Drug information pushing method and device, equipment and medium

Abstract

The embodiment of the invention discloses a medicine information pushing method and device, equipment and a medium. The method comprises the following steps: acquiring historical interaction data between a current user and each candidate drug; inputting the historical interaction data into a trained drug prediction model, obtaining a click rate reference value and a conversion rate reference value by the drug prediction model according to the drug deviation characteristics, and obtaining a basic click rate and a basic conversion rate according to the drug sequence characteristics and the drug numerical value characteristics; outputting the medicine click rate and the medicine conversion rate of the corresponding candidate medicine according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate; and according to the drug click rate and the drug conversion rate, determining an information push score of the corresponding candidate drug, and according to the information push score of each candidate drug, pushing drug information corresponding to each candidate drug. And the accuracy of the pushed medicine information is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of data processing technology, and in particular to a method, device, equipment and medium for pushing drug information. Background Art

[0002] With the rapid development of digital medical technology, more and more users can order medicines on digital medical platforms. In particular, digital medical platforms will push drug information to users, and users can select the drugs they need based on the pushed drug information. Therefore, it is crucial to improve the accuracy of the pushed drug information. Summary of the invention

[0003] The present invention provides a drug information push method, device, equipment and medium to improve the accuracy of pushed drug information.

[0004] According to one aspect of the present invention, a method for pushing drug information is provided, comprising:

[0005] Acquire historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features, and drug deviation features;

[0006] Inputting the historical interaction data into a trained drug prediction model, the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation feature, and obtains a basic click rate and a basic conversion rate according to the drug sequence feature and the drug numerical feature;

[0007] Outputting the drug click rate and drug conversion rate of the corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate;

[0008] The information push scores of the corresponding candidate drugs are determined according to the drug click rate and the drug conversion rate, and the drug information corresponding to each candidate drug is pushed according to the information push scores of each candidate drug.

[0009] According to another aspect of the present invention, there is provided a drug information push device, comprising:

[0010] A historical data acquisition module is used to acquire historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features;

[0011] A model input module, used for inputting the historical interaction data into a trained drug prediction model, wherein the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation feature, and obtains a basic click rate and a basic conversion rate according to the drug sequence feature and the drug numerical feature;

[0012] A model output module, configured to output a drug click rate and a drug conversion rate of a corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate;

[0013] The information push module is used to determine the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate, and push the drug information corresponding to each candidate drug according to the information push score of each candidate drug.

[0014] According to another aspect of the present invention, there is provided an electronic device, comprising:

[0015] one or more processors;

[0016] A memory for storing one or more programs;

[0017] When one or more programs are executed by one or more processors, the one or more processors can execute any one of the drug information push methods provided in the embodiments of the present invention.

[0018] According to another aspect of the present invention, a computer-readable storage medium is provided, which stores computer instructions, and the computer instructions are used to enable a processor to implement any one of the drug information push methods provided in the embodiments of the present invention when executed.

[0019] An embodiment of the present invention provides a drug information push solution, which obtains historical interaction data between a current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features; the historical interaction data is input into a trained drug prediction model, the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation features, and obtains a basic click rate and a basic conversion rate according to the drug sequence features and the drug numerical features; according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate, the drug click rate and the drug conversion rate of the corresponding candidate drug are output; according to the drug click rate and the drug conversion rate, the information push score of the corresponding candidate drug is determined, and according to the information push score of each candidate drug, the drug information corresponding to each candidate drug is pushed. The above scheme inputs historical interaction data into the drug prediction model, outputs the drug click-through rate and drug conversion rate, and determines the information push score based on the drug click-through rate and drug conversion rate. Based on the information push score, the drug information is pushed to the current user, thereby improving the accuracy of the pushed drug information; at the same time, in the embodiment of the present invention, by introducing a click-through rate reference value and a conversion rate reference value, the basic click-through rate and the basic conversion rate are corrected respectively, thereby improving the accuracy of the drug click-through rate and the drug conversion rate output by the model; and the information push score in the embodiment of the present invention comprehensively considers the click-through rate and the conversion rate, thereby improving the accuracy of the determined information push score, so that the pushed drug information is more in line with the needs of the current user.

