Advertisement recommendation method and system

Abstract

The present invention discloses an advertisement recommendation method and system. By constructing a recommendation tree, the judgment part can easily judge the advertisement information and obtain the estimated recommended advertisement information. The estimated recommended advertisement information is passed through a range control structure to obtain a recommended range vector. The recommendation neural network is trained by multiple advertisement information and corresponding product information in the blockchain, and the advertisement category probability value set is used as input, so that the accuracy of multiple information of the advertisement information category is improved during the training and detection process. The recommended range vector is used as a restriction to judge the recommendation neural network product vector to obtain the recommended advertisement information, and the recommended advertisement information includes one or more advertisement situations, and the advertisement situation includes information such as advertisement price, price adjustment, delivery platform, delivery direction, switch time period, etc. The accuracy of advertisement recommendation is improved, and the revenue efficiency of advertisement is improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an advertisement recommendation method and system. Background Art

[0002] Currently, there are many ways to analyze advertisements and recommend them, that is, to recommend products corresponding to the advertisements. The relationship between advertisement information and product information is analyzed to obtain the situation of the advertisement, such as the advertisement bid, price adjustment, etc., and then the advertisement is recommended in a targeted manner. However, the current advertisement recommendation is not accurate enough, resulting in low advertising efficiency. Summary of the invention

[0003] The purpose of the present invention is to provide an advertisement recommendation method and system to solve the above problems existing in the prior art.

[0004] In a first aspect, an embodiment of the present invention provides an advertisement recommendation method, comprising:

[0005] Obtaining advertisement information; the advertisement information indicates information related to advertisement and product recommendation;

[0006] Obtaining an advertisement blockchain category set; the advertisement blockchain category set represents the classification type of each type of advertisement information in the blockchain;

[0007] Based on the advertising blockchain category set, a set of advertising probabilities is obtained through a Gaussian function;

[0008] Based on the advertisement information and the advertisement probability set, recommended advertisement information is obtained through a recommendation model.

[0009] Optionally, the advertisement probability set is obtained by using a Gaussian function based on the advertisement blockchain category set, including:

[0010] Based on the advertising blockchain category set, a product category Gaussian set is obtained; the values ​​in the product category Gaussian set are distributed using a Gaussian function;

[0011] Based on the Gaussian set of product categories, an advertisement probability set is obtained through a Gaussian function.

[0012] Optionally, obtaining a Gaussian set of product categories based on the advertisement blockchain category set includes:

[0013] Based on the advertising blockchain category set, a product category reciprocal set is obtained; the value in the product category reciprocal set is the reciprocal of the value in the advertising blockchain category set;

[0014] Based on the inverse set of product categories, a total product number is obtained; the total product number is the product of multiplication;

[0015] The vector value in the reciprocal set of the product categories is multiplied by the total product number of the products to obtain a Gaussian set of product categories.

[0016] Optionally, the recommendation model includes a recommendation tree, a recommendation neural network and a range control structure:

[0017] The input of the recommendation tree is the advertisement probability set; the input of the recommendation neural network is the output of the recommendation tree; the input of the range control structure is the output of the recommendation tree and the output of the recommendation tree.

[0018] Optionally, obtaining recommended advertising information through a recommendation model based on the advertising information and the advertising probability set includes:

[0019] Based on the advertisement information, an advertisement category value set is obtained; a value of 1 in the advertisement category value set indicates that the advertisement has a certain type of information; a value of 0 in the advertisement category value set indicates that the advertisement does not have a certain type of information;

[0020] Based on the advertising category value set, the estimated recommended advertising information is obtained through the recommendation tree;

[0021] Based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information, the recommended advertisement information is obtained through a recommendation neural network structure.

[0022] Optionally, the construction of the recommendation tree includes:

[0023] Obtaining a recommendation root node; the input of the recommendation root node is the advertisement probability set;

[0024] Based on the recommendation root node, a plurality of recommendation child nodes are obtained; the recommendation child nodes are a collection of a plurality of products;

[0025] By judging the values ​​in the advertisement probability set multiple times, multiple recommendation sub-nodes are constructed until all values ​​in the advertisement probability set are traversed; the values ​​in the recommendation sub-nodes correspond to different products.

