Information matching model training method and information matching method
The training method for an information matching model using deep learning addresses the challenge of complex account-device mappings by improving the accuracy and reliability of user identification through conversion rate-based matching.
Patent Information
- Application Number
- CN202310654042.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-05
- Publication Date
- 2024-12-06
AI Technical Summary
Existing technologies face challenges in accurately and reliably determining the mapping relationship between user account information and device information due to complex and non-simple mappings, especially when multiple accounts are logged in on multiple devices or a single device.
A training method for an information matching model using deep learning to determine the mapping relationship between account and device information based on feature information, utilizing conversion rates to measure matching accuracy and reliability.
Improves the accuracy and reliability of matching between account and device information by determining the mapping relationship using a trained information matching model, enhancing the precision of user identification in e-commerce scenarios.
Smart Images

Figure CN119089208A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the field of information processing technology, and in particular to a training method for an information matching model and an information matching method. Background Art
[0002] At present, there are a variety of identity documents (ID) for identifying users in e-commerce scenarios, such as user business account information, where the user business account can be a personal identification number (PIN), and user device information, where the user device information can be an International Mobile Equipment Identity (IMEI). Account information and device information are usually a mapping relationship generated when an account is logged in to a certain device. For example, an account has only been logged in on one device, an account may have been logged in on multiple devices, and multiple accounts may have been logged in on the same device. Therefore, the mapping relationship between an account and a device is relatively complicated and is not a simple meaning mapping. Therefore, how to more accurately and reliably determine the mapping relationship between an account and a device has become a problem that needs to be solved urgently. Summary of the invention
[0003] The present disclosure provides a training method for an information matching model, an information matching method, an apparatus, an electronic device, a storage medium and a computer program product.
[0004] The first aspect of the present disclosure proposes a method for training an information matching model, the method comprising: obtaining sample login data, wherein the sample login data comprises sample account information and sample device information having a login relationship; obtaining feature information of the sample login data; and training an information matching model according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine a mapping relationship between account information and device information.
[0005] In an embodiment of the present disclosure, by obtaining sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship, and obtaining feature information of the sample login data, an information matching model is trained according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine a mapping relationship between account information and device information. The present disclosure is based on a deep learning method to train an information matching model, and the mapping relationship between account information and device information can be determined based on the trained target information matching model. At the same time, the conversion rate is used to measure the degree of matching between account information and device information, thereby improving the accuracy and reliability of the matching between account information and device information.
[0006] An embodiment of the second aspect of the present disclosure proposes an information matching method, the method comprising: obtaining an information set from an account information set to be processed and a device information set to be processed as the information set to be processed; inputting the information set to be processed into a target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set; wherein the target information matching model is a model trained using the method described in any one of the first aspects.
[0007] In an embodiment of the present disclosure, by obtaining an information set from the account information set to be processed and the device information set to be processed as the information set to be processed, the information set to be processed is input into the target information matching model to obtain a target information set that matches the information set to be processed, the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set. Based on the target information matching model, the present disclosure can match the corresponding device information set for the account information set to be processed and the corresponding account information set for the device information set to be processed, thereby improving the accuracy and reliability of the matching between account information and device information.
[0008] The third aspect embodiment of the present disclosure proposes a training device for an information matching model, comprising: a first acquisition module, used to acquire sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship; a second acquisition module, used to acquire feature information of the sample login data; a training module, used to train the information matching model according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine the mapping relationship between account information and device information.
[0009] The fourth aspect of the present disclosure proposes an information matching device, including: an acquisition module, used to acquire one of the account information set to be processed and the device information set to be processed, as the information set to be processed; a matching module, used to input the information set to be processed into a target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set; wherein the target information matching model is a model trained using the method described in the first aspect.
[0010] The fifth aspect embodiment of the present disclosure proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the training method of the information matching model of the first aspect embodiment or the information matching method of the second aspect embodiment as described above.
[0011] The sixth aspect embodiment of the present disclosure proposes a computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the training method of the information matching model of the first aspect embodiment or the information matching method of the second aspect embodiment as described above.
[0012] The seventh aspect embodiment of the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, implements the training method of the information matching model of the first aspect embodiment of the present disclosure or the information matching method of the second aspect embodiment.
[0013] Additional aspects and advantages of the present disclosure will be given in part in the following description and in part will be obvious from the following description or learned through practice of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] The above and / or additional aspects and advantages of the present disclosure will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:
[0015] Figure 1 A flowchart of a method for training an information matching model provided in one embodiment of the present disclosure;
[0016] Figure 2 A flowchart of a method for training an information matching model provided by another embodiment of the present disclosure;
[0017] Figure 3 A flowchart of a method for training an information matching model provided by another embodiment of the present disclosure;
[0018] Figure 4 A flowchart of a method for training an information matching model provided by another embodiment of the present disclosure;
[0019] Figure 5 A flowchart of an information matching method provided by an embodiment of the present disclosure;
[0020] Figure 6 A flowchart of an information matching method provided by another embodiment of the present disclosure;
[0021] Figure 7 A schematic diagram of the structure of a training device for an information matching model provided in one embodiment of the present disclosure;
[0022] Figure 8 A schematic diagram of the structure of an information matching device provided in one embodiment of the present disclosure;
[0023] Fig. 9 A block diagram of an electronic device provided according to an embodiment of the present disclosure. DETAILED DESCRIPTION
[0024] Embodiments of the present disclosure are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present disclosure, and should not be construed as limiting the present disclosure.