[0020] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0022] Figure 1 This is a flow chart of a method for pushing drug information provided by Embodiment 1 of the present invention;

[0023] Figure 2A This is a flow chart of a method for pushing drug information provided by Embodiment 2 of the present invention;

[0024] Figure 2B is a schematic diagram of a training process of a drug prediction model provided in Embodiment 2 of the present invention;

[0025] Figure 3 It is a structural schematic diagram of a drug information push device provided by Embodiment 3 of the present invention;

[0026] Figure 4 It is a structural schematic diagram of an electronic device for implementing a method for pushing drug information provided in a fourth embodiment of the present invention. DETAILED DESCRIPTION

[0027] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0028] The existing recommendation model architecture is mainly based on single-objective optimization, which indirectly increases the sales volume generated by the recommended products by optimizing the click-through rate. However, data shows that there is not a linear relationship between the click-through rate and the sales volume, and there are other factors that affect the increase in sales volume, such as the conversion rate of the product. Therefore, increasing the sales volume by optimizing the single goal of click-through rate often fails to achieve the optimal effect. Especially in the pharmaceutical field, due to the special properties of drugs, their recommendation effect is more significantly affected by different factors compared with ordinary products.

[0029] Embodiment 1

[0030] Figure 1 This is a flow chart of a drug information push method provided in Example 1 of the present invention. This embodiment can be applicable to situations where drug information is pushed to users. The method can be executed by a drug information push device, which can be implemented in software and / or hardware and can be configured in an electronic device that carries the drug information push function.

[0031] See also Figure 1 The drug information push method shown includes:

[0032] S110, obtaining historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features, and drug deviation features.

[0033] The current user refers to the user who needs to push drug information at the current moment. The candidate drug refers to the drug pushed to the current user. The embodiment of the present invention does not limit the method of obtaining the candidate drug, which can be set by the technician based on experience. For example, the candidate drug can be screened from the optional drugs according to the preset screening rules.

[0034] The embodiment of the present invention does not impose any limitation on the preset screening rules, which can be set by technicians based on experience. Optional drugs refer to drugs that can be pushed by the drug information push platform. The drug information push platform refers to a platform that pushes drug information to users.

[0035] It should be noted that each time before pushing drug information to a user, candidate drugs need to be re-screened from the optional drugs according to the preset screening rules. The preset screening rules can be updated periodically.

[0036] The historical interaction data refers to the interaction data between the current user and the candidate drug in the historical time period. The embodiment of the present invention does not impose any limitation on the length of the historical time period, which can be set by the technician based on experience.

[0037] The drug sequence feature refers to the sequence feature between the current user and the candidate drugs in the historical time period. For example, the candidate drugs can be sorted according to the number of times the current user clicks to view or places an order to purchase a candidate drug, and the drug sequence feature can be obtained.

[0038] The drug numerical feature refers to the numerical feature between the current user and the candidate drug in the historical time period. For example, the drug numerical feature may be the number of times the candidate drug is clicked and viewed by the current user in the historical time period.

[0039] Among them, the drug deviation feature refers to the deviation feature of the candidate drug within the historical time period. Exemplarily, the drug deviation feature may include the drug location information when the candidate drug is displayed to the current user, and the features of other candidate drugs adjacent to any candidate drug when the candidate drug is displayed. For example, the features of other candidate drugs adjacent to any candidate drug when the candidate drug is displayed may be the categories of other candidate drugs adjacent to the candidate drug. Drug location information refers to the location information of any candidate drug when it is displayed to the current user.

[0040] Specifically, based on preset screening rules, candidate drugs are screened from optional drugs; and historical interaction data between the current user and each candidate drug is obtained.

[0041] S120. Input the historical interaction data into the trained drug prediction model. The drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation characteristics, and obtains a basic click rate and a basic conversion rate according to the drug sequence characteristics and the drug numerical characteristics.

[0042] Among them, the drug prediction model can be used to predict the click-through rate and conversion rate of drug information. The click-through rate reference value can be used to calibrate the basic click-through rate. The basic click-through rate refers to the click-through rate of information for the initial prediction of the corresponding candidate drug based on the drug sequence characteristics and drug numerical characteristics.

[0043] The conversion rate reference value can be used to calibrate the basic conversion rate. The basic conversion rate refers to the information conversion rate of the initial prediction of the corresponding candidate drugs based on the drug sequence characteristics and drug numerical characteristics.

