Content recommendation method and device, computer equipment and medium

Abstract

The invention provides a content recommendation method and device, computer equipment and a medium, and the method comprises the steps: obtaining data information and an index proportion, calculating the push proportion of each category based on the index proportion and the data information, and recommending corresponding content to a client in a revision duration according to the push proportion of each category. The method has the beneficial effects that the user can comprehensively browse related contents, the phenomenon of over-fitting push is reduced, the user can contact diversified information and obtain information of different products, and the shopping experience and satisfaction of the user are improved.

Description

Technical Field

[0001] The present invention relates to the field of big data, and in particular to a content recommendation method, device, computer equipment and medium. Background Art

[0002] On e-commerce platforms, existing content recommendations generally recommend related content based on users' search information. After users browse these contents, related content will be recommended to users based on their browsing information. This will cause overfitting of the content recommended to users, that is, all recommended content is of the same category. In this way, users will ignore information of other categories and will not be able to push expanded product content, which limits users' opportunities to access diversified information and thus affects their shopping experience and satisfaction. Summary of the invention

[0003] The main purpose of the present invention is to provide a content recommendation method, device, computer equipment and medium, aiming to solve the problem that the existing technology cannot realize the extended push of commodity content.

[0004] The present invention provides a content recommendation method, comprising:

[0005] Detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period;

[0006] If the first preset time period is exceeded, obtaining the weight of the user's indicators for each category, and obtaining data information of the user browsing the contents of each category in multiple sub-time periods of the set time period;

[0007] Input each of the data information into each of the recurrent units in the long short-term memory network one by one to obtain a browsing state; wherein the long short-term memory network includes a plurality of recurrent units, the number of the recurrent units is the same as the number of the data information, and a recurrent unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous recurrent unit;

[0008] Inputting the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector;

[0009] Nonlinearly map the browsing state vector to the interval (0,1), and calculate the linear distance from 1;

[0010] A standard value is obtained based on the index proportion, and the linear distance is compared with the standard value to obtain a comparison value;

[0011] When the comparison value exceeds the preset value, one or more recurrent units are added to the long short-term memory network, and simulation data information is input into the newly added recurrent units and the simulation data information is continuously adjusted until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information;

[0012] Setting a revision duration according to the target simulation data information;

[0013] The push weight of each category is adjusted based on the target simulation data information, and corresponding content is recommended to the customer within the revised time period according to the push weight of each category.

[0014] Furthermore, the step of inputting each of the data information into each recurrent unit in the long short-term memory network one by one to obtain the browsing state includes:

[0015] Input each of the data information into the forget gate, input gate, new information, and output gate of each recurrent unit in the long short-term memory network;

[0016] According to formula F t =sigmoid(W xf |X t |+W hf H t-1 +b f )

[0017] I t =sigmoid(W xi |X t |+W hi H t-1 +b i )

[0018] O t =sigmoid(W xo |X t |+W ho H t-1 +b o )

[0019] N t =tanh(W xn |X t |+W hn H t-1 +b n )

[0020] C t =F t ⊙C t-1 +I t ⊙N t

[0021] Calculate the output state of each cycle unit in turn to obtain the browsing state of the last cycle unit output; where F t represents the tth forget gate, I t represents the tth input gate, N t represents the tth new information, O t represents the tth output gate, C t represents the tth output state, X t represents the tth data information, W xf , W hf 、b f , W xi , W hi 、b i , W xo , W ho 、b o , W xn , W hn 、b h These are preset parameters.

[0022] Furthermore, before the step of recommending corresponding content to the customer within the revised time period according to the push proportion of each category, the method further includes:

[0023] Obtaining preference information of each category of content according to the data information of each category of content;

[0024] According to the preference information of each category of content, content information related to the preference information is obtained to recommend to the user.

[0025] Furthermore, the step of obtaining the weight of the user's indicators for each category includes:

[0026] Obtain the user's personal information and the user's search content;

[0027] The weight of the index of each category of the user is set according to the search content and the personal information.

