Method and device for determining to push a product, storage medium and electronic device

Abstract

The application discloses a product pushing determination method and device, a storage medium and electronic equipment. The method comprises the following steps: obtaining interaction behavior information of a target object from an operation log of a target platform, wherein the interaction behavior information is used to represent a concerned product which has generated interaction behavior with the target object; selecting a product to be pushed from a plurality of preset products of the target platform according to the interaction behavior information, wherein the product to be pushed is selected from an unconcerned product which has not generated interaction behavior with the target object in the target platform; analyzing the product to be pushed by using a preset analysis model, predicting a first transaction probability of the product to be pushed in a preset time period; and determining that the product to be pushed with a first transaction probability higher than a preset transaction threshold as a target pushing product. The application solves the technical problem that product recommendation cannot be performed according to product transaction conditions of a seller.

Description

Technical Field

[0001] This invention relates to the field of data processing, and more specifically, to a method, apparatus, storage medium, and electronic device for determining push products. Background Technology

[0002] In recent years, with the rise of big data, research on consumer behavior analysis has flourished. Scholars from many fields, including databases and data mining, information systems and information management, image processing and computer vision, social network analysis, and e-commerce, have joined the ranks of consumer behavior researchers. At the same time, this research field has also attracted significant attention from enterprises operating in the digital economy, such as e-commerce and social networks. Consumer behavior analysis is considered one of the effective means for enterprises to understand their consumers and conduct marketing activities in the digital economy.

[0003] Existing product recommendation methods are based on user consumption behavior profiles, and cannot recommend products based on merchant behavior, which is not conducive to cultivating sellers' operational behavior.

[0004] There is currently no effective solution to the problem of not being able to recommend products based on the seller's product transaction history. Summary of the Invention

[0005] This invention provides a method, apparatus, storage medium, and electronic device for determining recommended products, in order to at least solve the technical problem of being unable to recommend products based on the seller's product transaction history.

[0006] According to one aspect of the present invention, a method for determining a product to be pushed is provided, characterized by comprising: obtaining interaction behavior information of a target object from the operation logs of a target platform, wherein the target platform is used to provide the target object with a plurality of preset products, and the interaction behavior information is used to represent the products that have interacted with the target object; selecting a product to be pushed from the plurality of preset products on the target platform according to the interaction behavior information, wherein the product to be pushed is selected from the unfollowed products on the target platform that have not interacted with the target object; analyzing the product to be pushed using a preset analysis model to predict the initial transaction probability of the product to be pushed completing its first transaction within a preset time period, wherein the preset analysis model is trained based on the initial transaction probability of the plurality of preset products on the target platform completing their first transaction within a historical time period, and the historical time period and the preset time period have the same period length; and determining the product to be pushed whose initial transaction probability is higher than a preset transaction threshold as the target product to be pushed.

[0007] Optionally, selecting a product to be pushed from a plurality of preset products on the target platform based on the interaction behavior information includes: analyzing the interaction behavior information to determine the target object's interaction behavior with the product it is interested in; evaluating the product it is interested in based on the interaction behavior to determine the evaluation value of the product it is interested in, wherein the evaluation value is used to represent the degree of interest of the target object in the product it is interested in; and predicting a product that the target object is interested in from the un-followed products based on the evaluation value of the product it is interested in.

[0008] Optionally, predicting products that the target object is interested in from the unfollowed products based on the evaluation value of the followed products includes: identifying followed products with evaluation values ​​higher than a preset evaluation threshold as the target object's favorite products; using the favorite products as training data to train a first interest model for the target object through machine learning, wherein the first interest model is used to represent the target product characteristics of the favorite products; and using the first interest model to perform feature analysis on the unfollowed products to determine unfollowed products that match the target product characteristics as the products to be pushed.

[0009] Optionally, predicting products that the target object is interested in from the unfollowed products based on the evaluation values ​​of the followed products includes: using the followed products and their evaluation values ​​as training data to train a second interest model for the target object through machine learning, wherein the second interest model is used to evaluate the evaluation values ​​of the preset products; using the second interest model to perform feature analysis on the unfollowed products to determine their evaluation values; and identifying unfollowed products with evaluation values ​​higher than a preset evaluation threshold as the products to be pushed.

[0010] Optionally, before analyzing the products to be pushed using a preset analysis model to determine the initial transaction probability of the products to be pushed, the method further includes: obtaining product transaction data from the operation logs of the target platform, wherein the product transaction data is used to represent the transaction results of a preset number of preset products provided by the target platform within a historical time period, the transaction results including: the preset product completing its first transaction within the historical time period, and the preset product not completing its first transaction within the historical time period; analyzing the product transaction data using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset products, wherein the preset initial transaction prediction model is used to determine the initial transaction probability of the preset products completing their first transaction within the historical time period and the non-transaction probability of not completing their first transaction within the historical time period based on the transaction results of a preset number of preset products within the historical time period; using the initial transaction probability of the preset products as positive samples and the non-transaction probability of the preset products as negative samples, training the preset analysis model through machine learning, wherein the preset analysis model is constructed based on a Bayesian algorithm.

[0011] Optionally, the product transaction data further includes: transaction factors affecting the transaction results. After obtaining product transaction data from the operation logs of the target platform, the method further includes: identifying the preset product that completed its first transaction within the historical time period as a sample product; assigning a preset evaluation coefficient based on the transaction factors of the sample product; and determining the initial transaction probability of the sample product based on the transaction factors and the preset evaluation coefficient.

[0012] Optionally, after determining that the product to be pushed with an initial transaction probability higher than a preset transaction threshold is a target push product, the method further includes: adding the target push product to a target push list; and when the number of the target push products in the target push list reaches a preset quantity threshold, pushing the target push products in the target push list to the target object.

[0013] According to another aspect of the present invention, a device for determining a product to be pushed is also provided, comprising: an acquisition module, configured to acquire interaction behavior information of a target object from the operation logs of a target platform, wherein the target platform is configured to provide the target object with a plurality of preset products, and the interaction behavior information is used to represent products that have interacted with the target object; a selection module, configured to select a product to be pushed from the plurality of preset products on the target platform according to the interaction behavior information, wherein the product to be pushed is selected from the unfollowed products on the target platform that have not interacted with the target object; a prediction module, configured to analyze the product to be pushed using a preset analysis model to predict the probability of the product to be pushed completing its first transaction within a preset time period, wherein the preset analysis model is trained based on the probability of the preset products on the target platform completing their first transaction within a historical time period, and the historical time period and the preset time period have the same period length; and a determination module, configured to determine the product to be pushed whose probability of the first transaction is higher than a preset transaction threshold as a target product to be pushed.

