CN113327133B

CN113327133B - Data recommendation method, data recommendation device, electronic equipment and readable storage medium

Info

Publication number: CN113327133B
Application number: CN202110663421.2A
Authority: CN
Inventors: 宋云蛟
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2021-06-15
Filing date: 2021-06-15
Publication date: 2024-06-21
Anticipated expiration: 2041-06-15
Also published as: CN113327133A

Abstract

The disclosure discloses a data recommendation method, a data recommendation device, electronic equipment and a readable storage medium, and relates to the field of big data, in particular to the field of recommendation. The specific implementation scheme is as follows: responding to a request aiming at abnormal recommendation, analyzing the request to obtain feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data; determining one or more replacement feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data; processing each piece of replacement feature data in the plurality of pieces of replacement feature data by using an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data; target feature dimension data that is considered to cause abnormal recommendation is determined from the plurality of feature dimension data based on the first predicted recommended result and the second predicted recommended result.

Description

Data recommendation method, data recommendation device, electronic equipment and readable storage medium

Technical Field

The disclosure relates to the technical field of big data, in particular to the field of recommendation.

Background

The presence and popularization of the Internet brings a large amount of information to users, and meets the requirements of the users on the information. There may be a problem in that it is difficult to obtain information of interest to the user from the user when the user is faced with a large amount of information.

For this reason, a recommendation system that can recommend information of interest to a user based on the characteristics of interest and behavior habits of the user, etc., has been developed. For example, a recommendation system for recommending advertisements, which recommends the appropriate advertisements to the user. However, in the recommendation process, there may be a case where the recommendation result does not conform to expectations.

Disclosure of Invention

The present disclosure provides a data recommendation method, a data recommendation device, an electronic apparatus, and a readable storage medium.

According to an aspect of the present disclosure, there is provided a data recommendation method including: responding to a request for abnormal recommendation, analyzing the request to obtain feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data; determining one or more alternative feature dimension data corresponding to each of the plurality of feature dimension data; processing each piece of replacement feature data in a plurality of pieces of replacement feature data by using an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data; and determining target feature dimension data considered to cause the abnormal recommendation from the plurality of feature dimension data based on the first predicted recommended result and the second predicted recommended result.

According to another aspect of the present disclosure, there is provided a feature recommendation apparatus including: the analysis module is used for responding to the request for abnormal recommendation, analyzing the request and obtaining feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data; a first determining module, configured to determine one or more alternative feature dimension data corresponding to each of the plurality of feature dimension data; the obtaining module is used for respectively processing each piece of replacement feature data in a plurality of pieces of replacement feature data by utilizing an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data; and a second determining module configured to determine target feature dimension data considered to cause the abnormal recommendation from the plurality of feature dimension data, based on the first predicted recommended result and the second predicted recommended result.

According to another aspect of the present disclosure, there is provided an electronic device including: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor to enable the at least one processor to perform the method as described above.

According to another aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method as described above.

According to another aspect of the present disclosure, there is provided a computer program product comprising a computer program which, when executed by a processor, implements a method as described above.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 schematically illustrates an exemplary system architecture to which data recommendation methods and apparatus may be applied, according to embodiments of the present disclosure;

FIG. 2 schematically illustrates a flow chart of a data recommendation method according to an embodiment of the present disclosure;

FIG. 3 schematically illustrates a flow chart of determining one or more alternative feature dimension data corresponding to each feature dimension data of a plurality of feature dimension data, according to an embodiment of the disclosure;

FIG. 4 schematically illustrates a flow chart of determining target feature dimension data from a plurality of feature dimension data that is believed to result in an abnormal recommendation based on a first predicted recommendation and a second predicted recommendation in accordance with an embodiment of the present disclosure;

FIG. 5 schematically illustrates a schematic diagram of a data recommendation process according to an embodiment of the present disclosure;

FIG. 6 schematically illustrates a block diagram of a data recommendation device, according to an embodiment of the present disclosure; and

Fig. 7 shows a block diagram of an electronic device that may be adapted for use with the data recommendation method according to an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

In the recommendation system, recommendation can be realized by using a recommendation model obtained by training the deep neural network model by using a training sample. Because the deep neural network model has stronger generalization capability, the recommendation model can be realized in a mode of updating the model by continuously updating the training sample under the condition of determining the model parameters, and has better prediction effect.

In the process of realizing the disclosed concept, the situation that the recommendation result does not accord with the expectation is found, and the situation affects the use experience of the user and may cause the waste of resources. In order to effectively ensure the use experience of users and reduce the waste of resources, the method can be realized by improving the prediction accuracy of the recommendation model. In order to improve the prediction accuracy of the recommendation model, the cause of the unexpected recommendation result may be analyzed to improve the recommendation model based on the analyzed cause.

The reasons for the unexpected recommended results may be analyzed by randomly replacing the value of each plaintext field, processing the replaced plaintext field using an offline recommendation model to obtain a predicted recommended result corresponding to the replaced plaintext field, comparing the predicted recommended result corresponding to the replaced plaintext field with the predicted recommended result corresponding to the original plaintext field, and determining a target plaintext field deemed to result in the unexpected recommended result based on the comparison result. Each plaintext field may understand user data associated with a recommendation that can be intuitively understood.

In the process of implementing the disclosed concept, it is found that, because the utilization mode of the feature dimension data of the recommendation model is complex, for example, a plurality of plaintext fields correspond to one feature dimension data or one plaintext field corresponds to a plurality of feature dimension data, a more accurate determination of a cause of the unexpected recommendation result is difficult to be implemented by using the plaintext field replacement mode. In addition, since the recommendation system actually uses the online recommendation model instead of the offline recommendation model, there is a case that the offline recommendation model and the online recommendation model are inconsistent, for example, some parameters that may have a large influence on the offline recommendation model may not have a large influence on the online recommendation model. Thus, it is difficult to more accurately determine the cause of the unexpected recommendation using the offline recommendation model.

