification strategy for bulk listing targets in the game accessories is as follows: S11 determines the number of additional price influencing factors in the game accessories based on the additional price influencing factors, and determines the difficult-to-identify accessories in the game accessories based on the number of additional price influencing factors; S12 Based on the distribution data of unlisted game items in different merchants, determine the merchants that require bulk listing of the difficult-to-identify items; S13 uses the data on difficult-to-identify accessories and the bulk listing requirements of different merchants for difficult-to-identify accessories to determine the identification strategy for bulk listing targets in the game accessories.

[0011] Furthermore, the method for determining the automatic listing control strategy for the game accessories is as follows: S41 Based on the batch listing identification results of the merged processing merchants in different game items, determine that there are merged processing merchants who have made secondary modifications in the game items, and identify them as secondary modification merchants. S42 determines the identification deviation risk coefficient of the game accessories based on the secondary modification merchant data and the batch listing identification data of the merchants. S43 determines the automatic listing control strategy for the game items based on the identification deviation risk coefficient of different game items and the identification deviation risk of game items in merchants where the game items are available.

[0012] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned fuzzy merging processing method for a trading platform when running the computer program.

[0013] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0014] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0015] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0016] Figure 1 This is a flowchart of a fuzzy merging processing method used in trading platforms; Figure 2 This is a flowchart illustrating the method for determining the identification strategy for bulk listing targets in game accessories; Figure 3This is a flowchart illustrating the method for determining the types of bias risks associated with the identification of different additional price influencing factors for bulk product listings; Figure 4 This is a flowchart illustrating the method for determining merchants in the merged processing of merchants. Detailed Implementation

[0017] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0018] Example 1 like Figure 1 As shown, this application provides a fuzzy merging processing method for a trading platform, specifically including: S1 uses the factors influencing the additional price of game items as a basis, and combines the distribution data of unlisted game items in different merchants to determine the identification strategy for batch listing targets of the game items. Based on the identification strategy, batch listing targets are determined. According to the fuzzy merging processing method, the degree of deviation between the image recognition result of the batch listing target and the secondary processing result of the merchant is determined. Using the degree of deviation, the identification deviation risk type of the batch listing target under different additional price influencing factors is determined. S2, based on the identified bias risk type and the merchant's pending listing data for different game items, determines the association between the pending listing game items of the target risk type and other merchants, and, in conjunction with the pending listing game item data of the target risk type, determines the merchants to be merged among the merchants; Based on the identification results of the batch listing targets of the merged processing merchants in different game items, S3 determines the identification deviation risk of different game items, and in combination with the identification deviation risk of the game items of the merchants that have the game items, determines the automatic listing control strategy of the game items.

[0019] Furthermore, the additional price-influencing factors include price-influencing factors other than wear and tear, including gradients and the type of print.

[0020] Furthermore, the distribution data of unlisted game items in different merchants is determined based on the number of game items in the unlisted game items of different merchants.

[0021] Specifically, such as Figure 2As shown, the method for determining the identification strategy for bulk listing targets in the game accessories is as follows: During the bulk listing of game items, the system dynamically determines the appropriate identification strategy to efficiently and accurately identify bulk listing targets based on the complexity of the game items and their distribution across different merchants' inventories. The core logic involves a layered approach: first, based on the number of potential additional price-influencing factors inherent in the game items themselves, items that are difficult to identify are identified. Then, combined with the distribution data of these items in merchants' unlisted inventory, items with bulk listing needs and their corresponding merchants are gradually filtered out. Finally, based on the scope of the demand, the difficulty and requirements of the current bulk listing identification process are determined. Based on this difficulty and requirements, a preset identification strategy or a secondary identification strategy—that is, a bulk listing target identification and control scheme—is adopted.

[0022] S11 determines the number of additional price influencing factors in the game accessories based on the additional price influencing factors, and determines the difficult-to-identify accessories in the game accessories based on the number of additional price influencing factors; It should be noted that the difficult-to-identify items are game items whose number of additional price-influencing factors exceeds a preset threshold. Since the number of image features that need to be recognized and processed by the image recognition model when they are listed in batches is large, they are classified as difficult-to-identify items.

[0023] It is understood that the above steps include the following: Obtain the number of difficult-to-identify items in the game items, and determine whether the proportion of difficult-to-identify items in the game items is greater than a preset proportion threshold. If so, use a preset identification strategy to determine the identification strategy for the batch listing target in the game items. If not, proceed to step S12.

[0024] Additional price-influencing factors: These refer to visual or attribute features, other than wear and tear, that can affect the price of in-game cosmetics. These include, but are not limited to, gradients (such as different skin grades or color gradient types) and types of stickers (such as sticker series, rarity, holographic effects, etc.). These factors increase the complexity of cosmetic images, requiring image recognition models to process more features.

[0025] Difficult-to-identify cosmetic items: These refer to cosmetic items categorized based on skin type, where the number of potential additional price-influencing factors exceeds a preset threshold. The emphasis here is on the total set of all additional price-influencing factors that the cosmetic item can incorporate into the game design, not the actual number carried by a single cosmetic item instance. Because these items may contain numerous variations in image features, image recognition models need to be able to identify and process a large number of feature types during batch release, making identification difficult. Therefore, these cosmetic items are classified as difficult to identify.

[0026] By quantifying the number of additional price-influencing factors that a particular piece of jewelry may contain, we can objectively determine whether the jewelry is generally complex in image recognition. If a high percentage of jewelry is difficult to recognize, it indicates that complex jewelry is the mainstream in the market, and the overall recognition strategy may need to be adjusted, prioritizing a more universal preset recognition strategy.

[0027] Specific example: Suppose the game cosmetics are weapon skins in *Counter-Strike: Global Offensive* (CS:GO). We categorize them based on the skin itself; for example, "AK-47 | Surface Hardened" is one cosmetic, and "M4A4 | Roar" is another. Additional price-influencing factors include the types of stickers (decals) that may appear on the skin and the gradation effects the skin may have. A preset threshold for the number of influencing factors is set to a certain value. For the "AK-47 | Surface Hardened" skin, there are dozens of possible sticker types in the game; therefore, the number of potential price-influencing factors for this skin far exceeds the preset threshold, making it a difficult-to-identify cosmetic. Conversely, for basic skins like "P2000 | Ivory," their simple design and fewer possible sticker types mean that the number of potential price-influencing factors is less than the preset threshold, and it is not considered a difficult-to-identify cosmetic.

[0028] When most jewelry items in the market are difficult to identify (e.g., greater than 0.3), it indicates that complex jewelry items are the mainstream. In this case, using a preset identification strategy can avoid tedious layer-by-layer judgment and directly use common features for batch identification, thereby improving efficiency.