[0025] The following describes the information matching model training method, device, electronic device and storage medium of the embodiments of the present disclosure with reference to the accompanying drawings.
[0026] Figure 1 A flowchart of a method for training an information matching model provided in accordance with an embodiment of the present disclosure.
[0027] like Figure 1 As shown, the method comprises the following steps:
[0028] S101, obtaining sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship.
[0029] It should be noted that when training the information matching model, for a large number of users and devices, in order to reduce the amount of model calculations, the rationality of the information matching model should be considered at the same time, that is, sample account information and sample device information that do not have a login relationship. There is no need to consider the degree of matching between the sample account information and the sample device information. Therefore, the sample account information and sample device information that have a login relationship can be used as sample data.
[0030] Optionally, historical data of a preset time window may be obtained, and sample account information and sample device information having a login relationship may be selected from the historical data as sample data.
[0031] It should be noted that the present disclosure does not limit the setting of the preset time window, which can be set according to actual conditions.
[0032] Optionally, the preset time window may be set to 3 years, 4 years, etc.
[0033] S102, obtaining characteristic information of sample login data.
[0034] In an embodiment of the present disclosure, when attempting to obtain characteristic information of sample login data, characteristic information of the sample login data is obtained based on the sample account dimension, the sample device dimension, and the cross-dimension of the sample account and the sample device. For each dimension, corresponding login characteristic information and behavioral characteristic information are obtained, so that the characteristic information of the sample login data can be obtained more comprehensively.
[0035] For example, for the sample account dimension, characteristic data corresponding to the sample account information can be obtained, for the sample device dimension, characteristic data corresponding to the sample device information can be obtained, and for the cross-dimension of the sample account and sample device, characteristic data corresponding to the sample account information and sample device information can be obtained. The sum of the characteristic data corresponding to the sample account information dimension, the characteristic data corresponding to the sample device information dimension, and the characteristic data corresponding to the cross-dimension of the sample account information and sample device information can be used as the characteristic information of the sample login data.
[0036] S103, training the information matching model according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine the mapping relationship between the account information and the device information.
[0037] Optionally, the information matching model is a first information matching model that is based on the first account information set and outputs a first device information set that matches the first account information set.
[0038] Optionally, the information matching model is a second information matching model that is based on the second device information set and outputs a second account information set that matches the second device.
[0039] In an embodiment of the present disclosure, characteristic information of sample login data may be input into an information matching model to obtain a predicted conversion rate value between sample account information and sample device information output by the information matching model. The conversion rate is used to indicate the degree of matching between the sample account information and the sample device information. The information matching model is adjusted based on the predicted conversion rate value and the true conversion rate value, and the adjusted information matching model is continuously trained until the training end conditions are met to obtain the target information matching model.
[0040] It should be noted that the present disclosure uses conversion rate to indicate the degree of match between sample account information and sample device information. In related technologies, matching is performed based on similarity. However, when multiple active devices share an account, the similarity between each device and the account is not high. The present disclosure uses conversion rate as an indicator, and can select the match between sample account information and sample device information to maximize business effects.
[0041] Optionally, the conversion rate may be an order conversion or click rate generated by the sample account on the sample device.
[0042] In an embodiment of the present disclosure, sample login data is obtained, wherein the sample login data includes sample account information and sample device information having a login relationship, and feature information of the sample login data is obtained, and an information matching model is trained according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine a mapping relationship between account information and device information. The present disclosure is based on a deep learning method to train an information matching model, and the mapping relationship between account information and device information can be determined based on the trained target information matching model. At the same time, the conversion rate is used to measure the degree of matching between account information and device information, thereby improving the accuracy and reliability of the matching between account information and device information.
[0043] Figure 2 A flowchart of a training method for an information matching model provided in an embodiment of the present disclosure is provided. Based on the above embodiment, further combined with Figure 2 , the process of training the information matching model according to the feature information to obtain the trained target information matching model is explained, including the following steps:
[0044] S201, inputting feature information of sample login data into an information matching model to obtain a predicted value of conversion rate between sample account information and sample device information output by the information matching model, wherein the conversion rate is used to indicate the degree of matching between the sample account information and the sample device information.
[0045] S202, adjusting the information matching model according to the predicted conversion rate value and the actual conversion rate value, and continuing to train the adjusted information matching model until the training end condition is met to obtain the target information matching model.
[0046] It should be noted that the present disclosure does not limit the specific method for obtaining the true value of the conversion rate, which can be selected according to actual conditions.