[0026] Optionally, obtaining the recommended advertising information by recommending a neural network structure based on the advertising category value set, the advertising probability set and the estimated recommended advertising information includes:

[0027] The estimated recommended advertisement information is input into a range control structure to obtain a recommended range vector; the number of vector values ​​of the recommended range vector is the same as the number of products used for recommendation; the vector value of the recommended range vector is 1, indicating that the corresponding product exists in the corresponding estimated recommended advertisement information; the vector value of the recommended range vector is 0, indicating that the corresponding product does not exist in the corresponding estimated recommended advertisement information;

[0028] Based on the advertisement probability set and the advertisement category value set, an advertisement category probability value set is obtained; the advertisement category probability value set is a set formed by multiplying advertisement category values ​​of multiple categories in the advertisement category value set by values ​​in the corresponding advertisement probability set;

[0029] Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification to obtain a recommendation neural network product vector;

[0030] Based on the recommended neural network product vector and the recommended range vector, recommended advertising information is obtained; the products in the recommended advertising information are products whose vector values ​​in the recommended neural network product vector and the corresponding recommended range vector are both 1.

[0031] Optionally, it is characterized in that a training method of a neural network model is recommended:

[0032] Obtaining a training set; the training set includes training information and annotated recommended products; the training information includes a set of training advertisement category probability values;

[0033] Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification, and obtain a training recommendation neural network product vector;

[0034] Obtaining a recommendation loss value based on the training recommendation neural network product vector and the labeled recommended product;

[0035] Obtaining the current number of training iterations of the recommendation neural network model and the preset maximum number of training iterations of the recommendation neural network model;

[0036] When the recommendation loss value is less than or equal to the threshold or the number of training iterations reaches the maximum number of iterations, the training is stopped to obtain a trained recommendation neural network model.

[0037] In a second aspect, an embodiment of the present invention provides an advertisement recommendation system, including:

[0038] Acquisition module: acquiring advertisement information; the advertisement information indicates information related to the advertisement itself and the product recommendation;

[0039] Gaussian probability module: based on the advertisement information, obtain an advertisement blockchain category set; the advertisement blockchain category set represents the classification type of each type of information in the advertisement information; based on the advertisement blockchain category set, obtain an advertisement probability set through a Gaussian function;

[0040] Recommendation module: based on the advertisement probability set, obtains recommended advertisement information through a recommendation model;

[0041] The recommendation model includes a recommendation tree, a product classification neural network and multiple recommendation subtrees.

[0042] Optionally, the advertisement probability set is obtained by using a Gaussian function based on the advertisement blockchain category set, including:

[0043] Based on the advertising blockchain category set, a product category Gaussian set is obtained; the product category Gaussian set is a vector of distribution probabilities obtained using a Gaussian function;

[0044] Based on the Gaussian set of product categories, an advertisement probability set is obtained through a Gaussian function.

[0045] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0046] The embodiment of the present invention also provides an advertisement recommendation method and system, which obtains the difficulty of each information classification through a Gaussian function. By constructing a recommendation tree, the judgment part is relatively easy to judge the advertisement information, and the estimated recommended advertisement information is obtained. The estimated recommended advertisement information passes through the range control structure to obtain a recommendation range vector. A recommendation neural network is constructed according to the advertisement probability set, and the higher the probability, the more neural network layers are passed. The recommendation neural network is trained through multiple advertisement information and corresponding product information in the blockchain, and the advertisement category probability value set is used as input, so that the accuracy of multiple information of the advertisement information category is improved during the training and detection process. The recommendation range vector is used as a restriction to judge the recommendation neural network product vector to obtain the recommended advertisement information. The accuracy of advertisement recommendation is improved, and the revenue efficiency of advertisement is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flow chart of an advertisement recommendation method provided by an embodiment of the present invention.

[0048] Figure 2 It is a structural schematic diagram of a recommendation neural network in an advertisement recommendation system provided by an embodiment of the present invention.