[0029] S12 Based on the distribution data of unlisted game items in different merchants, determine the merchants that require bulk listing of the difficult-to-identify items; In the above steps, the merchants requesting bulk listing of difficult-to-identify accessories are those whose number of difficult-to-identify accessories in the unlisted game accessories exceeds a preset threshold for the number of difficult-to-identify accessories.

[0030] The above steps include the following: S121 identifies the difficult-to-identify accessories from merchants with bulk listing needs as accessories with bulk listing needs, and determines whether the number of accessories with bulk listing needs is greater than the preset threshold for the number of accessories with listing needs. If so, a preset identification strategy is used to determine the identification strategy for the bulk listing target in the game accessories. If not, proceed to step S122. S122 determines whether there are any batch-listing demand products with a number of batch-listing demand merchants that are greater than a preset threshold for the number of batch-listing demand merchants based on the number of batch-listing demand merchants for the batch-listing demand products. If yes, proceed to step S13; otherwise, adopt the second identification strategy to determine the identification strategy for the batch-listing target in the game accessories.

[0031] Unlisted game items: These are game items that a merchant has not yet listed for sale in their store and are usually stored in the merchant's inventory.

[0032] Distribution data: refers to the quantity distribution of various game items among unlisted game items from different merchants, i.e., the list of unlisted items held by each merchant.

[0033] Merchants with bulk listing needs: For a certain difficult-to-identify accessory, if a merchant has not listed more than a preset threshold for the number of difficult-to-identify accessories, then that merchant is considered to have a bulk listing need for that difficult-to-identify accessory. Because the merchant holds a large number of this accessory but has not listed it, it indicates that they may have plans to list it in bulk in the future, and this requires close monitoring.

[0034] By analyzing data from merchants' unlisted inventory, we can identify which merchants have a potential demand for listing specific hard-to-identify jewelry items, thereby determining whether these items are of market importance and prioritizing them in subsequent strategies.

[0035] Specific example: A threshold for the number of difficult-to-list accessories is set to a certain value. Merchant A holds 15 unlisted "AK-47 | Surface Hardened" accessories, exceeding the threshold, so Merchant A is considered a merchant requesting bulk listing of this accessory. Merchant B holds 5 of these accessories, which does not meet the threshold, so they are not considered a requesting merchant.

[0036] Step S121: If there are a large number of merchants in the market who have a need to list difficult-to-identify jewelry in bulk (for example, when the number of jewelry items that need to be listed in bulk is more than 20), it means that there is a wide market demand for these jewelry items. The preset identification strategy can quickly cover the model verification needs of the difficult-to-identify jewelry items that have a need to be listed in bulk.

[0037] Step S122 further filters out those difficult-to-identify jewelry items that are of common interest to multiple merchants. These jewelry items have higher commonalities and may require comprehensive decision-making in S13 by combining merchant information. If the number of merchants demanding a single piece of jewelry is not large, it indicates that its market attention is scattered, and a second identification strategy based on similarity can be adopted to expand the scope.

[0038] Specific example: A threshold for the number of merchants requesting listings is set to a certain value. If there are 4 merchants requesting a batch of listings for the item "AK-47 | Surface Hardened", which is greater than the threshold, then the condition is met, and proceed to S13; if there are at most 2 merchants requesting all game items, which is less than the threshold, then the condition is not met, and the second identification strategy is directly adopted.

[0039] S13 uses the data on difficult-to-identify accessories and the bulk listing requirements of different merchants for difficult-to-identify accessories to determine the identification strategy for bulk listing targets in the game accessories.

[0040] In the above steps, batch listing demand products with a number of merchants whose batch listing demand exceeds a preset threshold are selected as filter demand products. Based on the batch listing demand merchants with the filter demand products, deduplication is performed to obtain duplicate merchants. The demand coefficient is determined based on the proportion of duplicate merchants among all merchants. It is determined whether the demand coefficient is greater than a preset demand coefficient threshold. If it is, a preset identification strategy is used to determine the identification strategy of the batch listing target in the game accessories. If not, a second identification strategy is used to determine the identification strategy of the batch listing target in the game accessories.

[0041] Filter demand products: Select jewelry items that are listed in bulk by more than the preset threshold for the number of merchants with demand for listing as the filter demand products. These are jewelry items that are difficult to identify and are of common interest to many merchants.

[0042] Deduplicated Merchants: Based on merchants who have batch listing needs for products with filtering requirements, these merchants are deduplicated to obtain a set of non-repeating merchants, that is, all merchants that have at least one product with filtering requirements.

[0043] Demand Coefficient: The proportion of unique merchants to all merchants, reflecting the coverage of merchants involved in filtering desired products. The higher the coefficient, the wider the distribution of popular and difficult-to-find accessories among merchants.

[0044] By calculating the percentage of merchants who possess popular, difficult-to-find accessories, we can assess whether these accessories are common. A high percentage indicates that these difficult-to-find accessories are in demand by most merchants, and using a pre-defined identification strategy can maximize effectiveness. A low percentage may require a secondary identification strategy for only a few merchants to avoid overgeneralization.

[0045] Specific example: Assume there are 50 merchants. After S122 filtering, there are 5 products with filtering requirements (e.g., "AK-47 | Surface Hardened", "M4A4 | Roar", "AWP | Medusa", "USP-S | Gunshot Kills", "Glock-18 | Water Spirit"). There are 20 merchants (after deduplication) with batch listing requirements for these products. Therefore, the demand coefficient is 20 / 50 = 0.4. The preset demand coefficient threshold is set to 0.5. If 0.4 is less than 0.5, the requirement is not met, so the second identification strategy is adopted.

[0046] Specifically, the preset identification strategy is to target game items that do not fall under the category of difficult-to-identify items but have the same additional price influencing factors as difficult-to-identify items for bulk listing.

[0047] Game items that are not considered difficult to identify but share the same additional price-influencing factors as those difficult to identify will be targeted for bulk listing. These shared additional price-influencing factors include those from the same game item series. For example, if "AK-47 | Surface Hardened" is identified as a difficult-to-identify item, and its potential additional price-influencing factor includes a "crown" print, then other non-difficult-to-identify items with the same "crown" print probability (such as "P2000 | Ivory," which may also have a "crown" print) will also be considered for bulk listing. This leverages commonalities to avoid the high risk of identification bias that can result from directly using items that are difficult to identify.

[0048] Specifically, the same additional price influencing factors include the additional price influencing factors corresponding to the same series of game accessories.