[0014] According to another aspect of the present invention, a non-volatile storage medium is also provided, wherein a program is stored in the non-volatile storage medium, wherein the program controls the device where the non-volatile storage medium is located to execute the above-described method for determining the push product when it is executed.

[0015] According to another aspect of the present invention, an electronic device is also provided, including: a memory and a processor, the processor being configured to run a program stored in the memory, wherein the program, when running, executes the above-described method for determining the push product.

[0016] In this embodiment of the invention, interaction behavior information of the target object is obtained from the operation logs of the target platform. The target platform provides multiple preset products to the target object, and the interaction behavior information represents the products the target object has interacted with. Based on the interaction behavior information, products to be pushed are selected from the multiple preset products on the target platform. These products are selected from the un-followed products on the target platform that have not interacted with the target object. A preset analysis model is used to analyze the products to be pushed, predicting the probability of the products to be pushed completing their first transaction within a preset time period. This preset analysis model is trained based on the probability of the first transaction of multiple preset products on the target platform within a historical time period, where the historical time period and the preset time period have the same length. Products with a probability of first transaction higher than a preset transaction threshold are identified as target push products. Therefore, during the product push process to the target object, un-followed products can be pushed to the target object based on the product's transaction status, achieving the technical effect of product recommendation based on the seller's product transaction status, thereby solving the technical problem of not being able to recommend products based on the seller's product transaction status. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0018] Figure 1 This is a flowchart of a method for determining a product to be pushed according to an embodiment of the present invention;

[0019] Figure 2 This is a schematic diagram of a device for determining the delivery of a product according to an embodiment of the present invention;

[0020] Figure 3 This is a structural block diagram of a computer terminal according to an embodiment of the present invention. Detailed Implementation

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0023] According to an embodiment of the present invention, a method for determining a product to be pushed is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0024] Figure 1 This is a flowchart of a method for determining a product to be pushed according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0025] Step S102: Obtain the interaction behavior information of the target object from the operation log of the target platform. The target platform is used to provide multiple preset products to the target object, and the interaction behavior information is used to indicate the products that have interacted with the target object.

[0026] Step S104: Select the product to be pushed from multiple preset products on the target platform based on the interaction behavior information. The product to be pushed is selected from the unfollowed products on the target platform that have not interacted with the target object.

[0027] Step S106: Use a preset analysis model to analyze the product to be pushed and predict the probability of the product to be pushed completing its first transaction within a preset time period. The preset analysis model is trained based on the probability of the first transaction of multiple preset products in the target platform completing their first transaction within a historical time period. The historical time period and the preset time period have the same period length.

[0028] Step S108: Identify products to be pushed that have an initial transaction probability higher than a preset transaction threshold as target push products.

[0029] Through the above steps, interaction behavior information of the target object is obtained from the operation logs of the target platform. The target platform provides multiple preset products to the target object, and the interaction behavior information represents the products the target object has interacted with and followed. Based on the interaction behavior information, products to be pushed are selected from the multiple preset products on the target platform. These products are selected from the unfollowed products on the target platform that have not interacted with the target object. A preset analysis model is used to analyze the products to be pushed, predicting the probability of initial transaction completion within a preset time period. This preset analysis model is trained based on the initial transaction probability of multiple preset products on the target platform within historical time periods, where the historical and preset time periods have the same length. Products with an initial transaction probability higher than a preset transaction threshold are identified as target push products. Therefore, during product push to the target object, unfollowed products can be pushed based on the product's transaction status, achieving the technical effect of product recommendation based on the seller's product transaction status, thus solving the technical problem of not being able to recommend products based on the seller's product transaction status.

[0030] In step S102 above, the target platform can be an online shopping platform, and the target object can be a consumer who accesses the online shopping platform, i.e., a consumer user. The target object can browse and trade preset products through the target platform.

[0031] In step S102 above, the exchange behavior includes: the target object clicking on the product link, browsing the product content, liking the product content, collecting the product content, downloading the product content, sharing the product content, and commenting on the product content, etc.

[0032] In step S102 above, the interaction behavior of the target object with the preset products of the target platform can generate products of interest.

[0033] For example, products followed can be products that the target audience clicks on, browses, downloads (such as products whose images have been downloaded), likes, favorites, shares, and comments on.

[0034] In step S104 above, the preset products that have not interacted with the target object on the target platform are not followed. The products to be pushed can be selected from the unfollowed products.

[0035] For example, products that the target audience is interested in can be selected from their unfollowed products as products to be promoted, or products with good historical transaction records can be selected from their unfollowed products as products to be promoted.

[0036] In step S106 above, the initial transaction probability is used to represent the probability that a certain preset product can complete at least one transaction within a preset time period of a predetermined length. The initial transaction probability is determined based on the initial transaction probability of each preset product in the historical time period.

[0037] Optionally, when it is necessary to determine the initial transaction probability of product A, it can be determined based on the transaction situation of product A in the historical time period, or it can be determined based on the transaction situation of product B, which is similar to product A, in the historical time period.

[0038] Optionally, by training the model based on the transaction data of multiple preset products within a historical time period, a preset analysis model can be obtained that can evaluate the initial transaction probability of each preset product. The preset analysis model represents the correspondence between the preset products and the initial transaction probability.

[0039] In step S108 above, if the probability of an initial transaction is higher than the preset transaction threshold, it means that the product to be pushed has a high probability of being sold within a preset time period in the future. Therefore, the product to be pushed can be pushed to the target to increase the transaction probability of the product to be pushed, thus achieving marketing assistance to the merchant.

[0040] As an optional embodiment, selecting products to be pushed from multiple preset products on the target platform based on interaction behavior information includes: analyzing interaction behavior information to determine the target object's interaction behavior with the products it is interested in; evaluating the products it is interested in based on the interaction behavior to determine the evaluation value of the products it is interested in, wherein the evaluation value is used to represent the target object's degree of interest in the products it is interested in; and predicting products that the target object is interested in from the products it is not interested in based on the evaluation value of the products it is interested in.

[0041] The above embodiments of the present invention can evaluate the products that the target object is interested in based on the target object's historical interaction behavior on the target platform. The evaluation value represents the target object's degree of interest in the products they are interested in. In this way, when pushing products to the target object, the target object can be pushed products that they are interested in, thereby increasing the probability of successful transactions for the products to be pushed.

[0042] As an optional embodiment, predicting products that a target object is interested in from unfollowed products based on the rating values ​​of followed products includes: identifying followed products with rating values ​​higher than a preset rating threshold as the target object's favorite products; using the favorite products as training data to train a first interest model for the target object through machine learning, wherein the first interest model is used to represent the target product characteristics of the favorite products; and using the first interest model to perform feature analysis on unfollowed products to determine unfollowed products that match the target product characteristics as products to be pushed.

