Item comparison method, item price prediction model training method and apparatus

By comparing item attribute information to construct virtual item data, the magnitude and degree of price impact can be determined, solving the problem that users cannot intuitively understand the contribution of attributes in existing technologies, and achieving more granular price display and a better user experience.

CN123118191APending Publication Date: 2026-11-17BEIJING WODONG TIANJUN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202510585295.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-05-07
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing item comparison methods can only display the total price difference and attribute differences between items. Users cannot intuitively understand the specific contribution of each attribute to the total price difference, resulting in a poor user experience.

Method used

By comparing the attribute information of each item to be compared, the target attribute information is determined, and virtual item data is constructed to determine the price impact of each target attribute information on the items to be compared. Based on the price impact, the degree of influence of each target attribute information on the total price difference between the items to be compared is determined, and the item comparison results are generated.

Benefits of technology

It can intuitively display the proportion of each attribute difference relative to the total price difference for users, improving the user experience, and improves the accuracy of price prediction and comparison efficiency through the item price prediction model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123118191A_ABST
    Figure CN123118191A_ABST
Patent Text Reader

Abstract

This invention discloses an item comparison method, an item price prediction model training method, and an apparatus, relating to the field of computer technology. A specific embodiment of the item comparison method includes: comparing attribute information corresponding to each item to be compared to determine target attribute information; constructing virtual item data for each target attribute information; determining the price impact range based on the virtual item data; determining the degree of impact on the total price difference based on the price impact range; and generating item comparison results for the items to be compared based on the degree of impact. This method can intuitively display the proportion of each attribute difference relative to the total price difference, improving the user experience. A specific embodiment of the item price prediction model training method includes: filling the obtained target item's category name and core attributes into a pre-configured training data template to obtain first training data; training an item price prediction model based on the first training data and item prices; this method can improve the accuracy of item price prediction and comparison efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method for comparing items and a training method and apparatus for an item price prediction model. Background Technology

[0002] In today's e-commerce and retail industry, users often need to compare the prices and attributes of different items to make the best purchasing decision. Currently, item comparison technology mainly reads the price and attribute information of items directly from the database and displays it on the page in a grid or similar format.

[0003] In the process of realizing this invention, the inventors discovered at least the following problems in the related technology:

[0004] Existing item comparison methods typically only display the total price difference and attribute differences between items. Users cannot intuitively understand the specific contribution of each attribute to the total price difference, resulting in a poor user experience. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide an item comparison method, an item price prediction model training method and apparatus, which construct virtual item data based on each different attribute between items, determine the impact of changes in each attribute on the item price, and break down the total price difference of items into each different attribute based on the price impact, which can intuitively show users the proportion of each attribute difference relative to the total price difference, thereby improving the user experience.

[0006] To achieve the above objectives, according to one aspect of the present invention, an article comparison method is provided, comprising:

[0007] Compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; the target attribute information is the attribute information that is different between the items to be compared.

[0008] For each target attribute, construct virtual item data, and determine the impact of each target attribute on the price of the comparable item based on the virtual item data;

[0009] Based on the magnitude of price impact, determine the degree of influence of each target attribute information on the total price difference between the items to be compared, and generate the item comparison results based on the degree of influence.

[0010] Optionally, virtual item data is constructed for each target attribute, including:

[0011] Repeat the following steps until all target attribute information has been replaced: replace one target attribute of the first item to be compared with one target attribute of the second item to be compared, and keep the other attribute information of the first item to be compared unchanged; wherein, the first item to be compared and the second item to be compared are different; form a virtual item data of the first item to be compared based on the replaced attribute information;

[0012] The virtual item data for each item to be compared is obtained based on the data of each virtual item.

[0013] Optionally, the impact of each target attribute on the price of the item to be compared is determined based on the virtual item data, including:

[0014] Predict the expected price of each item to be compared based on virtual item data;

[0015] The expected price and preset price of each item to be compared are compared one by one to obtain the impact of each target attribute information on the price of the item to be compared.

[0016] Optionally, the degree of influence of each target attribute information on the total price difference between the items to be compared is determined based on the magnitude of the price impact, including:

[0017] The summation of the impact of each target attribute on the price of the item under comparison is obtained;

[0018] Calculate the percentage of the impact of each target attribute information on the price of the item compared to the summation result;

[0019] The degree of influence of the total price difference between the items to be compared is determined based on the proportion of each target attribute information.

[0020] According to a second aspect of the present invention, a method for training a commodity price prediction model is provided, characterized in that it includes:

[0021] Retrieve the category name, core attributes, and item price of the target item;

[0022] The category name and core attributes corresponding to the target item are filled into the pre-configured training data template to obtain the first training data;

[0023] The first training data is used as the model input, and the item prices are used as the model output to train an item price prediction model; the item price prediction model is used to predict the expected price of each item to be compared.

[0024] Optionally, the pre-configured training data template is obtained through the following steps:

[0025] Obtain multiple categories and their corresponding item attributes, and determine the information entropy of each item attribute;

[0026] Sort the information entropy of each item's attributes and determine the core attributes of the category based on the sorting results;

[0027] The obtained category names and core attributes are concatenated to obtain the training data template.

[0028] According to a third aspect of the present invention, an article comparison device is provided, comprising:

[0029] The determination module is used to compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; the target attribute information is the attribute information that is different between the items to be compared.

