Method, device and equipment for recommending functions of application software, and storage medium

Abstract

The present disclosure provides a function recommendation method and device of application software, equipment and storage medium, which can be applied to the field of artificial intelligence technology. The method comprises the following steps: determining an evaluation feature corresponding to a business scenario, the evaluation feature being a transaction feature with the highest importance in the business scenario, the importance indicating the influence degree of the transaction feature in the business scenario; determining a score matrix for a target user according to the evaluation feature; determining M similar users from a plurality of candidate users according to the score matrix of the target user, M being greater than or equal to 2; and determining a to-be-recommended function matched with the target user according to historical behavior information of the M similar users.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence, specifically to a method, apparatus, device, and storage medium for recommending functions in application software. Background Technology

[0002] In recommendation scenarios, features are typically recommended to users based on their preferences for the functions they use in the application. Related technologies rely on questionnaires or telephone surveys to determine these preferences. However, multiple users may use different rating criteria for the same user experience, and users may not have rated all the functions they use, leading to biases in the determined user preferences.

[0003] In related technologies, the acquisition of user preferences for application software functions relies on subjective factors, which leads to biases in the function recommendations based on user preferences. This results in low recommendation accuracy and a poor user experience. Summary of the Invention

[0004] In view of the above problems, this disclosure provides a method, apparatus, device and storage medium for recommending application software functions.

[0005] According to a first aspect of this disclosure, a method for recommending features in application software is provided, comprising:

[0006] Identify the evaluation features corresponding to the business scenario. The evaluation features are the transaction features with the highest importance in the business scenario. The importance indicates the degree of influence of the transaction features in the business scenario.

[0007] Based on the evaluation characteristics, determine the rating matrix for the target users;

[0008] Based on the target user's rating matrix, identify M similar users from multiple candidate users, where M is greater than or equal to 2; and

[0009] Based on the historical behavior information of M similar users, identify the features to be recommended that match the target user.

[0010] According to embodiments of this disclosure, determining the rating matrix for the target user based on evaluation features includes:

[0011] Determine N scoring intervals that match the evaluation characteristics, where N is greater than or equal to 2;

[0012] Obtain the first feature value of the target user for the evaluation feature; and

[0013] The rating matrix for the target user is determined based on the first eigenvalue and the rating interval.

[0014] According to embodiments of this disclosure, determining the rating matrix for the target user based on a first feature value and a rating interval includes:

[0015] Generate an initial rating matrix according to the order of N rating intervals. The initial rating matrix includes N elements, and the value of each element is 0.

[0016] Determine the nth rating interval, which includes the first feature value, from N rating intervals, where n is greater than or equal to 1 and less than or equal to N; and

[0017] Modify the element value corresponding to the nth rating interval in the initial rating matrix to 1 to obtain the rating matrix.

[0018] According to embodiments of this disclosure, determining N scoring intervals that match the evaluation features includes:

[0019] Based on the evaluation features, obtain the second feature values ​​of multiple candidate users;

[0020] Find the minimum and maximum second eigenvalues ​​from a plurality of second eigenvalues; and

[0021] The minimum and maximum second eigenvalues ​​are divided into N scoring intervals by equal distances.

[0022] According to embodiments of this disclosure, determining M similar users from a plurality of candidate users based on a target user's rating matrix includes:

[0023] Based on the rating interval and the second feature values ​​of multiple candidate users, a rating matrix for multiple candidate users is determined.

[0024] Based on the rating matrix, calculate the similarity between the target user and multiple candidate users; and

[0025] The M candidate users whose similarity is greater than or equal to the first threshold are identified as M similar users.

[0026] According to embodiments of this disclosure, determining the recommended function that matches the target user based on the historical behavior information of M similar users includes:

[0027] Based on the historical behavior information, a score value is determined for P functions within the application scenario, where P is greater than or equal to 1;

[0028] Calculate the ratings for P functions based on the similarity between M similar users and the target user, and the ratings of the M similar users for the P functions.

[0029] The feature with the highest score among the P feature ratings is identified as the feature to be recommended.

[0030] According to embodiments of this disclosure, the calculation of P function ratings based on the similarity between M similar users and the target user, and the ratings of the M similar users for the P functions, includes:

[0031] Using similarity as a weight, calculate the weighted sum of the p-th function based on the ratings of M similar users for the p-th function, where p is greater than or equal to 1 and p is less than or equal to P;

[0032] Calculate the sum of similarities between the target user and the M similar users; and

[0033] The ratio of the weighted sum to the sum of similarities is used as the function score for the p-th function.