[0043] In the above embodiments of the present invention, when a product to be pushed to a target object is determined, the target object's favorite products can be selected from the target object's followed products based on the evaluation value. Then, a first interest model is trained based on the favorite products to summarize the target product characteristics of the favorite products. Furthermore, when screening products to be pushed from unfollowed products, the unfollowed products can be input into the first interest model to select unfollowed products that meet the target product characteristics as products to be pushed. Thus, it can be determined that the products to be pushed output by the first interest model are products that the target object is interested in, thereby increasing the transaction success probability of the products to be pushed.

[0044] As an optional embodiment, predicting products that a target object is interested in from unfollowed products based on the evaluation values ​​of followed products includes: using followed products and their evaluation values ​​as training data, training a second interest model for the target object through machine learning, wherein the second interest model is used to evaluate the evaluation values ​​of preset products; using the second interest model to perform feature analysis on unfollowed products to determine their evaluation values; and identifying unfollowed products with evaluation values ​​higher than a preset evaluation threshold as products to be pushed.

[0045] In the above embodiments of the present invention, when a product to be pushed to a target object is determined, a second interest model can be trained first based on the evaluation value of the products being followed, and the correspondence between preset products and evaluation values ​​can be summarized. Then, the unfollowed products are input into the second interest model, and the evaluation value of each unfollowed product can be predicted. Based on the predicted evaluation value, products that the target object is interested in can be selected from the unfollowed products, thereby increasing the probability of successful transactions of the products to be pushed.

[0046] Optionally, the rating value Rate i =w1*C click +w2*C view +w3*C like +w4*C collect +w5*C download +w6*Cshare+w7*C comment , where C click The number of times a product link is opened for the target audience, C view C represents the number of times the target audience views the product. like The number of times a product is liked by the target audience, C collect C represents the number of times the target audience clicks to add a product to their favorites. download C represents the number of times the product is downloaded by the target audience. share C is the number of times the product is shared with the target audience. comment The number of times the target audience evaluates the product. w1, w2, w3, w4, w5, w6, and w7 are the corresponding weights, and w1 <= w2 <= w3 <= w4 <= w5 <= w6 <= w7.

[0047] Optionally, select M user samples (i.e., M target objects) and N product samples (i.e. N products of interest) to form a matrix R, where the original value r in the u-th row and i-th column of R is... ui , representing the rating value (Rate) of target object u for product i. i .

[0048] Optionally, the evaluation value Ratei is normalized to limit its range to between 0 and 1. A preset evaluation threshold t is set; if r ui If the value is greater than or equal to t, it will be set to 1, indicating that the target object u likes product i and product i can be recommended to the target object u; otherwise, it will be set to 0, indicating that the target object u has a moderate interest in product i and it is not recommended to recommend it.

[0049] As an optional embodiment, before analyzing the products to be pushed using a preset analysis model to determine the initial transaction probability of the products to be pushed, the method further includes: obtaining product transaction data from the operation logs of the target platform, wherein the product transaction data is used to represent the transaction results of a preset number of preset products provided by the target platform within a historical time period, and the transaction results include: the preset products completing their first transaction within the historical time period, and the preset products not completing their first transaction within the historical time period; analyzing the product transaction data using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset products, wherein the preset initial transaction prediction model is used to determine the initial transaction probability of the preset products completing their first transaction within the historical time period and the non-transaction probability of not completing their first transaction within the historical time period based on the transaction results of a preset number of preset products within the historical time period; using the initial transaction probability of the preset products as positive samples and the non-transaction probability of the preset products as negative samples, training the preset analysis model through machine learning, wherein the preset analysis model is constructed based on a Bayesian algorithm.

[0050] In the above embodiments of the present invention, the transaction results include: a preset product completing its first transaction within a historical time period, and a preset product not completing its first transaction within a historical time period. By analyzing the product transaction data using a preset initial transaction prediction model, the initial transaction probability and non-transaction probability of each preset product within the historical time period can be determined. Then, the initial transaction probability of the preset product is used as a positive sample, and the non-transaction probability of the preset product is used as a negative sample. A preset analysis model based on a Bayesian algorithm is trained through machine learning, so that the trained preset analysis model can predict the initial transaction probability of products that are not being followed.

[0051] As an optional embodiment, the product transaction data also includes: transaction factors that affect the transaction results. After obtaining the product transaction data from the operation logs of the target platform, the method further includes: obtaining a preset product that completed its first transaction within a historical time period as a sample product; assigning a preset evaluation coefficient based on the transaction factors of the sample product; and determining the initial transaction probability of the sample product based on the transaction factors and the preset evaluation coefficient.

[0052] In the above embodiments of the present invention, each sample product of a preset product that completes its first transaction within a historical time period is assigned a preset evaluation coefficient according to the transaction factors that affect the transaction result. This can determine the linear impact of the transaction factors on the transaction result, and thus determine the initial transaction probability of each sample product. This realizes the transformation of 0-1 type variables based on transaction success and transaction failure into probabilistic variables.

[0053] As an optional implementation, the preset initial transaction prediction model is as follows: Where P(y=1) represents the probability of the first transaction, x i β represents the transaction factors of the sample products. i This represents the preset evaluation coefficient, where i is based on the preset quantity of sample products.

[0054] Alternatively, combining the two formulas above yields the widely used logistic regression algorithm:

[0055] Alternatively, the formula can be expressed as:

[0056]

[0057] Optionally, the parameter estimation method used in logistic regression is typically the maximum likelihood method. The parameter estimation using the maximum likelihood method usually involves the following steps:

[0058] Let y be a variable of type 0-1, and K = (x1, x2, ..., x...). p K = (x1, x2, ..., xp) is a variable related to y (i.e., trading factors), and n sets of observation data are (x i1 ,x i2 ,…,x ip ), yi and (x i1 ,x i2 ,…,x ip The relationship between ) is as follows:

[0059]

[0060] Here, the function f(x) is a monotonically increasing function with a range of [0,1]. For logistic regression, we have...

[0061] Optionally, Yi is a π-mean π i =f(β0+β1x) i1 +β2x i2 +…+β p x ip The probability function of y is: P(y) = 0-1 distribution. i =1)=π i , P(y i =0)=1-π i .

[0062] Alternatively, the likelihood function described above can be:

[0063]

[0064] Optionally, for logistic regression, Substituting into the above equation, we get:

[0065]

[0066] Optionally, a significance check can be performed: in, The estimated value of the regression coefficient β1 (i.e., the pre-set evaluation coefficient). The standard deviation of β1. If the estimated value of β1... The Wald test is significant. Generally speaking, if the p-value of the variable is less than 0.05, the independent variable (i.e., the transaction factor) can be considered to have a significant effect on the dependent variable (i.e., the transaction outcome); otherwise, the effect is not significant.