[0028] Furthermore, the indicator proportion is a data volume proportion or a browsing time proportion.

[0029] Furthermore, the indicator weight is the data volume weight, and after the step of adjusting the push weight of each category based on the target simulation data information and recommending corresponding content to the customer within the revised time period according to the push weight of each category, the method further includes:

[0030] Determining whether the time period during which the user has not placed an order after the revised time period exceeds a second preset time period;

[0031] If the second preset time period is exceeded, then the time information of the user browsing each category within the set revised time period is counted;

[0032] Setting the duration ratio of each category according to the duration information;

[0033] Setting a new data volume ratio according to the duration ratio;

[0034] Obtaining a new push weight based on the new data volume weight;

[0035] Corresponding content is recommended to the customer according to the new push proportion.

[0036] Furthermore, before the step of detecting whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period, the method further includes:

[0037] Get the user's search information;

[0038] Push relevant content to the user based on the search information.

[0039] The present invention also provides a content recommendation device, comprising:

[0040] A detection module, used to detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period;

[0041] An acquisition module, configured to acquire the weight of the user's indicators for each category if the first preset time period is exceeded, and acquire data information of the user browsing the contents of each category within a plurality of sub-time periods of a set time period;

[0042] A first input module is used to input each of the data information into each loop unit in the long short-term memory network in a one-to-one correspondence to obtain a browsing state; wherein the long short-term memory network includes a plurality of loop units, the number of the loop units is the same as the number of the data information, and a loop unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous loop unit;

[0043] A second input module is used to input the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector;

[0044] A mapping module, used for nonlinearly mapping the browsing state vector to the interval (0,1) and calculating the linear distance from 1;

[0045] A comparison module, configured to obtain a standard value based on the index proportion, compare the linear distance with the standard value, and obtain a comparison value;

[0046] An adding module is used for adding one or more recurrent units to the long short-term memory network when the comparison value exceeds the preset value, inputting simulation data information to the newly added recurrent units and continuously adjusting the simulation data information until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information;

[0047] A setting module, used for setting a revision time according to the target simulation data information;

[0048] The adjustment module is used to adjust the push weight of each category based on the target simulation data information, and recommend corresponding content to the customer within the revised time period according to the push weight of each category.

[0049] The present invention also provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any one of the above methods when executing the computer program.

[0050] The present invention also provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned methods are implemented.

[0051] The beneficial effects of the present invention are as follows: by obtaining data information and index proportions, and calculating the push proportions of each category based on the index proportions and data information, corresponding content is recommended to customers within the revised time period according to the push proportions of each category. This enables users to browse relevant content comprehensively, reduces the phenomenon of overfitting push, enables users to access diversified information, obtain information on different products, and improves users' shopping experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a flow chart of a content recommendation method according to an embodiment of the present invention;

[0053] Figure 2 is a schematic block diagram of the structure of a content recommendation device according to an embodiment of the present invention;

[0054] Figure 3 A schematic block diagram of the structure of a computer device according to an embodiment of the present application.

[0055] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] It should be noted that all directional indications in the embodiments of the present invention (such as up, down, left, right, front, back, etc.) are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly. The connection can be a direct connection or an indirect connection.

[0058] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there may be three relationships. For example, A and B can represent three situations: A exists alone, A and B exist at the same time, and B exists alone.

[0059] In addition, in the present invention, descriptions such as "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0060] Reference Figure 1 The present invention proposes a content recommendation method, comprising:

[0061] S1: Detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period;

[0062] S2: if the first preset time period is exceeded, obtaining the weight of the user's indicators for each category, and obtaining data information of the user browsing the contents of each category in multiple sub-time periods of the set time period;

[0063] S3: Input each of the data information into each of the recurrent units in the long short-term memory network one by one to obtain a browsing state; wherein the long short-term memory network includes a plurality of recurrent units, the number of the recurrent units is the same as the number of the data information, and one recurrent unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous recurrent unit;

[0064] S4: inputting the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector;

[0065] S5: nonlinearly map the browsing state vector to the interval (0, 1), and calculate the linear distance from 1;

[0066] S6: obtaining a standard value based on the index proportion, and comparing the linear distance with the standard value to obtain a comparison value;

[0067] S7: when the comparison value exceeds the preset value, one or more recurrent units are added to the long short-term memory network, and simulation data information is input to the newly added recurrent units and the simulation data information is continuously adjusted until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information;

[0068] S8: setting a revision duration according to the target simulation data information;

[0069] S9: adjusting the push weight of each category based on the target simulation data information, and recommending corresponding content to the customer within the revised time period according to the push weight of each category.