[0030] The construction module is used to build virtual item data for each target attribute information, and determine the impact of each target attribute information on the price of the item to be compared based on the virtual item data;

[0031] The generation module is used to determine the degree of influence of each target attribute information on the price difference between the items to be compared based on the magnitude of the price influence, and to generate the item comparison results of the items to be compared based on the degree of influence.

[0032] According to a fourth aspect of the present invention, a training apparatus for a commodity price prediction model is provided, comprising:

[0033] The acquisition module is used to obtain the category name, core attributes, and item price of the target item;

[0034] The input module is used to input the category name and core attributes corresponding to the target item into a pre-configured training data template to obtain the first training data;

[0035] The training module is used to train an item price prediction model by taking the first training data as the model input and the item price as the model output; the item price prediction model is used to predict the expected price of each item to be compared.

[0036] According to a fifth aspect of the present invention, an electronic device is provided, comprising:

[0037] One or more processors;

[0038] Storage device for storing one or more programs.

[0039] When one or more programs are executed by one or more processors, the one or more processors implement the methods of any of the above embodiments.

[0040] According to a sixth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method of any of the above embodiments.

[0041] According to a seventh aspect of the present invention, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method of any of the above embodiments.

[0042] One embodiment of the above invention has the following advantages or beneficial effects: by comparing the attribute information corresponding to each item to be compared, the target attribute information is determined based on the comparison results; the target attribute information is the different attribute information between the items to be compared; virtual item data is constructed for each target attribute information, and the price impact of each target attribute information on the items to be compared is determined based on the virtual item data; the degree of influence of each target attribute information on the total price difference between the items to be compared is determined based on the price impact, and the item comparison results of the items to be compared are generated based on the degree of influence; this embodiment constructs virtual item data based on each different attribute between items, and determines the impact of the change of each attribute on the item price, and breaks down the total price difference of the items into each different attribute based on the price impact, which can intuitively show the user the proportion of each attribute difference relative to the total price difference, thus improving the user experience. By obtaining the category name, core attributes, and item price of the target item; filling the category name and core attributes of the target item into a pre-configured training data template to obtain the first training data; using the first training data as the model input and the item price as the model output, a price prediction model is trained; the price prediction model is used to predict the expected price of each item to be compared; thus, the expected price of an item can be predicted based on its attributes, thereby improving the accuracy of price prediction and increasing the efficiency of item comparison.

[0043] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0044] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0045] Figure 1 This is a schematic diagram of the main process of the article comparison method according to an embodiment of the present invention;

[0046] Figure 2 This is a schematic diagram of the main flow of the training method for the item price prediction model according to an embodiment of the present invention;

[0047] Figure 3 This is a schematic diagram of the main flow of an article comparison method according to a preferred embodiment of the present invention;

[0048] Figure 4 This is a schematic diagram of the main flow of a training method for a commodity price prediction model according to a preferred embodiment of the present invention;

[0049] Figure 5 This is a schematic diagram of the main modules of the article comparison device according to an embodiment of the present invention;

[0050] Figure 6 This is a schematic diagram of the main modules of the training device for the item price prediction model according to an embodiment of the present invention;

[0051] Figure 7 This is an exemplary system architecture diagram in which embodiments of the present invention can be applied;

[0052] Figure 8 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0053] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0054] It should be noted that the acquisition, storage, and application of personal information involved in the embodiments of the present invention comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0055] In today's e-commerce and retail industry, users often need to compare the prices and attributes of different items to make the best purchasing decision. Current item comparison technology mainly reads item price and attribute information directly from a database and displays it on a page in a grid or similar format. This typically only shows the total price difference and attribute differences between items; users cannot intuitively understand the specific contribution of each attribute to the total price difference, resulting in a poor user experience.

[0056] In view of this, according to a first aspect of the present invention, an article comparison method is provided.

[0057] Figure 1 This is a schematic diagram illustrating the main flow of the item comparison method according to an embodiment of the present invention. Figure 1 As shown, the article comparison method according to an embodiment of the present invention mainly includes the following steps S101 to S103.

[0058] Step S101: Compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; the target attribute information is the different attribute information between the items to be compared.

[0059] The items to be compared refer to items that need to be compared within a specific category or scenario, such as two or more products in the same category (e.g., laptops, mobile phones). Attribute information refers to the attributes that describe the characteristics of an item and their specific values, such as "Brand: xx", "CPU Model: i7-14650HX", or "Hard Drive: 1TB", etc. Target attribute information refers to the attributes that differ between the items to be compared. For example, for two mobile phones, if they are identified as differing in screen resolution and battery capacity, these two attributes are the target attribute information.

[0060] Comparing the attribute information of the items to be compared can be done through attribute-by-attribute comparison. This involves comparing the attribute information of each item one by one, selecting all attributes with different values ​​as the target attribute information. For example, if the CPU models of two laptops are "i7-14650HX" and "i9-13900H" respectively, then that attribute information is marked as the target attribute information. Alternatively, a vectorized feature comparison method can be used. All attribute values ​​of the items to be compared are converted into feature vectors, and then distance metrics (such as cosine similarity or Euclidean distance) are used to calculate the overall difference between the items, extracting the target attribute information from the feature differences.

[0061] Step S102: Construct virtual item data for each target attribute information, and determine the price impact of each target attribute information on the item to be compared based on the virtual item data.