[0067] It should be noted that Bayes' theorem is the foundation of the Naive Bayesian Classification method. If a given dataset has M categories, Naive Bayes can predict whether a given observation belongs to the specific category with the highest posterior probability. In other words, Naive Bayes predicts that X belongs to category C if and only if: P(C... i |X)>P(C j |X)1≤j≤m,j≠i, in this case, if we maximize P(C j |X), whose P(C) j |X)(C j The class with the largest |X) is called the maximum a posteriori hypothesis, according to Bayes' theorem: It can be seen that since P(X) is equal for all categories, it is only necessary to have P(X|C) i )P(C i Take the maximum value.

[0068] Optionally, in order to predict the class of an unknown sample X, for each class C... i Estimate the corresponding P(X|C) i )P(C i ).

[0069] As an optional embodiment, after determining that the product to be pushed has an initial transaction probability higher than a preset transaction threshold as the target push product, the method further includes: adding the target push product to the target push list; and pushing the target push product in the target push list to the target object when the number of target push products in the target push list reaches a preset quantity threshold.

[0070] In the above embodiments of the present invention, the target push products are first added to the target push list, and then the target push products in the target push list are pushed to the target object when the number of target push products in the target push list reaches a preset quantity threshold, thereby realizing the batch push of target push products to the target object.

[0071] The present invention also provides a preferred embodiment, which provides a method for stock management analysis based on convolutional neural network training.

[0072] Regression analysis, primarily referring to techniques such as logistic regression and multiple linear regression, is one of the most widely used analytical tools in quantitative statistics and a widely applied analytical method (technique) in data analysis and mining practice. Although, narrowly defined, regression analysis falls under the category of statistical analysis, drawing a strict line between statistical analysis and data mining is meaningless for data analysis and mining practice. As long as it solves practical business problems and improves operational efficiency, it is a good technique, especially since regression analysis is currently widely used in data mining practice.

[0073] This invention addresses digital analysis in existing business operations by combining AI with regression analysis. First, it uses the probability-discretionary ratio (ODDS) (e.g., a pre-defined initial transaction prediction model) to analyze the probability of an event occurring divided by the probability of it not occurring, thus predicting the probability of business data. Second, it uses logistic regression algorithms (e.g., a pre-defined analysis model) to determine if they can effectively and scalably facilitate online transactions between buyers and sellers and ensure continued transactions. This data analysis identifies the seller group most likely to achieve initial transactions in the short term, analyzes their typical characteristics, and allows operators to conduct tiered and refined operations based on this. The ultimate goal is to effectively increase the number of sellers achieving initial transactions within a given time period through data-driven operations, and to identify key operational strategies for future seller development to help them grow effectively.

[0074] Step 1: Obtain data (registration, product information posting, online orders, and online transactions) from seller logs and compile this user behavior data into a dataset. Then, use the ODDS algorithm to construct a pre-defined initial transaction prediction model and analyze the dataset to obtain the initial transaction to non-transaction ratio.

[0075] As an optional implementation, for sellers, from the initial registration and posting of product information to the subsequent continuous online order acquisition and online transactions, there is one crucial and breakthrough point: the first online transaction, also known as the initial transaction conversion. The value of this first transaction to the seller's success experience and incentive is self-evident.

[0076] In addition, from the perspective of the website platform operator, the seller's first transaction is also an important evaluation link and indicator of the website operation. Only when the number of sellers with the first transaction is larger and the cycle is shorter can the possibility of continuous and large-scale online transactions in the later stage be effectively guaranteed.

[0077] As an optional implementation, the preset initial transaction prediction model is as follows: Where P(y=1) represents the probability of the first transaction, x i β represents the transaction factors of the sample products. i This represents the preset evaluation coefficient, where i is based on the preset quantity of sample products.

[0078] Alternatively, combining the two formulas above yields the widely used logistic regression algorithm:

[0079] Alternatively, the formula can be expressed as:

[0080]

[0081] Optionally, the parameter estimation method used in logistic regression is typically the maximum likelihood method. The parameter estimation using the maximum likelihood method usually involves the following steps:

[0082] Let y be a variable of type 0-1, and K = (x1, x2, ..., x...). p K = (x1, x2, ..., xp) is a variable related to y (i.e., trading factors), and n sets of observation data are (x i1 ,x i2 ,…,x ip ), yi and (x i1 ,x i2 ,…,x ip The relationship between ) is as follows:

[0083]

[0084] Here, the function f(x) is a monotonically increasing function with a range of [0,1]. For logistic regression, we have...

[0085] Optionally, Yi is a π-mean π i =f(β0+β1x) i1 +β2x i2 +…+β p x ip The probability function of y is: P(y) = 0-1 distribution. i =1)=π i , P(y i =0)=1-π i .

[0086] Alternatively, the likelihood function described above can be:

[0087]

[0088] Optionally, for logistic regression, Substituting into the above equation, we get:

[0089]

[0090] Optionally, a significance check can be performed: in, The estimated value of the regression coefficient β1 (i.e., the pre-set evaluation coefficient). The standard deviation of β1. If the estimated value of β1... The Wald test is significant. Generally speaking, if the p-value of the variable is less than 0.05, the independent variable (i.e., the transaction factor) can be considered to have a significant effect on the dependent variable (i.e., the transaction outcome); otherwise, the effect is not significant.

[0091] Step two involves retrieving data from the seller logs related to non-initial transactions (registrations, product listings, online orders, and online sales) – specifically, data on products not yet followed – and compiling this user behavior data into a dataset. A Bayesian algorithm is then used to construct a pre-defined analytical model to determine the probability of the product achieving its first transaction in the future.

[0092] As an optional implementation, Bayes' theorem is the foundation of the Naive Bayesian Classifier. If a given dataset contains M categories, the Naive Bayesian Classifier can predict whether a given observation belongs to the specific category with the highest posterior probability. In other words, the Naive Bayesian Classifier predicts that X belongs to category C if and only if: P(C... i |X)>P(C j|X)1≤j≤m,j≠i, in this case, if we maximize P(C j |X), whose P(C) j |X)(C j The class with the largest |X) is called the maximum a posteriori hypothesis, according to Bayes' theorem: It can be seen that since P(X) is equal for all categories, it is only necessary to have P(X|C) i )P(C i Take the maximum value.

[0093] Optionally, in order to predict the class of an unknown sample X, for each class C... i Estimate the corresponding P(X|C) i )P(C i ).