For this purpose, a solution is proposed for implementing the determination of the cause of the unexpected recommendation result based on the predicted recommendation result obtained by processing the data of the feature level using the online recommendation model. That is, the embodiments of the present disclosure provide a data recommendation method, a data recommendation apparatus, an electronic device, a non-transitory computer-readable storage medium storing computer instructions, and a computer program product. The data recommendation method comprises the following steps: and responding to the request for abnormal recommendation, analyzing the request to obtain feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data, determining one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data, respectively processing each piece of replacement feature data in the plurality of pieces of replacement feature data by using an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data, and determining target feature dimension data which is considered to cause abnormal recommendation from the plurality of pieces of feature dimension data according to the first prediction recommendation result and the second prediction recommendation result.

Fig. 1 schematically illustrates an exemplary system architecture 100 in which data recommendation methods and apparatus may be applied, according to embodiments of the present disclosure.

It should be noted that fig. 1 is only an example of a system architecture to which embodiments of the present disclosure may be applied to assist those skilled in the art in understanding the technical content of the present disclosure, but does not mean that embodiments of the present disclosure may not be used in other devices, systems, environments, or scenarios. For example, in another embodiment, the exemplary system architecture 100 to which the data recommendation method and apparatus may be applied may include a terminal device, but the terminal device may implement the data recommendation method and apparatus provided by the embodiments of the present disclosure without interaction with a server.

As shown in fig. 1, a system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and/or wireless communication links, and the like.

The user may interact with the server 105 via the network 104 using the terminal devices 101, 102, 103 to receive or send messages or the like. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as a knowledge reading class application, a web browser application, a search class application, an instant messaging tool, a mailbox client and/or social platform software, etc. (as examples only).

The terminal devices 101, 102, 103 may be a variety of electronic devices having a display screen and supporting web browsing, including but not limited to smartphones, tablets, laptop and desktop computers, and the like.

The server 105 may be a server providing various services, such as a background management server (by way of example only) providing support for content browsed by the user using the terminal devices 101, 102, 103. The background management server may analyze and process the received data such as the user request, and feed back the processing result (e.g., the web page, information, or data obtained or generated according to the user request) to the terminal device.

It should be noted that, the data recommendation method provided by the embodiments of the present disclosure may be generally performed by the terminal device 101, 102, or 103. Accordingly, the data recommendation device provided by the embodiment of the present disclosure may also be provided in the terminal device 101, 102, or 103.

Or the data recommendation method provided by the embodiments of the present disclosure may be generally performed by the server 105. Accordingly, the data recommendation device provided in the embodiments of the present disclosure may be generally disposed in the server 105. The data recommendation method provided by the embodiments of the present disclosure may also be performed by a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105. Accordingly, the data recommendation device provided by the embodiments of the present disclosure may also be provided in a server or a server cluster that is different from the server 105 and is capable of communicating with the terminal devices 101, 102, 103 and/or the server 105.

For example, the server 105 parses the request in response to the request for the abnormal recommendation, obtains feature data and a first predicted recommended result corresponding to the feature data, determines one or more pieces of replacement feature dimension data corresponding to each of the plurality of pieces of feature dimension data, processes each piece of replacement feature data of the plurality of pieces of replacement feature data with the online recommendation model, respectively, obtains a second predicted recommended result corresponding to each piece of replacement feature data, and determines target feature dimension data considered to cause the abnormal recommendation from the plurality of pieces of feature dimension data according to the first predicted recommended result and the second predicted recommended result. Or by a server or cluster of servers capable of communicating with the terminal devices 101, 102, 103 and/or server 105, and ultimately achieve the determination of target feature dimension data that is believed to result in abnormal recommendations.

It should be understood that the number of terminal devices, networks and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.

Fig. 2 schematically illustrates a flow chart of a data recommendation method 200 according to an embodiment of the present disclosure.

As shown in fig. 2, the method includes operations S210 to S240.

In operation S210, in response to the request for abnormal recommendation, the request is parsed to obtain feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data includes a plurality of feature dimension data.

In operation S220, one or more replacement feature dimension data corresponding to each of the plurality of feature dimension data is determined.

In operation S230, each of the plurality of replacement feature data obtained based on the feature data and the plurality of replacement feature dimension data is processed by using the online recommendation model, respectively, to obtain a second prediction recommendation result corresponding to each of the replacement feature data.

In operation S240, target feature dimension data considered to cause abnormal recommendation is determined from the plurality of feature dimension data according to the first predicted recommended result and the second predicted recommended result.

According to embodiments of the present disclosure, an abnormal recommendation may refer to a recommendation in which the recommendation result does not conform to an expected recommendation, i.e., a recommendation in which the recommendation result and the expected recommendation result do not conform. The feature data may be data obtained by extracting features of user data related to the recommendation. The feature data may comprise at least two feature dimension data, i.e. the dimensions of the feature data may comprise at least two, the feature data of each dimension may be referred to as feature dimension data. Each feature dimension data may have one or more alternative feature dimension data corresponding to the feature dimension data. The feature dimension data and the replacement feature dimension data corresponding to the feature dimension data differ in value, i.e., the feature dimension data and the replacement feature dimension data corresponding to the feature dimension data characterize the same feature dimension, but differ in value.

According to the embodiment of the disclosure, for each feature dimension data, the replacement feature data corresponding to the feature dimension data can be obtained according to the feature data corresponding to the feature dimension data and the replacement feature dimension data.