[0049] It can also be understood that the second identification strategy is to target game accessories that do not fall under the category of difficult-to-identify accessories and have the same additional price influencing factors as the products required for bulk listing.

[0050] Furthermore, the fuzzy merging processing method involves classifying products whose image similarity meets the fuzzy merging similarity threshold into the same product based on the image recognition results, and then performing batch uploading processing.

[0051] Furthermore, the degree of deviation of the merchant's secondary processing result is determined based on the merchant's secondary modification data, which is based on the recognition result of the additional price influencing factors given by the image recognition result.

[0052] Furthermore, such as Figure 3 As shown, the method for determining the identification bias risk type of the batch listing target under different additional price influencing factors is as follows: During the bulk listing of game items, the system utilizes data from merchants' secondary modifications to the identification results to quantify the degree of identification deviation for each additional price-influencing factor. Based on this, risk levels are categorized, with Category I deviation representing the highest risk and requiring priority intervention, Category II deviation the next highest, and Category III deviation the lowest risk. This allows for the priority handling of high-risk factors and the automated release of low-risk factors. The core logic involves statistically analyzing the number of targets modified for each factor and calculating its proportion among all targets containing that factor, forming an identification deviation coefficient. A higher coefficient indicates higher risk, and deviations are categorized into Category I, Category II, and Category III based on the coefficient from highest to lowest.

[0053] S21 determines the secondary modification data of the additional price influencing factors in different batch listing targets based on the degree of deviation; Deviation level: refers to the extent or frequency with which merchants correct the results of additional price influencing factors identified by the image recognition model. Specifically, it is reflected in the modification records of each additional price influencing factor in each batch of listing targets, including the type of factor modified, the number of modifications, and the content before and after the modification.

[0054] Secondary modification data: This refers to summarizing all batch-up targets and, for each additional price influencing factor, counting the number of times it was modified by the merchant and the corresponding list of modified targets, i.e., which targets modified that factor.

[0055] Merchants' secondary modifications provide direct feedback on the accuracy of image recognition, representing the difference between the model's recognition results and the actual product characteristics. By collecting this modification data, we can provide the initial basis for quantifying the recognition bias of each factor, thereby identifying factors where the model performs poorly, especially high-risk factors that are frequently modified.

[0056] Specific example: Suppose there are 200 game items to be listed in a batch. The system identifies 50 items with a "crown" sticker, 30 items with a "gradient" color scheme, and 80 items with a "glitter" sticker. During the review process, the merchant modifies the "crown" sticker 10 times, the "gradient" color scheme 12 times, and the "glitter" sticker 5 times. For each factor, the data for the secondary modifications includes the identifiers of the modified targets and the modified content, such as target ID, the factor before modification, and the factor after modification.

[0057] S22 will take the batch listing target where the additional price influencing factors have been modified twice as the modification target; Modification Target: Refers to the batch listing targets that have been modified by the merchant for a specific additional price influencing factor. The same target may be listed as multiple modification targets because it includes multiple factors.

[0058] To focus on the specific objectives of modifications for each factor, and thus to calculate the percentage of modifications for that factor and assess the severity of identification bias, it is essential to clearly define the modification objectives. Only by clarifying the modification objectives can the modification frequency for each factor be accurately calculated.

[0059] For example: Among 200 batch-uploaded targets, for the "crown" print factor, there are 50 targets with this factor, of which 10 have been modified and are the targets for modification of this factor; for the "gradient color" gradient factor, there are 30 targets with this factor, of which 12 have been modified and are the targets for modification of this factor; for the "glitter" print factor, there are 80 targets with this factor, of which 5 have been modified and are the targets for modification of this factor.

[0060] S23 determines the identification bias risk type of the additional price influencing factors based on the modified target data.

[0061] Specifically, based on the modified target data, the identification bias risk type of the additional price influencing factors is determined, including: Based on the proportion of the modified target among the batch listing targets with the additional price influencing factors, the identification deviation coefficient of the additional price influencing factors is determined; Based on the preset deviation coefficient range in which the identification deviation coefficient is located, the identification deviation risk type of the additional price influencing factor is determined.

[0062] Specifically, the identified bias risk types are classified into three types: Type I bias, Type II bias, and Type III bias.

[0063] Batch listing targets with the aforementioned additional price influencing factors: refers to all batch listing targets identified by the image recognition model as containing this factor, regardless of whether the merchant subsequently modifies them. This is the denominator in calculating the recognition deviation coefficient.

[0064] Identification Bias Coefficient: For a given additional price-influencing factor, the ratio of the number of modification targets to the total number of batch-listed targets containing that factor. This coefficient reflects the frequency with which merchants correct their identification results for that factor; a higher value indicates a greater identification bias and a higher risk.

[0065] Identifying Deviation Risk Types: Based on the preset deviation coefficient range in which the identified deviation coefficient falls, factors are categorized into different risk levels. Type I deviation corresponds to the highest deviation coefficient range, representing the highest risk; Type II deviation corresponds to the medium deviation coefficient range, representing medium risk; and Type III deviation corresponds to the lowest deviation coefficient range, representing the lowest risk. This categorization method allows the platform to prioritize limited resources on high-risk factors of Type I deviation.

[0066] By setting deviation coefficient ranges and classifying risk types from high to low, the reliability of different factors can be graded. This allows limited resources to be prioritized for optimizing or manually reviewing high-risk factors, improving overall identification accuracy and shelf-readiness. Factors with the first type of deviation require the most urgent intervention, followed by the second type, while those with the third type can be considered relatively reliable.

[0067] Specific example: The preset deviation coefficient ranges are as follows: Type I deviation type corresponds to a coefficient between 0.3 and 1 (high-risk range); Type II deviation type corresponds to a coefficient between 0.1 and 0.3 (medium-risk range); and Type III deviation type corresponds to a coefficient between 0 and 0.1 (low-risk range). For the "gradient color" gradient, 12 targets were modified, and 30 targets had this factor. The identification deviation coefficient = 12 / 30 = 0.4, belonging to Type I deviation type. For the "crown" print, the coefficient = 10 / 50 = 0.2, belonging to Type II deviation type. For the "glitter" print, the coefficient = 5 / 80 = 0.0625, belonging to Type III deviation type. This method assesses the identification risk of each factor and clarifies that Type I deviation type is the highest risk level. This provides clear priority decision support for subsequent model iterations and manual review strategy formulation, enabling continuous optimization of the identification system.