[0044] Specifically, the historical interaction data is input into the trained drug prediction model, so that the click rate deviation calibration network and the conversion rate deviation calibration network in the drug prediction model can obtain the click rate reference value and the conversion rate reference value respectively according to the drug deviation characteristics, and the self-attention layer, the normalization layer and the probability prediction network in the drug prediction model can obtain the basic click rate and the basic conversion rate according to the drug sequence characteristics and the drug numerical characteristics.

[0045] S130. Output the drug click rate and drug conversion rate of the corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate.

[0046] The drug click rate refers to the attractiveness of the drug information of the candidate drug predicted by the drug prediction model to the current user. The drug conversion rate refers to the possibility of the current user purchasing the corresponding candidate drug based on the drug information predicted by the drug prediction model.

[0047] Exemplarily, the click rate reference value is added to the basic click rate to obtain the drug click rate; the conversion rate reference value is added to the basic conversion rate to obtain the drug conversion rate; the drug prediction model outputs the drug click rate and the drug conversion rate.

[0048] S140. Determine the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate, and push the drug information corresponding to each candidate drug according to the information push score of each candidate drug.

[0049] Among them, the information push score can be used to quantify the push effect of drug information of candidate drugs.

[0050] In an optional embodiment, the information push score of the corresponding candidate drug is determined based on the drug click rate and the drug conversion rate, including: determining the click rate weight and conversion rate weight of the current user; determining the information push score of the corresponding candidate drug based on the click rate weight, the conversion rate weight, the drug click rate and the drug conversion rate.

[0051] Among them, the click rate weight can be used to quantify the importance of the drug click rate in the information push score. The conversion rate weight can be used to quantify the importance of the drug conversion rate in the information push score. The embodiment of the present invention does not impose any restrictions on the size of the click rate weight and the conversion rate weight, which can be set by technicians based on experience or needs. It should be noted that the click rate weight and the conversion rate weight can be the same or different, and the embodiment of the present invention does not specifically limit this.

[0052] It should be noted that different current users may correspond to different click rate weights and conversion rate weights; when pushing drug information to the same current user at different times, the click rate weights and conversion rate weights may also be different. The click rate weights and conversion rate weights can be updated as needed.

[0053] Specifically, the click rate weight and conversion rate weight preset by the current user are obtained; the drug click rate and drug conversion rate are weighted and summed according to the click rate weight and the conversion rate weight, respectively, to obtain the information push score of the corresponding candidate drug.

[0054] It can be understood that by introducing click-through rate weight and conversion rate weight, the information push score of the corresponding candidate drug is determined according to the click-through rate weight, conversion rate weight, drug click rate and drug conversion rate, thereby improving the accuracy of the determined information push score.

[0055] In an optional embodiment, drug information corresponding to each candidate drug is pushed according to the information push score of each candidate drug, including: determining the information push order according to the information push score; and pushing drug information of the corresponding candidate drug according to the information push order.

[0056] Among them, the information push order can be used to determine the push order of drug information.

[0057] Specifically, the candidate drugs are sorted from large to small according to the information push score to obtain the information push order, and the drug information of each candidate drug is pushed to the current user in turn according to the information push order.

[0058] It can be understood that by introducing the information push sequence, the drug information of the candidate drugs is pushed according to the information push sequence, thereby improving the accuracy of the pushed drug information.

[0059] An embodiment of the present invention provides a drug information push solution, which obtains historical interaction data between a current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features; the historical interaction data is input into a trained drug prediction model, the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation features, and obtains a basic click rate and a basic conversion rate according to the drug sequence features and the drug numerical features; according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate, the drug click rate and the drug conversion rate of the corresponding candidate drug are output; according to the drug click rate and the drug conversion rate, the information push score of the corresponding candidate drug is determined, and according to the information push score of each candidate drug, the drug information corresponding to each candidate drug is pushed. The above scheme inputs historical interaction data into the drug prediction model, outputs the drug click-through rate and drug conversion rate, and determines the information push score based on the drug click-through rate and drug conversion rate. Based on the information push score, the drug information is pushed to the current user, thereby improving the accuracy of the pushed drug information; at the same time, in the embodiment of the present invention, by introducing a click-through rate reference value and a conversion rate reference value, the basic click-through rate and the basic conversion rate are corrected respectively, thereby improving the accuracy of the drug click-through rate and the drug conversion rate output by the model; and the information push score in the embodiment of the present invention comprehensively considers the click-through rate and the conversion rate, thereby improving the accuracy of the determined information push score, so that the pushed drug information is more in line with the needs of the current user.