[0047] Optionally, the true value of the conversion rate between the sample account information and the sample device information may be obtained based on the characteristic information of the sample login data.
[0048] For example, the actual value of the conversion rate between the sample account information and the sample device information can be obtained based on the behavioral characteristic information of the sample login data.
[0049] In the embodiment of the present disclosure, after obtaining the predicted conversion rate and the actual conversion rate, the loss function of the information matching model can be determined based on the predicted conversion rate and the actual conversion rate, and the loss function can be adjusted according to the hyperparameters of the information matching model until the training end conditions are met to obtain a trained target information matching model.
[0050] Optionally, the information matching model may be a fusion model of a regression model and a binary classification model.
[0051] For example, after obtaining the conversion rate prediction value, it can be normalized to the first label (label1), and the data with the first label can be used to train the regression model. Linear regression, random forest regression, spline regression, etc. can be used. The maximum conversion rate prediction value is used as the second label (positive sample), and the remaining conversion rate prediction values are used as the second label (negative sample). The data with the second label is trained with a binary classification model. Specifically, logistic regression, random forest, network model, etc. can be used.
[0052] Optionally, for the fusion of the models, the square loss function is used for the loss function loss1 of the regression model, and the cross entropy loss function is used for the loss function loss2 of the binary classification model. The loss function of the fusion model (information matching model) is:
[0053] loss=a*loss1+(1-a)*loss2
[0054] Among them, a is the hyperparameter of the information matching model, loss1 is the loss function of the regression model, and loss2 is the loss function of the binary classification model.
[0055] In an embodiment of the present disclosure, feature information of sample login data is input into an information matching model to obtain a predicted conversion rate value between sample account information and sample device information output by the information matching model. The conversion rate is used to indicate the degree of matching between the sample account information and the sample device information. The information matching model is adjusted according to the predicted conversion rate value and the true conversion rate value, and the adjusted information matching model is continuously trained until the training end condition is met to obtain the target information matching model. Thus, based on a deep learning method, the information matching model is trained. Based on the trained target information matching model, the mapping relationship between the account information and the device information can be determined. At the same time, the conversion rate is used to measure the degree of matching between the account information and the device information, thereby improving the accuracy and reliability of the matching between the account information and the device information.
[0056] Figure 3A flowchart of a training method for an information matching model provided in an embodiment of the present disclosure is provided. Based on the above embodiment, further combined with Figure 3 , the process of obtaining sample login data is explained, including the following steps:
[0057] S301, obtaining initial sample login data.
[0058] It should be noted that this application does not limit the specific method of obtaining the initial sample login data, which can be selected according to actual circumstances.
[0059] Optionally, initial sample login data may be obtained based on a historical login database.
[0060] S302, determining sample login data required by the information matching model from the initial sample login data according to the type of the information matching model.
[0061] Optionally, the number of mappings between accounts and devices in the initial sample login data may be obtained, and the sample login data required for the information matching model may be determined based on the type and number of mappings of the information matching model.
[0062] For example, when the information matching model is the first information matching model, the sample login data required for determining the first information matching model is sample login data of a single account corresponding to multiple devices and sample login data of multiple accounts corresponding to multiple devices.
[0063] For example, when the information matching model is the second information matching model, the sample login data required for the second information matching model is determined to be sample login data of multiple accounts corresponding to a single device and sample login data of multiple accounts corresponding to multiple devices.
[0064] In the embodiments of the present disclosure, initial sample login data can be obtained, and sample login data required for the information matching model can be determined from the initial sample login data based on the type of the information matching model. The sample login data can be obtained more accurately, laying the foundation for subsequent accurate training of the information matching model.
[0065] Figure 4 A flowchart of a training method for an information matching model provided in an embodiment of the present disclosure is provided. Based on the above embodiment, further combined with Figure 4 , the process of obtaining characteristic information of sample login data is explained, including the following steps:
[0066] S401, obtaining the initial sample login data feature hive table.
[0067] Optionally, the account features of the initial sample login data may be obtained, and the account features may be stored in the form of a hive data table to obtain an account feature hive table, wherein the account feature hive table is used to store the correspondence between the account primary key and the account feature.
[0068] Among them, account characteristics may include account login characteristics and account behavior characteristics.
[0069] For example, for account login features, the number of days, times, duration, etc. of account logins in different time windows (for example, in the past year, 180 days, 30 days, 7 days, 3 days, and 1 day) of the accounts in the initial sample login data can be obtained. 18 login feature information can be obtained. For account behavior features, such as browsing, clicking, adding to cart, placing orders, etc., that is, the order volume, number of purchased goods, consumption amount, etc. of the account in different time windows (for example, in the past 180 days, 90 days, 30 days, and 7 days), that is, the number of visits (Page View, PV for short) and unique visitors (Unique Visitor, UV for short) of the account in different time windows (in the past 30 days, 3 days, and 1 day), the initial sample login data is cleaned, processed, and aggregated to the account granularity, and the account features are stored in the form of a hive table. The account feature hive table is used to store the correspondence between the account primary key and the account features.