[0070] As described in step S1 above, it is detected whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period. The first preset time period is a preset time period, which is usually determined based on business needs and user behavior data. For example, it can be set that if the user has not placed an order within X hours after viewing the recommended content, it is considered that the user's interest in the category of content may have weakened or shifted.

[0071] As described in step S2 above, if the first preset time is exceeded, the user's index proportion for each category is obtained, and the data information of the user browsing the content of each category in multiple sub-time periods of the set time period is obtained. That is, after the time period is exceeded, it can be seen that the pushed content cannot satisfy the user. Therefore, the user's index proportion for each category can be obtained. The index can specifically be the amount of data or the browsing time. The method of obtaining can be obtained by the user's own input, or one or more index proportions can be formulated for the user according to the user's information, allowing the user to choose, and the data information of the user browsing the content of each category in multiple sub-time periods of the set time period is obtained. The specific method of obtaining the data information can be to obtain the information of the data browsed by the user and record it in the background. For example, when the user browses the content, a time information will be generated. At this time, the category information and time information of the browsed content can be obtained, and then statistics can be performed based on the category, so that the data information of the user browsing the content of each category in multiple sub-time periods of the set time period can be obtained. The set time is a preset time, such as the first preset time after the user inputs the search information, or it can be other customized time. In addition, the category is a predefined category, which can be the type, size, origin, price, etc. of the product.

[0072] As described in the above steps S3-S5, each of the data information is input one by one into each of the recurrent units in the long short-term memory network to obtain a browsing state; wherein the long short-term memory network includes a plurality of recurrent units, the number of the recurrent units is the same as the number of the data information, and a recurrent unit includes an output state, new information, an input gate, an output gate and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate and the output state of the previous recurrent unit; the browsing state is input into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector; the browsing state vector is nonlinearly mapped to the (0,1) interval, and the linear distance from 1 is calculated.

[0073] Specifically, the long short-term memory network includes multiple recurrent units, a softmax layer and a fully connected layer. The multiple recurrent units are connected in sequence until the last recurrent unit. The last recurrent unit is connected to the fully connected layer, and then the softmax layer is connected to the fully connected layer. A recurrent unit includes an output state, new information, an input gate, an output gate and a forget gate. The output state is respectively connected to the forget gate, the new information, the input gate, the output gate and the output state of the previous recurrent unit. The activation function of the new information layer is the tanh function, and the activation functions of the forget gate, the input gate and the output gate are the sigmoid function. After inputting the data information one by one, the corresponding browsing status can be obtained using the long short-term memory network. Due to the characteristics of the long short-term memory network, it can handle long-term dependencies and solve the gradient vanishing problem in the simple neural network model. By controlling the information flow through the gating mechanism, more complex dependencies can be captured. Therefore, the browsing status obtained can accurately summarize the user's browsing situation, and then it is input into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector. The fully connected layer can convert the input browsing state into a browsing state vector of the specified dimension, and then perform sigmoid mapping on it to the (0,1) interval, perform normalization processing, and calculate the linear distance between it and 1, specifically, subtracting the mapping value from 1.