[0068] Furthermore, such as Figure 4 As shown, the method for determining the merchants involved in the merged processing is as follows: During the bulk listing of game items, for items held by different merchants, the system dynamically determines whether a merchant can be considered for consolidation based on the number of high-risk price-influencing factors (i.e., type 1 deviation) and the distribution of medium-risk items (target risk type) among different merchants. The core logic is multi-level screening: first, risk levels are classified based on the number of high-risk factors for a single item, and it is determined whether a merchant is directly excluded due to an excessively high proportion of severely risky items; if not excluded, the system further considers the proportion of reliable items, the number of target risk items, and the popularity of target risk items in the overall market to comprehensively determine whether the merchant is a merchant eligible for consolidation. A large variety of target risk items means that the diversity can be used to identify problems and verify reliability, thus favoring consolidation; while a high popularity of target risk items in the overall market (i.e., a high number of items) means that if a deviation occurs, the overall impact risk is too high, thus favoring non-consolidation.

[0069] S31 takes the game items to be listed by the merchant as items to be listed, and determines the number of additional price influencing factors of one type of deviation of the items to be listed based on the identification deviation risk type of the additional price influencing factors of the items to be listed. Items awaiting listing: These refer to game items that the merchant is currently preparing to list but have not yet completed the listing process; that is, items in the merchant's inventory that need to be listed in bulk.

[0070] Additional price influencing factors of a type of deviation: These are the factors with the highest risk level among the identification deviation risk types determined by the preceding steps. They are the factors with the largest identification deviation coefficients and the most likely to be misjudged by the image recognition model, such as a specific print or gradient.

[0071] Number of additional price influencing factors of type one deviation: For each accessory to be listed, count the number of factors of type one deviation among all the additional price influencing factors it contains.

[0072] Type 1 bias factors are the highest-risk factors most likely to lead to identification errors. The number of such factors in a piece of jewelry directly reflects the identification risk that may arise when the jewelry is listed in batches. The higher the number, the more manual intervention is required for the jewelry, and the less suitable it is for fully automated batch merging and listing.

[0073] Specific example: Suppose that after the previous steps, it has been determined that the "gradient color" gradient is a Type I deviation type (high risk), the "crown" print is a Type II deviation type, and the "glitter" print is a Type III deviation type. A merchant's upcoming item includes "AK-47 | Surface Hardened," which may have a "gradient color" gradient and may also have a "crown" print and a "glitter" print. The number of Type I deviation factors is 1 (only the "gradient color" gradient), because "crown" and "glitter" are not in the same category. Another item, "M4A4 | Roar," may have a "gradient color" gradient and multiple prints. If it contains two Type I factors, the number is 2.

[0074] S32 determines the batch listing risk type of the jewelry to be listed based on the number of additional price influencing factors of the aforementioned deviation type; Specifically, the risk type of the batch listing of the jewelry to be listed is determined based on the quantity range of the additional price influencing factors of the first type of deviation, specifically including severe risk type, target risk type and general risk type.

[0075] Bulk Listing Risk Types: Based on the number of deviation type factors contained in the items to be listed, the items are divided into different risk levels to determine their handling method in bulk listing. These typically include severe risk types, target risk types, and general risk types, corresponding to high, medium, and low risk, respectively. Among these, the severe risk type corresponds to the largest number of deviation type factors, followed by the target risk type, while the general risk type has the fewest or none.

[0076] By setting quantity ranges, jewelry can be categorized according to risk level, allowing for differentiated batch listing strategies for jewelry of different risk types. Jewelry of severe risk requires the highest level of human intervention, with the target risk type serving as an observation sample for the model's identification capabilities, while general risk types can be considered for automated processing.

[0077] Specific examples: The preset quantity ranges are as follows: For severe risk types, the number of Category 1 factors is ≥3; for target risk types, the number of Category 1 factors is 1 or 2; and for general risk types, the number of Category 1 factors is 0. For the above "AK-47 | Surface Hardening" (1 Category 1 factor), it belongs to the target risk type; if an item "AWP | Medusa" contains 3 Category 1 factors, it belongs to the severe risk type; if "P2000 | Ivory" does not contain any Category 1 factors, it belongs to the general risk type.

[0078] Specifically, the severe risk type is greater than the target risk type, and the target risk type is greater than the general risk type.

[0079] It is understandable that, in the above steps, if the proportion of the number of high-risk items to be listed in the merchant's inventory is greater than the preset threshold for the proportion of items to be listed, then the merchant is determined not to be a merchant subject to merging. If the proportion of the number of high-risk items to be listed in the merchant's inventory is not greater than the preset threshold for the proportion of items to be listed, then proceed to step S33.

[0080] High-risk jewelry items are a frequent source of identification errors. If a merchant holds too many of these items, merging and listing them in bulk will lead to numerous errors, which is counterproductive. Therefore, a simple percentage threshold should be used to quickly exclude high-risk merchants, avoiding complex calculations later.

[0081] Specific example: The preset threshold for the percentage of items to be listed is 10%. A merchant has 100 items to be listed, of which 15 are of the high-risk type, accounting for 15% > 10%. In this case, the merchant is directly determined not to be a merchant subject to consolidation processing, and the process ends. If there are only 8 high-risk items, accounting for 8% ≤ 10%, then proceed to step S33 for further evaluation.

[0082] S33 determines whether a merchant belongs to the merged processing merchant based on the data of pending jewelry items that do not have additional price influencing factors of a certain type of deviation, the distribution data of pending jewelry items of the target risk type in different merchants, and the data of pending jewelry items of the target risk type.

[0083] Jewelry items awaiting listing that do not have additional price-influencing factors of a certain type of deviation: namely, jewelry of the general risk type. They do not contain any high-risk factors, have high identification reliability, and can be called reliable jewelry items for listing.

[0084] Reliability Ratio: The proportion of reliable listed accessories to the total number of accessories to be listed by the merchant.

[0085] Number of items to be listed for the target risk type: This refers to the number of different skin types (based on skin) of the items belonging to the target risk type held by the merchant, reflecting the diversity of the merchant in medium-risk items.

[0086] Popular Jewelry Categories Across the Market: This refers to jewelry categories that are ranked in descending order of total quantity, based on the target risk type of all merchants. These categories are ranked within a predetermined range (e.g., top 10). Because these categories are listed in large quantities across the entire market, any misidentification will impact the bulk listing operations of numerous merchants, resulting in a wide-ranging influence.

[0087] Popular Category Threshold: This refers to a pre-set value used to determine whether a merchant has too many popular categories of accessories in the entire market, and thus decide whether to exclude them from the merged processing.