According to an embodiment of the present disclosure, the first predicted recommended result corresponding to the feature data may be a result of processing the feature data using an online recommendation model, that is, the first predicted recommended result is a predicted recommended result obtained using the online recommendation model.

According to an embodiment of the present disclosure, the reason for processing the feature data to obtain the first predicted recommended result using the online recommendation model is that: in the recommendation system, the user who inputs the cost compares the intentional conversion effect, but because the conversion data is less, the generalization capability of the online recommendation model is insufficient to be fully born, so in order to solve the above problem, the prediction recommendation result obtained by using the online recommendation model can be calibrated by using the feedback coefficient, and therefore, the prediction recommendation result may be obtained after the original prediction recommendation result is calibrated by using the feedback coefficient. The feedback coefficient may be understood as being determined from posterior data relating to the effect of the transformation.

And since the prediction recommendation result obtained by the online recommendation model needs to be analyzed for determining the reasons for the non-expected recommendation result, if the prediction recommendation result is not the prediction recommendation result obtained by the online recommendation model but the prediction recommendation result obtained by calibrating the prediction recommendation result obtained by the online recommendation model, the determined reasons for the non-expected recommendation result may not be the same as the actual situation due to the addition of other factors. Therefore, in order to ensure that the determined reasons for the unexpected recommended results are accurate as much as possible, the first predicted recommended results corresponding to the feature data may be predicted recommended results obtained by processing the feature data using the online recommendation model instead of calibrating the first predicted recommended results obtained using the online recommendation model.

According to embodiments of the present disclosure, an online recommendation model may be understood as a recommendation model that is actually utilized by a recommendation system. The online recommendation model may be used to make object recommendations to a user. The online recommendation model may be obtained by training the deep neural network model with training samples.

According to an embodiment of the present disclosure, in case an abnormal recommendation is detected, a request for the abnormal recommendation may be generated, which may include feature data and a first predicted recommendation result corresponding to the feature data. Generating the request for the exception recommendation may include: and acquiring data related to the abnormal recommendation from the target file, loading the data related to the abnormal recommendation into a memory, and processing the data related to the abnormal recommendation to obtain a request for the abnormal recommendation with a target format, wherein the target format can comprise Proto Buffer. In the case that a request for abnormal recommendation is acquired, the request may be parsed in response to the request to obtain feature data and a first predicted recommendation result corresponding to the feature data. The above-described retrieval of requests for exception recommendations may be understood as being implemented by an online request module. The means for generating and sending the request for the exception recommendation may send the request for the exception recommendation to the online request module via a remote procedure call.

According to embodiments of the present disclosure, since it is more difficult for an online request module to fulfill a request to receive a feature level, the online request module may be modified so that it can support receiving a request of a feature level. In an embodiment of the present disclosure, the online request module for executing the request for exception recommendation is an online request module capable of receiving a request of a feature level.

According to the embodiment of the disclosure, in order to ensure that the determined reasons for the unexpected recommendation result are accurate as much as possible, the consistency of the online request module and the online recommendation model needs to be ensured as much as possible, and in order to ensure the consistency of the online request module and the online recommendation model as much as possible, the codes of the online request module and the online recommendation model can be subjected to homologous management.

In addition, the switch module may be further used to control whether to activate the online request module capable of receiving the request of the feature level, that is, if the switch state of the switch module is in the on state, it may be explained that the online request module capable of receiving the request of the feature level is activated. If the switch state of the switch model is in the off state, it may be stated that the on-line request module switch module that is capable of receiving the request for the feature level may include a switch module in the on state and in the off state to use the switch control model module to determine whether the request for the feature level may be received.

According to an embodiment of the present disclosure, after parsing the request to obtain feature data, one or more substitute dimension feature data corresponding to each feature dimension data included in the feature data may be determined, and each substitute feature data may be determined according to the feature data and each substitute feature dimension data corresponding to each feature dimension data, that is, each of the plurality of substitute feature data is obtained based on the feature data and each substitute feature dimension data corresponding to each feature dimension data. That is, for each feature dimension data corresponding to the feature data, the replacement feature dimension data is replaced with the feature dimension data in the feature data for each replacement feature dimension data corresponding to the feature dimension data, and is combined with other feature dimension data than the feature dimension data in the feature data as one replacement feature data corresponding to the feature data.

According to an embodiment of the present disclosure, determining one or more alternative feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data may include: one or more replacement feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data is determined using a random replacement method.

Alternatively, a training sample corresponding to the historical online recommendation model is determined, and the training sample corresponding to the historical online recommendation model is processed to obtain one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data.

Alternatively, a training sample corresponding to the online recommendation model is determined, and the training sample corresponding to the online recommendation model is processed to obtain one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data.

According to an embodiment of the present disclosure, the online recommendation model described above may be understood as an online recommendation model that generates abnormal recommendations. The online recommendation model can be updated according to actual business requirements, so that multiple versions of online recommendation models can be generated, and the historically utilized online recommendation model can be called a historical online recommendation model.

According to an embodiment of the present disclosure, determining one or more replacement feature dimension data corresponding to each feature dimension data of a plurality of feature dimension data using a random replacement method may include: for each feature dimension data in the feature dimension data, determining each value corresponding to the feature dimension data, and randomly selecting one or more values from the value set to serve as one piece of replacement feature dimension data corresponding to the feature dimension data.

