E-commerce live broadcast intelligent item selection method and system

Abstract

The invention relates to an e-commerce live broadcast intelligent item selection method and system. The method comprises the following steps: S1, constructing a multi-dimensional commodity selection criterion system, and collecting original data of a to-be-recommended commodity under each criterion; step S2, subjective weight calculation; s3, objective weight calculation; and S4, carrying out weight fusion and scheme evaluation. According to the method, weight quantification verification is performed through the analytic hierarchy process and the entropy weight method, and the fusion coefficient is dynamically adjusted in combination with the business scene, so that the problems of subjective experience contradiction, poor cold start adaptability and low decision transparency of traditional product selection are solved, and an experience quantification-data verification-weight fusion decision chain is formed; and finally, the product selection scientificity is improved, the scene adaptability is enhanced, and the decision transparency is improved.

Description

Technical Field

[0001] This invention relates to the field of e-commerce live streaming technology, specifically to an intelligent product selection method and system for e-commerce live streaming. Background Technology

[0002] With the rapid development of e-commerce live streaming, product selection is a core element of live streaming operations, and the quality of its decision-making directly affects the conversion rate of the live stream.

[0003] Traditional decision-making techniques face numerous technical bottlenecks, resulting in insufficient scientific rigor and scenario adaptability in product selection.

[0004] First, regarding subjective experience-based decision-making, existing methods rely on manually setting the weights of product selection metrics. For example, livestreamers might subjectively rank products based on experience, such as "appearance > cost-effectiveness > practicality," but this lacks a systematic logical verification mechanism. This experience-driven model often leads to contradictory metric importance, such as "cost-effectiveness is more important than brand power, brand power is more important than user satisfaction rate, and user satisfaction rate surpasses cost-effectiveness," resulting in logical confusion and a lack of scientific basis for weight allocation.

[0005] Secondly, in terms of data-driven models, pure machine learning methods rely too heavily on objective data such as historical sales and user interactions. They are poorly adapted to cold start scenarios that lack historical data, such as new products and customized models, and are prone to overlooking emerging products with market potential.

[0006] More critically, existing decision-making models generally suffer from low logical transparency. Whether it's human experience-based decision-making or algorithmic recommendations, the decision-making process lacks interpretability. This hinders the iterative optimization of product selection strategies and makes it impossible to achieve end-to-end traceability of the decision-making process. The root cause of these technical bottlenecks lies in the lack of a deep integration mechanism that combines subjective logical verification, objective data quantification, and dynamic scenario adaptation. This has kept e-commerce livestream product selection in a trial-and-error stage for a long time, making it difficult to achieve scientific and systematic quantitative decision-making.

[0007] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0008] In view of the above-mentioned deficiencies of the prior art, the first aspect of the present invention provides an intelligent product selection method for e-commerce live streaming, comprising the following steps: Step S1: Construct a multi-dimensional product selection criterion system and collect raw data on the products to be recommended under each criterion; among which, the Analytic Hierarchy Process (AHP) is used to perform AHP judgment matrix analysis on subjective criteria. The construction of the entropy weight method is based on the entropy weight method to perform entropy weight data matrix on objective criteria. The construction of the criteria is as follows: m is the number of products, and n is the number of criteria; the subjective criteria include at least one of the following elements: anchor experience, product attributes, live streaming scenario, and brand awareness; the objective criteria include at least one of the following elements: product sales, user interaction data, and inventory turnover rate. Step S2: Subjective weight calculation; wherein, for the AHP judgment matrix Perform eigenvalue decomposition to obtain the largest eigenvalue. and corresponding feature vectors After normalization, the subjective weight vector is obtained and a consistency check is performed. Step S3: Objective weight calculation; wherein, for the entropy weight method data matrix After normalization, the probability matrix is ​​obtained. And calculate the entropy value of each criterion. By mining the discriminative value of data, the objective weights of each criterion, i.e., entropy weights, can be obtained. ; Step S4: Weight fusion and scheme evaluation; wherein, a linear weighted fusion of subjective and objective weights is adopted. The fusion coefficient is dynamically adjusted according to the live streaming business scenario, and finally the comprehensive score of each product is calculated. This enables the selection of the best products.

[0009] In the e-commerce live streaming intelligent product selection method described above, optionally, in step S2, the normalized subjective weight vector is calculated based on the following formula: ; When the random consistency ratio CR < 0.1, the judgment matrix is ​​determined. Reasonable.