[0062] Virtual item data is a dataset constructed by modifying the attribute values ​​of items to be compared, used to simulate the impact of changes in the target attribute on the item's price. Methods for constructing virtual item data include direct replacement and incremental adjustment. Specifically, direct replacement involves fixing other attribute values ​​and replacing the target attribute value with the corresponding value in other items to be compared. For example, for two laptops, if the target attribute is "CPU model," then while keeping all other attributes of item A unchanged, its "CPU model" is replaced from "i7-14650HX" to "i9-13900H" of item B, generating virtual item data; and vice versa. Incremental adjustment involves setting multiple possible value ranges for a target attribute to generate a series of gradually changing virtual item data. For example, "screen resolution" can be gradually increased from "1080p" to "2.5k" or "4k," generating a virtual data record with each adjustment.

[0063] The magnitude of price impact refers to the specific effect of a change in a target attribute on the price of an item, which can be expressed as the absolute value of the price change or the relative percentage change. Determining the magnitude of price impact based on virtual item data can be achieved using methods such as the difference method or regression analysis. Specifically, the difference method directly compares the input and output of the item price prediction model to calculate the difference in price prediction results before and after replacing the target attribute. For example, if the "CPU model" is replaced from "i7-14650HX" to "i9-13900H", and the expected price increases from 8000 yuan to 8500 yuan, then the magnitude of the price impact of the "CPU model" is +500 yuan. Regression analysis, based on multiple sets of virtual item data, uses a regression model to analyze the relationship between changes in the target attribute and price changes, calculating the magnitude of the price impact of the target attribute. For example, for the target attribute "graphics card model", multiple sets of virtual item data for different graphics card models are constructed. Combined with the expected price output of the item price prediction model, a relationship curve between the attribute value and the expected price is fitted to extract the specific magnitude of the impact of the target attribute change. This method can capture nonlinear or interactive effects and is more suitable for complex attribute relationship analysis.

[0064] Step S103: Determine the degree of influence of each target attribute information on the total price difference between the items to be compared based on the price influence range, and generate the item comparison results of the items to be compared based on the degree of influence.

[0065] The degree of influence refers to the proportion or weight of the price impact of each target attribute in the total price difference of the compared items, reflecting the degree of contribution of different attributes to the price difference between two items. This degree of influence can be determined through methods such as normalization or allocation ratios. The normalization method divides the price impact of each target attribute by the sum of the price impacts of all target attributes to obtain the relative contribution of each attribute to the total price difference. For example, if the impact of "CPU model" on the price is 600 yuan and "graphics card model" is 400 yuan, then the degree of influence of "CPU model" is 600 / 1000 = 60%, and that of "graphics card model" is 400 / 1000 = 40%. The allocation ratio method assigns a specific contribution value to the total price difference based on the direct proportional relationship between the absolute price impact of each target attribute and the total price difference. For example, if the impact of "screen resolution" is 200 yuan and the total price difference is 1000 yuan, then its degree of influence can be considered as 200 yuan directly accounting for 20% of the total price difference.

[0066] The item comparison results are generated by comprehensively analyzing the price impact of each target attribute and other auxiliary information. The comparison results can be ranked according to the degree of influence of each target attribute on the total price difference, generating a ranking of the attribute's contribution to the price difference. For example, "CPU model" has the highest impact on the price difference, followed by "graphics card model," and finally "memory stick," which clearly demonstrates the importance of each attribute. Alternatively, attributes can be weighted and scored based on multiple indicators, such as combining factors like the degree of influence, user preferences, and the importance of the attribute, to quantitatively evaluate the target attribute and ultimately generate the overall item comparison results. For example, given that users pay more attention to the "graphics card model" attribute, it can be given a higher weight, highlighting its contribution in the final comparison results. Through these methods, the item comparison results can effectively present the differences in key attributes of the items being compared and intuitively demonstrate the impact of these differences on the price, providing more granular price breakdown and clear guidance for users' choices.

[0067] This invention, through comparison of attribute information corresponding to each item to be compared, determines target attribute information based on the comparison results. The target attribute information consists of the different attributes between the items to be compared. Virtual item data is constructed for each target attribute information, and the price impact of each target attribute information on the items to be compared is determined based on the virtual item data. The degree of influence of each target attribute information on the total price difference between the items to be compared is determined based on the price impact, and the item comparison results are generated based on the degree of influence. This embodiment constructs virtual item data based on each different attribute between items and determines the impact of changes in each attribute on the item price. The total price difference is then broken down into each different attribute based on the price impact, allowing users to intuitively see the proportion of each attribute difference relative to the total price difference, thus improving the user experience.

[0068] Optionally, constructing virtual item data for each target attribute information includes: repeatedly executing the following steps until all target attribute information is replaced: replacing one target attribute information of the first item to be compared with one target attribute information corresponding to the second item to be compared, while keeping other attribute information of the first item to be compared unchanged; wherein the first item to be compared and the second item to be compared are different; forming a virtual item data of the first item to be compared based on the replaced attribute information; and obtaining virtual item data for each item to be compared based on each virtual item data.