According to an embodiment of the present disclosure, processing a training sample corresponding to a historical online recommendation model to obtain one or more alternative feature dimension data corresponding to each feature dimension data of a plurality of feature dimension data may include: and processing the training samples corresponding to the historical online recommendation model to obtain values of each feature dimension data in the historical training samples, and processing each value to obtain each piece of replacement feature dimension data corresponding to each piece of feature dimension data. Processing each value to obtain each piece of replacement feature dimension data corresponding to each piece of feature dimension data, which may include: and combining the values to obtain one or more combined values, and taking each combined value as each piece of replacement feature dimension data corresponding to each piece of feature dimension data. Alternatively, each value is taken as each piece of replacement feature dimension data corresponding to each piece of feature dimension data. The historical training samples may be understood as training samples corresponding to the historical online recommendation model.

According to the embodiment of the disclosure, after each piece of replacement feature data corresponding to the feature data is obtained, each piece of replacement feature data may be processed by using the online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, that is, each piece of replacement feature data may be input into the online recommendation model to obtain the second prediction recommendation result corresponding to each piece of replacement feature data. Because the online recommendation model is used for processing the replacement characteristic data, the online recommendation model is actually used by a recommendation system, and therefore the accuracy of the prediction recommendation result can be effectively ensured.

According to the embodiment of the disclosure, after the first prediction recommendation result corresponding to the feature data and the second prediction recommendation result corresponding to each piece of replacement feature data are obtained, the first prediction recommendation result may be compared with each of the plurality of second prediction recommendation results, respectively, to obtain a comparison result, and the target feature dimension data may be determined from the plurality of feature dimension data according to the comparison result. The target feature dimension data may be understood as feature dimension data that is considered to lead to abnormal recommendations. The target feature dimension data may include one or more.

According to an embodiment of the present disclosure, comparing the first predicted recommended result with each of the plurality of second predicted recommended results may include: and determining a prediction recommendation corresponding to each of the replacement feature data sets according to the second prediction recommendation corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets. And comparing the first predicted recommended result with the predicted recommended result corresponding to each replacement characteristic data set to obtain a comparison result.

According to an embodiment of the present disclosure, determining the predicted recommended result corresponding to each of the replacement feature data sets according to the second predicted recommended result corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets may include: and determining the maximum value, wherein the maximum value is the maximum value of the prediction recommendation result corresponding to each piece of replacement characteristic data included in each piece of replacement characteristic data set, and the maximum value comprises the maximum value or the minimum value. The most value is determined as a predicted recommendation corresponding to each of the replacement feature data sets. Alternatively, an average value is determined, wherein the average value is an average value of predicted recommended results corresponding to respective replacement feature data included in each of the replacement feature data sets. The average value is determined as a predicted recommendation corresponding to each of the replacement feature data sets.

According to the embodiment of the disclosure, since the target feature dimension data causing abnormal recommendation is determined, the target feature dimension data can be considered as the data to be recommended, so that the direction in which the online recommendation model can be optimized is determined according to the target feature dimension data, and the probability that the recommendation result does not meet the expectations is effectively reduced.

It should be noted that, in the technical solution of the embodiment of the disclosure, the related acquisition, storage, application, etc. of the feature data and the replacement feature dimension data all conform to the rules of related laws and regulations, and necessary security measures are adopted without violating the public order colloquial.

According to the embodiment of the disclosure, the request is analyzed to obtain feature data and a first prediction recommendation result corresponding to the feature data, one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data are determined, each piece of replacement feature data in the plurality of pieces of replacement feature data is processed by using an online recommendation model respectively to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, and target feature dimension data which is considered to cause abnormal recommendation is determined from the plurality of pieces of feature dimension data according to the first prediction recommendation result and the second prediction recommendation result. Since the replacement feature data is obtained based on the feature data and each of the replacement feature dimension data corresponding to each of the feature dimension data, the identification of the feature level is realized. On the basis, the online recommendation model is used for processing the replacement feature data to obtain a second prediction recommendation result corresponding to the replacement feature data, and the online recommendation model is actually used by a recommendation system, so that the accuracy of the prediction recommendation result can be effectively ensured, further, the fact that target feature dimension data which is considered to cause abnormal recommendation is determined to be more accurate from a plurality of feature dimension data according to the first prediction recommendation result corresponding to the feature data and each second prediction recommendation result can be effectively ensured, and therefore, the technical problem that reasons for unexpected recommendation results are difficult to determine more accurately is at least partially overcome, the direction in which the online recommendation model can be optimized can be determined according to the target feature dimension data, and the probability that the recommendation results are inconsistent with expectations is reduced.

According to an embodiment of the present disclosure, the above data recommendation method may further include the following operations.

Generating a request for abnormal recommendation under the condition that the first prediction recommendation result is detected to meet the preset conditions, wherein the preset conditions comprise one of the following: the first predicted recommended result is greater than or equal to a first threshold. The first predicted recommended result is less than or equal to a second threshold, wherein the second threshold is less than or equal to the first threshold. The first difference between the first predicted recommended result and the historical recommended result is greater than or equal to a first difference threshold.

According to embodiments of the present disclosure, the preset condition may be used as a basis for determining whether the recommendation is an abnormal recommendation. The values of the first threshold, the second threshold, and the first difference threshold may be configured according to actual service requirements, which are not limited herein.

According to the embodiments of the present disclosure, after obtaining the first predicted recommended result corresponding to the feature data, it may be determined whether the first predicted recommended result satisfies the prediction condition, and if it is determined that the first predicted recommended result satisfies the preset condition, it may be explained that the recommendation for generating the first predicted recommended result is an abnormal recommendation. If it is determined that the first predicted recommended result does not satisfy the preset condition, it may be explained that the recommendation for generating the first predicted recommended result is a non-abnormal recommendation. If it is determined that the recommendation that generated the first predicted recommendation result is an abnormal recommendation, a request for the abnormal recommendation may be generated.