[0010] In the e-commerce live streaming intelligent product selection method described above, optionally, in step S3, the probability matrix... of ; Entropy values ​​of each criterion Calculated using the following formula: ; in, It should be a very small positive number to avoid logarithmic divergence; The objective weights of each criterion, i.e., entropy weights. Calculated using the following formula: .

[0011] In the intelligent product selection method for e-commerce live streaming described above, a linear weighted fusion of subjective and objective weights may be optionally adopted. Calculated using the following formula: , Where α and β are fusion coefficients, which are dynamically adjusted according to the live streaming business scenario or product life cycle stage, α+β=1, α,β∈[0,1]; The overall score for each product is calculated using the following formula: , The product with the highest score is selected as the recommendation result.

[0012] In the intelligent product selection method for e-commerce live streaming described above, optionally, in an experience-driven scenario, the subjective weight α=0.6, and in a data-driven scenario, β=0.7.

[0013] In the e-commerce live streaming intelligent product selection method described above, optionally, the life cycle stages include the introduction stage, growth stage, maturity stage, and decline stage. The corresponding fusion coefficient configuration rules are as follows: Introduction phase: α=0.7, β=0.3, dominated by expert experience, dependent on brand and trend matching degree; Growth stage: α=0.5, β=0.5, balancing subjective judgment and objective data feedback; Maturity stage: α=0.3, β=0.7, dominated by historical objective data, focusing on sales volume and profit stability; Recession phase: α=0.6, β=0.4, expert decision on whether to liquidate or restart, with historical objective data as a supplementary reference.

[0014] To achieve the above objectives, a second aspect of the present invention provides an intelligent product selection system for e-commerce live streaming, using the intelligent product selection method for e-commerce live streaming as described in any of the embodiments of the first aspect, including: The criterion construction and data acquisition module includes an AHP judgment matrix for subjective criteria based on the analytic hierarchy process. The construction of the entropy weight method is based on the entropy weight method to perform entropy weight data matrix on objective criteria. The construction of the criteria is as follows: m is the number of products, and n is the number of criteria; the subjective criteria include at least one of the following elements: anchor experience, product attributes, live streaming scenario, and brand awareness; the objective criteria include at least one of the following elements: product sales, user interaction data, and inventory turnover rate. Subjective weight calculation module; wherein, for the AHP judgment matrix Perform eigenvalue decomposition to obtain the largest eigenvalue. and corresponding feature vectors After normalization, the subjective weight vector is obtained and a consistency check is performed. Objective weight calculation module; wherein, for the entropy weight method data matrix After normalization, the probability matrix is ​​obtained. And calculate the entropy value of each criterion. By mining the discriminative value of data, the objective weights of each criterion, i.e., entropy weights, can be obtained. ; The weight fusion and scheme evaluation module; among which, a linear weighted fusion of subjective and objective weights is adopted. The fusion coefficient is dynamically adjusted according to the live streaming business scenario, and finally the comprehensive score of each product is calculated. This enables the selection of the best products.

[0015] To achieve the above objectives, a third aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when running the program, implements the e-commerce live streaming intelligent product selection method as described in any one of the first aspects above.

[0016] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium storing computer-executable instructions or a computer program, which, when processed and executed, implement the e-commerce live streaming intelligent product selection method as described in any one of the first aspects above.

[0017] This invention provides an intelligent product selection method and system for e-commerce live streaming. By constructing a weight calculation system that integrates subjective and objective factors, and combining the analytic hierarchy process (AHP) and entropy weight method to achieve quantitative verification of weight allocation, the system dynamically adjusts the integration coefficient based on business scenarios. This effectively solves the problems of subjective experience contradictions, poor cold start adaptability, and low decision-making transparency in traditional product selection methods. It deeply integrates subjective logic verification, objective data quantification, and dynamic scenario adaptation, forming a complete decision-making chain of "experience quantification - data verification - weight integration." This method has the advantages of improving the scientific nature of product selection decisions, enhancing scenario adaptability, and increasing the transparency of the decision-making process.

[0018] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating an embodiment of an intelligent product selection method for e-commerce live streaming provided by the present invention. Detailed Implementation

[0020] To make the technical means, inventive features, objectives, and effects of the invention readily understandable, the invention is further illustrated below with reference to specific figures. However, the invention is not limited to the embodiments described below.