[0074] As described in the above steps S6-S7, a standard value is obtained based on the index proportion, and the linear distance is compared with the standard value to obtain a comparison value. Among them, the standard value is obtained according to the index proportion, that is, the corresponding relationship table between the index proportion and the standard value can be pre-set, and then after the index proportion is obtained, the corresponding standard value can be obtained. The specific corresponding relationship is not limited and can be determined in advance by humans. It should be noted that the standard value needs to be determined in combination with the linear distance, that is, a determined index proportion needs to correspond to a standard linear distance, and the standard value is the linear distance of the standard. Then the linear distance is compared with the standard value to obtain a comparison value. When the comparison value exceeds the preset value, one or more recurrent units are added to the long short-term memory network, and the simulation data information is input into the newly added recurrent unit and the simulation data information is continuously adjusted until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information. Among them, the preset value is a pre-set value. When it exceeds the preset value, it is considered that the weights of each category of content recommended to the user are unreasonable. When it is less than or equal to the preset value, it is considered that the weights of each category of content recommended to the user are reasonable. At this time, one or more loop units are added thereto, and the newly added loop units input simulation data information and continuously adjust the simulation data information until the comparison value is less than or equal to the preset value. In fact, this is to simulate the data of the content category recommended to the user subsequently. When the comparison value is less than or equal to the preset value, the target simulation data information can be obtained.

[0075] As described in the above steps S8-S9, since there are more or less target simulation data information, it is necessary to determine the revised time according to the specific data volume, that is, the time given to users to browse, and then adjust the push weight of each category based on the target simulation data information, and recommend corresponding content to customers within the revised time according to the push weight of each category. This allows users to browse relevant content comprehensively, reduces the phenomenon of overfitting push, allows users to access diversified information, obtain information on different products, and improves users' shopping experience and satisfaction.

[0076] In one embodiment, the step S3 of inputting each piece of data information into each recurrent unit in the long short-term memory network one by one to obtain the browsing state includes:

[0077] S301: Input each of the data information into the forget gate, input gate, new information, and output gate of each recurrent unit in the long short-term memory network;

[0078] S302: According to formula F t =sigmoid(W xf |X t |+W hf H t-1 +bf )

[0079] I t =sigmoid(W xi |X t |+W hi H t-1 +b i )

[0080] O t =sigmoid(W xo |X t |+W ho H t-1 +b o )

[0081] N t =tanh(W xn |X t |+W hn H t-1 +b n )

[0082] C t =F t ⊙C t-1 +I t ⊙N t

[0083] Calculate the output state of each cycle unit in turn to obtain the browsing state of the last cycle unit output; where F t represents the tth forget gate, I t represents the tth input gate, N t represents the tth new information, O t represents the tth output gate, C t represents the tth output state, X t represents the tth data information, W xf , W hf 、b f , W xi , W hi 、b i , W xo , W ho 、b o , W xn , W hn 、b n These are preset parameters.

[0084] As described in the above steps S301-S302, the browsing status is obtained. Since each output state is a combination of the output state of the previous cycle unit and the newly added information, of course its weight is determined by the input gate and the forget gate, so that the current cycle unit can obtain information about the data of the previous cycle unit, so that the browsing status can accurately summarize the user's browsing situation.

[0085] In one embodiment, before the step S9 of recommending corresponding content to the customer within the revised time period according to the push proportion of each category, the method further includes:

[0086] S801: Obtaining preference information of each category of content according to the data information of each category of content;

[0087] S802: According to the preference information of each category of content, obtain content information related to the preference information to recommend to the user.

[0088] As described in the above steps S801-S802, even for the same category, users may have different preferences for recommended content. For example, a user may prefer a certain host or a certain brand. Therefore, the preference information of each category of content can be obtained from previous data information, and then the preference information of each category of content can be obtained to obtain content information related to the preference information for recommendation to the user.

[0089] In one embodiment, the step S2 of obtaining the weight of the user's indicators for each category includes:

[0090] S201: Obtaining the user's personal information and the user's search content;

[0091] S202: Setting the weight of the user's indicators for each category according to the search content and the personal information.

[0092] As described in the above steps S201-S202, the indicator weights are determined, and the user's personal information and the user's search content are obtained. The user's personal information may contain the user's preferences. For example, when buying snack products, the user may prefer spicy strips produced in Guizhou, but the user searches for spicy strips. Therefore, different spicy strips can have different indicator weights, and more weights can be set for spicy strips that users often buy.