[0034] According to embodiments of this disclosure, determining the evaluation features corresponding to the business scenario includes:

[0035] For a given business scenario, the input to a pre-trained prediction model is obtained. This model predicts whether a user will use the target function within that scenario. The input to the prediction model includes the user's transaction characteristics, which include behavioral information, user profiles, and asset information. The output of the prediction model represents whether the user uses the target function. The prediction model includes a decision tree model.

[0036] The decision tree model's preset function is invoked to calculate the importance of transaction features; and

[0037] The transaction features with the highest importance are used as evaluation features.

[0038] A second aspect of this disclosure provides a function recommendation device for application software, comprising:

[0039] The feature determination module is used to determine the evaluation features corresponding to the business scenario. The evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario.

[0040] The matrix determination module is used to determine the rating matrix for the target user based on the evaluation characteristics.

[0041] The user identification module is used to identify M similar users from multiple candidate users based on the target user's rating matrix, where M is greater than or equal to 2; and

[0042] The function determination module is used to determine the recommended functions that match the target user based on the historical behavior information of M similar users.

[0043] A third aspect of this disclosure provides an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the recommended method for the functions of the application software described above.

[0044] A fourth aspect of this disclosure also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the recommended method for the functions of the aforementioned application software.

[0045] The fifth aspect of this disclosure also provides a computer program product, including a computer program that, when executed by a processor, implements the recommended method for the functions of the aforementioned application software.

[0046] This disclosure achieves application software function recommendation by identifying evaluation features corresponding to business scenarios, where the evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario; determining a rating matrix for target users based on the evaluation features; identifying M similar users from multiple candidate users based on the target user's rating matrix; and determining the recommended functions that match the target user based on the historical behavior information of the M similar users. Since the evaluation of user preferences in the embodiments of this disclosure is based on actual behavioral features and does not rely on the user's subjective evaluation factors, the determined rating matrix is ​​more accurate, the similar users identified based on the rating matrix are more accurate, and the recommended functions identified based on similar users are more accurate, thus improving the overall accuracy of the recommendation function and enhancing the user experience. Attached Figure Description

[0047] The foregoing contents, as well as other objects, features, and advantages of this disclosure, will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0048] Figure 1 This illustration schematically depicts an application scenario of the function recommendation method for application software according to embodiments of the present disclosure;

[0049] Figure 2 A flowchart illustrating a function recommendation method for application software according to an embodiment of the present disclosure is shown schematically.

[0050] Figure 3 A flowchart illustrating a method for determining a scoring matrix according to an embodiment of the present disclosure is shown schematically;

[0051] Figure 4 A flowchart illustrating a method for determining similar users according to an embodiment of the present disclosure is shown schematically.

[0052] Figure 5The illustration shows a schematic diagram of a method for determining a function to be recommended according to a specific embodiment of the present disclosure;

[0053] Figure 6A This illustration schematically shows a diagram illustrating the importance of transaction features in a remittance scenario according to embodiments of the present disclosure;

[0054] Figure 6B This illustration schematically shows a diagram illustrating the importance of transaction features in a balance query scenario according to an embodiment of the present disclosure;

[0055] Figure 6C This illustration schematically shows the importance of transaction features in a financial management scenario according to embodiments of the present disclosure;

[0056] Figure 6D This illustration schematically shows a diagram illustrating the importance of transaction features in a loan scenario according to an embodiment of the present disclosure;

[0057] Figure 6E This illustration schematically shows a diagram illustrating the importance of transaction features in a credit card scenario according to an embodiment of the present disclosure;

[0058] Figure 7 This schematically illustrates a structural block diagram of an application software function recommendation device according to embodiments of the present disclosure; and

[0059] Figure 8 A block diagram of an electronic device for a function recommendation method suitable for application software according to an embodiment of the present disclosure is illustrated schematically. Detailed Implementation

[0060] The embodiments of the present disclosure will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the disclosure. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the present disclosure for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concepts of the present disclosure.

[0061] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit this disclosure. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0062] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0063] When using expressions such as "at least one of A, B, and C", they should generally be interpreted in accordance with the meaning that is commonly understood by a person skilled in the art (e.g., "a system having at least one of A, B, and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B, and C, etc.).