[0094] Step 3: Substitute the optimized dataset predicted by the Bayesian classification (i.e., the pre-defined analysis model) into a one-dimensional convolutional neural network for training. This allows for data analysis to identify the seller group most likely to achieve initial sales in the short term, analyze their typical characteristics, and enable the operator to conduct tiered and refined operations based on this information.

[0095] As an optional implementation, the generation of the training dataset includes: first, calculating the user's evaluation of the content (i.e., evaluation value) based on the user's interaction with content (such as a product) (i.e., interaction behavior information). User interactions with content (i.e., interaction behaviors) include clicking content links, browsing content, liking content, saving content, downloading content, sharing content, and commenting on content. All these interactions are ranked in ascending order of their contribution to the content evaluation. The platform needs to statistically analyze the number of all user interactions with the content.

[0096] Optionally, Ratei = w1*Cclick + w2*Cview + w3*Clike + w4*Ccollect + w5*Cdownload + w6*Cshare + w7*Ccomment, where Cclick is the number of times a user opens a content link, Cview is the number of times a user views the content, Clike is the number of times a user likes the content, Ccollect is the number of times a user saves the content, Cdownload is the number of times a user downloads the content, Cshare is the number of times a user shares the content, and Ccomment is the number of times a user rates the content. w1, w2, w3, w4, w5, w6, and w7 are the corresponding weights, and w1 <= w2 <= w3 <= w4 <= w5 <= w6 <= w7.

[0097] Optionally, M user samples and N content samples are selected from the platform users and content library to form a matrix R. The original value rui in the u-th row and i-th column of R represents the rating value Ratei of user u for content i.

[0098] Optionally, Ratei is normalized to limit the value to between 0 and 1. A threshold (i.e., a preset evaluation threshold) t is set; if r ui If the value is greater than or equal to t, it will be set to 1, indicating that user u likes content i and content i can be recommended to user u; otherwise, it will be set to 0, indicating that user u has little interest in content i and it is not recommended to recommend it.

[0099] Optionally, matrix R constitutes the training dataset for the recommendation algorithm.

[0100] As an optional embodiment, the overall recommendation algorithm's network architecture includes: First, generating user embeddings and content embeddings of depth 1 from the user ID and content ID, respectively. Then, concatenating the user embeddings and content embeddings. Next, inputting them into a one-dimensional convolutional neural network. Multiple convolutional kernels can be used. One ReLU layer and one pooling layer are added. Finally, the results are passed through a three-layer fully connected network and a Softmax activation function to obtain the classification result.

[0101] As an optional implementation, model training includes: dividing the training dataset into training data, validation data, and test data according to a certain ratio; and using the training data to train the one-dimensional convolutional neural network to obtain a classification model.

[0102] As an optional implementation, content recommendation for users includes: deploying a model on a platform to provide AI inference services. When a user logs into the platform, content is selected sequentially from those the user has not clicked, processed by the AI ​​inference service, and a classification result is obtained. A threshold (such as a preset quantity threshold) is set for the model's output; if the inference result exceeds the threshold, the content is considered suitable for recommendation to the user and added to the recommendation list. When the length of the recommendation list reaches its maximum value (i.e., when the number of target push products in the target push list reaches the preset quantity threshold), this batch of content is sent to the user as the recommendation result.

[0103] According to an embodiment of the present invention, an embodiment of a device for determining a push product is also provided. It should be noted that the device for determining a push product can be used to execute the method for determining a push product in the embodiment of the present invention, and the method for determining a push product in the embodiment of the present invention can be executed in the device for determining a push product.

[0104] Figure 2This is a schematic diagram of a product-determining device according to an embodiment of the present invention, such as... Figure 2 As shown, the device may include: an acquisition module 22, used to acquire interaction behavior information of a target object from the operation logs of the target platform, wherein the target platform provides multiple preset products to the target object, and the interaction behavior information is used to represent the products that have interacted with the target object; a selection module 24, used to select a product to be pushed from the multiple preset products of the target platform based on the interaction behavior information, wherein the product to be pushed is selected from the unfollowed products of the target platform that have not interacted with the target object; a prediction module 26, used to analyze the product to be pushed using a preset analysis model to predict the probability of the product to be pushed completing its first transaction within a preset time period, wherein the preset analysis model is trained based on the probability of the first transaction of multiple preset products in the target platform completing their first transaction within a historical time period, and the historical time period and the preset time period have the same period length; and a determination module 28, used to determine the product to be pushed with an initial transaction probability higher than a preset transaction threshold as the target push product.

[0105] It should be noted that the acquisition module 22 in this embodiment can be used to execute step S102 in this application embodiment, the selection module 24 in this embodiment can be used to execute step S104 in this application embodiment, the prediction module 26 in this embodiment can be used to execute step S106 in this application embodiment, and the determination module 28 in this embodiment can be used to execute step S108 in this application embodiment. The examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments.

[0106] In the above embodiments of the present invention, interaction behavior information of a target object is obtained from the operation logs of a target platform. The target platform provides multiple preset products to the target object, and the interaction behavior information represents the products the target object has interacted with. Based on the interaction behavior information, products to be pushed are selected from the multiple preset products on the target platform. These products are selected from the un-followed products on the target platform that have not interacted with the target object. A preset analysis model is used to analyze the products to be pushed, predicting the probability of the products completing their first transaction within a preset time period. This preset analysis model is trained based on the probability of the first transaction of multiple preset products on the target platform within a historical time period, where the historical time period and the preset time period have the same length. Products with a probability of first transaction higher than a preset transaction threshold are identified as target push products. Therefore, during the product push process to the target object, un-followed products can be pushed to the target object based on the product's transaction status, achieving the technical effect of product recommendation based on the seller's product transaction status, thereby solving the technical problem of not being able to recommend products based on the seller's product transaction status.

[0107] As an optional embodiment, the selection module includes: an analysis unit, used to analyze interaction behavior information and determine the interaction behavior of the target object towards the product of interest; a first determination unit, used to evaluate the product of interest based on the interaction behavior and determine the evaluation value of the product of interest, wherein the evaluation value is used to represent the degree of interest of the target object towards the product of interest; and a prediction unit, used to predict the product to be pushed that the target object is interested in from the unfollowed products based on the evaluation value of the product of interest.

[0108] As an optional embodiment, the prediction unit includes: a second determining unit, used to determine the products with evaluation values ​​higher than a preset evaluation threshold as the target object's favorite products; a first training unit, used to use the favorite products as training data to train a first interest model of the target object through machine learning, wherein the first interest model is used to represent the target product characteristics of the favorite products; and a third determining unit, used to use the first interest model to perform feature analysis on unfollowed products and determine the unfollowed products that meet the target product characteristics as products to be pushed.