[0093] In one embodiment, the indicator weight is a data volume weight or a browsing time weight.

[0094] In one embodiment, the indicator weight is the data volume weight, and after the step S9 of adjusting the push weight of each category based on the target simulation data information and recommending corresponding content to the customer within the revised time period according to the push weight of each category, the method further includes:

[0095] S1001: Determine whether the time period during which the user has not placed an order after the revised time period exceeds a second preset time period;

[0096] S1002: If the second preset time period is exceeded, then statistics are collected on the time information of the user browsing each category within the set revised time period;

[0097] S1003: Setting the duration proportion of each category according to the duration information;

[0098] S1004: setting a new data volume ratio according to the duration ratio;

[0099] S1005: Obtaining a new push weight based on the new data volume weight;

[0100] S1006: Recommending corresponding content to the customer according to the new push proportion.

[0101] As described in the above steps S1001-S1006, a secondary push of content is implemented. If the time period during which the user has not placed an order after the revised time period exceeds the second preset time period, it means that the user is still not satisfied with the recommendation. Therefore, the time information of the user browsing each category can be obtained, and the time proportion of each category can be set according to the time information, and the time proportion is used as the new data volume proportion. Then, a new push proportion is obtained based on the new data volume proportion. The method of obtaining the new push proportion based on the data volume proportion is the same as the method of the above steps S3-S9, which will not be repeated here. The corresponding content is recommended to the customer based on the new push proportion, so that on the premise of expanding the information for the user before, the expanded information is further fitted, so that the secondary recommended content can be closer to the user, thereby improving the user experience.

[0102] In one embodiment, before the step S1 of detecting whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period, the step further includes:

[0103] S001: Obtain the user's search information;

[0104] S002: Pushing relevant content to the user based on the search information.

[0105] As described in the above steps S001-S002, that is, after obtaining the user's search information, relevant content can be pushed to the user first, and the information of the pushed content can be a push method pre-set in the e-commerce platform.

[0106] Reference Figure 2 The present invention also provides a content recommendation device, comprising:

[0107] A detection module 10 is used to detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period;

[0108] The acquisition module 20 is used to acquire the weight of the user's indicators for each category if the first preset time period is exceeded, and to acquire data information of the user browsing the contents of each category within multiple sub-time periods of the set time period;

[0109] A first input module 30 is used to input each of the data information into each loop unit in the long short-term memory network in a one-to-one correspondence to obtain a browsing state; wherein the long short-term memory network includes a plurality of loop units, the number of the loop units is the same as the number of the data information, and a loop unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous loop unit;

[0110] A second input module 40 is used to input the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector;

[0111] A mapping module 50, used for non-linearly mapping the browsing state vector to the interval (0, 1) and calculating the linear distance from 1;

[0112] A comparison module 60, configured to obtain a standard value based on the index proportion, compare the linear distance with the standard value, and obtain a comparison value;

[0113] An adding module 70 is used to add one or more recurrent units to the long short-term memory network when the comparison value exceeds the preset value, input simulation data information to the newly added recurrent units, and continuously adjust the simulation data information until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information;

[0114] A setting module 80, configured to set a revision duration according to the target simulation data information;

[0115] The adjustment module 90 is used to adjust the push weight of each category based on the target simulation data information, and recommend corresponding content to the customer within the revised time period according to the push weight of each category.

[0116] In one embodiment, the first input module 30 includes:

[0117] A data information input submodule, used to input each of the data information into the forget gate, input gate, new information, and output gate of each recurrent unit in the long short-term memory network;

[0118] The output state calculation submodule is used to calculate the output state according to the formula F t =sigmoid(W xf |X t |+W hf H t-1 +b f )

[0119] I t =sigmoid(W xi |X t |+W hi H t-1 +b i )

[0120] O t =sigmoid(W xo |X t |+W ho H t-1 +b o )

[0121] N t =tanh(W xn |X t |+W hn H t-1 +b n )

[0122] C t =F t ⊙C t-1 +I t ⊙N t

[0123] Calculate the output state of each cycle unit in turn to obtain the browsing state of the last cycle unit output; where F t represents the tth forget gate, I t represents the tth input gate, N t represents the tth new information, O t represents the tth output gate, C t represents the tth output state, X t represents the tth data information, W xf , W hf 、b f , W xi , W hi 、b i , W xo , W ho 、bo , W xn , W hn 、b h These are preset parameters.