[0049] Figure 3 It is a schematic diagram of the block structure of an electronic device provided by an embodiment of the present invention.

[0050] Labels in the figure: bus 500; receiver 501; processor 502; transmitter 503; memory 504; bus interface 505. DETAILED DESCRIPTION

[0051] The present invention will be described in detail below in conjunction with the accompanying drawings.

[0052] Example 1

[0053] like Figure 1As shown, an embodiment of the present invention provides an advertisement recommendation method, the method comprising:

[0054] S101: Obtaining advertisement information. The advertisement information refers to information related to advertisement and product recommendation.

[0055] In this embodiment, the advertisement information includes the usage location, color preference, demand time and the type of product purchased.

[0056] S102: Obtain an advertisement blockchain category set. The advertisement blockchain category set represents the classification type of each type of advertisement information in the blockchain.

[0057] S103: Based on the advertisement blockchain category set, an advertisement probability set is obtained through a Gaussian function.

[0058] S104: Based on the advertisement information and the advertisement probability set, obtain recommended advertisement information through a recommendation model.

[0059] In the embodiment of the present invention, the recommended advertisement information includes one or more advertisement conditions, and the advertisement conditions include information such as advertisement price, price adjustment, delivery platform, delivery direction, and switch time period.

[0060] Optionally, the advertisement probability set is obtained by using a Gaussian function based on the advertisement blockchain category set, including:

[0061] Based on the advertisement blockchain category set, a product category Gaussian set is obtained. The product category Gaussian set is a distribution probability obtained by using a Gaussian function.

[0062] In this embodiment, the color preferences in the advertisement information are divided into white, black, red, green, yellow, blue and colorful, and the product category quantity value corresponding to the color is 7.

[0063] Based on the product category quantity value set, an advertising probability set is obtained through a Gaussian function.

[0064] The parameters of the Gaussian function include expectation and standard deviation. The closer to the expectation, that is, the smaller the standard deviation, the greater the probability; the farther from the expectation, that is, the larger the standard deviation, the smaller the probability.

[0065] Through the above method, the number of types of each type of information in the advertising information is used, and the Gaussian function is used to calculate the number of categories of each type of advertising information, so that the probability of more advertising information categories is greater, and the probability of fewer advertising information categories is smaller.

[0066] Optionally, obtaining a Gaussian set of product categories based on the advertisement blockchain category set includes:

[0067] Based on the advertising blockchain category set, a product category reciprocal set is obtained; the value in the product category reciprocal set is the reciprocal of the value in the advertising blockchain category set.

[0068] Based on the inverse set of product categories, a total product number is obtained, which is the product of multiplication.

[0069] The vector value in the reciprocal set of the product categories is multiplied by the total product number to obtain a Gaussian set of product categories.

[0070] Among them, as described in this embodiment, the advertising blockchain category set value is {7, 5, 3, 9}, then the product category inverse set is The total product number is 7×5×3×9=945, and the Gaussian set of product categories is {135, 189, 315, 105}.

[0071] Through the above method, the numerical value is made to be smaller when the number of advertising information categories is greater, and larger when the number of advertising information categories is less.

[0072] Optionally, the recommendation model includes a recommendation tree, a recommendation neural network and a range control structure:

[0073] The input of the recommendation tree is the advertisement probability set; the input of the recommendation neural network is the output of the recommendation tree; the input of the range control structure is the output of the recommendation tree and the output of the recommendation tree;

[0074] The recommendation neural network is constructed based on an advertisement probability set;

[0075] Among them, the higher the probability, the more neural network layers it passes through.

[0076] Optionally, obtaining recommended advertising information through a recommendation model based on the advertising information and the advertising probability set includes:

[0077] Based on the advertisement information, an advertisement category value set is obtained; a value of 1 in the advertisement category value set indicates that the advertisement has a certain type of information; a value of 0 in the advertisement category value set indicates that the advertisement does not have a certain type of information.

[0078] The advertisement category value set represents the status of the advertisement.