According to an embodiment of the present disclosure, in a case where it is detected that the first predicted recommended result satisfies the preset condition, generating the request for the abnormal recommendation may include the following operations.

And generating a request for abnormal recommendation in a preset time period under the condition that the first prediction recommendation result is detected to meet the preset condition.

According to embodiments of the present disclosure, in the case of an online recommendation system, there is a high demand for efficiency in determining the cause of the unexpected recommendation. In order to improve the efficiency of determining the reason for the unexpected recommended result, when the first predicted recommended result is detected to meet the preset condition, a request for abnormal recommendation is generated in time, so that the reason for abnormal recommendation can be analyzed in time. For example, a request for an abnormal recommendation may be generated within a preset period of time. The value of the preset time period can be configured according to the actual service requirement, and is not limited herein.

According to the embodiment of the disclosure, by generating the request for abnormal recommendation in the preset time period under the condition that the first prediction recommendation result is detected to meet the preset condition, the reason for the abnormal recommendation can be analyzed in time, so that the recommendation meeting the expectations can be provided for the user, and the use experience of the user is improved.

According to an embodiment of the present disclosure, in response to a request for abnormal recommendation, parsing the request to obtain feature data and a first prediction recommendation result corresponding to the feature data may include the following operations.

And analyzing the request in the expected analysis time period of the received request to obtain the characteristic data and the first prediction recommendation result corresponding to the characteristic data.

According to the embodiment of the disclosure, in order to improve the efficiency of determining the reasons for the reasons which do not meet the expected recommendation results, the analysis of the request is completed within the expected analysis time period when the request is received, so that the reasons for abnormal recommendation can be analyzed in time.

According to an embodiment of the present disclosure, the obtaining the second predicted recommended result is achieved through multithreading by processing each of the plurality of replacement feature data separately using an online recommendation model.

In accordance with embodiments of the present disclosure, to increase the efficiency of determining the cause of the unexpected recommended result, a multi-threaded implementation may be utilized, i.e., multiple replacement feature data may be processed by at least two threads. Furthermore, different threads may be deployed on the same electronic device or on different electronic devices.

According to an embodiment of the present disclosure, the request further includes an online model version identification.

Operation S230 may include the following operations.

And respectively processing the plurality of replacement feature data by using an online recommendation model corresponding to the online model version identifier to obtain a second prediction recommendation result corresponding to each replacement feature data.

According to the embodiment of the present disclosure, since the online recommendation model may be updated according to actual business requirements, there may be multiple versions of the online recommendation model, and thus, there may be a case where the online recommendation model currently being utilized is inconsistent with the online recommendation model that causes recommendation results not to conform to expectations. In order to ensure that the determined reasons for the unexpected recommendation result are accurate as much as possible, each of the plurality of replacement feature data needs to be processed by using the online recommendation model that causes the recommendation result to be unexpected.

To enable processing of each of the plurality of replacement feature data with an online recommendation model that results in the recommendation not conforming to expectations, each version of the online recommendation model may be stored for subsequent utilization. The online recommendation model may be characterized with an online model version identification. On this basis, generating the request for the exception recommendation may also include online model version identification. The online model version identification included in the request may be an online model version identification of an online recommendation model that generated the abnormal recommendation.

According to the embodiment of the disclosure, an online recommendation model corresponding to the online model version identifier can be determined, namely, an online recommendation model generating abnormal recommendation is determined, each piece of replacement feature data in the plurality of pieces of replacement feature data is respectively processed by using the online recommendation model generating abnormal recommendation, and a second prediction recommendation result corresponding to each piece of replacement feature data is obtained.

The method shown in fig. 2 is further described below with reference to fig. 3-5 in conjunction with the exemplary embodiment.

Fig. 3 schematically illustrates a flow chart of determining one or more alternative feature dimension data 300 corresponding to each of a plurality of feature dimension data, according to an embodiment of the disclosure.

As shown in fig. 3, the method includes operations S321 to S322.

In operation S321, a training sample corresponding to the online recommendation model is determined.

In operation S322, the training samples corresponding to the online recommendation model are processed to obtain one or more alternative feature dimension data corresponding to each of the plurality of feature dimension data.

According to embodiments of the present disclosure, a training sample corresponding to an online recommendation model may be understood as a training sample for training an online recommendation model that yields abnormal recommendations.

According to the embodiment of the disclosure, a training sample corresponding to an online recommendation model is processed to obtain values of each feature dimension data in the training sample, and each value is processed to obtain each piece of replacement feature dimension data corresponding to each piece of feature dimension data. Processing each value to obtain each piece of replacement feature dimension data corresponding to each piece of feature dimension data, which may include: and combining the values to obtain one or more combined values, and taking each combined value as each piece of replacement feature dimension data corresponding to each piece of feature dimension data. Alternatively, each value is taken as each piece of replacement feature dimension data corresponding to each piece of feature dimension data.

According to the embodiment of the disclosure, the training samples for training the online recommendation model are utilized to obtain one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data, so that the obtained replacement feature dimension data is meaningful, complete and time-efficient, and therefore higher-quality data support can be provided for ensuring the determined reasons which cause the unexpected recommendation results as much as possible.

According to an embodiment of the present disclosure, in order to improve efficiency in determining a cause of the unexpected recommended result, the replacement feature dimension data may be generated in advance and stored in the cache.

Fig. 4 schematically illustrates a flow chart of determining target feature dimension data 400 from a plurality of feature dimension data that is believed to result in an abnormal recommendation according to a first predicted recommended result and a second predicted recommended result in accordance with an embodiment of the present disclosure.

As shown in fig. 4, the method includes operations S441 to S442.