[0109] As an optional embodiment, the prediction unit includes: a second training unit, used to train a second interest model of the target object by using the products being followed and their evaluation values ​​as training data, wherein the second interest model is used to evaluate the evaluation values ​​of preset products; a fourth determination unit, used to perform feature analysis on unfollowed products using the second interest model to determine the evaluation values ​​of the unfollowed products; and a fifth determination unit, used to determine unfollowed products with evaluation values ​​higher than a preset evaluation threshold as products to be pushed.

[0110] As an optional embodiment, the apparatus further includes: a first acquisition submodule, configured to acquire product transaction data from the operation logs of the target platform before analyzing the product to be pushed using a preset analysis model to determine the initial transaction probability of the product to be pushed, wherein the product transaction data represents the transaction results of a preset number of preset products provided by the target platform within a historical time period, and the transaction results include: the preset product completing its first transaction within the historical time period, and the preset product not completing its first transaction within the historical time period; a first determination submodule, configured to analyze the product transaction data using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset product, wherein the preset initial transaction prediction model is used to determine the initial transaction probability of the preset product completing its first transaction within the historical time period and the non-transaction probability of not completing its first transaction within the historical time period based on the transaction results of a preset number of preset products within the historical time period; and a training submodule, configured to train the preset analysis model using the initial transaction probability of the preset product as a positive sample and the non-transaction probability of the preset product as a negative sample through machine learning, wherein the preset analysis model is constructed based on a Bayesian algorithm.

[0111] As an optional embodiment, the product transaction data further includes: transaction factors that affect the transaction results. The device further includes: a second acquisition submodule, used to acquire a preset product that has completed its first transaction within a historical time period as a sample product after acquiring product transaction data from the operation log of the target platform; an allocation submodule, used to allocate a preset evaluation coefficient according to the transaction factors of the sample product; and a second determination submodule, used to determine the initial transaction probability of the sample product according to the transaction factors and the preset evaluation coefficient.

[0112] As an optional embodiment, the device further includes: a storage module, used to add the target push product to the target push list after determining that the product to be pushed has an initial transaction probability higher than a preset transaction threshold as the target push product; and a push module, used to push the target push product in the target push list to the target object when the number of target push products in the target push list reaches a preset quantity threshold.

[0113] Embodiments of the present invention can provide a computer terminal, which can be any computer terminal device in a group of computer terminals. Optionally, in this embodiment, the computer terminal can also be replaced by a mobile terminal or other terminal device.

[0114] Optionally, in this embodiment, the computer terminal may be located in at least one of a plurality of network devices in a computer network.

[0115] In this embodiment, the computer terminal described above can execute the program code for the following steps in the method for determining the push product: obtaining the interaction behavior information of the target object from the operation log of the target platform, wherein the target platform is used to provide multiple preset products to the target object, and the interaction behavior information is used to represent the products that have interacted with the target object; selecting the product to be pushed from the multiple preset products of the target platform according to the interaction behavior information, wherein the product to be pushed is selected from the unfollowed products of the target platform that have not interacted with the target object; analyzing the product to be pushed using a preset analysis model to predict the initial transaction probability of the product to be pushed completing its first transaction within a preset time period, wherein the preset analysis model is trained based on the initial transaction probability of multiple preset products in the target platform completing their first transaction within a historical time period, and the historical time period and the preset time period have the same period length; determining the product to be pushed with an initial transaction probability higher than a preset transaction threshold as the target push product.

[0116] Optionally, Figure 3 This is a structural block diagram of a computer terminal according to an embodiment of the present invention. Figure 3 As shown, the computer terminal 30 may include one or more (only one is shown in the figure) processors 32 and memory 34.

[0117] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the product push determination method and apparatus in this embodiment of the invention. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned product push determination method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal 30 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0118] The processor can access information and applications stored in memory via a transmission device to perform the following steps: Obtaining interaction behavior information of the target object from the target platform's operational logs, wherein the target platform provides multiple preset products to the target object, and the interaction behavior information represents the products the target object has interacted with; selecting products to be pushed from the multiple preset products on the target platform based on the interaction behavior information, wherein the products to be pushed are selected from the un-followed products on the target platform that have not interacted with the target object; analyzing the products to be pushed using a preset analysis model to predict the initial transaction probability of the products to be pushed within a preset time period, wherein the preset analysis model is trained based on the initial transaction probability of multiple preset products on the target platform within a historical time period, and the historical time period and the preset time period have the same period length; determining the products to be pushed with an initial transaction probability higher than a preset transaction threshold as target push products.

[0119] Optionally, the processor may also execute program code that performs the following steps: analyzes interaction behavior information to determine the target object's interaction behavior with the product of interest; evaluates the product of interest based on the interaction behavior to determine the evaluation value of the product of interest, wherein the evaluation value is used to represent the target object's degree of interest in the product of interest; and predicts the product to be pushed to the target object from among the products of interest that the target object is interested in based on the evaluation value of the product of interest.

[0120] Optionally, the processor may also execute program code for the following steps: identifying products with evaluation values ​​higher than a preset evaluation threshold as favorite products of the target object; using favorite products as training data to train a first interest model of the target object through machine learning, wherein the first interest model is used to represent the target product characteristics of favorite products; and using the first interest model to perform feature analysis on unfollowed products to determine unfollowed products that meet the target product characteristics as products to be pushed.

[0121] Optionally, the processor may also execute program code for the following steps: using the products being followed and their ratings as training data, training a second interest model for the target object through machine learning, wherein the second interest model is used to evaluate the ratings of preset products; using the second interest model to perform feature analysis on unfollowed products to determine their ratings; and identifying unfollowed products with ratings higher than a preset rating threshold as products to be pushed to the system.

[0122] Optionally, the processor may also execute program code for the following steps: Before analyzing the product to be pushed using a preset analysis model to determine the initial transaction probability of the product to be pushed, obtain product transaction data from the target platform's operation logs. The product transaction data represents the transaction results of a preset number of preset products provided by the target platform within a historical time period. The transaction results include: the preset product completing its first transaction within the historical time period, and the preset product not completing its first transaction within the historical time period. Analyze the product transaction data using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset product. The preset initial transaction prediction model is used to determine the initial transaction probability of the preset product completing its first transaction within the historical time period and the non-transaction probability of not completing its first transaction within the historical time period, based on the transaction results of a preset number of preset products within the historical time period. Train the preset analysis model using machine learning, with the initial transaction probability of the preset product as a positive sample and the non-transaction probability of the preset product as a negative sample. The preset analysis model is constructed based on a Bayesian algorithm.

[0123] Optionally, the product transaction data also includes: transaction factors that affect the transaction results. The processor can also execute program code that performs the following steps: after obtaining product transaction data from the target platform's operation logs, obtain a preset product that completed its first transaction within a historical time period as a sample product; assign a preset evaluation coefficient based on the transaction factors of the sample product; and determine the initial transaction probability of the sample product based on the transaction factors and the preset evaluation coefficient.