[0079] Based on the advertising category value set, the estimated recommended advertising information is obtained through the recommendation tree.

[0080] Based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information, the recommended advertisement information is obtained through a recommendation neural network structure.

[0081] Through the above method, a recommendation tree is first used to distinguish certain types of advertising information to obtain estimated recommended advertising information. Under the constraints of the estimated recommended advertising information, a neural network is used to obtain recommended products.

[0082] Optionally, the recommendation tree is constructed as follows:

[0083] A recommendation root node is obtained; the input of the recommendation root node is the advertisement probability set.

[0084] A plurality of recommended sub-nodes are obtained; the recommended sub-nodes are a collection of a plurality of products.

[0085] By judging the values ​​in the advertising probability set multiple times, multiple child nodes are constructed until all values ​​in the advertising probability set are traversed; the values ​​in the child nodes correspond to different products.

[0086] Among them, in the process of tree construction, the cost complexity pruning method in post-pruning is used to prune the recommendation tree.

[0087] Among them, let the non-leaf nodes of the recommendation tree be {T1, T2, T3, ..., T n}. Calculate the surface error rate gain values ​​of all non-leaf nodes, select the non-leaf node with the smallest surface error rate gain value, if multiple non-leaf nodes have the same small surface error rate gain value, select the non-leaf node with the largest number of nodes. Prune the selected non-leaf nodes

[0088] The calculation formula of the surface error rate gain value is:

[0089]

[0090] Among them, R(t) represents the error cost of the leaf node, R(t) = r(t) × p(t), r(t) is the error rate of the node, and p(t) is the proportion of the node data volume. R(T) represents the error cost of the subtree, r i (t) is the error rate of child node i, p i (t) represents the percentage of data nodes of node i. N(T) represents the number of subtree nodes.

[0091] Through the above method, without knowing the advertising probability set, the constructed tree should be able to judge and classify all the values ​​in the advertising probability set. The values ​​in the classification subset are the values ​​in the specified area after classifying the values ​​in the advertising probability set, so as to recommend advertisements of different products for different types.

[0092] Optionally, obtaining the recommended advertising information by recommending a neural network structure based on the advertising category value set, the advertising probability set and the estimated recommended advertising information includes:

[0093] The recommendation neural network structure includes a recommendation neural network input layer and a range control structure.

[0094] The estimated recommended advertisement information is input into a range control structure to obtain a recommended range vector. The number of vector values ​​of the recommended range vector is the same as the number of products used for recommendation. A vector value of the recommended range vector of 1 indicates that the corresponding product exists in the corresponding estimated recommended advertisement information. A vector value of the recommended range vector of 0 indicates that the corresponding product does not exist in the corresponding estimated recommended advertisement information.

[0095] The range control structure is used to construct a vector with the same size as all product categories, and to modify the value in the vector according to the median value of the estimated recommended advertising information.

[0096] Based on the advertisement probability set and the advertisement category value set, an advertisement category probability value set is obtained. The advertisement category probability value set is a set formed by multiplying advertisement category values ​​of multiple categories in the advertisement category value set by values ​​in the corresponding advertisement probability set.

[0097] Among them, such as {1, 0, 0, 0, 0, 0, 1, 0, 0}, the first three values ​​{1, 0, 0} represent a category of advertising information, and the last six values ​​{0, 0, 0, 1, 0, 0} represent a category of advertising information. The advertising probability set is {0.5, 0.1}. Multiply {1, 0, 0} by 0.5 to get {0.5, 0, 0}, multiply {0, 0, 0, 1, 0, 0} by 0.1 to get {0, 0, 0, 0.1, 0, 0}, and then form a new advertising category probability value set {0.5, 0, 0, 0, 0, 0.1, 0, 0}.

[0098] The advertisement category probability value set is input into a recommendation neural network for classification to obtain a recommendation neural network product vector.

[0099] Based on the recommendation neural network product vector and the recommendation range vector, recommended advertising information is obtained. The products in the recommended advertising information are products whose vector values ​​in the recommendation neural network product vector and the corresponding recommendation range vector are both 1.