In operation S441, a third predicted recommended result corresponding to each of the replacement feature data sets is determined according to the second predicted recommended result corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets, wherein each of the feature dimension data corresponds to one of the replacement feature dimension data sets.

In operation S442, target feature dimension data that is considered to cause abnormal recommendation is determined from the plurality of feature dimension data according to the first predicted recommended result and each of the third predicted recommended results.

According to embodiments of the present disclosure, for each feature dimension data, one alternative feature dimension data set corresponding to the feature dimension data may be determined, each alternative feature dimension data set may include one or more alternative feature dimension data.

According to the embodiment of the disclosure, the first prediction result may be compared with each third prediction result respectively to obtain a comparison result, and target feature dimension data considered to cause abnormal recommendation may be determined from the plurality of feature dimension data according to the comparison result.

According to an embodiment of the present disclosure, operation S441 may include the following operations.

An average value is determined, wherein the average value is an average value of predicted recommended results corresponding to the respective replacement feature data included in each of the replacement feature data sets. The average value is determined as a third predicted recommendation corresponding to each of the replacement feature data sets.

According to the embodiment of the disclosure, for each replacement feature data set, the prediction recommendation results corresponding to the replacement feature data set are added to obtain a sum value, and a ratio of the sum value to the number of the replacement feature data included in the replacement feature data set is determined, wherein the ratio is an average value. The average value is determined as a third predicted recommendation corresponding to the replacement feature dataset.

According to an embodiment of the present disclosure, determining target feature dimension data that is considered to cause abnormal recommendation from among a plurality of feature dimension data according to a first predicted recommended result and each third predicted recommended result may include the following operations.

A second difference between the first predicted recommendation and each of the third predicted recommendations is determined. Target feature dimension data that is considered to cause abnormal recommendation is determined from the plurality of feature dimension data based on the plurality of second differences.

According to an embodiment of the present disclosure, a second difference between the first predicted recommended result and each of the third predicted recommended results may be determined, resulting in a plurality of second differences. Determining target feature dimension data that is considered to result in an abnormal recommendation from the plurality of feature dimension data based on the plurality of second differences may include: target feature dimension data that is considered to result in an abnormal recommendation may be determined from the plurality of feature dimension data based on the second difference threshold and the plurality of second differences. Alternatively, the plurality of second differences are ranked to obtain a ranking result, and target feature dimension data which is considered to cause abnormal recommendation is determined from the plurality of feature dimension data according to the ranking result.

According to an embodiment of the present disclosure, determining target feature dimension data that is considered to cause abnormal recommendation from among a plurality of feature dimension data according to a plurality of second differences may include the following operations.

For each feature dimension data, in a case where it is determined that the second difference value corresponding to the feature dimension data is greater than or equal to the second difference threshold value, the feature dimension data is determined as target feature dimension data that is considered to cause abnormal recommendation.

According to an embodiment of the present disclosure, the second difference threshold may be used as one of the basis for determining the target feature dimension data from the plurality of feature dimension data. The value of the second difference threshold may be configured according to the actual service requirement, which is not limited herein.

And sequencing the plurality of second differences to obtain a sequencing result. And determining target feature dimension data which is considered to cause abnormal recommendation from the feature dimension data according to the sorting result.

According to the embodiment of the disclosure, the plurality of second differences are ranked to obtain a ranking result, and target feature dimension data which is considered to cause abnormal recommendation is determined from the plurality of feature dimension data according to the ranking result. The ordering may include ordering in order of the second difference from small to large or ordering in order of the second difference from large to small.

Fig. 5 schematically illustrates a schematic diagram of a data recommendation process 500 according to an embodiment of the disclosure.

As shown in fig. 5, in the case where it is detected that the first predicted-recommended result 503 satisfies the preset condition, a request 501 for abnormal recommendation is generated. In response to the request for abnormal recommendation 501, the request for abnormal recommendation 501 is parsed, resulting in feature data 502 and a first predicted recommendation result 503 corresponding to the feature data 502. The feature data 503 may include a plurality of feature dimension data.

One or more alternative feature dimension data corresponding to each feature dimension data is determined. And determining each piece of replacement feature data according to the feature data and each piece of replacement feature dimension data corresponding to each piece of feature dimension data. From the one or more replacement feature data corresponding to each feature dimension data, a replacement feature data set 504 corresponding to each feature dimension data is derived.

Each piece of replacement feature data included in each set of replacement feature data 504 is processed using the online recommendation model to obtain a second predicted recommendation result 505 corresponding to each piece of replacement feature data.

The first predicted recommended result 503 and each second predicted recommended result 505 are compared to obtain a comparison result, and target feature dimension data 507 which is considered to cause abnormal recommendation is determined from the plurality of feature dimension data according to the comparison result.

Fig. 6 schematically illustrates a block diagram of a data recommendation device 600 according to an embodiment of the disclosure.

As shown in fig. 6, the data recommendation apparatus 600 may include a parsing module 610, a first determining module 620, an obtaining module 630, and a second determining module 640.

The parsing module 610 is configured to parse the request in response to the request for abnormal recommendation, to obtain feature data and a first prediction recommendation result corresponding to the feature data, where the feature data includes a plurality of feature dimension data.

The first determining module 620 is configured to determine one or more alternative feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data.

And an obtaining module 630, configured to process each piece of replacement feature data in the plurality of pieces of replacement feature data by using the online recommendation model, to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, where the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data.

The second determining module 640 is configured to determine target feature dimension data that is considered to cause abnormal recommendation from the plurality of feature dimension data according to the first predicted recommended result and the second predicted recommended result.

According to an embodiment of the present disclosure, the first determination module 620 may include a first determination sub-module and a second processing sub-module.