[0021] It should be noted that the structures, proportions, sizes, etc., illustrated in the accompanying drawings of this specification are only used to complement the content disclosed in the specification for those skilled in the art to understand and read, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0022] Terms such as “comprising” and “including” indicate that, in addition to the components that are directly and explicitly stated in the specification and claims, the technical solution of the present invention does not exclude the presence of other components that are not directly or explicitly stated.

[0023] Furthermore, the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0024] In traditional e-commerce live-streaming product selection decision-making systems, the disconnect between subjective experience-based judgment and objective data evaluation leads to structural flaws in the decision-making logic. When product selection metrics are manually weighted, the lack of a systematic logical verification mechanism often results in contradictory weighting of metrics. For example, the weight of cost-effectiveness may be higher than brand power, brand power may be higher than user positive review rate, but user positive review rate may be weighted higher than cost-effectiveness, creating a paradoxical situation. This chaotic weighting directly causes the ranking of recommended products to deviate from business objectives, especially in cold start scenarios. New products lack historical sales data and cannot be effectively evaluated by machine learning models, while manual experience-based decisions lack quantitative support and are difficult to accurately predict product potential.

[0025] like Figure 1 As shown, the intelligent product selection method for e-commerce live streaming according to this application may include the following steps: Step S1: Construct a multi-dimensional product selection criterion system and collect the original data of the products to be recommended under each criterion.

[0026] In step S1, based primarily on e-commerce business needs, various evaluation criteria are determined, such as user preferences, product prices, and inventory quantities. Specifically, domain experts construct an AHP judgment matrix based on the Analytic Hierarchy Process (AHP) to determine the relative importance of the criteria. The construction of the m×n-dimensional entropy weight data matrix is ​​based on the entropy weight method for objective criteria. The construction of , where m is the quantity of goods and n is the quantity of criteria.

[0027] Subjective criteria can include elements such as the streamer's experience, product attributes, livestreaming scenario, and brand awareness. Objective criteria can include elements such as product sales volume, user interaction data, and inventory turnover rate.

[0028] In some of the solutions mentioned above in this application, an AHP judgment matrix based on the analytic hierarchy process is proposed to determine the weights of indicators. However, if there is a lack of a unified importance scaling rule when constructing the judgment matrix, it may lead to subjective arbitrariness in the importance judgment of different decision-makers on the same criterion. For example, the importance of indicator i relative to j may be arbitrarily assigned a non-standardized value, which may lead to logical contradictions in weight allocation and affect the effectiveness of subsequent consistency checks. Table 1 For example, taking product selection decisions in an e-commerce live streaming scenario, we select three subjective criteria as upper-level indicators: "host experience (U1)," "product attributes (U2)," and "live streaming scenario (U3)." Based on this, a three-dimensional judgment matrix can be constructed. Specifically .

[0029] Among them, a 12 =3 indicates that the streamer's experience is slightly more important than product attributes; a 13 =5 indicates that the streamer's experience is significantly more important than the live streaming scenario; a 23 =3 indicates that product attributes are slightly more important than the live streaming scenario; the remaining elements are based on a ij = 1 / a ji Relationship filling.

[0030] For example, when a decision-maker considers a criterion to be between slightly important and significantly important, they can only choose the value 4 instead of any decimal. This discretization method matches the integer matrix operation characteristics of subsequent eigenvalue decomposition algorithms, avoiding errors in eigenvector calculation caused by non-standardized input. In this way, the decision-maker's subjective judgment can be transformed into quantifiable numerical values, providing standardized input data for subsequent weight calculations.

[0031] Through the above technical solution, this application achieves a standardized setting of the importance ratio of indicators in the AHP judgment matrix. This avoids subjective arbitrariness in the importance judgments of different decision-makers regarding the same criterion, reducing the risk of logical contradictions in weight allocation. Furthermore, this standardized setting provides a reliable basis for subsequent consistency checks, improving the credibility and scientific rigor of the weight calculation results. Specifically, by adopting the 1-9 scaling method, decision-makers are confined to a preset quantification level for judgment, which not only simplifies the judgment process but also ensures the comparability of judgment results among different decision-makers. For example, when multiple decision-makers evaluate the same set of indicators, their results can be directly compared and synthesized because the same scaling rules are used. In addition, this standardized method facilitates sensitivity analysis; by adjusting certain judgment values, changes in the final weights can be observed, thereby assessing the stability and reliability of the judgment.