[0100] Through the above method, the tree is used to judge some values ​​and obtain the values ​​corresponding to certain advertising information categories, but for problems that cannot be classified, a neural network is used to judge. The recommended neural network structure increases the flexibility of judgment.

[0101] Optional, recommended training method for neural network model:

[0102] A training set is obtained; the training set includes training information and annotated recommended products; the training information includes a set of training advertisement category probability values.

[0103] The advertisement category probability value set is input into the recommendation neural network for classification to obtain a training recommendation neural network product vector.

[0104] Among them, the recommended neural network part structure is as follows Figure 2 shown.

[0105] A recommendation loss value is obtained based on the training recommendation neural network product vector and the labeled recommended product.

[0106] Among them, the cross entropy loss function is used to calculate the recommendation loss value.

[0107] The current training iteration number of the recommendation neural network model and the preset maximum iteration number of the recommendation neural network model training are obtained.

[0108] The maximum number of iterations in this embodiment is 8000.

[0109] When the recommendation loss value is less than or equal to the threshold or the number of training iterations reaches the maximum number of iterations, the training is stopped to obtain a trained recommendation neural network model.

[0110] Through the above method, through the category of advertising information, through the Gaussian function, the difficulty of each information classification is obtained. The advertising information is judged by part of the relatively easy advertising information, and the estimated recommended advertising information is obtained through the judgment of the constructed recommendation tree. The recommendation neural network is constructed according to the advertising probability set. The higher the probability, the more neural network layers are passed. The neural network is trained by the product information corresponding to multiple advertising information in the blockchain, and the advertising category probability value set is obtained through the Gaussian function, so that the accuracy of the neural network with more categories is improved during the training process. The recommended range vector is obtained by estimating the recommended advertising information through the range control structure. The recommended range vector is used as a restriction to judge the recommended neural network product vector to obtain the recommended advertising information.

[0111] Example 2

[0112] Based on the above-mentioned advertisement recommendation method, an embodiment of the present invention further provides an advertisement recommendation system, the system comprising an acquisition module, a Gaussian probability module and a recommendation module:

[0113] The acquisition module is used to obtain advertisement information, wherein the advertisement information refers to information related to the advertisement itself and the product recommendation.

[0114] The Gaussian probability module is used to obtain an advertisement blockchain category set based on the advertisement information. The advertisement blockchain category set represents the classification type of each type of information in the advertisement information; based on the advertisement blockchain category set, an advertisement probability set is obtained through a Gaussian function.

[0115] The recommendation module is used to obtain recommended advertisement information based on the advertisement probability set through a recommendation model.

[0116] The recommendation model includes a recommendation tree, a product classification neural network and multiple recommendation subtrees.

[0117] Regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0118] The embodiment of the present invention further provides an electronic device, such as Figure 3 As shown, it includes a memory 504, a processor 502, and a computer program stored in the memory 504 and executable on the processor 502. When the processor 502 executes the program, the steps of any one of the above-mentioned advertisement recommendation methods are implemented.

[0119] Among them, Figure 3 In the embodiment of the present invention, a bus architecture (represented by bus 500) is shown, which may include any number of interconnected buses and bridges, and bus 500 links various circuits including one or more processors represented by processor 502 and memory represented by memory 504. Bus 500 may also link various other circuits such as peripherals, voltage regulators, and power management circuits, which are well known in the art and are therefore not further described herein. Bus interface 505 provides an interface between bus 500 and receiver 501 and transmitter 503. Receiver 501 and transmitter 503 may be the same element, namely a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 502 is responsible for managing bus 500 and general processing, while memory 504 may be used to store data used by processor 502 when performing operations.

[0120] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the program implements the steps of any one of the above-mentioned advertisement recommendation methods and the above-mentioned data involved.

[0121] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the present invention is not directed to any specific programming language either. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description of the above specific languages ​​is for disclosing the best mode of the present invention.