[0124] In one embodiment, the content recommendation device further includes:

[0125] A preference information acquisition module, used to acquire preference information of each category of content according to the data information of each category of content;

[0126] The content information acquisition module is used to acquire content information related to the preference information according to the preference information of each category of content, so as to recommend it to the user.

[0127] In one embodiment, the acquisition module 20 includes:

[0128] The search content acquisition submodule is used to obtain the user's personal information and the user's search content;

[0129] The indicator weight setting submodule is used to set the indicator weight of each category of the user according to the search content and the personal information.

[0130] In one embodiment, the indicator weight is a data volume weight or a browsing time weight.

[0131] In one embodiment, the indicator weight is the data volume weight, and the content recommendation device further includes:

[0132] A duration determination module, used to determine whether the time during which the user has not placed an order after the revised time exceeds a second preset time;

[0133] A duration information statistics module, for counting the duration information of the user browsing each category within the set revised duration if the second preset duration is exceeded;

[0134] A duration weight setting module, used to set the duration weight of each category according to the duration information;

[0135] A data volume weight setting module, used to set a new data volume weight according to the duration weight;

[0136] A new push weight acquisition module, used for acquiring a new push weight based on the new data volume weight;

[0137] The content recommendation module is used to recommend corresponding content to customers according to the new push proportion.

[0138] In one embodiment, the content recommendation device further includes:

[0139] A search information acquisition module is used to obtain the user's search information;

[0140] The related content push module is used to push related content to the user based on the search information.

[0141] The beneficial effects of the present invention are as follows: by obtaining data information and index proportions, and calculating the push proportions of each category based on the index proportions and data information, corresponding content is recommended to customers within the revised time period according to the push proportions of each category. This enables users to browse relevant content comprehensively, reduces the phenomenon of overfitting push, enables users to access diversified information, obtain information on different products, and improves users' shopping experience and satisfaction.

[0142] Reference Figure 3 In an embodiment of the present application, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a network interface and a database connected via a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The database of the computer device is used to store various data information, etc. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the content recommendation method described in any of the above embodiments can be implemented.

[0143] Those skilled in the art will understand that Figure 3 The structure shown in is merely a block diagram of a portion of the structure related to the present application solution and does not constitute a limitation on the computer device to which the present application solution is applied.

[0144] An embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the content recommendation method described in any of the above embodiments can be implemented.

[0145] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided in this application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0146] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, device, article or method including the element.

[0147] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0148] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0149] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of the claims of the present invention.

Claims

1. A content recommendation method, characterized in that: include: Detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period; If the first preset time period is exceeded, obtaining the weight of the user's indicators for each category, and obtaining data information of the user browsing the contents of each category in multiple sub-time periods of the set time period; Input each of the data information into each of the recurrent units in the long short-term memory network one by one to obtain a browsing state; wherein the long short-term memory network includes a plurality of recurrent units, the number of the recurrent units is the same as the number of the data information, and a recurrent unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous recurrent unit; Inputting the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector; Nonlinearly map the browsing state vector to the interval (0,1), and calculate the linear distance from 1; A standard value is obtained based on the index proportion, and the linear distance is compared with the standard value to obtain a comparison value; When the comparison value exceeds the preset value, one or more recurrent units are added to the long short-term memory network, and simulation data information is input into the newly added recurrent units and the simulation data information is continuously adjusted until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information; Setting a revision duration according to the target simulation data information; The push weight of each category is adjusted based on the target simulation data information, and corresponding content is recommended to the customer within the revised time period according to the push weight of each category.