[0069] Building virtual item data can be accomplished by iteratively replacing each target attribute. First, select two items to be compared, for example, a first item and a second item. The target attribute information between the first and second items includes CPU model, graphics card model, etc. During the replacement process, keep the other attribute information of the first item unchanged, and only replace one target attribute with the corresponding attribute value of the second item. For example, for two laptop products, the attributes of the first item to be compared are "Laptop [SEP] Brand: xx [SEP] CPU Model: i5-12500H [SEP] Graphics Card Model: RTX3050 [SEP] Hard Drive: 512GB [SEP] Memory: 16GB [SEP] Resolution: 1080p [SEP] Refresh Rate: 60Hz", and the attributes of the second item to be compared are "Laptop [SEP] Brand: xx [SEP] CPU Model: i7-12650H [SEP] Graphics Card Model: RTX3060 [SEP] Hard Drive: 1TB [SEP] Memory: 32GB [SEP] Resolution: 1440p [SEP] Refresh Rate: 120Hz". For the target attribute information "CPU model", the "i5-12500H" of the first comparison item is replaced with "i7-12650H" of the second comparison item, generating virtual item data: "Laptop [SEP] Brand: xx [SEP] CPU Model: i7-12650H [SEP] Graphics Card Model: RTX3050 [SEP] Hard Drive: 512GB [SEP] Memory: 16GB [SEP] Resolution: 1080p [SEP] Refresh Rate: 60Hz". Similarly, for the target attribute information "graphics card model", the "RTX3050" of the first comparison item is replaced with "RTX3060", generating another set of virtual item data. This process is repeated iteratively, replacing all target attribute information one by one, until all target attributes have been replaced. For each replacement step, a virtual item dataset is formed. Finally, complete virtual item datasets can be constructed based on these virtual item datasets for the first and second comparison items respectively. This virtual item data can be used for further analysis, such as calculating the impact of each target attribute on the price, or assessing the differences between two items to be compared.

[0070] Optionally, the impact of each target attribute information on the price of the item to be compared can be determined based on the virtual item data, including: predicting the expected price of each item to be compared based on the virtual item data; comparing the expected price and the preset price of each item to be compared one by one to obtain the impact of each target attribute information on the price of the item to be compared.

[0071] In practice, for each target attribute, virtual item data is constructed, and a trained item price prediction model or other methods are used to calculate the expected price of each item to be compared. The item price prediction model predicts prices based on the input attribute information, combined with historical transaction data or market data. For example, if the first item to be compared is a certain model of laptop, and the constructed virtual item data is a version with only the "CPU model" replaced, the model can output the expected price of the first item to be compared based on the virtual attribute combination. By comparing the expected prices of each item to be compared with their corresponding actual preset prices, the impact of specific target attribute information on prices is analyzed. For example, assuming that the "graphics card model" of a certain item to be compared is replaced with a high-performance model, while other attributes remain unchanged, its expected price increases from 5000 yuan to 6000 yuan, then the impact of the change in "graphics card model" on the price can be considered to be 1000 yuan. Similarly, for the replacement of other target attribute information and changes in item prices, comparisons are made one by one, and the price impact of each target attribute is recorded.

[0072] The magnitude of price impact can be determined in different ways. For example, it can be achieved by directly calculating the absolute value of the price difference—the difference between the expected price after the target attribute information is replaced and the price before replacement—to represent the specific contribution of the target attribute to the price. Alternatively, the relative proportion of price change can be calculated, such as the percentage increase in price relative to the preset price, to measure the relative impact of the target attribute on the price. This method is suitable for comparing goods in different price ranges and can eliminate biases introduced by price benchmarks. By completing the above processes one by one, the specific impact of each target attribute information on the price of the compared items can be comprehensively evaluated, providing basic data for subsequent analysis and improving the accuracy of item comparisons.

[0073] Optionally, determining the degree of influence of each target attribute information on the total price difference between the comparable items based on the magnitude of price influence includes: summing the magnitude of price influence of each target attribute information on the comparable items to obtain a summation result; calculating the proportion of the magnitude of price influence of each target attribute information on the comparable items relative to the summation result; and determining the degree of influence of each target attribute information on the total price difference between the comparable items based on the proportion of each target attribute information.

[0074] In this embodiment, the price impact magnitude corresponding to each target attribute is summed to obtain the total price impact magnitude of all target attribute information. Assuming the price difference between product A and product B is 100 yuan, and products A and B differ in attributes A and B respectively, the price impact magnitude of attribute A is 80 yuan, and the price impact magnitude of attribute B is 40 yuan. Therefore, the total sum is 80 + 40 = 120 yuan. Based on the proportion of the price impact magnitude of each target attribute information to the total sum, the relative contribution of each attribute to the total price difference between the compared items is calculated. Taking attribute A as an example, its price impact magnitude proportion is 80 / 120 ≈ 66.67%, while that of attribute B is 40 / 120 ≈ 33.33%. Finally, based on the proportion of each target attribute information and combined with the actual total price difference between the compared items, its specific degree of influence on the total price difference is determined. Continuing with the example above, if the total price difference between product A and product B is 100 yuan, then the influence of attribute A on the total price difference is approximately 67 yuan (100 × 66.67%), while the influence of attribute B is approximately 33 yuan (100 × 33.33%). This indicates that of the total price difference between product A and product B, 67 yuan can be attributed to the difference in attribute A, and 33 yuan to the difference in attribute B. This method quantifies the specific contribution of the difference attributes between the items to be compared, thereby clarifying the degree of influence of each target attribute information on the total price difference, providing support for the subsequent generation of product comparison results and the optimization of pricing strategies.

[0075] According to a second aspect of the present invention, a method for training a commodity price prediction model is provided.

[0076] Figure 2 This is a schematic diagram illustrating the main flow of the training method for the commodity price prediction model according to an embodiment of the present invention; as follows: Figure 2 As shown, the training method of the item price prediction model according to an embodiment of the present invention mainly includes the following steps S201 to S203.

[0077] Step S201: Obtain the category name, core attributes, and item price corresponding to the target item.