[0122] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

Claims

1. An advertisement recommendation method, characterized in that: include: Obtain advertising information; The advertisement information indicates information related to advertisement and product recommendation; Obtaining an advertisement blockchain category set; the advertisement blockchain category set represents the classification type of each type of advertisement information in the blockchain; Based on the advertising blockchain category set, a set of advertising probabilities is obtained through a Gaussian function; Based on the advertisement information and the advertisement probability set, obtaining recommended advertisement information through a recommendation model; The advertisement probability set is obtained based on the advertisement blockchain category set through a Gaussian function, including: Based on the advertising blockchain category set, a product category Gaussian set is obtained; the values ​​in the product category Gaussian set are distributed using a Gaussian function; Based on the Gaussian set of product categories, an advertising probability set is obtained through a Gaussian function; The recommendation model includes a recommendation tree, a recommendation neural network and a range control structure: The input of the recommendation tree is the advertisement information; the input of the recommendation neural network is the advertisement probability set and the advertisement information; the input of the range control structure is the output of the recommendation tree; The recommendation neural network is constructed based on an advertisement probability set; The step of obtaining recommended advertising information based on the advertising information and the advertising probability set through a recommendation model includes: Based on the advertisement information, an advertisement category value set is obtained; a value of 1 in the advertisement category value set indicates that the advertisement has a certain type of information; a value of 0 in the advertisement category value set indicates that the advertisement does not have a certain type of information; Based on the advertising category value set, the estimated recommended advertising information is obtained through the recommendation tree; Based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information, obtaining the recommended advertisement information through a recommendation neural network; The construction of the recommendation tree includes: Obtaining a recommendation root node; the input of the recommendation root node is the advertisement probability set; Based on the recommendation root node, a plurality of recommendation child nodes are obtained; the recommendation child nodes are a collection of a plurality of products; By judging the values ​​in the advertising probability set multiple times, multiple recommendation sub-nodes are constructed until all values ​​in the advertising probability set are traversed; the values ​​in the recommendation sub-nodes correspond to different products; The method of obtaining the recommended advertisement information based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information through a recommendation neural network includes: The estimated recommended advertisement information is input into a range control structure to obtain a recommended range vector; the number of vector values ​​of the recommended range vector is the same as the number of products used for recommendation; the vector value of the recommended range vector is 1, indicating that the corresponding product exists in the corresponding estimated recommended advertisement information; the vector value of the recommended range vector is 0, indicating that the corresponding product does not exist in the corresponding estimated recommended advertisement information; Based on the advertisement probability set and the advertisement category value set, an advertisement category probability value set is obtained; the advertisement category probability value set is a set formed by multiplying advertisement category values ​​of multiple categories in the advertisement category value set by values ​​in the corresponding advertisement probability set; Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification to obtain a recommendation neural network product vector; Based on the recommendation neural network product vector and the recommendation range vector, recommended advertising information is obtained; the products in the recommended advertising information are products whose vector values ​​in the recommendation neural network product vector and the corresponding recommendation range vector are both 1; Recommended neural network training method: Obtaining a training set; the training set includes training information and annotated recommended products; the training information includes a set of training advertisement category probability values; Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification, and obtain a training recommendation neural network product vector; Obtaining a recommendation loss value based on the training recommendation neural network product vector and the labeled recommended product; Obtaining the current number of training iterations of the recommendation neural network and a preset maximum number of training iterations of the recommendation neural network; When the recommendation loss value is less than or equal to the threshold or the number of training iterations reaches the maximum number of iterations, the training is stopped to obtain a trained recommendation neural network.

2. The advertisement recommendation method according to claim 1, characterized in that: The Gaussian set of product categories is obtained based on the advertising blockchain category set, including: Based on the advertising blockchain category set, a product category reciprocal set is obtained; the value in the product category reciprocal set is the reciprocal of the value in the advertising blockchain category set; Based on the inverse set of product categories, a total product number is obtained; the total product number is the product of multiplication; The vector value in the reciprocal set of the product categories is multiplied by the total product number to obtain a Gaussian set of product categories.