[0088] When a merchant is not directly excluded and has a low reliability rate, it's necessary to examine the composition of their target risk items. If the number of target risk items is large, their diversity can be used to test the model's recognition reliability in different scenarios, helping to identify recognition problems. Therefore, such merchants are more likely to be identified for merging. If the number of items is small, it's necessary to further determine whether these items are popular items across the entire market. If the merchant's target risk items include many popular items, the impact of misidentifying these items would be too large, making it unsuitable to bear the risk through batch merging. Therefore, such merchants should not be identified for merging. Conversely, if the merchant's target risk items are all unpopular items, even if a deviation occurs, the impact will be relatively limited, and batch merging and listing can be allowed.

[0089] The above steps include the following: S331 designates any jewelry items to be listed that do not have additional price-influencing factors of a certain type of deviation as reliable listing items. The proportion of reliable listing items among the jewelry items to be listed in the merchant is taken as the reliable proportion. It is determined whether the reliable proportion of the merchant is greater than a preset reliable proportion threshold. If so, it is determined that the merchant belongs to the merged processing merchant. If not, proceed to step S332. Reliable listed jewelry: refers to jewelry that does not contain any type of bias factor, and its identification results are highly reliable and less prone to error.

[0090] Reliability Ratio: The ratio of the number of reliable listed accessories to the total number of accessories to be listed by the merchant reflects the overall controllability of the merchant's risk.

[0091] A high reliability rate indicates that the vast majority of the merchant's accessories are low-risk and can be safely listed in batches. Therefore, the merchant is directly identified as eligible for batch processing without further evaluation. Low-risk merchants are given quick approval to improve efficiency.

[0092] Specific example: The preset reliable percentage threshold is 80%. A merchant has 100 accessories to be listed, of which 85 are reliable. The reliable percentage is 85% > 80%, so this merchant is directly subject to merged processing.

[0093] S332: Based on the data of jewelry to be listed for the target risk type, determine the number of types of jewelry to be listed for the target risk type, and determine whether the number of types of jewelry to be listed for the target risk type is greater than a preset threshold. If yes, determine that the merchant belongs to the merged processing merchant. If no, proceed to step S333.

[0094] When the reliability percentage is low, it indicates that the merchant has a certain number of medium-risk items. If there are many types of these medium-risk items, it means that the merchant holds various types of medium-risk items. This provides the platform with a rich sample of test samples, making it easy to identify recognition problems of the image recognition model on different items and verify the model's true recognition reliability in medium-risk scenarios. Therefore, even if there is some risk, it is worthwhile to obtain more feedback data by merging and uploading items in batches to optimize the model. Hence, the merchant is determined to be subject to merge processing.

[0095] Specific example: The preset threshold for the number of product categories is 5. If a merchant has 8 different skins for the target risk type of accessories (such as "AK-47 | Surface Hardened", "M4A4 | Roar", "USP-S | Gunshot Death", etc.), and the number of categories is 8 > 5, then this merchant belongs to the category of merchants subject to merged processing.

[0096] S333 determines the ranking of the target risk type of the jewelry to be listed among all merchants based on the number of items to be listed from most to least. The target risk type of the jewelry to be listed that is ranked before the target order is regarded as risk jewelry. It is then determined whether the types of risk jewelry of the merchant are within the preset range of types and quantities. If so, the merchant is determined to be a merchant that is subject to merging. If not, the merchant is determined not to be a merchant that is subject to merging.

[0097] Popular Jewelry Categories Across the Market: This refers to jewelry categories ranked in descending order of the total number of listings for the target risk type across the entire market, and those ranking within a predetermined range (e.g., top 10). These jewelry categories represent mainstream market demand and are broadly representative.

[0098] Popular Category Threshold: A pre-set value used to measure whether the number of popular accessories held by a merchant reaches a level that requires exclusion and merging. If it exceeds this threshold, it indicates that the merchant is involved in a large number of high-impact accessories, and if the identification is incorrect, the impact will be significant, so it will not be merged.

[0099] If the number of target risk accessories is small, it's necessary to further examine their popularity across the entire market. If a merchant holds a large number of target risk accessories that are popular across the market, a misidentification of these accessories could lead to numerous incorrect listings, resulting in a large-scale impact and excessively high risk. In this case, merging listings in bulk should not be used to bear such a large-scale risk, and the merchant should not be considered for merging. Conversely, if a merchant holds very few or no popular target risk accessories, it means that even if a misidentification occurs, it will only affect a few merchants or a few products, and the risk is controllable. Therefore, the merchant can be considered for merging.

[0100] Achieving a balance between risk control and efficiency, avoiding losing more than one gains, and minimizing potential large-scale impacts.

[0101] Specific example: Suppose all merchants' target risk items are sorted from most to least numerous in the entire market, and the top 10 are considered the most popular items. A merchant has 3 target risk items, 2 of which are among the top 10 most popular items (e.g., "AK-47 | Surface Hardened" and "M4A4 | Roar"). The preset threshold for popular items is set to 1 (i.e., if more than 1 type of popular item is considered too risky). Therefore, if 2 > 1, the merchant is not eligible for consolidation. If only 1 of the merchant's 3 target risk items is considered popular, and the threshold is set to 1 (where 1 is not greater than 1), then the merchant is eligible for consolidation. The actual threshold can be adjusted according to the platform's risk tolerance; for example, it could be set to 0, meaning that holding any popular item would prevent consolidation, thus strictly avoiding large-scale risk.

[0102] Suppose a game item platform has already identified the risk types of bias for each additional price-influencing factor through previous steps. The "Gradient Color" gradient and "Holographic" prints are classified as Type I bias (high risk), the "Crown" print as Type II bias, and the "Flash" print as Type III bias. Now, the platform needs to merge the item listings of 50 merchants. The preset thresholds are as follows: Preset percentage of items to be listed = 10%, preset reliable percentage threshold = 80%, preset number of categories threshold = 5, target order (popular ranking) = top 10, preset popular category threshold = 0 (i.e., if any item is listed as popular across the entire market, it will not be merged).

[0103] S31: Determine the number of one type of deviation factors for the jewelry to be listed. Merchant A has a total of 100 jewelry items to be listed, including: 20 pieces of "AK-47 | Surface Hardened": including "Gradient Color" gradient (Category 1) and "Crown" print (Category 2), with 1 element of Category 1.

[0104] "M4A4 | Roar" 15 pieces: includes "Gradient Colors" gradient (one type) and "Holographic" print (one type), with one type factor quantity = 2.

[0105] "AWP | Medusa" 10 pieces: includes "Gradient Colors" gradient (one type), "Holographic" print (one type) and another type of print, with one type factor quantity = 3.

[0106] “P2000 | Ivory” 30 pieces: Contains no factor of any kind, the number of factors of one kind = 0.

[0107] The remaining 25 items are other accessories, and the number of each of their respective categories is counted.