[0070] Optionally, the device characteristics of the initial sample login data may be obtained, and the device characteristics may be stored in the form of a hive data table to obtain a device characteristic hive table, wherein the device characteristic hive table is used to store the correspondence between the device primary key and the device characteristics.
[0071] Among them, device characteristics may include device login characteristics and device behavior characteristics.
[0072] For example, for device login characteristics, the process of obtaining is similar to that of account characteristics (no further details will be given), that is, the number of days, times, duration of logins in different time windows, and behavioral characteristics such as browsing, clicking, adding to cart, and placing orders in different time windows. The initial sample login data is cleaned, processed, and aggregated to the device granularity, and the device characteristics are stored in the form of a hive table. The device feature hive table is used to store the correspondence between the device primary key and the device characteristics.
[0073] Optionally, the account and device cross-features of the initial sample login data may be obtained, and the cross-features may be stored in the form of a hive data table to obtain a cross-feature hive table, wherein the cross-feature hive table is used to store the correspondence between the device primary key and the account and device cross-features.
[0074] Among them, the account and device intersection characteristics may include account and device intersection login characteristics and account and device intersection behavior characteristics.
[0075] For example, for the cross-features of account and device, the process of obtaining account features is similar to that of obtaining account features (no further details will be given), that is, the login features such as the number of days, number of times, and duration of logins in different time windows, and the behavioral features such as browsing, clicking, adding to cart, and placing orders in different time windows, are used to clean and process the initial sample login data. However, when processing the initial sample login data, it is necessary to retain the account and device as dual primary keys, and when counting data such as the number of logins and the number of orders, only the login features and behavioral features of the account on the device are counted, and finally the hive table is stored with the device as the primary key.
[0076] Furthermore, the account feature hive table, the device feature hive table and the cross feature hive table may be concatenated to obtain the feature hive table corresponding to the initial sample login data.
[0077] Optionally, the primary key of the feature hive table corresponding to the initial sample login data is the initial sample login data. First, the account feature hive table is matched to the initial sample login data with the sample account information as the primary key, and then the device feature hive table is matched to the initial sample login data with the sample device information as the primary key. Finally, the cross-feature hive table is matched to the initial sample login data with the sample account information and the sample device information as the double primary keys, so as to obtain the feature hive table corresponding to the initial sample login data.
[0078] S402, using the sample login data as a query primary key, querying a feature hive table to obtain feature information of the sample login data.
[0079] In an embodiment of the present application, after obtaining the feature hive table corresponding to the initial sample login data, the sample login data can be used as a query primary key to query the feature hive table to obtain feature information of the sample login data.
[0080] In the disclosed embodiment, by obtaining the initial sample login data feature hive table, using the sample login data as the query primary key, and querying the feature hive table to obtain the feature information of the sample login data, the feature information of the sample login data can be obtained more comprehensively, laying the foundation for the subsequent accurate training of the information matching model.
[0081] Figure 5 A flowchart of an information matching method provided in accordance with an embodiment of the present disclosure.
[0082] like Figure 5 As shown, the method comprises the following steps:
[0083] S501, obtaining one of the to-be-processed account information set and the to-be-processed device information set as the to-be-processed information set.
[0084] Optionally, the information set to be processed may be an account set to be processed; optionally, the information set to be processed may be a device set to be processed.
[0085] S502, input the information set to be processed into the target information matching model to obtain the target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set.
[0086] For example, when the information set to be processed is an account information set to be processed, the account information set to be processed is input into the target information matching model to obtain a device information set that matches the account information set to be processed.
[0087] For example, when the information set to be processed is a device number information set to be processed, the device information set to be processed is input into the target information matching model to obtain an account information set that matches the device information set to be processed.
[0088] In an embodiment of the present disclosure, by obtaining an information set from the account information set to be processed and the device information set to be processed, as the information set to be processed, the information set to be processed is input into the target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set. The present disclosure can determine the mapping relationship between account information and device information based on the trained target information matching model, thereby improving the accuracy and reliability of the matching between the account information set and the device information set.
[0089] Figure 6 A flowchart of an information matching method provided in an embodiment of the present disclosure is provided. Based on the above embodiment, further combined with Figure 6 , the process of determining the target information set from the candidate target information set output by the target information matching model is explained, including the following steps:
[0090] S601, obtaining an information matching strategy, wherein the information matching strategy includes one-to-one conversion between account information and device information, equal conversion between account information and device information, and expanded conversion between account information and device information.
[0091] There is a one-to-one conversion between account information and device information, that is, each account in the input account information set matches one device, and the number of matched device information sets is not limited.
[0092] The account information and the device information are converted equally, that is, each account in the input account information set can match multiple devices, and the number of matched device information sets is consistent with the number of account information sets.
[0093] Among them, the conversion between account information and device information is expanded, that is, each account in the input account information set can match multiple devices and match a device information set of a specified order.
[0094] S602: Determine a target information set from candidate target information sets output by a target information matching model according to an information matching strategy.