[0032] Step S2: Calculate subjective weights.

[0033] In step S2, the AHP judgment matrix is... Perform eigenvalue decomposition to obtain the largest eigenvalue. and corresponding feature vectors After normalization, the subjective weight vector is obtained and a consistency check is performed.

[0034] In some of the solutions mentioned above in this application, subjective weight vectors are obtained through eigenvalue decomposition and normalization and consistency checks are performed. However, if there is a lack of effective verification of the logical consistency of the judgment matrix during the subjective weight calculation process, the weight allocation results may have inherent contradictions, such as the indicator importance ratio violating transitivity, which in turn affects the scientificity and reliability of product selection decisions.

[0035] In this regard, this application further proposes that in step S2, the normalized subjective weight vector is calculated based on the following formula: .

[0036] When the random consistency ratio CR < 0.1, the decision matrix is ​​determined. Reasonable.

[0037] Specifically, after constructing the AHP judgment matrix, the largest eigenvalue and its corresponding eigenvector are solved using the power iteration method or the Jacobian method. Eigenvector normalization eliminates dimensional differences; for example, when eigenvector elements have orders of magnitude differences, normalization compresses each element into the 0-1 interval, ensuring the rationality and comparability of the weight values. During the consistency check, the CI value reflects the degree to which the judgment matrix deviates from consistency, and the RI value is obtained by randomly generating a large number of matrices and calculating their average CI value.

[0038] Through the above technical solution, this application achieves normalization and consistency verification of the subjective weight vector. Normalization ensures that the sum of all elements in the weight vector is 1, making the weights comparable across different criteria. The consistency verification mechanism evaluates the logical consistency of the judgment matrix using the quantitative indicator CR, effectively identifying and eliminating unreasonable weight allocations caused by contradictions in subjective judgments. This method improves the scientific rigor and reliability of subjective weight calculation, providing more accurate input for subsequent weight fusion, thereby enhancing the rationality of the overall product selection decision.

[0039] Step S3: Calculate objective weights.

[0040] In step S3, the entropy weighting data matrix is ​​processed. After normalization, the probability matrix is ​​obtained. And calculate the entropy value of each criterion. By mining the discriminative value of data, the objective weights of each criterion, i.e., entropy weights, can be obtained. .

[0041] In some of the solutions described above in this application, directly calculating the entropy value of the original data may lead to distortion of the probability matrix due to uneven data distribution or extreme values. At the same time, the traditional entropy weight method does not consider the problem of logarithmic operation divergence when the data discrimination is insufficient, which leads to the objective weight allocation deviating from the actual business needs.

[0042] To address this, this application further proposes that the entropy value of each criterion be calculated by adding a correction term with a very small positive number, and that the objective weight be obtained by converting it using the entropy weight formula. The normalization of the probability matrix can be achieved using the range method or the standard deviation method, mapping the original data of different dimensions to a unified interval, for example, linearly transforming the data to the range [0,1].

[0043] In step S3, the probability matrix of ; Entropy values ​​of each criterion Calculated using the following formula: ; in, It is a very small positive number to avoid logarithmic divergence.

[0044] The objective weights of each criterion, i.e., entropy weights. Calculated using the following formula: .

[0045] Specifically, in the data preprocessing stage, the original data matrix is ​​normalized using the range method to eliminate the impact of differences in the dimensions of different indicators on the weight calculation. Subsequently, when calculating the entropy value of each criterion, a minimum positive number is introduced as a smoothing factor to correct for possible zero or minimum values ​​in the probability matrix, ensuring the numerical stability of the logarithmic operation. Finally, the entropy value of each criterion is converted into a weight coefficient using the entropy weight formula, so that criteria with higher data discrimination are assigned higher weight values.

[0046] Step S4: Weight fusion and scheme evaluation.

[0047] In step S4, a linear weighted fusion of subjective and objective weights is adopted. The fusion coefficient is dynamically adjusted according to the live streaming business scenario, and finally the comprehensive score of each product is calculated. This enables the selection of the best products.

[0048] In an optional embodiment, a linear weighted fusion of subjective and objective weights is employed. It can be calculated using the following formula: ; Where α and β are fusion coefficients, which are dynamically adjusted according to the live streaming business scenario or product life cycle stage, α+β=1, α,β∈[0,1].