[0078] Step S202: Fill the category name and core attributes corresponding to the target item into the pre-configured training data template to obtain the first training data.

[0079] Step S203: The first training data is used as the model input and the item price is used as the model output to train the item price prediction model; wherein, the item price prediction model is used to predict the expected price of each item to be compared.

[0080] The target item refers to the specific product that needs to be compared and analyzed during model training. For example, the target item could be an electronic product, a home appliance, or a piece of furniture. The category name describes the category to which the target item belongs, clearly identifying the item's major category. For example, "smartphone" in electronic products, "washing machine" in home appliances, and "office chair" in furniture. Core attributes are a set of attributes reflecting the key characteristics of the target item. For example, the core attributes of a smartphone might include "brand," "screen size," "processor model," "memory capacity," "storage capacity," and "camera pixels." The item price is the market price of the target item, used as the model's predictive output. Specifically, this information can be obtained through existing product databases, e-commerce platform interfaces, or data collection tools.

[0081] When filling the target item information into a pre-configured training data template, a fixed format can be used to standardize the data. For example, the template can be designed as "Category Name [SEP] Core Attribute 1 [SEP] Core Attribute 2... [SEP] Core Attribute n". The first training data is generated by replacing the placeholders in the template with the target item's category name and core attributes in a predetermined order. For example, for the smartphone product "iPhone 15 Pro Max", the information is "Smartphone [SEP] Apple [SEP] 6.7-inch [SEP] A17 Pro [SEP] 8GB [SEP] 256GB [SEP] 48MP". The first training data after filling is structured text input. During model training, BERT or other pre-trained language models can be used to encode the input data, mapping the text information into a high-dimensional vector space. The model parameters are then updated via backpropagation, combined with the supervision signal of the item price. A fully trained item price prediction model can effectively capture the combined impact of product category and attribute features on price, and can be used for subsequent price analysis and item comparison tasks.

[0082] For example, in predicting laptop prices, the target item can be laptops of different brands and configurations. First, obtain the category name "laptop," core attributes (such as "brand," "processor model," "memory capacity," "hard drive capacity," and "graphics card model"), and item price from an e-commerce platform. For instance, the product information might be "Lenovo, i5-13500H, 16GB, 512GB, RTX4060," with a price of "6999 yuan." This information is then entered into the template "Category Name [SEP] Core Attribute 1 [SEP] Core Attribute 2... [SEP] Core Attribute n," generating the first training data "Laptop [SEP] Lenovo [SEP] i5-13500H [SEP] 16GB [SEP] 512GB [SEP] RTX4060," with the price "6999 yuan" as the model output. By collecting data on laptops with various configurations, a training set is constructed and input into a deep learning-based price prediction model (such as LSTM or Transformer). Through iterative optimization, a high-precision item price prediction model is obtained. Through the above process, a dedicated item price prediction model can be trained according to different product categories. The model can effectively capture the complex relationship between the category characteristics and attributes of the target item, and achieve accurate prediction of the expected price of the product, which can be used for price analysis, comparison tasks or other pricing scenarios.

[0083] This invention provides an embodiment of the invention that obtains the category name, core attributes, and item price of a target item; fills the category name and core attributes of the target item into a pre-configured training data template to obtain first training data; uses the first training data as model input and the item price as model output to train an item price prediction model; wherein, the item price prediction model is used to predict the expected price of each item to be compared; thereby, the expected price of an item can be predicted based on its attributes, thus improving the accuracy of item price prediction and increasing the efficiency of item comparison.

[0084] Optionally, the pre-configured training data template is obtained through the following steps: acquiring multiple categories and corresponding item attributes, and determining the information entropy of each item attribute; sorting the information entropy of each item attribute, and determining the core attribute of the category based on the sorting result; and concatenating the acquired category name and the core attribute of the category to obtain the training data template.

[0085] In this embodiment, multiple categories and their corresponding item attributes are acquired. For each category, all relevant attributes of all items under that category are listed. For example, the electronics category may include multiple attributes such as brand, model, screen size, processor, and memory capacity. The home appliance category may include power, energy efficiency rating, brand, and type. The information entropy of each item attribute is calculated. Information entropy is a measure of information uncertainty, reflecting the distribution of an attribute value. The higher the information entropy, the more dispersed and uncertain the attribute value; conversely, the lower the information entropy, the more concentrated and certain the attribute value. Information entropy can be calculated by statistically analyzing the frequency of each attribute value in each item and then using the entropy formula or other methods. The information entropy of each item attribute is sorted to obtain an order from high to low. Generally, attributes with higher information entropy contribute more to the item's recognizability and are therefore more likely to be core attributes. Based on the sorting results, attributes with higher information entropy are selected as the core attributes of the category. These core attributes provide more valuable information, helping the prediction model better understand and process the price and characteristics of items. A training data template is constructed by concatenating the category name and its core attributes. This template specifies the core attributes for each category and the format in which these attributes should be presented in the model input. For example, for the smartphone category, core attributes might include "brand," "processor," "memory," and "screen size"; for the washing machine category, core attributes might include "capacity," "speed," and "energy efficiency rating." The template can be concatenated as: "Category Name [SEP]Core Attribute 1 [SEP]Core Attribute 2 [SEP]...". This method generates structured training data, enabling the model to learn the relationship between item prices and their core attributes during subsequent training.

[0086] Through the above steps, this invention can accurately select the core attributes that have a significant impact on item price prediction by sorting them according to the entropy value of the category attributes. Finally, it constructs a training data template with efficient prediction capabilities. This template will be used as the model input, which is conducive to training a price prediction model with high accuracy and strong generalization ability.