[0060] Embodiment 2

[0061] Figure 2A It is a flow chart of a drug information push method provided in the second embodiment of the present invention. Based on the above embodiments, this embodiment further adds "obtaining sample interaction data between sample users and each sample drug, and inputting the sample interaction data into the constructed drug prediction model to obtain the click rate deviation value, conversion rate deviation value, initial click rate and initial conversion rate; determining the predicted click rate output by the drug prediction model according to the click rate deviation value and the initial click rate, and determining the predicted conversion rate output by the drug prediction model according to the conversion rate deviation value and the initial conversion rate; determining the model loss according to the predicted click rate, predicted conversion rate, and the actual click rate and actual conversion rate corresponding to the corresponding sample drugs, and training the drug prediction model according to the model loss" to improve the training mechanism of the drug prediction model. It should be noted that for the parts not described in detail in the embodiments of the present invention, reference can be made to the descriptions of other embodiments.

[0062] See also Figure 2A The drug information push method shown includes:

[0063] S210, obtaining sample interaction data between sample users and each sample drug, and inputting the sample interaction data into the constructed drug prediction model to obtain a click rate deviation value, a conversion rate deviation value, an initial click rate, and an initial conversion rate.

[0064] The sample user may be a user who has previously pushed drug information. The sample drug refers to a drug that has been pushed to the sample user in the past. The embodiment of the present invention does not limit the method of obtaining the sample drug, which may be set by a technician based on experience. For example, the sample drug may be screened from the optional drugs according to a preset screening rule.

[0065] The sample interaction data in the embodiment of the present invention may be generated by an old drug prediction model.

[0066] The sample interaction data refers to the interaction data between the sample user and the sample drug within the sampling time period. The embodiment of the present invention does not impose any limitation on the sampling time period, which can be set by technicians based on experience.

[0067] The click rate deviation value can be used to calibrate the initial click rate. The conversion rate deviation value can be used to calibrate the initial conversion rate.

[0068] The initial click rate refers to the click rate of information when the drug prediction model makes the initial prediction for the corresponding sample drug. The initial conversion rate refers to the conversion rate of information when the drug prediction model makes the initial prediction for the corresponding sample drug.

[0069] In an optional embodiment, if the sample interaction data includes sample sequence features, sample numerical features and sample deviation features, the sample interaction data is input into the constructed drug prediction model to obtain a click rate deviation value, a conversion rate deviation value, an initial click rate and an initial conversion rate, including: inputting the sample deviation feature into the click rate deviation calibration network in the drug prediction model to obtain a click rate deviation value, and inputting the sample deviation feature into the conversion rate deviation calibration network in the drug prediction model to obtain a conversion rate deviation value; inputting the sample sequence feature into the self-attention layer in the drug prediction model to obtain a sequence key vector, and inputting the sample numerical feature into the normalization layer in the drug prediction model to obtain a numerical vector; according to the sequence key vector and the numerical vector, obtaining the initial click rate and the initial conversion rate.

[0070] Among them, the sample sequence feature refers to the sequence feature between the sample users and the sample drugs during the sampling period. The sample numerical feature refers to the numerical feature between the sample users and the sample drugs during the sampling period. The sample deviation feature refers to the deviation feature of the sample drugs during the historical period.

[0071] Among them, the click rate deviation calibration network can be used to determine the click rate deviation value based on the sample deviation characteristics. The conversion rate deviation calibration network can be used to determine the conversion rate deviation value based on the sample deviation characteristics. The click rate deviation calibration network and the conversion rate deviation calibration network can both be composed of linear neural networks, and the purpose is to calibrate the user's deviation in drug selection. It should be noted that the network structures of the click rate deviation calibration network and the conversion rate deviation calibration network can be the same, but the functions are different.