[0124] Optionally, the processor may also execute program code that performs the following steps: after determining that the product to be pushed has an initial transaction probability higher than a preset transaction threshold as the target push product, the target push product is added to the target push list; when the number of target push products in the target push list reaches a preset quantity threshold, the target push product in the target push list is pushed to the target object.

[0125] This invention provides a scheme for determining products to be pushed. Interaction behavior information of a target object is obtained from the operational logs of a target platform. The target platform provides multiple preset products to the target object, and the interaction behavior information represents products that have interacted with the target object and are considered "followed" products. Based on the interaction behavior information, products to be pushed are selected from the multiple preset products on the target platform. These products are selected from "unfollowed" products on the target platform that have not interacted with the target object. A preset analysis model is used to analyze the products to be pushed, predicting the probability of the products completing their first transaction within a preset time period. This preset analysis model is trained based on the probability of the first transaction of multiple preset products on the target platform within a historical time period, where the historical time period and the preset time period have the same length. Products with a probability of first transaction higher than a preset transaction threshold are determined as target push products. Therefore, during the product push process to the target object, unfollowed products can be pushed to the target object based on the product's transaction status, achieving the technical effect of product recommendation based on the seller's product transaction status, thus solving the technical problem of not being able to recommend products based on the seller's product transaction status.

[0126] Those skilled in the art will understand that Figure 3 The structure shown is for illustrative purposes only. The computer terminal can also be a smartphone (such as an Android phone, an iOS phone, etc.), a tablet computer, a mobile internet device (MID), a PAD, and other terminal devices. Figure 3 This does not limit the structure of the aforementioned electronic device. For example, computer terminal 30 may also include components that are more advanced than those described above. Figure 3 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 3 The different configurations shown.

[0127] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0128] Embodiments of the present invention also provide a non-volatile storage medium. Optionally, in this embodiment, the storage medium can be used to store the program code executed by the method for determining the push product provided in the above embodiments.

[0129] Optionally, in this embodiment, the non-volatile storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0130] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: obtaining interaction behavior information of the target object from the operation logs of the target platform, wherein the target platform provides multiple preset products to the target object, and the interaction behavior information is used to represent the products that have interacted with the target object; selecting products to be pushed from the multiple preset products of the target platform according to the interaction behavior information, wherein the products to be pushed are selected from the unfollowed products of the target platform that have not interacted with the target object; analyzing the products to be pushed using a preset analysis model to predict the initial transaction probability of the products to be pushed completing their first transaction within a preset time period, wherein the preset analysis model is trained based on the initial transaction probability of multiple preset products in the target platform completing their first transaction within a historical time period, and the historical time period and the preset time period have the same period length; determining the products to be pushed with an initial transaction probability higher than a preset transaction threshold as target push products.

[0131] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: analyzing interaction behavior information to determine the target object's interaction behavior with the product of interest; evaluating the product of interest based on the interaction behavior to determine the evaluation value of the product of interest, wherein the evaluation value is used to represent the degree of interest of the target object in the product of interest;

[0132] Based on the ratings of products that the target audience has followed, predict which products the target audience might be interested in from among the products they have not followed.

[0133] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: identifying products with evaluation values ​​higher than a preset evaluation threshold as favorite products of the target object; using favorite products as training data to train a first interest model of the target object through machine learning, wherein the first interest model is used to represent the target product characteristics of favorite products; and using the first interest model to perform feature analysis on unfollowed products to determine unfollowed products that meet the target product characteristics as products to be pushed.

[0134] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: using the products being followed and their rating values ​​as training data, a second interest model for the target object is trained through machine learning, wherein the second interest model is used to evaluate the rating values ​​of preset products; the second interest model is used to perform feature analysis on unfollowed products to determine their rating values; and unfollowed products with rating values ​​higher than a preset rating threshold are identified as products to be pushed.

[0135] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: before analyzing the product to be pushed using a preset analysis model to determine the initial transaction probability of the product to be pushed, product transaction data is obtained from the operation logs of the target platform. The product transaction data represents the transaction results of a preset number of preset products provided by the target platform within a historical time period. The transaction results include: the preset product completing its first transaction within the historical time period, and the preset product not completing its first transaction within the historical time period. The product transaction data is analyzed using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset product. The preset initial transaction prediction model is used to determine the initial transaction probability of the preset product completing its first transaction within the historical time period and the non-transaction probability of not completing its first transaction within the historical time period based on the transaction results of a preset number of preset products within the historical time period. The initial transaction probability of the preset product is used as a positive sample, and the non-transaction probability of the preset product is used as a negative sample. The preset analysis model is trained using machine learning. The preset analysis model is constructed based on a Bayesian algorithm.

[0136] Optionally, in this embodiment, the product transaction data further includes: transaction factors that affect the transaction results, and the non-volatile storage medium is configured to store program code for performing the following steps: after obtaining product transaction data from the operation logs of the target platform, obtaining a preset product that has completed its first transaction within a historical time period as a sample product; assigning a preset evaluation coefficient according to the transaction factors of the sample product; and determining the initial transaction probability of the sample product according to the transaction factors and the preset evaluation coefficient.

[0137] Optionally, in this embodiment, the non-volatile storage medium is configured to store program code for performing the following steps: after determining that the product to be pushed has an initial transaction probability higher than a preset transaction threshold as the target push product, the target push product is added to the target push list; when the number of target push products in the target push list reaches a preset quantity threshold, the target push product in the target push list is pushed to the target object.