[0108] S32: Determine the risk type of bulk product listing: Based on the preset ranges: severe risk type (number of factors in one category ≥ 3), target risk type (number of factors in one category = 1 or 2), and general risk type (number of factors in one category = 0). Therefore: Ten items under the "AWP | Medusa" category are classified as high-risk.

[0109] The target risk types are 20 pieces of "AK-47 | Surface Hardened" and 15 pieces of "M4A4 | Roaring", totaling 35 pieces.

[0110] The "P2000 | Ivory" category (30 pieces) falls under the general risk category.

[0111] Of the remaining 25 items, assuming there are 5 serious risks, 10 target risks, and 10 general risks, then Merchant A has a total of 15 serious risks, 45 target risks, and 40 general risks.

[0112] The percentage of serious risks is 15 / 100 = 15% > 10%, therefore Merchant A is directly determined not to be a merchant subject to merger processing, and the process ends.

[0113] Merchant B: There are 120 accessories to be listed, of which 8 are of high risk, accounting for 8% ≤ 10%, and will be transferred to S33.

[0114] S331: Calculate the reliability percentage: Merchant B has 70 items of general risk type (reliable listed accessories), and the reliable percentage is 70 / 120 ≈ 58.3% < 80%, so it is transferred to S332.

[0115] S332: Number of target risk jewelry types: Merchant B has 6 different skins for its target risk type accessories (such as "AK-47 | Surface Hardened", "M4A4 | Roar", "USP-S | Gunshot Kills", etc.). Since the number of types (6 > 5) is greater than 5, Merchant B falls under the category of merchants requiring merged processing. Here, due to the large number of types, the platform can use Merchant B's diverse medium-risk accessories to test the model's recognition performance in different scenarios and promptly identify recognition problems; therefore, it is determined that Merchant B is a merchant requiring merged processing.

[0116] Merchant C: There are 100 jewelry items to be listed. The percentage of serious risks is 6% ≤10%, the percentage of reliable risks is 60% <80%, and the number of target risk jewelry items is 4 types ≤5. Transfer to S333.

[0117] S333: Assess the holdings of popular jewelry items across the entire market: All merchants holding target risk items are sorted by total market quantity from most to least, and the top 10 are considered the most popular items in the market. Assume the top 10 include: "AK-47 | Surface Hardened" (most numerous), "M4A4 | Roar," "AWP | Medusa," "USP-S | Gunshot Kill," "Glock-18 | Water Spirit," etc. Merchant C holds 4 target risk items, 3 of which are among the top 10 most popular items (e.g., "AK-47 | Surface Hardened," "M4A4 | Roar," "USP-S | Gunshot Kill"). The preset threshold for popular items is 0, meaning that holding any popular item will not result in merging. 3 > 0, therefore Merchant C is not eligible for merging.

[0118] If merchant D holds 4 types of target risk accessories, and only 1 of them is among the top 10 most popular accessories, then 1 > 0, and is therefore not considered a merchant subject to merger processing. If merchant E holds 4 types of target risk accessories that are not among the top 10, then the number of popular accessory types is 0, which is not greater than the threshold of 0, therefore merchant E is considered a merchant subject to merger processing.

[0119] This invention combines the risk profile of a merchant's pending jewelry listings with market distribution through a multi-level progressive judgment logic, enabling accurate identification of merchants for merged processing. It quickly eliminates high-risk merchants by assessing the proportion of seriously risky jewelry, avoiding complex subsequent calculations and improving judgment efficiency. By combining the reliable proportion, the number of target risky jewelry types, and market popularity, it comprehensively considers the overall risk level and potential value of the merchant.

[0120] When there are many types of target risk items, they are identified as merchants to be merged. Their diversity can be used to identify problems in model identification and verify reliability, thus promoting model optimization. When a merchant holds a large number of popular items in the entire market, they are not identified as merchants to be merged, in order to avoid large-scale listing errors due to identification bias and to achieve effective risk control.

[0121] Furthermore, the method for determining the automatic listing control strategy for the game accessories is as follows: During the bulk listing of game items, the actual listing results of merchants already identified for merged processing are used to statistically analyze the secondary modifications of each item across different merchants. The identification deviation risk coefficient for each item is calculated, and based on this coefficient and the risk distribution of the item in the overall market, it is determined whether the item is allowed to be automatically listed and the scale of automatic listing. The core logic is to assess the reliability of item level identification through merchant secondary modification data. High-risk items are directly prohibited from automatic listing; for low-risk items, different automatic listing batches are set according to the proportion of high-risk items in the entire market, the item's own risk coefficient, and the merchant environment: if the proportion of high-risk items in the entire market is high, it indicates that most items are already strictly controlled, and the remaining low-risk items have extremely low risk, allowing for larger batch automatic listing; if the proportion of high-risk items in the entire market is low, then further adjustments are made based on the item's own risk coefficient and the risk environment of the merchant, using medium batch automatic listing or prohibiting automatic listing altogether.

[0122] S41 Based on the batch listing identification results of the merged processing merchants in different game items, determine that there are merged processing merchants who have made secondary modifications in the game items, and identify them as secondary modification merchants. Merged Merchants: refers to merchants who have been determined through the previous steps to be eligible for batch merging and listing. Their jewelry items to be listed are automatically identified and merged by the system during the batch listing process.

[0123] Batch listing recognition results: This refers to the additional price influencing factors identified by the system after image recognition of each game item during the batch listing process, as well as the record of subsequent modifications made by the merchant to the results.

[0124] Merchants who made secondary modifications: For a specific game item, this refers to merchants who, after being identified in batches, made secondary modifications to the identification results of that item during the merging process.

[0125] Consolidating the processing of secondary modifications made by merchants provides direct feedback on the accuracy of identifying the jewelry. By identifying which merchants have modified a particular piece of jewelry, we can pinpoint the source of evidence that the jewelry may have identification problems. This links merchant-level modifications to specific pieces of jewelry, providing foundational data for subsequently calculating the risk coefficient of each piece of jewelry.

[0126] Specific example: Suppose 50 merchants participated in the batch listing process. For the game item "AK-47 | Surface Hardened", after system identification, 8 merchants modified the identification result of the item (e.g., changed the print type). These 8 merchants are considered to have made secondary modifications to the item. The remaining 42 merchants did not make any modifications, and their identification results are considered accepted.

[0127] S42 determines the identification deviation risk coefficient of the game accessories based on the secondary modification merchant data and the batch listing identification data of the merchants. Merchant data for bulk listing identification and processing: refers to the total number of merchants involved in the bulk listing identification and processing of this accessory, i.e., how many merchants have listed this accessory.

[0128] Number of merchants who have made two modifications to this item: This refers to the number of merchants who have made two modifications to this item.