[0072] It should be noted that the connection between drugs is not necessarily established. For example, some drugs can form a prescription, but this prescription is not suitable for all users. If these drugs are displayed to all users at the same time, the overall push effect may decrease. Therefore, the click rate deviation calibration network and the conversion rate deviation calibration network are introduced. The click rate deviation calibration network and the conversion rate deviation calibration network can capture the complex relationship between drugs in the recommendation scenario, so as to make the predicted click rate and conversion rate more accurate.

[0073] Among them, the self-attention layer can be used to extract key features from the sample sequence features and perform vector conversion on the key features to obtain the sequence key vector. The sequence key vector refers to the vector obtained by vectorizing the key features in the sample sequence features.

[0074] The normalization layer can be used to normalize the sample numerical features and convert the normalized sample numerical features into numerical vectors. The numerical vector refers to a vector obtained by converting the sample numerical features into vectors.

[0075] It can be understood that by introducing the click rate deviation calibration network, the conversion rate deviation calibration network, the self-attention layer and the normalization layer, the click rate deviation value, the conversion rate deviation value, the initial click rate and the initial conversion rate are determined respectively, thereby achieving task concurrency and improving the efficiency of determining data; at the same time, sample sequence features, sample numerical features and sample deviation features are introduced to improve the comprehensiveness of sample interaction data.

[0076] In an optional embodiment, the initial click rate and the initial conversion rate are obtained based on the sequence key vector and the numerical vector, including: splicing the sequence key vector and the numerical vector to obtain a sample splicing vector; inputting the sample splicing vector into the probability prediction network in the drug prediction model to obtain the initial click rate and the initial conversion rate.

[0077] Among them, the sample splicing vector refers to the vector obtained by merging the sequence key vector and the numerical vector of the sample drug. The probability prediction network can also be called a multi-expert neural network, which can be used to predict the initial click rate and initial conversion rate based on the sample splicing vector. The embodiment of the present invention does not impose any limitation on the network structure of the probability prediction network, which can be set by technicians based on experience. Exemplarily, the probability prediction network can be composed of multiple linear neural networks.

[0078] It should be noted that the probabilistic prediction network can capture the relationship between click-through rate and conversion rate, making the predicted click-through rate and conversion rate more accurate, thereby indirectly achieving the goal of increasing sales.

[0079] It can be understood that by introducing the probability prediction network and inputting the sample concatenation vector into the probability prediction network, the initial click rate and the initial conversion rate are obtained, thereby improving the accuracy of the determined initial click rate and the initial conversion rate.

[0080] The drug prediction model in the embodiment of the present invention may include a click rate deviation calibration network, a conversion rate deviation calibration network, a self-attention layer, a normalization layer and a probability prediction network.

[0081] S220. Determine the predicted click rate output by the drug prediction model according to the click rate deviation value and the initial click rate, and determine the predicted conversion rate output by the drug prediction model according to the conversion rate deviation value and the initial conversion rate.

[0082] Among them, the predicted click rate refers to the attractiveness of the drug information of the sample drug predicted by the drug prediction model to the sample user. The predicted conversion rate refers to the possibility of the sample user predicting by the drug prediction model to purchase the corresponding sample drug based on the drug information.

[0083] Specifically, the click rate deviation value is added to the initial click rate to obtain the predicted click rate; the conversion rate deviation value is added to the initial conversion rate to obtain the predicted conversion rate; the drug prediction model outputs the predicted click rate and the predicted conversion rate.

[0084] S230: Determine the model loss based on the predicted click rate, the predicted conversion rate, and the actual click rate and actual conversion rate corresponding to the corresponding sample drugs, and train the drug prediction model based on the model loss.

[0085] Among them, the actual click rate refers to the actual click rate of the drug information corresponding to the sample drug during the sampling period. The actual conversion rate refers to the actual conversion rate of the drug information corresponding to the sample drug during the sampling period. The model loss refers to the model loss value of the drug prediction model. The embodiment of the present invention does not impose any limitation on the loss function used, which can be set by the technician based on experience. Exemplarily, a binary cross entropy loss function can be used.

[0086] In the embodiment of the present invention, if the model loss no longer changes, it is determined that the drug prediction model has completed training.

[0087] Specifically, the actual click rate and actual conversion rate of the sample drugs are obtained; the model loss is determined based on the predicted click rate, predicted conversion rate, actual click rate and actual conversion rate of the corresponding sample drugs; and the drug prediction model is trained based on the model loss.