[0087] Figure 3 This is a schematic diagram of the main flow of an article comparison method according to a preferred embodiment of the present invention. Figure 4 This is a schematic diagram illustrating the main flow of a training method for a commodity price prediction model according to a preferred embodiment of the present invention. Figure 3 and Figure 4As shown in the illustration, the scenario of this embodiment is that when a user compares products, based on the price and attribute information of two products selected by the user, a price difference splitting algorithm is used to break down the price difference between the two products into different attributes. This allows the user to intuitively display the differences in each product attribute, assisting the user in making a product selection. The product price difference splitting algorithm disclosed in this embodiment successfully solves the problems of existing product comparison systems lacking fine-grained price comparison information and users being unable to intuitively perceive the differences in product attributes, resulting in a poor user experience. It can split the price difference of products into different attributes, enriching the completeness of product comparison display information and improving the user experience.

[0088] Specifically, such as Figure 4 As shown, a commodity price prediction model first calculates the 20 attributes with the lowest information entropy for each third-level category (such as flat-screen TVs, laptops, gaming laptops, mobile phones, etc.) as the core attributes of that category. Then, it filters highly correlated attributes through correlation analysis, retaining only the attribute with the lowest information entropy in the highly correlated attribute group. Finally, it concatenates the category name and the filtered commodity core attribute information to form a training data template, and fills the template with the category name and the corresponding core attributes of each commodity to construct training data. For example, for the laptop category, the training data template is constructed as "Laptop [SEP] Brand [SEP] CPU Model [SEP] Graphics Card Model [SEP] Hard Drive [SEP] Memory [SEP] Resolution [SEP] Refresh Rate". Then, the information corresponding to the commodity is filled into the template as training data, such as "Laptop [SEP] xx [SEP] i7-14650HX [SEP] RTX4060 [SEP] 1t [SEP] 16g [SEP] 2.5k [SEP] 165hz". Then, a price prediction model is built based on the BERT model and trained using the data constructed above. The input to the price prediction model is the data constructed above. First, the original text input data is encoded using the BERT model, mapping the input text into a high-dimensional space vector. Then, the dimension of the vector is compressed through an average pooling layer. Finally, the price of the product is decoded through two fully connected layers.

[0089] like Figure 3As shown, the virtual product data construction involves identifying the different attribute information of two products selected for comparison. Then, these different product attributes are iterated through sequentially. While keeping other product attributes constant, the corresponding attributes are replaced with those of the comparison product, constructing data in a manner similar to the steps described above. For example, given two products under the "Laptop" category, product A has a CPU model of 12400f, and product B has a CPU model of 12600kf. Based on product A's attributes, data is constructed as "Laptop [SEP] Brand [SEP] 12600kf [SEP] Graphics Card Model [SEP] Hard Drive [SEP] Memory [SEP] Resolution [SEP] Refresh Rate". Based on product B's attributes, data is constructed as "Laptop [SEP] Brand [SEP] 12400f [SEP] Graphics Card Model [SEP] Hard Drive [SEP] Memory [SEP] Resolution [SEP] Refresh Rate". Essentially, the CPU models of the two products are replaced to construct the data.

[0090] The product price difference breakdown involves feeding the constructed virtual product data into a trained product price prediction model to predict the expected price of each product. Comparing this prediction with the actual price of the product reveals the price change after attribute replacement. Assigning a price difference based on the magnitude of the price change after each attribute replacement yields the product price difference breakdown. For example, if product A and product B differ by two attributes, with a price difference of 100 yuan, replacing attribute A in both products increases the price of product A by 50 yuan and decreases the price of product B by 30 yuan. The impact of attribute A on the price is 80 yuan. Similarly, if the impact of attribute B on the price is 40 yuan, then the total price difference between product A and product B is 100 yuan (67 yuan due to attribute A and 33 yuan due to attribute B).

[0091] This invention constructs virtual product data based on each different attribute of a product pair, feeds it into a product price prediction model to predict the price of the constructed virtual products, then calculates the impact of changes in each attribute on the product price, and finally breaks down the price difference of the product pair into different attributes based on the magnitude of the impact, using the positive or negative value of the attribute to represent its superiority or inferiority. Addressing the problem that users cannot intuitively perceive the differences between attributes, this invention, by breaking down the product price difference into each attribute, visually displays the proportion of each attribute's difference in the overall price, helping users understand which attributes in the product pair have the largest differences.

[0092] According to a third aspect of the present invention, an article comparison device is provided.

[0093] Figure 5 This is a schematic diagram of the main modules of the article comparison device according to an embodiment of the present invention, such as... Figure 5 As shown, an article comparison device 500 includes:

[0094] The determination module 501 is used to compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; the target attribute information is the attribute information that is different between the items to be compared.

[0095] Module 502 is used to construct virtual item data for each target attribute information and determine the price impact of each target attribute information on the item to be compared based on the virtual item data;

[0096] The generation module 503 is used to determine the degree of influence of each target attribute information on the price difference between the items to be compared based on the price influence range, and to generate the item comparison results of the items to be compared based on the degree of influence.

[0097] Optionally, the building module 502 is also used for:

[0098] Repeat the following steps until all target attribute information has been replaced: replace one target attribute of the first item to be compared with one target attribute of the second item to be compared, and keep the other attribute information of the first item to be compared unchanged; wherein, the first item to be compared and the second item to be compared are different; form a virtual item data of the first item to be compared based on the replaced attribute information;

[0099] The virtual item data for each item to be compared is obtained based on the data of each virtual item.