[0049] The overall score for each product is calculated using the following formula: , The product with the highest score is selected as the recommendation result.

[0050] In the linear weighted fusion formula, the fusion coefficients α and β are configured as dynamically adjustable parameters, with their values ​​limited to between 0 and 1 and satisfying the constraint α + β = 1. The dynamic adjustment rules for the fusion coefficients can be preset based on the type of live streaming business scenario or the product's lifecycle stage.

[0051] In an alternative embodiment, the subjective weight α = 0.6 in an experience-driven scenario and β = 0.7 in a data-driven scenario.

[0052] In the experience-driven scenario, the weighting configuration enhances expert experience by increasing the subjective weight to 0.6. For example, in the product cold start phase or new product promotion, subjective criteria such as anchor experience and brand awareness are given higher priority, thereby reducing the risk of misjudgment due to insufficient objective data. In the data-driven scenario, the weighting configuration prioritizes data indicators by increasing the objective weight to 0.7. For example, when selecting products in the mature stage, objective criteria such as product sales and user interaction data are given higher weight, ensuring that the decision-making results align with actual business performance. The benchmark setting and dynamic adjustment mechanism of the two scenarios complement each other, preserving the flexibility of weight integration while improving execution efficiency through parameter solidification in typical scenarios.

[0053] Specifically, when the live-streaming business scenario is identified as experience-driven, the subjective weight α is set to 0.6. In this case, subjective criteria such as the streamer's experience and brand awareness account for a higher proportion in the overall score calculation, making product selection decisions more reliant on expert judgment and effectively addressing decision-making needs in scenarios with missing data. When the scenario switches to data-driven, the objective weight β is set to 0.7. In this case, the weight of objective criteria such as product sales and user interaction data increases significantly, ensuring that product selection results are closely linked to real-time business data. By setting benchmark values ​​for typical scenarios, the weight fusion process obtains clear quantitative configuration basis. For example, the proportion of subjective weight increases by more than 20% in experience-driven scenarios, and the proportion of objective weight increases by more than 40% in data-driven scenarios, thereby achieving simultaneous optimization of product selection decision accuracy and execution efficiency.

[0054] In some of the solutions mentioned above in this application, the lack of specific configuration rules for the fusion coefficient at different stages of the product life cycle makes it impossible to accurately adapt the product selection strategy to the characteristics of the market stage in which the product is located, and it is difficult to balance the weight of subjective experience and objective data at different stages.

[0055] In this regard, this application further proposes that the life cycle stages include the introduction stage, growth stage, maturity stage, and decline stage, and the corresponding fusion coefficient configuration rules can be as follows: Introduction phase: α=0.7, β=0.3, dominated by expert experience, dependent on brand and trend matching degree; Growth stage: α=0.5, β=0.5, balancing subjective judgment and objective data feedback; Maturity stage: α=0.3, β=0.7, dominated by historical objective data, focusing on sales volume and profit stability; Recession phase: α=0.6, β=0.4, expert decision on whether to liquidate or restart, with historical objective data as a supplementary reference.

[0056] The lifecycle stage division is based on a combination of product market performance data and expert experience. For example, the stage classification is determined by a combination of indicators such as product sales growth rate and market share change rate. A dynamic mapping relationship is formed between the fusion coefficient configuration rules and the subjective weights calculated by the analytic hierarchy process (AHP) and the objective weights calculated by the entropy weight method. For instance, in the introduction stage, the brand matching index weight is increased to 70% of the total weight by increasing the proportion of subjective weights. The trigger conditions for adjusting the coefficients at each stage can be set to a preset period or a key indicator threshold. For example, when the product's sales growth rate is below 5% for three consecutive weeks, the decline stage coefficient switch is triggered.

[0057] Specifically, in the introduction phase, where historical data is lacking, an α=0.7 is set to give subjective weights a 70% weighting, emphasizing expert judgment on brand fit and market trends. For example, the weight of the trend prediction sub-item in the anchor experience criteria is increased to 0.45. As the product enters the growth phase, with accumulated sales data, subjective and objective weights are each 50%. For instance, the entropy weight calculation result of the user interaction data criterion begins to influence the overall score. In the maturity phase, the entropy weight calculation result of objective criteria such as inventory turnover rate increases to 70%. For example, products with stable sales for more than three months are automatically switched to this phase. In the decline phase, the subjective decision-making weight is increased again by α=0.6. For example, when the product profit margin drops to 10%, experts can use this coefficient to determine the priority of clearance strategies. This phased dynamic adjustment mechanism allows the weight fusion process to automatically adapt to changes in the product's market status, solving the problem that a single fusion coefficient cannot adapt to the product's development cycle.