[0095] In the embodiment of the present disclosure, when the information matching strategy is a one-to-one conversion strategy, the first target information in which the account information and the device information have a one-to-one correspondence can be determined from the candidate target information set, and the second target information with the highest predicted conversion rate value can be determined from the candidate target information set. The first target information and the second target information can be merged and deduplicated to obtain the target information set.
[0096] In the embodiment of the present disclosure, when the information matching strategy is an equal conversion strategy or an expanded conversion strategy, the number of target information sets corresponding to the information matching strategy can be obtained, the first candidate target information set that needs to be fused can be determined from the candidate target information set output by the target information matching model, the conversion rate prediction value of the fused first candidate target information set can be obtained, and the target information set can be determined based on the conversion rate prediction value and the number of target information sets.
[0097] For example, for an account information set to be processed as input and a device information set matching the account information set to be processed as output, the conversion rate prediction value of the fused first candidate target information set is obtained:
[0098]
[0099] Among them, score j is the predicted conversion rate of the jth device in the first candidate target information set after fusion, Nj is the number of device information sets to be processed that include device j, score ij It is the predicted value of the conversion rate between the i-th account in the account information set to be processed and the j-th device in the account information set to be processed.
[0100] Optionally, when trying to obtain the number of target information sets corresponding to the information matching strategy, you can obtain the information matching amplification multiple and the number of information sets to be processed, obtain the number of first target information and the number of second target information, and obtain the number of target information sets corresponding to the information matching strategy based on the amplification multiple, the number of information sets to be processed, the number of first target information and the number of second target information.
[0101] k=λ*N input -N I -N II
[0102] Where λ is the expansion multiple (when the information matching strategy is the equal conversion strategy, λ = 1), N input is the number of information sets to be processed, N I is the number of first target information, N II is the number of the second target information.
[0103] In an embodiment of the present disclosure, by obtaining an information matching strategy, wherein the information matching strategy includes a one-to-one conversion between account information and device information, an equal conversion between account information and device information, and an expanded conversion between account information and device information, according to the information matching strategy, a target information set is determined from a candidate target information set output by a target information matching model. The present disclosure can match a specified number of account information sets or device information sets to achieve the effect of expanding the matching quantity, and solves the problem in related technologies that the number of output account information sets / device information sets cannot be guaranteed, and can meet the business requirements for the number of output account information sets / device information sets. At the same time, the output account information sets / device information sets have better business effects.
[0104] In order to implement the information matching model training method of the first aspect of the embodiment, the present disclosure proposes an information matching model training device. Figure 7 FIG. 1 is a schematic diagram of the structure of a training device for an information matching model according to an embodiment of the present disclosure. Figure 7 As shown, the information matching model training device 700 includes:
[0105] A first acquisition module 710 is used to acquire sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship;
[0106] A second acquisition module 720 is used to acquire feature information of the sample login data;
[0107] The training module 730 is used to train the information matching model according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine the mapping relationship between the account information and the device information.
[0108] In one embodiment of the present disclosure, the training module 730 is further used to: input the feature information of the sample login data into the information matching model to obtain the information matching model outputting a predicted conversion rate value between the sample account information and the sample device information, wherein the conversion rate is used to indicate the degree of matching between the sample account information and the sample device information; adjust the information matching model according to the predicted conversion rate value and the true conversion rate value, and continue to train the adjusted information matching model until the training end condition is met to obtain the target information matching model.
[0109] In one embodiment of the present disclosure, the process of obtaining the true value of the conversion rate includes: obtaining the true value of the conversion rate between the sample account information and the sample device information according to the feature information of the sample login data.
[0110] In one embodiment of the present disclosure, the training module 730 is also used to: determine the loss function of the information matching model based on the predicted conversion rate value and the actual conversion rate value; adjust the loss function according to the hyperparameters of the information matching model until the training end condition is met to obtain the trained target information matching model.
[0111] In one embodiment of the present disclosure, the information matching model is a first information matching model based on a first account information set, outputting a first device information set that matches the first account information set; or, the information matching model is a second information matching model based on a second device information set, outputting a second account information set that matches the second device.
[0112] In one embodiment of the present disclosure, the first acquisition module 710 is further used to: acquire initial sample login data; and determine the sample login data required by the information matching model from the initial sample login data according to the type of the information matching model.
[0113] In one embodiment of the present disclosure, the first acquisition module 710 is further used to: obtain the number of mappings between accounts and devices in the initial sample login data; and determine the sample login data required by the information matching model based on the type of the information matching model and the number of mappings.
[0114] In one embodiment of the present disclosure, the first acquisition module 710 is further used to: when the information matching model is the first information matching model, determine that the sample login data required by the first information matching model is the sample login data of a single account corresponding to multiple devices and the sample login data of multiple accounts corresponding to multiple devices; or, when the information matching model is the second information matching model, determine that the sample login data required by the second information matching model is the sample login data of multiple accounts corresponding to a single device and the sample login data of multiple accounts corresponding to multiple devices.