[0100] Optionally, the building module 502 is also used for:

[0101] Predict the expected price of each item to be compared based on virtual item data;

[0102] The expected price and preset price of each item to be compared are compared one by one to obtain the impact of each target attribute information on the price of the item to be compared.

[0103] Optionally, the generation module 503 is also used for:

[0104] The summation of the impact of each target attribute on the price of the item under comparison is obtained;

[0105] Calculate the percentage of the impact of each target attribute information on the price of the item compared to the summation result;

[0106] The degree of influence of the total price difference between the items to be compared is determined based on the proportion of each target attribute information.

[0107] It should be noted that the specific implementation details of the item comparison device in the embodiments of the present invention have been described in detail in the item comparison method above, so the details will not be repeated here.

[0108] According to a fourth aspect of the present invention, a training apparatus for a commodity price prediction model is provided.

[0109] Figure 6 This is a schematic diagram of the main modules of a training device for a commodity price prediction model according to an embodiment of the present invention, as shown below. Figure 6 As shown, a training device 600 for a commodity price prediction model mainly includes:

[0110] Module 601 is used to obtain the category name, core attributes, and item price of the target item.

[0111] The input module 602 is used to input the category name and core attributes corresponding to the target item into the pre-configured training data template to obtain the first training data;

[0112] The training module 603 is used to train an item price prediction model by taking the first training data as the model input and the item price as the model output; wherein, the item price prediction model is used to predict the expected price of each item to be compared.

[0113] Optionally, the training device 600 also includes a configuration module, which is used for:

[0114] Obtain multiple categories and their corresponding item attributes, and determine the information entropy of each item attribute;

[0115] Sort the information entropy of each item's attributes and determine the core attributes of the category based on the sorting results;

[0116] The obtained category names and core attributes are concatenated to obtain the training data template.

[0117] It should be noted that the specific implementation details of the training device for the item price prediction model in this embodiment of the invention have been described in detail in the above-described training method for the item price prediction model, so the details will not be repeated here.

[0118] According to a fifth aspect of the present invention, an electronic device is provided, comprising:

[0119] One or more processors;

[0120] Storage device for storing one or more programs.

[0121] When one or more programs are executed by one or more processors, the one or more processors implement the methods provided by the first aspect and / or the second aspect of the embodiments of the present invention.

[0122] According to a sixth aspect of the present invention, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the methods provided in the first aspect and / or the second aspect of the present invention.

[0123] According to a seventh aspect of the present invention, a computer program product is provided, including a computer program that, when executed by a processor, implements the methods provided by the first aspect and / or the second aspect of the present invention.

[0124] Figure 7 An exemplary system architecture 700 is shown that can be applied to the article comparison method or article comparison device of the present invention. Figure 7 An exemplary system architecture 700 is shown, which can be used to train a product price prediction model according to embodiments of the present invention, or to train a product price prediction model.

[0125] like Figure 7 As shown, system architecture 700 may include terminal devices 701, 702, and 703, a network 704, and a server 705. Network 704 serves as the medium for providing communication links between terminal devices 701, 702, and 703 and server 705. Network 704 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0126] Users can use terminal devices 701, 702, and 703 to interact with server 705 via network 704 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 701, 702, and 703, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0127] Terminal devices 701, 702, and 703 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0128] Server 705 can be a server that provides various services, such as a backend management server that supports shopping websites browsed by users using terminal devices 701, 702, and 703 (for example only). The backend management server can analyze and process data such as received item comparison requests, and feed back the processing results (such as item comparison results - for example only) to the terminal devices.

[0129] It should be noted that the item comparison method provided in the embodiments of the present invention is generally run by server 705, and correspondingly, the item comparison device is generally set in server 705.

[0130] It should be noted that the training method of the item price prediction model provided in the embodiments of the present invention is generally run by the server 705, and correspondingly, the training device of the item price prediction model is generally set in the server 705.

[0131] It should be understood that Figure 7 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0132] The following is for reference. Figure 8 It shows a schematic diagram of the structure of a computer system 800 suitable for implementing a terminal device of the present invention. Figure 8 The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0133] like Figure 8 As shown, the computer system 800 includes a central processing unit (CPU) 801, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 802 or programs loaded from storage section 808 into random access memory (RAM) 803. The RAM 803 also stores various programs and data required for the operation of the system 800. The CPU 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0134] The following components are connected to I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to I / O interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 810 as needed so that computer programs read from it can be installed into storage section 808 as needed.

[0135] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is run by central processing unit (CPU) 801, it performs the functions defined above in the system of this invention.

[0136] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium, a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit programs for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0137] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more operable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually operate substantially in parallel, and they may sometimes operate in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0138] The modules described in the embodiments of the present invention can be implemented in software or hardware. The described modules can also be housed in a processor; for example, a processor may include a determining module, a building module, and a generating module. The names of these modules do not necessarily limit the module itself; for example, the determining module may also be described as "a module for comparing the attribute information corresponding to each item to be compared and determining the target attribute information based on the comparison results." Alternatively, a processor may include an acquiring module, an input module, and a training module, where the names of these modules do not necessarily limit the module itself; for example, the acquiring module may also be described as "a module for acquiring the category name, core attributes, and item price corresponding to the target item."