And the first determining submodule is used for determining training samples corresponding to the online recommendation model.

And the first processing sub-module is used for processing the training samples corresponding to the online recommendation model to obtain one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data.

The data recommendation device 600 may further include a generating module according to an embodiment of the present disclosure.

The generating module is used for generating a request for abnormal recommendation under the condition that the first prediction recommendation result is detected to meet the preset condition, wherein the preset condition comprises one of the following: the predicted recommended result is greater than or equal to a first threshold. The first predicted recommended result is less than or equal to a second threshold, wherein the second threshold is less than or equal to the first threshold. The first difference between the first predicted recommended result and the historical recommended result is greater than or equal to a first difference threshold.

According to an embodiment of the present disclosure, the generation module may include a generation sub-module.

And the generation sub-module is used for generating a request for abnormal recommendation in a preset time period under the condition that the first prediction recommendation result is detected to meet the preset condition.

The acquisition module 630 may include an acquisition sub-module.

And the obtaining submodule is used for respectively processing the plurality of replacement characteristic data by utilizing the online recommendation model corresponding to the online model version identifier to obtain a second prediction recommendation result corresponding to each replacement characteristic data.

According to an embodiment of the present disclosure, each of the plurality of replacement feature data is processed separately using an online recommendation model, and obtaining the second predicted recommendation result corresponding to each of the replacement feature data is achieved through multithreading.

According to an embodiment of the present disclosure, the second determination module 640 may include a second determination sub-module and a third determination sub-module.

And the second determining submodule is used for determining a third prediction recommendation result corresponding to each alternative feature data set according to the second prediction recommendation result corresponding to each alternative feature data included in each alternative feature dimension data set, wherein each feature dimension data corresponds to one alternative feature dimension data set.

And a third determination sub-module for determining target feature dimension data considered to cause abnormal recommendation from the plurality of feature dimension data according to the first predicted recommended result and each third predicted recommended result.

According to an embodiment of the present disclosure, the second determination sub-module may include a first determination unit and a second determination unit.

And a first determining unit configured to determine an average value, wherein the average value is an average value of prediction recommendation results corresponding to the respective replacement feature data included in each of the replacement feature data sets.

And a second determining unit configured to determine an average value as a third prediction recommendation result corresponding to each of the replacement feature data sets.

According to an embodiment of the present disclosure, the third determination sub-module may include a third determination unit and a fourth determination unit.

And a third determining unit for determining a second difference between the first predicted recommended result and each of the third predicted recommended results.

And a fourth determining unit configured to determine target feature dimension data considered to cause abnormal recommendation from the plurality of feature dimension data based on the plurality of second differences.

According to an embodiment of the present disclosure, the fourth determination unit may include a first determination subunit.

A first determination subunit configured to determine, for each feature dimension data, the feature dimension data as target feature dimension data that is considered to cause abnormal recommendation, in a case where it is determined that a second difference value corresponding to the feature dimension data is greater than or equal to a second difference threshold.

According to an embodiment of the present disclosure, the fourth determination unit may include a sorting subunit and a second determination subunit.

And the sequencing subunit is used for sequencing the second differences to obtain a sequencing result.

And a second determination subunit configured to determine, from the plurality of feature dimension data, target feature dimension data that is considered to cause abnormal recommendation, based on the ranking result.

According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

According to an embodiment of the present disclosure, an electronic device includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executable by the at least one processor to enable the at least one processor to perform the method as described above.

According to an embodiment of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform the method as described above.

According to an embodiment of the present disclosure, a computer program product comprising a computer program which, when executed by a processor, implements a method as described above.

Fig. 7 illustrates a block diagram of an electronic device 700 that may be suitable for use with the data recommendation method, according to an embodiment of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 7, the apparatus 700 includes a computing unit 701 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 702 or a computer program loaded from a storage unit 708 into a Random Access Memory (RAM) 703. In the RAM703, various programs and data required for the operation of the device 700 may also be stored. The computing unit 701, the ROM702, and the RAM703 are connected to each other through a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.

Various components in device 700 are connected to I/O interface 705, including: an input unit 706 such as a keyboard, a mouse, etc.; an output unit 707 such as various types of displays, speakers, and the like; a storage unit 708 such as a magnetic disk, an optical disk, or the like; and a communication unit 709 such as a network card, modem, wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information/data with other devices via a computer network, such as the internet, and/or various telecommunication networks.

The computing unit 701 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 701 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the respective methods and processes described above, such as a data recommendation method. For example, in some embodiments, the data recommendation method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 700 via ROM702 and/or communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the data recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 701 may be configured to perform the XXX method by any other suitable means (e.g. by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, 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 disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, 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, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. A data recommendation method, comprising:

Responding to a request for abnormal recommendation, analyzing the request to obtain feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data;

Determining one or more replacement feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data;

Processing each piece of replacement feature data in a plurality of pieces of replacement feature data by using an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data; and

Determining a third prediction recommendation corresponding to each of the replacement feature data sets according to a second prediction recommendation corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets, wherein each of the feature dimension data corresponds to one of the replacement feature dimension data sets;

determining target feature dimension data considered to cause the abnormal recommendation from the plurality of feature dimension data according to the comparison result of the first prediction recommendation result and each of the third prediction recommendation results;

wherein the determining one or more alternative feature dimension data corresponding to each feature dimension data of the plurality of feature dimension data comprises:

Determining a training sample corresponding to the online recommendation model, wherein the training sample is used for training to obtain the online recommendation model generating abnormal recommendation; and

Processing a training sample corresponding to the online recommendation model to obtain a value of each feature dimension data in the training sample;

combining the values to obtain one or more combined values;

And taking each combined value as each piece of replacement feature dimension data corresponding to each piece of feature dimension data, and obtaining one or more pieces of replacement feature dimension data corresponding to each piece of feature dimension data in the plurality of pieces of feature dimension data.