[0088] S240. Obtain historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features, and drug deviation features.

[0089] S250. Input the historical interaction data into the trained drug prediction model. The drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation characteristics, and obtains a basic click rate and a basic conversion rate according to the drug sequence characteristics and the drug numerical characteristics.

[0090] S260. Output the drug click rate and drug conversion rate of the corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate.

[0091] S270. Determine the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate, and push the drug information corresponding to each candidate drug according to the information push score of each candidate drug.

[0092] The embodiment of the present invention provides a drug information push solution, which obtains sample interaction data between sample users and each sample drug by adding and obtaining the sample interaction data, and inputs the sample interaction data into the constructed drug prediction model to obtain the click rate deviation value, conversion rate deviation value, initial click rate and initial conversion rate; according to the click rate deviation value and the initial click rate, the predicted click rate output by the drug prediction model is determined, and according to the conversion rate deviation value and the initial conversion rate, the predicted conversion rate output by the drug prediction model is determined; according to the predicted click rate, predicted conversion rate, and the actual click rate and actual conversion rate corresponding to the corresponding sample drugs, the model loss is determined, and according to the model loss, the drug prediction model is trained, thereby improving the training mechanism of the drug prediction model. The above scheme determines the model loss by predicting the click rate, predicted conversion rate, and the actual click rate and actual conversion rate corresponding to the corresponding sample drugs, thereby improving the accuracy of the determined model loss, and thereby improving the accuracy of the model training.

[0093] Based on the above embodiments, the embodiments of the present invention provide a specific method for determining the predicted click rate and the predicted conversion rate.

[0094] For example, see Figure 2BSchematic diagram of the training process of the drug prediction model shown in FIG. The sample deviation features are respectively input into the click rate deviation calibration network and the conversion rate deviation calibration network to obtain the click rate deviation value and the conversion rate deviation value; the sample sequence features are input into the self-attention layer to obtain the sequence key vector; the sample numerical features are input into the normalization layer to obtain the numerical vector; the sequence key vector and the numerical vector are merged to obtain the sample splicing vector; the sample splicing vector is input into the probability prediction network to obtain the initial click rate and the initial conversion rate; the initial click rate is added to the click rate deviation value to obtain the predicted click rate output by the model; the initial conversion rate is added to the conversion rate deviation value to obtain the predicted conversion rate output by the model.

[0095] Embodiment 3

[0096] Figure 3 This is a schematic diagram of the structure of a drug information push device provided in Embodiment 3 of the present invention. This embodiment is applicable to the case of pushing drug information to users. The method can be executed by a drug information push device, which can be implemented in software and / or hardware and can be configured in an electronic device that carries the drug information push function.

[0097] like Figure 3 As shown, the device includes: a historical data acquisition module 310, a model input module 320, a model output module 330 and an information push module 340. Among them,

[0098] The historical data acquisition module 310 is used to acquire the historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features;

[0099] A model input module 320, for inputting the historical interaction data into a trained drug prediction model, wherein the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation feature, and obtains a basic click rate and a basic conversion rate according to the drug sequence feature and the drug numerical feature;

[0100] A model output module 330, configured to output a drug click rate and a drug conversion rate of a corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate;

[0101] The information push module 340 is used to determine the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate, and push the drug information corresponding to each candidate drug according to the information push score of each candidate drug.

[0102] An embodiment of the present invention provides a drug information push solution, which obtains historical interaction data between a current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features; the historical interaction data is input into a trained drug prediction model, the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation features, and obtains a basic click rate and a basic conversion rate according to the drug sequence features and the drug numerical features; according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate, the drug click rate and the drug conversion rate of the corresponding candidate drug are output; according to the drug click rate and the drug conversion rate, the information push score of the corresponding candidate drug is determined, and according to the information push score of each candidate drug, the drug information corresponding to each candidate drug is pushed. The above scheme inputs historical interaction data into the drug prediction model, outputs the drug click-through rate and drug conversion rate, and determines the information push score based on the drug click-through rate and drug conversion rate. Based on the information push score, the drug information is pushed to the current user, thereby improving the accuracy of the pushed drug information; at the same time, in the embodiment of the present invention, by introducing a click-through rate reference value and a conversion rate reference value, the basic click-through rate and the basic conversion rate are corrected respectively, thereby improving the accuracy of the drug click-through rate and the drug conversion rate output by the model; and the information push score in the embodiment of the present invention comprehensively considers the click-through rate and the conversion rate, thereby improving the accuracy of the determined information push score, so that the pushed drug information is more in line with the needs of the current user.