[0139] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include: comparing attribute information corresponding to each item to be compared, determining target attribute information based on the comparison results; the target attribute information being the different attribute information between the items to be compared; constructing virtual item data for each target attribute information, determining the price influence range of each target attribute information on the items to be compared based on the virtual item data; determining the degree of influence of each target attribute information on the total price difference between the items to be compared based on the price influence range, and generating an item comparison result for the items to be compared based on the degree of influence. Alternatively, the device implements the following method: obtaining the category name, core attributes, and item price corresponding to the target item; filling the category name and core attributes corresponding to the target item into a pre-configured training data template to obtain first training data; using the first training data as model input and the item price as model output to train an item price prediction model; wherein the item price prediction model is used to predict the expected price of each item to be compared.

[0140] The computer program product provided in the embodiments of the present invention includes a computer program that, when executed by a processor, implements the item comparison method in the first aspect of the present invention and / or the item price prediction model training method in the second aspect of the present invention.

[0141] According to the technical solution of the present invention, the following advantages or beneficial effects are achieved: by comparing the attribute information corresponding to each item to be compared, target attribute information is determined based on the comparison results; the target attribute information consists of different attribute information between the items to be compared; virtual item data is constructed for each target attribute information, and the price impact of each target attribute information on the items to be compared is determined based on the virtual item data; the degree of influence of each target attribute information on the total price difference between the items to be compared is determined based on the price impact, and the item comparison results of the items to be compared are generated based on the degree of influence; this embodiment constructs virtual item data based on each different attribute between items, and determines the impact of changes in each attribute on the item price, and breaks down the total price difference of the items into each different attribute based on the price impact, which can intuitively show users the proportion of each attribute difference relative to the total price difference, thus improving the user experience. By obtaining the category name, core attributes, and item price of the target item; filling the category name and core attributes of the target item into a pre-configured training data template to obtain the first training data; using the first training data as the model input and the item price as the model output, a price prediction model is trained; the price prediction model is used to predict the expected price of each item to be compared; thus, the expected price of an item can be predicted based on its attributes, thereby improving the accuracy of price prediction and increasing the efficiency of item comparison.

[0142] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

[0143] It should be noted that the acquisition, storage, and application of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

Claims

1. A method for comparing items, characterized in that, include: Compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; The target attribute information refers to the different attribute information between the items to be compared; Virtual item data is constructed for each of the target attribute information, and the impact of each target attribute information on the price of the comparison item is determined based on the virtual item data; Based on the magnitude of the price impact, determine the degree of influence of each target attribute information on the total price difference between the items to be compared, and generate the item comparison results of the items to be compared based on the degree of influence.

2. The method according to claim 1, characterized in that, For each of the target attribute information, virtual item data is constructed, including: The following steps are repeated until all target attribute information is replaced: replace one target attribute information of the first item to be compared with one target attribute information of the second item to be compared, and keep the other attribute information of the first item to be compared unchanged; wherein, the first item to be compared and the second item to be compared are different; and form a virtual item data of the first item to be compared based on the replaced attribute information. The virtual item data for each item to be compared is obtained based on the data of each virtual item.

3. The method according to claim 1, characterized in that, Determining the impact of each target attribute on the price of the comparison item based on the virtual item data includes: Predict the expected price of each of the items to be compared based on the virtual item data; By comparing the expected price and the preset price of each item to be compared, the influence of each target attribute information on the price of the item to be compared is obtained.

4. The method according to claim 1, characterized in that, Determining the degree of influence of each of the target attribute pieces of information on the total price difference between the comparable items based on the price impact magnitude includes: The summation result is obtained by summing the impact of each target attribute information on the price of the item to be compared. Calculate the percentage of the impact of each target attribute on the price of the item to be compared relative to the summation result; The degree of influence on the total price difference between the items to be compared is determined based on the proportion of each of the target attribute information.

5. A training method for a commodity price prediction model, characterized in that, include: Retrieve the category name, core attributes, and item price of the target item; The category name and core attributes corresponding to the target item are filled into the pre-configured training data template to obtain the first training data; The first training data is used as the model input, and the item price is used as the model output to train the item price prediction model; wherein, the item price prediction model is used to predict the expected price of each of the items to be compared.

6. The method according to claim 5, characterized in that, The pre-configured training data template is obtained through the following steps: Obtain multiple categories and the corresponding item attributes of the categories, and determine the information entropy of each item attribute; The information entropy of each of the item attributes is sorted, and the core attributes of the category are determined based on the sorting results. The category name and core attributes of the obtained category are concatenated to obtain the training data template.

7. An item comparison device, characterized in that, include: The determination module is used to compare the attribute information corresponding to each item to be compared, and determine the target attribute information based on the comparison results; The target attribute information refers to the different attribute information between the items to be compared; The construction module is used to construct virtual item data for each of the target attribute information, and determine the impact of each of the target attribute information on the price of the comparison item based on the virtual item data; The generation module is used to determine the degree of influence of each target attribute information on the price difference between the items to be compared based on the price influence range, and to generate the item comparison results of the items to be compared according to the degree of influence.

8. A training device for a commodity price prediction model, characterized in that, include: The acquisition module is used to obtain the category name, core attributes, and item price of the target item; The input module is used to input the category name and core attributes corresponding to the target item into a pre-configured training data template to obtain the first training data; The training module is used to train the item price prediction model by taking the first training data as model input and the item price as model output; wherein the item price prediction model is used to predict the expected price of each of the items to be compared.

9. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.

10. A computer-readable medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-6.