[0129] Identification Bias Risk Coefficient: A value between 0 and 1 used to quantify the reliability of identifying the item among merchants undergoing batch processing. The formula is: Identification Bias Risk Coefficient = Number of merchants undergoing secondary modifications / Number of merchants undergoing batch listing and identification processing. A higher coefficient indicates a greater proportion of the item being modified, and a higher risk of identification bias.

[0130] By statistically analyzing the proportion of merchants whose products were modified out of all processed products, we can objectively reflect the frequency with which the product was misidentified in real-world listing scenarios. A higher proportion indicates a worse recognition performance of the existing model for that product, and a greater risk.

[0131] For example, for the item "AK-47 | Surface Hardened," 50 vendors listed it, and 8 of them made modifications. Therefore, the identification bias risk coefficient is 8 / 50 = 0.16. For another item, "P2000 | Ivory," 40 vendors listed it, and only 1 made modifications. The coefficient is 1 / 40 = 0.025.

[0132] S43 determines the automatic listing control strategy for the game items based on the identification deviation risk coefficient of different game items and the identification deviation risk of game items in merchants where the game items are available.

[0133] It should be noted that the identification deviation risk coefficient of the game accessories is related to the number of merchants who modify the items twice and the number of merchants who batch upload the items for identification processing. The more merchants who modify the items twice and the fewer merchants who batch upload the items for identification processing, the higher the identification deviation risk coefficient will be. Its value is between 0 and 1.

[0134] Specifically, if the identification deviation risk coefficient of the game item is greater than the preset deviation risk coefficient threshold, then the automatic listing control strategy for the game item is determined to be unable to automatically perform batch listing processing.

[0135] Preset Deviation Risk Coefficient Threshold: A pre-defined threshold value used to determine whether the risk of a certain accessory is high enough to warrant prohibiting its automatic listing. Accessories with a deviation risk coefficient greater than this threshold are defined as restricted game accessories.

[0136] Restricted game items: These are game items whose identification deviation risk coefficient is greater than the preset deviation risk coefficient threshold. These are high-risk items and are prohibited from being automatically listed. They must be manually reviewed.

[0137] Restricted item type percentage: The percentage of restricted game item types out of the total number of all game item types.

[0138] Preset threshold for the proportion of restricted jewelry categories: This is used to determine whether high-risk jewelry dominates the market. If the proportion is high, it means that most jewelry has been strictly controlled, and the remaining unrestricted jewelry has extremely low risk, so automatic batch listing can be relaxed.

[0139] Risk coefficient preset value: An intermediate value for further subdividing the risk level of unrestricted jewelry, provided that the identification deviation risk coefficient is not greater than the preset threshold, to distinguish between medium and low risk.

[0140] The percentage of restricted game items among merchants that sell the game items: For a specific non-restricted item, count all merchants that hold the item, calculate the percentage of restricted game items held by each merchant relative to their total number of items held, and take the average (or other representative statistics) to assess the merchant environment risk of the item.

[0141] Preset threshold for restricted item percentage: Used to determine whether the non-restricted item is mainly found in high-risk merchant environments. If the merchant environment risk is low, it can be automatically listed; if the merchant environment risk is high, it may be prohibited from automatic listing due to collateral effects.

[0142] First, determine whether the identification deviation risk coefficient of the game item exceeds the preset deviation risk coefficient threshold. If so, the item is determined to be a restricted game item, and its automatic listing control strategy is that it cannot be automatically batch listed, meaning that the item must be manually reviewed before it can be listed.

[0143] If the coefficient exceeds the threshold, it means that the item has been frequently modified by a large number of merchants, resulting in an excessively high error rate. If automatic listing is allowed to continue, it will lead to a large number of incorrect listings, damaging the platform's reputation and the interests of merchants. Therefore, manual intervention is necessary.

[0144] Specific example: The preset deviation risk coefficient threshold is set to 0.3. If the identification deviation risk coefficient of a certain accessory "AWP | Medusa" is 0.35 > 0.3, then the accessory is judged to be a restricted game accessory and cannot be automatically batch listed. Each listing requires manual review.

[0145] If the identification deviation risk coefficient is not greater than the preset deviation risk coefficient threshold, then the item is an unrestricted game item and will enter one of the following three scenarios, determining the automatic batch listing based on the overall risk distribution and its own characteristics: Additionally, it is understood that if the identification deviation risk coefficient of the game accessory is not greater than a preset deviation risk coefficient threshold, then the following content is included: Case 1: Game items with an identification deviation risk coefficient greater than the preset deviation risk coefficient threshold are designated as restricted game items. If the proportion of the restricted game items among all game items is greater than the preset restricted item type proportion threshold, then the automatic listing control strategy for the game items is determined to be that regardless of how the additional price factors of the game items change, the batch listing process will be automatically carried out each time the preset quantity is reached. When the proportion of restricted game items among all game items exceeds a preset threshold for restricted item types, it indicates that most items on the market are high-risk and strictly controlled. In this case, unrestricted items are a selected minority of low-risk items, making their identification highly reliable. Therefore, the automatic listing batch size can be relaxed. The automatic listing control strategy for these unrestricted items is determined as follows: regardless of changes in their additional price influencing factors, automatic batch listing is performed each time a preset quantity is reached during the batch listing process. Here, the preset quantity is a relatively large batch value.

[0146] If most accessories require manual review, then the accessories that can be automatically listed must be of extremely low risk. Using large-scale automatic listing can improve efficiency without introducing significant risks. This is based on the consideration that "the remaining risk is extremely low under strict control."

[0147] Specific example: The preset threshold for the proportion of restricted item types is 50%. Assuming there are 100 types of items, of which 60 are restricted items, accounting for 60% > 50%, then for any non-restricted item (such as "P2000 | Ivory", with a coefficient of 0.02), the control strategy is to automatically batch list it each time a preset quantity (such as 10 pieces) is reached.

[0148] Scenario 2: If the proportion of restricted game items among all game items is not greater than the preset threshold for the proportion of restricted game item types, and if the identification deviation risk coefficient of the game item is less than the preset risk coefficient value, then the automatic listing control strategy for the game item is determined to be that regardless of how the additional price influencing factors of the game item change, the batch listing process will be automatically carried out each time the second preset quantity is reached during the batch listing process. It is understandable that the second preset quantity is less than the preset quantity.

[0149] If the proportion of restricted game item types does not exceed the preset threshold for restricted item types, and the identification deviation risk coefficient of the unrestricted item is less than the preset risk coefficient value, it indicates that the item is low-risk and the overall market risk is not high. Therefore, a medium-scale batch automatic listing can be adopted. The automatic listing control strategy for this item is determined as follows: automatically batch listing is performed each time the second preset quantity is reached. The second preset quantity is less than the preset quantity.