[0103] Optionally, the information push module 340 includes:

[0104] A weight determination unit, used to determine the click rate weight and conversion rate weight of the current user;

[0105] The push score determination unit is used to determine the information push score of the corresponding candidate drug according to the click rate weight, the conversion rate weight, the drug click rate and the drug conversion rate.

[0106] Optionally, the drug prediction model is trained based on the following means:

[0107] A sample data input module is used to obtain sample interaction data between sample users and each sample drug, and input the sample interaction data into the constructed drug prediction model to obtain a click rate deviation value, a conversion rate deviation value, an initial click rate, and an initial conversion rate;

[0108] A sample data output module, used to determine the predicted click rate output by the drug prediction model according to the click rate deviation value and the initial click rate, and to determine the predicted conversion rate output by the drug prediction model according to the conversion rate deviation value and the initial conversion rate;

[0109] The model training module is used to determine the model loss based on the predicted click-through rate, the predicted conversion rate, and the actual click-through rate and actual conversion rate corresponding to the corresponding sample drugs, and train the drug prediction model based on the model loss.

[0110] Optionally, if the sample interaction data includes sample sequence features, sample numerical features and sample deviation features, the sample data input module includes:

[0111] A deviation value determination unit, used for inputting the sample deviation feature into a click rate deviation calibration network in the drug prediction model to obtain a click rate deviation value, and inputting the sample deviation feature into a conversion rate deviation calibration network in the drug prediction model to obtain a conversion rate deviation value;

[0112] A vector determination unit, used for inputting the sample sequence features into the self-attention layer in the drug prediction model to obtain a sequence key vector, and inputting the sample numerical features into the normalization layer in the drug prediction model to obtain a numerical vector;

[0113] The initial rate determination unit is used to obtain the initial click rate and the initial conversion rate according to the sequence key vector and the numerical vector.

[0114] Optionally, the initial rate determination unit is specifically configured to:

[0115] Concatenate the sequence key vector and the numerical vector to obtain a sample concatenation vector;

[0116] The sample concatenation vector is input into the probability prediction network in the drug prediction model to obtain the initial click rate and the initial conversion rate.

[0117] Optionally, the information push module 340 includes:

[0118] A push order determination unit, used to determine the information push order according to the information push score;

[0119] The drug information pushing unit is used to push the drug information of the corresponding candidate drugs according to the information pushing order.

[0120] The drug information push device provided in the embodiment of the present invention can execute the drug information push method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects for executing each drug information push method.

[0121] In the technical solution of the present invention, the collection, storage, use, processing, transmission, provision and disclosure of historical interaction data and sample interaction data involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0122] Embodiment 4

[0123] Figure 4 It is a structural diagram of an electronic device for implementing a method for pushing drug information provided by Embodiment 4 of the present invention. Electronic device 410 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0124] like Figure 4 As shown, the electronic device 410 includes at least one processor 411, and a memory connected to the at least one processor 411 in communication, such as a read-only memory (ROM) 412, a random access memory (RAM) 413, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 411 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 412 or the computer program loaded from the storage unit 418 to the random access memory (RAM) 413. In the RAM 413, various programs and data required for the operation of the electronic device 410 can also be stored. The processor 411, the ROM 412, and the RAM 413 are connected to each other via a bus 414. An input / output (I / O) interface 415 is also connected to the bus 414.

[0125] Multiple components in the electronic device 410 are connected to the I / O interface 415, including: an input unit 416, such as a keyboard, a mouse, etc.; an output unit 417, such as various types of displays, speakers, etc.; a storage unit 418, such as a disk, an optical disk, etc.; and a communication unit 419, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 419 allows the electronic device 410 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0126] The processor 411 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 411 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 411 executes the various methods and processes described above, such as a drug information push method.

[0127] In some embodiments, the drug information push method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 418. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 410 via the ROM 412 and / or the communication unit 419. When the computer program is loaded into the RAM 413 and executed by the processor 411, one or more steps of the drug information push method described above may be performed. Alternatively, in other embodiments, the processor 411 may be configured to execute the drug information push method in any other appropriate manner (e.g., by means of firmware).