[0115] In one embodiment of the present disclosure, the second acquisition module 720 is further used to: acquire the feature hive table of the initial sample login data; use the sample login data as the query primary key to query the feature hive table to acquire the feature information of the sample login data.
[0116] In one embodiment of the present disclosure, the second acquisition module 720 is further used to: acquire account characteristics of the initial sample login data, store the account characteristics in the form of a hive data table, and obtain an account characteristics hive table, wherein the account characteristics hive table is used to store the correspondence between the account primary key and the account characteristics; acquire device characteristics of the initial sample login data, store the device characteristics in the form of a hive data table, and obtain a device characteristics hive table, wherein the device characteristics hive table is used to store the correspondence between the device primary key and the device characteristics; acquire account and device cross characteristics of the initial sample login data, store the cross characteristics in the form of a hive data table, and obtain a cross characteristics hive table, wherein the cross characteristics hive table is used to store the correspondence between the device primary key and the account and device cross characteristics; concatenate the account characteristics hive table, the device characteristics hive table and the cross characteristics hive table to obtain a feature hive table corresponding to the initial sample login data.
[0117] It should be noted that the above explanation of the embodiment of the training method of the information matching model of the first aspect is also applicable to the training device of the information matching model of the embodiment of the present disclosure, and the specific process will not be repeated here.
[0118] In an embodiment of the present disclosure, sample login data is obtained, wherein the sample login data includes sample account information and sample device information having a login relationship, and feature information of the sample login data is obtained, and an information matching model is trained according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine a mapping relationship between account information and device information. The present disclosure is based on a deep learning method to train an information matching model, and the mapping relationship between account information and device information can be determined based on the trained target information matching model. At the same time, the conversion rate is used to measure the degree of matching between account information and device information, thereby improving the accuracy and reliability of the matching between account information and device information.
[0119] In order to implement the information matching method of the second aspect of the embodiment, the present disclosure proposes an information matching device. Figure 8 FIG. 1 is a schematic diagram of the structure of an information matching device according to an embodiment of the present disclosure. Figure 8 As shown, the information matching device 800 includes:
[0120] An acquisition module 810 is used to acquire one of the to-be-processed account information set and the to-be-processed device information set as the to-be-processed information set;
[0121] The matching module 820 is used to input the information set to be processed into the target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set; wherein the target information matching model is a model trained using the method described in any one of the first aspects.
[0122] In one embodiment of the present disclosure, the device 800 is also used to: obtain an information matching strategy, wherein the information matching strategy includes a one-to-one conversion between account information and device information, an equal amount conversion between account information and device information, and an extended amount conversion between account information and device information; according to the information matching strategy, determine the target information set from the candidate target information set output by the target information matching model.
[0123] In one embodiment of the present disclosure, if the information matching strategy is a one-to-one conversion strategy, the device 800 is also used to: determine the first target information in which the account information and the device information have a one-to-one correspondence from the candidate target information set; determine the second target information with the highest conversion rate prediction value from the candidate target information set; merge the first target information and the second target information, and perform deduplication processing to obtain the target information set.
[0124] In one embodiment of the present disclosure, if the information matching strategy is an equal conversion strategy or an expanded conversion strategy, the device 800 is also used to: obtain the number of target information sets corresponding to the information matching strategy; determine the first candidate target information set that needs to be fused from the candidate target information set output by the target information matching model, and obtain the conversion rate prediction value of the fused first candidate target information set; determine the target information set based on the conversion rate prediction value and the number of target information sets.
[0125] In one embodiment of the present disclosure, the device 800 is also used to: obtain the expansion multiple of the information matching and the number of the information sets to be processed; obtain the number of the first target information and the number of the second target information; and obtain the number of target information sets corresponding to the information matching strategy based on the expansion multiple, the number of the information sets to be processed, the number of the first target information and the number of the second target information.
[0126] In an embodiment of the present disclosure, by obtaining an information set from the account information set to be processed and the device information set to be processed, as the information set to be processed, the information set to be processed is input into the target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set. The present disclosure can determine the mapping relationship between account information and device information based on the trained target information matching model, thereby improving the accuracy and reliability of the matching between the account information set and the device information set.
[0127] It should be noted that the above explanation of the information matching method embodiment of the second aspect is also applicable to the information matching device of the embodiment of the present disclosure, and the specific process will not be repeated here.
[0128] like Fig. 9 , is a block diagram of an electronic device according to a training method of an information matching model or an information matching method according to an embodiment of the present disclosure. The electronic device 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 can also represent various forms of mobile devices, such as intelligent voice interaction devices, personal digital processing, cellular phones, smart phones, wearable devices, 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 disclosure described and / or required herein.
[0129] like Fig. 9As shown, the electronic device includes: one or more processors 901, a memory 902, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are interconnected using different buses and can be installed on a common mainboard or installed in other ways as needed. The processor 901 can process instructions executed in the electronic device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In other embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple electronic devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig. 9 A processor 901 is taken as an example.