[0058] Through the aforementioned technical solution, this application achieves dynamic optimization of product selection strategies at different stages of the product lifecycle. By setting differentiated fusion coefficients for each lifecycle stage, the system can perform refined adaptation based on the market status of the product. This method effectively balances the weight of subjective experience and objective data at different stages, improving the scientific nature and scenario adaptability of product selection decisions. For example, in the product introduction stage, by strengthening the weight of expert experience, the system can effectively avoid the cold start problem for new products; in the maturity stage, by increasing the weight of objective data, the system can better ensure the stability of product sales and profits; in the decline stage, by reinforcing the role of expert decision-making, the system can adjust product strategies in a timely manner, extending the product lifecycle or achieving effective clearance. This dynamic adjustment mechanism allows product selection strategies to more flexibly adapt to the characteristics of products at different market stages, thereby improving the overall product selection effect.

[0059] Here, a specific practical example will be used to illustrate the entire process of the method in this application: First, step S1: Construct a multi-dimensional product selection criterion system and collect raw data on the products to be recommended under each criterion: Establish evaluation criteria, including commodity price (U2, yuan) and commodity inventory (U3, units); Establish the AHP judgment matrix: This judgment matrix is ​​established based on the 1-9 scale method mentioned above, indicating that, according to expert judgment, the importance of commodity price is three times that of commodity inventory.

[0060] The logic is: product price > product inventory, which aligns with the recommendation strategy of "user needs as the core".

[0061] Establish an entropy weighted data matrix (products A and B): (Common sense: the lower the price, the better; the more inventory, the better.) Step S2: Subjective weight calculation: Calculation of eigenvalues ​​and eigenvectors: Calculate the largest eigenvalue of the above AHP judgment matrix. =2, the eigenvector corresponding to the largest eigenvalue is [0.9487, 0.3162]. The feature vectors are normalized (to eliminate the influence of dimensions) to obtain the subjective weight vector. =[0.75,0.25].

[0062] Consistency check: For a second-order matrix, the random consistency index RI = 0. Consistency Indicators n=2, Therefore, CI=0.

[0063] Step S3: Calculation of objective weights: Supports differentiated processing of positive and negative indicators (unifying indicator directionality). 1) Handling positive indicators: Directly normalize using the specific gravity method: , in, The original data, This is the normalized data.

[0064] 2) Handling negative indicators: First take the reciprocal, then normalize using the proportion method: , in, =10 -10 To prevent small amounts from being divided by zero.

[0065] Normalization treatment (gravity method): The corresponding data processing results are as follows: .

[0066] Protection against zero values: This includes special value treatments, such as: 0 in stock (no recommendation required), 0 price (free item). Due to the simplified requirements of the reference data and examples, scenarios with 0 inventory and 0 price are not included for the time being.

[0067] Entropy calculation (2 samples): , For the "Goods Inventory" indicator, the entropy value is: The calculation yields: 1 ≈ 0.988.

[0068] For the "commodity price" indicator, the entropy value is: Indicator 2 (Negative) Entropy Value: , The calculation yields: E2≈0.971.

[0069] Entropy weight calculation: , Indicator 1: ; Indicator 2: .

[0070] Two indicators, m=2: ; Overall coefficient of variation: 0.012 + 0.029 = 0.041 Indicator 1 Weight: ; Indicator 2 Weight: .

[0071] Step S4: Weight fusion and scheme evaluation: Weighted fusion: According to expert assessment, the ratio of subjective quantification to objective quantification is 0.5 (for demonstration purposes only). Commodity price ratio coefficient: 0.75*0.5 + 0.707*0.5 = 0.7285; Inventory ratio coefficient: 0.25*0.5+0.293*0.5=0.2715.

[0072] Overall rating: Product A: 0.7285 * 20 + 0.2715 * 50 = 14.57 + 13.575 = 28.145 Product B: 0.7285*30 + 0.2715*70 = 21.855 + 19.005 = 40.86 Product B scored higher than Product A overall. The final conclusion is that Product B scored higher than Product A (in terms of factors such as product price and inventory) and is more worthy of recommendation.