[0138] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0139] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0140] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0142] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0143] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of 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 a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0144] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for determining a product to be pushed, characterized in that, include: The interaction behavior information of the target object is obtained from the operation logs of the target platform. The target platform is used to provide the target object with multiple preset products, and the interaction behavior information is used to indicate the products that have interacted with the target object. Based on the interaction behavior information, a product to be pushed is selected from multiple preset products on the target platform, wherein the product to be pushed is selected from unfollowed products on the target platform that have not interacted with the target object. The product to be pushed is analyzed using a preset analysis model to predict the probability of the product to be pushed completing its first transaction within a preset time period. The preset analysis model is trained based on the probability of the first transaction of multiple preset products in the target platform completing their first transaction within a historical time period. The historical time period and the preset time period have the same period length. The products to be pushed that have an initial transaction probability higher than a preset transaction threshold are identified as target push products. Before analyzing the product to be pushed using a preset analysis model to determine the initial transaction probability of the product to be pushed, the method further includes: Product transaction data is obtained from the operation logs of the target platform. The product transaction data is used to represent the transaction results of a preset number of preset products provided by the target platform within a historical time period. The transaction results include: the preset product completed its first transaction within the historical time period, and the preset product did not complete its first transaction within the historical time period. The product transaction data is analyzed using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset product. The preset initial transaction prediction model is used to determine the initial transaction probability of the preset product completing the transaction for the first time in the historical time period and the non-transaction probability of not completing the transaction for the first time in the historical time period based on the transaction results of a preset number of preset products in the historical time period. The initial transaction probability of the preset product is used as a positive sample, and the non-transaction probability of the preset product is used as a negative sample. The preset analysis model is trained by machine learning, wherein the preset analysis model is constructed based on the Bayesian algorithm. The product transaction data further includes: transaction factors affecting the transaction results. After obtaining the product transaction data from the target platform's operational logs, the method further includes: The preset product that completes its first transaction within the historical time period is selected as the sample product. Preset evaluation coefficients are assigned based on the transaction factors of the sample products; The initial transaction probability of the sample product is determined based on the transaction factors and preset evaluation coefficients of the sample product. The preset initial transaction prediction model is as follows: , ,in, This indicates the probability of the initial transaction. The transaction factors described in the sample products, The preset evaluation coefficient, i, is determined based on the preset quantity of the sample products; The selection of products to be pushed from among the multiple preset products on the target platform based on the interaction behavior information includes: Analyze the interaction behavior information to determine the target object's interaction behavior with the product of interest; The product of interest is evaluated based on the interaction behavior to determine the evaluation value of the product of interest, wherein the evaluation value is used to represent the degree of interest of the target object in the product of interest; Based on the evaluation value of the products that are being followed, predict the products that the target object is interested in from the products that are not being followed. Specifically, based on the evaluation value of the followed products, the products to be pushed to the target audience from the unfollowed products include: Products with evaluation values ​​higher than a preset evaluation threshold are identified as the target object's favorite products. The preferred products are used as training data to train a first interest model for the target object through machine learning, wherein the first interest model is used to represent the target product features of the preferred products. The first interest model is used to perform feature analysis on the unfollowed products, and the unfollowed products that match the characteristics of the target product are identified as the products to be pushed.

2. The method according to claim 1, characterized in that, Based on the rating of the products the target audience is interested in, the following products are predicted to be among the untargeted products: Using the products of interest and their ratings as training data, a second interest model for the target object is trained through machine learning. The second interest model is used to evaluate the ratings of the preset products. The second interest model is used to perform feature analysis on the uninterested products to determine the evaluation value of the uninterested products. Unfollowed products whose evaluation values ​​are higher than a preset evaluation threshold are identified as the products to be pushed to.

3. The method according to claim 1, characterized in that, After determining that the product to be pushed has an initial transaction probability higher than a preset transaction threshold as the target product to be pushed, the method further includes: Add the target push product to the target push list; When the number of target push products in the target push list reaches a preset threshold, the target push products in the target push list are pushed to the target object.

4. A device for determining the delivery of a product, characterized in that, include: The acquisition module is used to acquire interaction behavior information of the target object from the operation logs of the target platform. The target platform is used to provide the target object with multiple preset products, and the interaction behavior information is used to represent the products that have interacted with the target object. The selection module is used to select a product to be pushed from a plurality of preset products on the target platform based on the interaction behavior information, wherein the product to be pushed is selected from unfollowed products on the target platform that have not interacted with the target object. The prediction module is used to analyze the product to be pushed using a preset analysis model and predict the probability of the product to be pushed completing its first transaction within a preset time period. The preset analysis model is trained based on the probability of the preset products completing their first transaction within a historical time period in the target platform. The historical time period and the preset time period have the same period length. The determination module is used to determine the products to be pushed that have an initial transaction probability higher than a preset transaction threshold as target push products; The device further includes: The first acquisition submodule is used to acquire product transaction data from the operation log of the target platform before analyzing the product to be pushed using a preset analysis model and determining the initial transaction probability of the product to be pushed. The product transaction data is used to represent the transaction results of a preset number of preset products provided by the target platform within a historical time period. The transaction results include: the preset product completed its first transaction within the historical time period, and the preset product did not complete its first transaction within the historical time period. The first determining submodule is used to analyze the product transaction data using a preset initial transaction prediction model to determine the initial transaction probability and non-transaction probability of the preset product. The preset initial transaction prediction model is used to determine the initial transaction probability of the preset product completing the transaction for the first time in the historical time period and the non-transaction probability of not completing the transaction for the first time in the historical time period based on the transaction results of a preset number of preset products in the historical time period. The training submodule is used to train the preset analysis model by taking the initial transaction probability of the preset product as a positive sample and the non-transaction probability of the preset product as a negative sample, and using machine learning. The preset analysis model is constructed based on the Bayesian algorithm. The product transaction data further includes: transaction factors that affect the transaction results; the device further includes: The second acquisition submodule is used to acquire, after acquiring product transaction data from the operation logs of the target platform, the preset product that completed its first transaction within the historical time period as a sample product. The allocation submodule is used to allocate preset evaluation coefficients based on the transaction factors of the sample products. The second determining submodule is used to determine the initial transaction probability of the sample product based on the transaction factors and preset evaluation coefficients of the sample product. The preset initial transaction prediction model is as follows: , ,in, This indicates the probability of the initial transaction. The transaction factors described in the sample products, The preset evaluation coefficient, i, is determined based on the preset quantity of the sample products; The selection module includes: Analysis unit, used to analyze the interaction behavior information and determine the interaction behavior of the target object with respect to the product of interest; The first determining unit is configured to evaluate the product of interest based on the interaction behavior and determine the evaluation value of the product of interest, wherein the evaluation value is used to represent the degree of interest of the target object in the product of interest; The prediction unit is used to predict, based on the evaluation value of the product being followed, a product that the target object is interested in from the unfollowed products to be pushed to; The prediction unit includes: The second determining unit is used to determine the products of interest whose evaluation value is higher than a preset evaluation threshold as the favorite products of the target object; The first training unit is used to train the first interest model of the target object by using the favorite products as training data and machine learning, wherein the first interest model is used to represent the target product features of the favorite products. The third determining unit is used to perform feature analysis on the unfollowed products using the first interest model, and determine the unfollowed products that match the characteristics of the target product as the products to be pushed.

5. A non-volatile storage medium, characterized in that, The non-volatile storage medium stores a program, wherein when the program is executed, it controls the device containing the non-volatile storage medium to execute the method for determining the push product as described in any one of claims 1 to 3.

6. An electronic device, characterized in that, include: A memory and a processor, the processor being configured to run a program stored in the memory, wherein the program, when running, executes the method for determining a push product as described in any one of claims 1 to 3.

Images