[0130] The memory 902 is a non-transitory computer-readable storage medium provided by the present disclosure. The memory stores instructions executable by at least one processor to enable at least one processor to execute the information matching model training method provided by the first aspect embodiment of the present disclosure or the information matching method provided by the second aspect embodiment. The non-transitory computer-readable storage medium of the present disclosure stores computer instructions, which are used to enable a computer to execute the information matching model training method provided by the first aspect embodiment of the present disclosure or the information matching method provided by the second aspect embodiment.
[0131] The memory 902, as a non-transient computer-readable storage medium, can be used to store non-transient software programs, non-transient computer executable programs and modules, such as the information matching model training method provided in the first aspect of the present disclosure or the information matching method provided in the second aspect of the present disclosure. The processor 901 executes various functional applications and data processing of the server by running the non-transient software programs, instructions and modules stored in the memory 902, that is, implementing the information matching model training method or information matching method in the above method embodiment.
[0132] The memory 902 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the training method of the information matching model or the use of the electronic device of the information matching method, etc. In addition, the memory 902 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory 902 may optionally include a memory remotely arranged relative to the processor 901, and these remote memories may be connected to the training method of the information matching model or the electronic device of the information matching method via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0133] The information matching model training method or the electronic device of the information matching method may further include: an input device 903 and an output device 904. The processor 901, the memory 902, the input device 903 and the output device 904 may be connected via a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.
[0134] The input device 903 can receive input digital or character information, and generate key signal input related to the training method of the information matching model or the user setting and function control of the electronic device of the information matching method, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator rod, one or more mouse buttons, a trackball, a joystick and other input devices. The output device 904 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The display device may include, but is not limited to, a liquid crystal display (LCD), a light emitting diode (LED) display and a plasma display. In some embodiments, the display device may be a touch screen.
[0135] In order to implement the above-mentioned embodiments, the present disclosure also proposes a non-temporary computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the training method of the information matching model proposed in the first aspect embodiment of the present disclosure or the information matching method proposed in the second aspect embodiment.
[0136] In order to implement the above embodiments, the present disclosure proposes a computer program product, including a computer program, which, when executed by a processor, implements the information matching model training method provided in the first aspect embodiment of the present disclosure or the information matching method provided in the second aspect embodiment.
[0137] Various implementations of the systems and techniques described herein can be realized in digital electronic circuit systems, integrated circuit systems, dedicated ASICs (application specific integrated circuits), 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.
[0138] These computer programs (also referred to as programs, software, software applications, or code) include machine instructions for programmable processors and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or means (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.
[0139] To provide interaction with a user, the systems and techniques described herein 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 the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).
[0140] The systems and techniques described herein may be implemented in a computing system that includes back-end 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 front-end 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 back-end components, middleware components, or front-end 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), and the Internet.
[0141] A computer 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 relationship between the client and the server 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 ("Virtual Private Server", or "VPS" for short).
[0142] 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 recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.
[0143] In the description of this specification, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this disclosure, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.
[0144] Although the embodiments of the present disclosure have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present disclosure. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present disclosure.
Claims
1. A training method for an information matching model, characterized in that: The method comprises: Acquire sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship; Acquire characteristic information of the sample login data; The information matching model is trained according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine a mapping relationship between the account information and the device information.
2. The method according to claim 1, characterized in that The step of training the information matching model according to the feature information to obtain a trained target information matching model includes: Inputting the characteristic information of the sample login data into an information matching model to obtain a predicted conversion rate between the sample account information and the sample device information output by the information matching model, wherein the conversion rate is used to indicate the degree of matching between the sample account information and the sample device information; The information matching model is adjusted according to the predicted conversion rate value and the actual conversion rate value, and the adjusted information matching model is continuously trained until a training end condition is met to obtain the target information matching model.
3. The method according to claim 2, characterized in that The process of obtaining the true value of the conversion rate includes: According to the characteristic information of the sample login data, the real value of the conversion rate between the sample account information and the sample device information is obtained.
4. The method according to claim 2, characterized in that: The method further comprises: Determining a loss function of the information matching model according to the predicted conversion rate value and the actual conversion rate value; The loss function is adjusted according to the hyperparameters of the information matching model until the training end condition is met to obtain the trained target information matching model.
5. The method according to any one of claims 1 to 4, characterized in that The information matching model is a first information matching model that is based on a first account information set and outputs a first device information set that matches the first account information set; or, the information matching model is a second information matching model that is based on a second device information set and outputs a second account information set that matches the second device.
6. The method according to claim 5, characterized in that The obtaining of sample login data includes: Get initial sample login data; The sample login data required by the information matching model is determined from the initial sample login data according to the type of the information matching model.
7. The method according to claim 5, characterized in that Determining the sample login data required by the information matching model from the initial sample login data according to the type of the information matching model includes: Obtaining the number of mappings between accounts and devices in the initial sample login data; The sample login data required for the information matching model is determined according to the type of the information matching model and the mapping quantity.