2. The method of claim 1, further comprising:

generating the request for abnormal recommendation under the condition that the first prediction recommendation result is detected to meet a preset condition,

Wherein the preset condition includes one of the following:

the first predicted recommended result is greater than or equal to a first threshold;

the first predicted recommended result is less than or equal to a second threshold, wherein the second threshold is less than or equal to the first threshold;

the first difference between the first predicted recommended result and the historical recommended result is greater than or equal to a first difference threshold.

3. The method of claim 2, wherein the generating the request for abnormal recommendation if the first predicted recommended result is detected to satisfy a preset condition comprises:

and under the condition that the first prediction recommendation result is detected to meet a preset condition, generating the request for abnormal recommendation in a preset time period.

4. The method of any of claims 1-3, wherein the request further comprises an online model version identification;

the processing each piece of replacement feature data in the plurality of pieces of replacement feature data by using the online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data includes:

and respectively processing the plurality of replacement feature data by using an online recommendation model corresponding to the online model version identifier to obtain a second prediction recommendation result corresponding to each piece of replacement feature data.

5. The method according to any one of claims 1 to 3, wherein the processing each of the plurality of replacement feature data by the online recommendation model to obtain the second prediction recommendation result corresponding to each of the replacement feature data is implemented through multithreading.

6. The method of claim 1, wherein the determining a third predicted recommendation corresponding to each of the replacement feature data sets based on the second predicted recommendation corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets comprises:

Determining an average value, wherein the average value is an average value of prediction recommendation results corresponding to the replacement feature data included in each replacement feature data set; and

And determining the average value as a third prediction recommendation corresponding to each of the replacement feature data sets.

7. The method of claim 1 or 6, wherein said determining target feature dimension data from said plurality of feature dimension data that is considered to result in said abnormal recommendation based on said first predicted recommendation and each of said third predicted recommendations comprises:

Determining a second difference between the first predicted recommended result and each of the third predicted recommended results; and

And determining target feature dimension data which is considered to cause the abnormal recommendation from the feature dimension data according to the second difference values.

8. The method of claim 7, wherein the determining target feature dimension data from the plurality of feature dimension data that is considered to result in the anomaly recommendation based on the plurality of the second differences comprises:

For each of the feature dimension data, in a case where it is determined that a second difference value corresponding to the feature dimension data is greater than or equal to a second difference threshold value, the feature dimension data is determined as target feature dimension data considered to cause the abnormal recommendation.

9. The method of claim 7, wherein the determining target feature dimension data from the plurality of feature dimension data that is considered to result in the anomaly recommendation based on the plurality of the second differences comprises:

sequencing the second difference values to obtain a sequencing result; and

And determining target feature dimension data which is considered to cause the abnormal recommendation from the feature dimension data according to the sorting result.

10. A feature recommendation device, comprising:

The analysis module is used for responding to a request for abnormal recommendation, analyzing the request and obtaining feature data and a first prediction recommendation result corresponding to the feature data, wherein the feature data comprises a plurality of feature dimension data;

a first determining module configured to determine one or more alternative feature dimension data corresponding to each of the plurality of feature dimension data;

The obtaining module is used for respectively processing each piece of replacement feature data in a plurality of pieces of replacement feature data by utilizing an online recommendation model to obtain a second prediction recommendation result corresponding to each piece of replacement feature data, wherein the plurality of pieces of replacement feature data are obtained based on the feature data and the plurality of pieces of replacement feature dimension data; and

A second determining sub-module, configured to determine a third prediction recommendation corresponding to each of the replacement feature data sets according to a second prediction recommendation corresponding to each of the replacement feature data included in each of the replacement feature dimension data sets, where each of the feature dimension data corresponds to one of the replacement feature dimension data sets;

a third determination sub-module for determining target feature dimension data considered to cause the abnormal recommendation from the plurality of feature dimension data based on a comparison result of the first predicted recommended result and each of the third predicted recommended results;

Wherein the first determining module includes:

The first determining submodule is used for determining a training sample corresponding to the online recommendation model, wherein the training sample is used for training to obtain the online recommendation model generating abnormal recommendation; and

The first processing sub-module is used for processing training samples corresponding to the online recommendation model to obtain values of each feature dimension data in the training samples;

combining the values to obtain one or more combined values;

11. The apparatus of claim 10, further comprising:

A generation module for generating the request for abnormal recommendation under the condition that the first prediction recommendation result is detected to meet the preset condition,

Wherein the preset condition includes one of the following:

the predicted recommended result is greater than or equal to a first threshold;

12. The apparatus of claim 11, wherein the generating module comprises:

And the generation sub-module is used for generating the request for abnormal recommendation in a preset time period under the condition that the first prediction recommendation result is detected to meet a preset condition.

13. The apparatus of any of claims 10-12, wherein the request further comprises an online model version identification;

The obtaining module comprises:

And the obtaining submodule is used for respectively processing the plurality of replacement feature data by utilizing the online recommendation model corresponding to the online model version identifier to obtain a second prediction recommendation result corresponding to each piece of replacement feature data.

14. The apparatus of claim 10, wherein the processing each of the plurality of replacement feature data with the online recommendation model to obtain the second predicted recommendation corresponding to each of the replacement feature data is performed by multithreading.

15. An electronic device, comprising:

At least one processor; and

A memory communicatively coupled to the at least one processor; wherein,

The memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

16. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-9.

17. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any one of claims 1 to 9.