[0073] Through the implementation of the aforementioned technical solutions and specific cases, this application achieves a dynamic fusion of subjective and objective weights, improving the flexibility and scenario adaptability of product selection strategies. Specifically, by introducing an adjustable fusion coefficient, the system can adaptively adjust the weight ratio of subjective experience and objective data according to changes in the live streaming business scenario and product lifecycle stage. This dynamic adjustment mechanism overcomes the limitations of fixed weight allocation schemes, making product selection decisions more aligned with actual business needs. Furthermore, this solution retains the guiding role of subjective experience while fully utilizing the support of objective data, demonstrating its advantages in different types of live streaming scenarios. Therefore, the technical solution of this application improves the accuracy and adaptability of intelligent product selection in e-commerce live streaming, providing technical support for enhancing the effectiveness of live streaming sales.

[0074] To achieve the above objectives, the present invention also provides an intelligent product selection system for e-commerce live streaming, wherein the intelligent product selection method for e-commerce live streaming described in any of the foregoing embodiments includes: a criterion construction and data collection module, a subjective weight calculation module, an objective weight calculation module, and a weight fusion and scheme evaluation module.

[0075] In the criteria construction and data acquisition module, the Analytic Hierarchy Process (AHP) is used to construct the AHP judgment matrix for subjective criteria. The construction of the entropy weight method is based on the entropy weight method to perform entropy weight data matrix on objective criteria. The framework is constructed where m represents the number of products and n represents the number of criteria. Elements of subjective criteria may include: streamer experience, product attributes, livestreaming scenario, brand awareness, etc.; elements of objective criteria may include: product sales volume, user interaction data, inventory turnover rate, etc.

[0076] In the subjective weight calculation module, the AHP judgment matrix is... Perform eigenvalue decomposition to obtain the largest eigenvalue. and corresponding feature vectors After normalization, the subjective weight vector is obtained and a consistency check is performed.

[0077] In the objective weight calculation module, the entropy weight method data matrix is... After normalization, the probability matrix is ​​obtained. And calculate the entropy value of each criterion. By mining the discriminative value of data, the objective weights of each criterion, i.e., entropy weights, can be obtained. .

[0078] In the weight fusion and scheme evaluation module, a linear weighted fusion of subjective and objective weights is adopted. The fusion coefficient is dynamically adjusted according to the live streaming business scenario, and finally the comprehensive score of each product is calculated. This enables the selection of the best products.

[0079] In this embodiment, the present application provides a dynamic weight adjustment interface: a reserved fusion coefficient configuration interface, which supports optimization through business rules or machine learning algorithms to adapt to the differences in recommendation strategies for different categories and user groups.

[0080] The specific implementation method has been described in detail above, and will not be repeated here.

[0081] To achieve the above objectives, the present invention also provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when running the program, implements the e-commerce live streaming intelligent product selection method as described in any of the foregoing embodiments. The processor and memory can be configured separately or integrated together, for example, integrated on a system-on-chip (SOC) of the terminal device.

[0082] To achieve the above objectives, the present invention also provides a computer-readable storage medium storing computer-executable instructions or computer programs, which, when processed and executed, implement the intelligent product selection method for e-commerce live streaming as described above.

[0083] The computer-readable storage medium is, for example, memory. Memory can be volatile or non-volatile, or it can include both volatile and non-volatile memory. Non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Volatile memory can be random access memory (RAM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (SLDRAM), and direct rambus RAM (DRRAM).

[0084] If the integrated units in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in the aforementioned computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause one or more computer devices (which may be personal computers, servers, or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0085] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. An intelligent product selection method for e-commerce live broadcast, characterized in that, The method comprises the following steps: Step S1: constructing a multi-dimensional product selection criterion system and collecting original data of the to-be-recommended goods under each criterion; wherein, the subjective criteria are subjected to AHP judgment matrix construction based on the analytic hierarchy process (AHP) , and the objective criteria are subjected to entropy weight data matrix construction based on the entropy weight method , wherein m is the number of goods, and n is the number of criteria; the elements of the subjective criteria at least include one of the following: anchor experience, product attribute, live scene, brand awareness; the elements of the objective criteria at least include one of the following: product sales, user interaction data, inventory turnover rate; Step S2: subjective weight calculation; wherein, the AHP judgment matrix Eigenvalue decomposition is performed to obtain the maximum eigenvalue And the corresponding eigenvector , After normalization, the subjective weight vector is obtained and consistency test is performed; Step S3: objective weight calculation; wherein the entropy weight method data matrix is normalized to obtain a probability matrix , and the entropy value of each criterion is calculated to mine the data differentiation value, and then the objective weight of each criterion, i.e. the entropy weight, is obtained ; Step S4: weight fusion and scheme evaluation; wherein linear weighted fusion of subjective and objective weights is adopted , and the fusion coefficient is dynamically adjusted according to the live broadcast service scene, and finally the comprehensive score of each commodity is calculated , and the optimal commodity is realized.