8. The method according to claim 7, characterized in that The determining, according to the type of the information matching model and the mapping quantity, the sample login data required by the information matching model comprises: When the information matching model is the first information matching model, determining that the sample login data required by the first information matching model is sample login data of a single account corresponding to multiple devices and sample login data of multiple accounts corresponding to multiple devices; or When the information matching model is the second information matching model, it is determined that the sample login data required by the second information matching model is sample login data of multiple accounts corresponding to a single device and sample login data of multiple accounts corresponding to multiple devices.
9. The method according to claim 1, characterized in that: The obtaining characteristic information of the sample login data includes: Get the initial sample login data feature hive table; The sample login data is used as a query primary key to query the feature hive table to obtain feature information of the sample login data.
10. The method according to claim 9, characterized in that The step of obtaining the feature hive table corresponding to the initial sample login data includes: Acquire the account features of the initial sample login data, store the account features in the form of a hive data table, and obtain an account feature hive table, wherein the account feature hive table is used to store the corresponding relationship between the account primary key and the account feature; Acquire device features of the initial sample login data, store the device features in the form of a hive data table, and obtain a device feature hive table, wherein the device feature hive table is used to store the corresponding relationship between the device primary key and the device features; Obtaining the account and device cross-features of the initial sample login data, storing the cross-features in the form of a hive data table to obtain a cross-feature hive table, wherein the cross-feature hive table is used to store the corresponding relationship between the device primary key and the account and device cross-features; The account feature hive table, the device feature hive table and the cross feature hive table are concatenated to obtain a feature hive table corresponding to the initial sample login data.
11. An information matching method, characterized in that: The method comprises: Obtain one of the account information set to be processed and the device information set to be processed as the information set to be processed; Input the information set to be processed into the target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set; Wherein, the target information matching model is a model trained using the method described in any one of claims 1-10.
12. The method according to claim 11, characterized in that The method comprises: Obtaining an information matching strategy, wherein the information matching strategy includes one-to-one conversion between account information and device information, equal conversion between account information and device information, and expanded conversion between account information and device information; According to the information matching strategy, the target information set is determined from the candidate target information sets output by the target information matching model.
13. The method according to claim 12, characterized in that If the information matching strategy is a one-to-one conversion strategy, wherein determining the target information set from the candidate target information set output by the target information matching model according to the information matching strategy includes: Determine, from the candidate target information set, first target information having a one-to-one correspondence between account information and device information; Determine, from the candidate target information set, second target information with the highest predicted conversion rate value; The first target information and the second target information are merged and deduplicated to obtain the target information set.
14. The method according to claim 12, characterized in that If the information matching strategy is an equal amount conversion strategy or an expanded amount conversion strategy, wherein, according to the information matching strategy, determining the target information set from the candidate target information set output by the target information matching model includes: Obtaining the number of target information sets corresponding to the information matching strategy; Determine a first candidate target information set to be fused from the candidate target information sets output by the target information matching model, and obtain a conversion rate prediction value of the fused first candidate target information set; The target information set is determined according to the predicted conversion rate value and the number of the target information sets.
15. The method according to any one of claims 12 to 14, characterized in that: If the information matching strategy is an equal amount conversion strategy or an expanded amount conversion strategy, wherein obtaining the number of target information sets corresponding to the information matching strategy includes: Obtaining the expansion multiple of information matching and the number of the information sets to be processed; Acquire the quantity of the first target information and the quantity of the second target information; The number of target information sets corresponding to the information matching strategy is acquired according to the expansion multiple, the number of the information sets to be processed, the number of the first target information, and the number of the second target information.
16. A training device for an information matching model, characterized in that: The device comprises: A first acquisition module, used to acquire sample login data, wherein the sample login data includes sample account information and sample device information having a login relationship; A second acquisition module, used to acquire feature information of the sample login data; The training module is used to train the information matching model according to the feature information to obtain a trained target information matching model, wherein the target information matching model is used to determine the mapping relationship between the account information and the device information.
17. An information matching device, characterized in that: The device comprises: An acquisition module, used to acquire one of the account information set to be processed and the device information set to be processed as the information set to be processed; A matching module, used for inputting the information set to be processed into a target information matching model to obtain a target information set that matches the information set to be processed; the information set to be processed is the account information set to be processed, the target information set is the matched device information set, the information set to be processed is the device information set to be processed, and the target information set is the matched account information set; wherein the target information matching model is a model trained using the method described in any one of claims 1 to 10.
18. An electronic device, characterized in that: including memory and processor; The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, and the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method described in any one of claims 1-10 or claims 11-15.
19. A computer-readable storage medium storing a computer program, wherein when the computer program is run on a computer, the method of any one of claims 1 to 10 or claims 11 to 15 is executed.
20. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1-10 or claims 11-15.
Citation Information
Patent Citations
Training method, device, storage medium and electronic device for vehicle-cargo matching model
CN109242044A
Information identification method and device
CN110555451A
Mapping relationship determination method and device
CN111340062A
Method and equipment for training matching degree detection model and storage medium
CN113269262A