[0128] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0129] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0130] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0131] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0132] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0133] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0134] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0135] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for pushing drug information, characterized in that: include: Acquire historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features, and drug deviation features; Inputting the historical interaction data into a trained drug prediction model, the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation feature, and obtains a basic click rate and a basic conversion rate according to the drug sequence feature and the drug numerical feature; Outputting the drug click rate and drug conversion rate of the corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate; The information push scores of the corresponding candidate drugs are determined according to the drug click rate and the drug conversion rate, and the drug information corresponding to each candidate drug is pushed according to the information push scores of each candidate drug.

2. The method according to claim 1, characterized in that Determining the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate includes: Determine the click rate weight and conversion rate weight of the current user; An information push score for a corresponding candidate drug is determined according to the click-through rate weight, the conversion rate weight, the drug click-through rate, and the drug conversion rate.

3. The method according to claim 1, characterized in that The drug prediction model is trained based on the following method: Obtain sample interaction data between sample users and each sample drug, and input the sample interaction data into the constructed drug prediction model to obtain a click rate deviation value, a conversion rate deviation value, an initial click rate, and an initial conversion rate; Determining the predicted click rate output by the drug prediction model according to the click rate deviation value and the initial click rate, and determining the predicted conversion rate output by the drug prediction model according to the conversion rate deviation value and the initial conversion rate; The model loss is determined based on the predicted click-through rate, the predicted conversion rate, and the actual click-through rate and actual conversion rate corresponding to the corresponding sample drugs, and the drug prediction model is trained based on the model loss.

4. The method according to claim 3, characterized in that If the sample interaction data includes sample sequence features, sample numerical features and sample deviation features, the sample interaction data is input into the constructed drug prediction model to obtain a click rate deviation value, a conversion rate deviation value, an initial click rate and an initial conversion rate, including: Inputting the sample deviation feature into the click rate deviation calibration network in the drug prediction model to obtain a click rate deviation value, and inputting the sample deviation feature into the conversion rate deviation calibration network in the drug prediction model to obtain a conversion rate deviation value; Inputting the sample sequence features into the self-attention layer in the drug prediction model to obtain a sequence key vector, and inputting the sample numerical features into the normalization layer in the drug prediction model to obtain a numerical vector; An initial click rate and an initial conversion rate are obtained according to the sequence key vector and the numerical vector.

5. The method according to claim 4, characterized in that The obtaining of the initial click rate and the initial conversion rate according to the sequence key vector and the numerical vector includes: Concatenate the sequence key vector and the numerical vector to obtain a sample concatenation vector; The sample concatenation vector is input into the probability prediction network in the drug prediction model to obtain the initial click rate and initial conversion rate.

6. The method according to any one of claims 1 to 5, characterized in that The pushing of drug information corresponding to each candidate drug according to the information push score of each candidate drug includes: Determining the order of information push according to the information push score; According to the information pushing order, the drug information of the corresponding candidate drugs is pushed.

7. A drug information push device, characterized in that: include: A historical data acquisition module is used to acquire historical interaction data between the current user and each candidate drug; wherein the historical interaction data includes drug sequence features, drug numerical features and drug deviation features; A model input module, used for inputting the historical interaction data into a trained drug prediction model, wherein the drug prediction model obtains a click rate reference value and a conversion rate reference value according to the drug deviation feature, and obtains a basic click rate and a basic conversion rate according to the drug sequence feature and the drug numerical feature; A model output module, configured to output a drug click rate and a drug conversion rate of a corresponding candidate drug according to the click rate reference value, the conversion rate reference value, the basic click rate and the basic conversion rate; The information push module is used to determine the information push score of the corresponding candidate drug according to the drug click rate and the drug conversion rate, and push the drug information corresponding to each candidate drug according to the information push score of each candidate drug.

8. The device according to claim 7, characterized in that The information push module includes: A weight determination unit, used to determine the click rate weight and conversion rate weight of the current user; The push score determination unit is used to determine the information push score of the corresponding candidate drug according to the click rate weight, the conversion rate weight, the drug click rate and the drug conversion rate.

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

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, a drug information push method as described in any one of claims 1-6 is implemented.

Images