2. The content recommendation method according to claim 1, characterized in that: The step of inputting each of the data information into each recurrent unit in the long short-term memory network one by one to obtain the browsing state includes: Input each of the data information into the forget gate, input gate, new information, and output gate of each recurrent unit in the long short-term memory network; According to formula F t =sigmoid(W xf |X t |+W hf H t-1 +b f ) I t =sigmoid(W xi |X t |+W hi H t-1 +b i ) O t =sigmoid(W xo |X t |+W ho H t-1 +b o ) N t =tanh(W xn |X t |+W hn H t-1 +b n ) C t =F t ⊙C t-1 +I t ⊙N t Calculate the output state of each cycle unit in turn to obtain the browsing state of the last cycle unit output; where F t represents the tth forget gate, I t represents the tth input gate, N t represents the tth new information, O t represents the tth output gate, C t represents the tth output state, X t represents the tth data information, W xf , W hf , b f , W xi , W hi , b i , W xo , W ho , b o , W xn , W hn , b n These are the preset parameters.

3. The content recommendation method according to claim 1, characterized in that: Before the step of recommending corresponding content to the customer within the revised time period according to the push proportion of each category, the method further includes: Obtaining preference information of each category of content according to the data information of each category of content; According to the preference information of each category of content, content information related to the preference information is obtained to recommend to the user.

4. The content recommendation method according to claim 1, characterized in that: The step of obtaining the weight of the user's indicators for each category includes: Obtain the user's personal information and the user's search content; The weight of the index of each category of the user is set according to the search content and the personal information.

5. The content recommendation method according to claim 1, characterized in that: The indicator proportion is the data volume proportion or the browsing time proportion.

6. The content recommendation method according to claim 1, characterized in that: The indicator weight is the data volume weight, and after the step of adjusting the push weight of each category based on the target simulation data information and recommending corresponding content to the customer within the revised time period according to the push weight of each category, the method further includes: Determining whether the time period during which the user has not placed an order after the revised time period exceeds a second preset time period; If the second preset time period is exceeded, then the time information of the user browsing each category within the set revised time period is counted; Setting the duration ratio of each category according to the duration information; Setting a new data volume ratio according to the duration ratio; Obtaining a new push weight based on the new data volume weight; Corresponding content is recommended to the customer according to the new push proportion.

7. The content recommendation method according to claim 1, characterized in that: Before the step of detecting whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period, the method further includes: Get the user's search information; Push relevant content to the user based on the search information.

8. A content recommendation device, characterized in that: include: A detection module, used to detect whether the time period during which the user has not placed an order after the content is pushed exceeds a first preset time period; An acquisition module, configured to acquire the weight of the user's indicators for each category if the first preset time period is exceeded, and acquire data information of the user browsing the contents of each category within a plurality of sub-time periods of a set time period; A first input module is used to input each of the data information into each recurrent unit in the long short-term memory network in a one-to-one correspondence to obtain a browsing state; wherein the long short-term memory network includes a plurality of recurrent units, the number of the recurrent units is the same as the number of the data information, and a recurrent unit includes an output state, new information, an input gate, an output gate, and a forget gate, and the output state is respectively connected to the forget gate, the new information, the input gate, the output gate, and the output state of the previous recurrent unit; A second input module is used to input the browsing state into the fully connected layer of the long short-term memory network to convert it into a specified dimension to obtain a browsing state vector; A mapping module, used for nonlinearly mapping the browsing state vector to the interval (0,1) and calculating the linear distance from 1; A comparison module, configured to obtain a standard value based on the index proportion, compare the linear distance with the standard value, and obtain a comparison value; An adding module is used for adding one or more recurrent units to the long short-term memory network when the comparison value exceeds the preset value, inputting simulation data information to the newly added recurrent units and continuously adjusting the simulation data information until the comparison value is less than or equal to the preset value, thereby obtaining the final adjusted target simulation data information; A setting module, used for setting a revision time according to the target simulation data information; The adjustment module is used to adjust the push weight of each category based on the target simulation data information, and recommend corresponding content to the customer within the revised time period according to the push weight of each category.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images