3. An advertisement recommendation system, characterized in that: include: Acquisition module: obtain advertising information; The advertisement information refers to information related to the advertisement itself and the product recommendation; Gaussian probability module: based on the advertisement information, obtain an advertisement blockchain category set; the advertisement blockchain category set represents the classification type of each type of information in the advertisement information; Based on the advertising blockchain category set, a set of advertising probabilities is obtained through a Gaussian function; Recommendation module: based on the advertisement probability set, obtains recommended advertisement information through a recommendation model; The advertisement probability set is obtained based on the advertisement blockchain category set through a Gaussian function, including: Based on the advertising blockchain category set, a product category Gaussian set is obtained; the values ​​in the product category Gaussian set are distributed using a Gaussian function; Based on the Gaussian set of product categories, an advertising probability set is obtained through a Gaussian function; The recommendation model includes a recommendation tree, a recommendation neural network and a range control structure: The input of the recommendation tree is the advertisement information; the input of the recommendation neural network is the advertisement probability set and the advertisement information; the input of the range control structure is the output of the recommendation tree; The recommendation neural network is constructed based on an advertisement probability set; The step of obtaining recommended advertising information based on the advertising information and the advertising probability set through a recommendation model includes: Based on the advertisement information, an advertisement category value set is obtained; a value of 1 in the advertisement category value set indicates that the advertisement has a certain type of information; a value of 0 in the advertisement category value set indicates that the advertisement does not have a certain type of information; Based on the advertising category value set, the estimated recommended advertising information is obtained through the recommendation tree; Based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information, obtaining the recommended advertisement information through a recommendation neural network; The construction of the recommendation tree includes: Obtaining a recommendation root node; the input of the recommendation root node is the advertisement probability set; Based on the recommendation root node, a plurality of recommendation child nodes are obtained; the recommendation child nodes are a collection of a plurality of products; By judging the values ​​in the advertising probability set multiple times, multiple recommendation sub-nodes are constructed until all values ​​in the advertising probability set are traversed; the values ​​in the recommendation sub-nodes correspond to different products; The method of obtaining the recommended advertisement information based on the advertisement category value set, the advertisement probability set and the estimated recommended advertisement information through a recommendation neural network includes: The estimated recommended advertisement information is input into a range control structure to obtain a recommended range vector; the number of vector values ​​of the recommended range vector is the same as the number of products used for recommendation; the vector value of the recommended range vector is 1, indicating that the corresponding product exists in the corresponding estimated recommended advertisement information; the vector value of the recommended range vector is 0, indicating that the corresponding product does not exist in the corresponding estimated recommended advertisement information; Based on the advertisement probability set and the advertisement category value set, an advertisement category probability value set is obtained; the advertisement category probability value set is a set formed by multiplying advertisement category values ​​of multiple categories in the advertisement category value set by values ​​in the corresponding advertisement probability set; Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification to obtain a recommendation neural network product vector; Based on the recommendation neural network product vector and the recommendation range vector, recommended advertising information is obtained; the products in the recommended advertising information are products whose vector values ​​in the recommendation neural network product vector and the corresponding recommendation range vector are both 1; Recommended neural network training method: Obtaining a training set; the training set includes training information and annotated recommended products; the training information includes a set of training advertisement category probability values; Input the set of probability values ​​of the advertising categories into the recommendation neural network for classification, and obtain a training recommendation neural network product vector; Obtaining a recommendation loss value based on the training recommendation neural network product vector and the labeled recommended product; Obtaining the current number of training iterations of the recommendation neural network and a preset maximum number of training iterations of the recommendation neural network; When the recommendation loss value is less than or equal to the threshold or the number of training iterations reaches the maximum number of iterations, the training is stopped to obtain a trained recommendation neural network.

4. The advertisement recommendation system according to claim 3, characterized in that: The advertisement probability set is obtained based on the advertisement blockchain category set through a Gaussian function, including: Based on the advertising blockchain category set, a product category Gaussian set is obtained; the product category Gaussian set is a vector of distribution probabilities obtained using a Gaussian function; Based on the Gaussian set of product categories, an advertisement probability set is obtained through a Gaussian function.

Images