2. The intelligent product selection method for live e-commerce according to claim 1, wherein, In the step S1, the AHP judgment matrix is the proportion of the importance of the index i based on j, which is set according to the 1-9 scale method. 3.The intelligent product selection method for live broadcast of e-commerce according to claim 1, characterized in that, In the step S2, the subjective weight vector obtained after normalization is calculated based on the following formula: ； The consistency check comprises calculating a consistency indicator and a random consistency ratio, wherein RI is the average random consistency indicator. When the random consistency ratio CR < 0.1, it is determined that the judgment matrix reasonable.

4. The intelligent product selection method for live e-commerce according to claim 1, wherein, In said step S3, said probability matrix of ; Entropy values for each criterion are calculated by the formula: ； wherein is a very small positive number to avoid log divergence; The objective weight of each criterion, i.e. the entropy weight is calculated by the following formula: 。 5. The intelligent product selection method for live e-commerce according to claim 1, wherein, The linear weighted fusion of subjective and objective weights is adopted The calculation is performed by the following formula: ， Wherein, α and β are fusion coefficients, which are dynamically adjusted according to the live broadcast service scene or the commodity life cycle stage, α+β=1, α, β∈[0,1]; The comprehensive score of each commodity is calculated by the following formula: ， The commodity with the highest score is selected as the recommendation result.

6. The intelligent product selection method for live e-commerce according to claim 5, wherein, In the experience-dominated scene, the subjective weight α=0.6, and in the data-driven scene, β=0.

7.

7. The intelligent product selection method for live e-commerce according to claim 5, wherein, The life cycle stage includes introduction period, growth period, mature period and recession period, The corresponding fusion coefficient configuration rules are as follows: Introduction period: α=0.7, β=0.3, dominated by expert experience, dependent on brand and trend matching degree; Growth period: α=0.5, β=0.5, balance subjective judgment and objective data feedback; Mature period: α=0.3, β=0.7, dominated by historical objective data, focusing on sales and profit stability; Recession period: α=0.6, β=0.4, whether to clear or restart is decided by experts, and historical objective data is used as a reference.

8. An intelligent product selection system for live e-commerce, characterized in that, The e-commerce live broadcast intelligent product selection method according to any one of claims 1 to 7 comprises: The criterion construction and data collection module comprises an AHP judgment matrix of subjective criteria based on an analytic hierarchy process (AHP) , and an entropy weight data matrix of objective criteria based on an entropy weight method ; m is the number of commodities, and n is the number of criteria; the elements of the subjective criteria at least include one of anchor experience, commodity attribute, live broadcast scene, and brand cognition; and the elements of the objective criteria at least include one of commodity sales, user interaction data, and inventory turnover rate. A subjective weight calculation module; wherein the AHP judgment matrix is subjected to eigenvalue decomposition to obtain a maximum eigenvalue and a corresponding eigenvector , and a subjective weight vector is obtained after normalization and consistency check; An objective weight calculation module; wherein the entropy weight method data matrix is normalized to obtain a probability matrix , and the entropy value of each criterion is calculated to mine the data differentiation value, and then the objective weight of each criterion, i.e. the entropy weight, is obtained ; The weight fusion and scheme evaluation module; wherein, linearly weighted fusion subjective and objective weights And dynamically adjust the fusion coefficient according to the live broadcast service scene, finally calculate the comprehensive score of each commodity , Realize the optimization of goods.

9. A terminal device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor runs the program to realize the e-commerce live broadcast intelligent product selection method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions or computer programs, which are processed and executed by the processor to realize the e-commerce live broadcast intelligent product selection method according to any one of claims 1 to 7.

Images