[0150] When there are not many high-risk items, low-risk items can be automatically listed. However, to be on the safe side, a smaller batch size than in scenario 1 should be used to control the risk per transaction.

[0151] Specific example: The preset risk coefficient is 0.1. For a certain jewelry item "P2000 | Ivory" with a coefficient of 0.02 < 0.1, and the category ratio of jewelry items is limited to ≤50%, the control strategy is to automatically process the batch listing each time the second preset quantity (e.g., 5 pieces) is reached.

[0152] Scenario 3: If the identification deviation risk coefficient of the game accessories is not less than the preset risk coefficient value, obtain the proportion of restricted game accessories among the merchants that have the game accessories, and determine whether the proportion of restricted game accessories among the merchants that have the game accessories is greater than the preset restricted accessory proportion threshold. If yes, then determine that the automatic listing control strategy of the game accessories is to automatically perform batch listing processing each time the second preset quantity is reached during the batch listing process, regardless of how the additional price factors of the game accessories change. If no, then determine that the automatic listing control strategy of the game accessories is that automatic batch listing processing cannot be performed.

[0153] If the proportion of restricted game item types is not greater than the preset threshold for restricted game item types, and the identification deviation risk coefficient of the non-restricted item is not less than the preset risk coefficient value (i.e., it is in the medium risk range), then it is necessary to further examine the merchant environment in which the item exists. Obtain the merchants that possess the item and calculate the proportion of restricted game items held by each merchant. If this proportion is greater than the preset threshold for restricted game item proportion, it indicates that the item mainly exists in low-risk merchant environments. These merchants have low overall risk, and due to the numerous restrictions, the overall risk of bulk listing is low. Therefore, automatic listing using the second preset quantity is allowed. If the average proportion is not greater than the threshold, it indicates that the merchant environment risk is high, and automatic listing is prohibited.

[0154] Medium-risk jewelry may not pose a significant risk on its own, but if the merchants in question generally deal in high-risk jewelry, these merchants may be more cautious in their operations or more likely to make modifications, leading to inconsistent recognition results for the jewelry. Therefore, a final judgment needs to be made based on the environmental risks.

[0155] Specific example: A preset risk coefficient of 0.1 and a preset restricted item percentage threshold of 30% are used. For example, the item "AK-47 | Surface Hardened" has a coefficient of 0.16 (between 0.1 and 0.3), and the restricted item percentage is ≤50%. If all merchants holding this item are counted, and the percentage of restricted game items held by the merchant is calculated to be 25%, which is less than 30%, then the merchant's environmental risk is considered high, and this item cannot be automatically batch-listed; manual review is required. Conversely, if another item, "M4A4 | Roar," has a coefficient of 0.18, and the restricted item percentage held by the merchant is greater than 30%, then the merchant's environmental risk is considered low, and the second preset quantity (5 items) is used for automatic listing.

[0156] Example 2 In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned fuzzy merging processing method for a trading platform when running the computer program.

[0157] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0158] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0159] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A fuzzy merging processing method for a trading platform, characterized in that, Specifically, it includes: Based on the factors influencing the additional price of game items, and combined with the distribution data of unlisted game items from different merchants, a strategy for identifying batch listing targets of the game items is determined. Based on the identification strategy, batch listing targets are identified. According to the fuzzy merging processing method, the degree of deviation between the image recognition results of the batch listing targets and the secondary processing results of the merchants is determined. Using the degree of deviation, the identification deviation risk type of the batch listing targets under different additional price influencing factors is determined. Based on the identified risk type and the merchant's pending listing data for different game items, the association between the pending listing game items of the target risk type and other merchants is determined, and the merchants to be merged are determined by combining the pending listing data of the target risk type. Based on the identification results of the batch listing targets of the merged processing merchants in different game items, the identification deviation risk of different game items is determined, and combined with the identification deviation risk of the game items of the merchants that have the game items, the automatic listing control strategy of the game items is determined. The degree of deviation refers to the extent or frequency at which merchants correct the results of additional price-influencing factors identified by the image recognition model. The fuzzy merging processing method involves classifying products whose image similarity meets the fuzzy merging similarity threshold into the same product based on the image recognition results, and then listing them in batches.

2. The fuzzy merging processing method for a trading platform as described in claim 1, characterized in that, The additional price-influencing factors include price-influencing factors other than wear and tear, including gradients and the type of print.

3. The fuzzy merging processing method for a trading platform as described in claim 1, characterized in that, The distribution data of unlisted game items in different merchants is determined based on the number of game items in the unlisted game items of different merchants.

4. The fuzzy merging processing method for a trading platform as described in claim 1, characterized in that, The method for determining the identification strategy for bulk listing targets in the aforementioned game accessories is as follows: Based on the factors influencing the added price of game accessories, determine the number of factors influencing the added price of the game accessories, and determine the difficult-to-identify accessories among the game accessories based on the number of factors influencing the added price; Based on the distribution data of unlisted game items in different merchants, identify the merchants that require bulk listing of the difficult-to-identify game items; By utilizing the data on difficult-to-identify game items and the bulk listing requests from different merchants for these items, a strategy for identifying bulk listing targets within the game items is determined.

5. The fuzzy merging processing method for a trading platform as described in claim 4, characterized in that, The difficult-to-identify items are game items whose number of additional price-influencing factors exceeds a preset threshold.

6. The fuzzy merging processing method for a trading platform as described in claim 1, characterized in that, The method for determining the identification bias risk type of the target for bulk listing under different additional price influencing factors is as follows: Based on the degree of deviation, determine the secondary modification data of the additional price influencing factors in different batch listing targets; The batch listing targets that require secondary modifications to the additional price influencing factors will be used as the modification targets; Based on the modified target data, determine the identification bias risk type of the additional price influencing factors.

7. The fuzzy merging processing method for a trading platform as described in claim 1, characterized in that, The method for determining the automatic listing control strategy for game accessories is as follows: Based on the batch listing identification results of the merged processing merchants in different game items, merged processing merchants that have undergone secondary modifications in the game items are identified and identified as secondary modification merchants. Based on the secondary modification merchant data and the batch listing identification data of the game accessories, the identification deviation risk coefficient of the game accessories is determined; Based on the identification deviation risk coefficients of different game items, and the identification deviation risk of game items in merchants that sell such game items, an automatic listing control strategy for the game items is determined.

8. A computer system, comprising: A memory and processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a fuzzy merging processing method for a trading platform as described in any one of claims 1-7.

Images