[0064] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision, disclosure, and application of data (including but not limited to user personal information) comply with the provisions of relevant laws and regulations, necessary confidentiality measures have been taken, and they do not violate public order and good morals.

[0065] In recommendation scenarios, user preferences for application features are typically gathered through surveys or phone calls. However, different users may use different rating standards for the same experience. For example, User A might use a perfect score to express liking, User B might use a score one point lower, and User C might tend to use the median score, either one point higher or lower. Therefore, different users will express their liking differently for the same level of enjoyment.

[0066] Regarding all the functions within the application software, some users only rated the functions of module A, without rating the functions of other modules; others only rated the functions of module B, without rating the functions of other modules. This means that user ratings cannot determine the full range of user preferences.

[0067] In addition, the scoring rules and scoring ranges for questionnaires or telephone surveys are determined by business personnel based on their experience, and the determination of the scoring rules and scoring ranges will also affect the final distribution of preferences.

[0068] In summary, the relevant technologies rely on data determined by subjective factors for recommendations, which can lead to biases in the final recommended functions, resulting in low recommendation accuracy and a poor user experience.

[0069] Embodiments of this disclosure provide a method for recommending application software features, including: determining evaluation features corresponding to a business scenario, wherein the evaluation features are transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario; determining a rating matrix for a target user based on the evaluation features; determining M similar users from multiple candidate users based on the rating matrix of the target user, where M is greater than or equal to 2; and determining recommended features that match the target user based on the historical behavior information of the M similar users.

[0070] Figure 1 The illustration depicts an application scenario of the function recommendation method for application software according to an embodiment of the present disclosure.

[0071] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0072] Users can interact with server 105 via network 104 using at least one of the first terminal device 101, second terminal device 102, and third terminal device 103 to receive or send messages, etc. Various application software can be installed on the first terminal device 101, second terminal device 102, and third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only). The application software includes multiple modules and performs various functions.

[0073] For example, a bank's application client includes multiple functions such as balance inquiry, wealth management, and loans.

[0074] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0075] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0076] It should be noted that the application software function recommendation method provided in this embodiment can generally be executed by server 105. Correspondingly, the application software function recommendation device provided in this embodiment can generally be located in server 105. The application software function recommendation method provided in this embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the application software function recommendation device provided in this embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0077] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0078] The following will be based on Figure 1 The described scene, through Figures 2 to 6E The method for recommending the functionality of application software according to the disclosed embodiments is described in detail.

[0079] It should be noted that all the processes involved in the following steps, including the collection, storage, use, processing, transmission, provision, disclosure, and application of historical behavioral information and historical evaluation data, are conducted with the user's permission.

[0080] Figure 2 A flowchart illustrating a function recommendation method for application software according to an embodiment of this disclosure is shown schematically.

[0081] like Figure 2 As shown, the method 200 includes operations S210 to S240.

[0082] In operation S210, the evaluation features corresponding to the business scenario are determined.

[0083] According to embodiments of this disclosure, the evaluation feature is the transaction feature with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction feature in the business scenario.

[0084] According to embodiments of this disclosure, multiple business scenarios include multiple transaction features. For example, transaction features in a balance inquiry scenario include: number of debit card transactions, debit card transaction amount, number of credit card transactions, credit card transaction amount, and offline Automated Teller Machine (ATM) usage behavior. Transaction features in a loan scenario include: loan amount, number of loans, and loan address. The evaluation feature can be the feature with the highest importance among the aforementioned transaction features.

[0085] According to embodiments of this disclosure, transaction features in this disclosure refer to features based on the user's actual transaction behavior, such as asset information and transaction behavior, rather than features generated based on the user's rating behavior.

[0086] According to embodiments of this disclosure, importance indicates the degree of influence of a transaction feature in a business scenario. The greater the influence of a transaction feature, the higher its importance; conversely, the smaller the influence of a transaction feature, the lower its importance.

[0087] According to embodiments of this disclosure, importance can be determined based on historical behavioral information that has occurred in the application software. For example, the importance of multiple transaction features can be determined based on the frequency of occurrence, transaction amount, time, or other data of the transaction features, so that the transaction feature with the highest importance can be identified as the evaluation feature.

[0088] According to embodiments of this disclosure, in this business scenario, the importance of input features in the trained model can also be used as the importance score to determine the importance of multiple transaction features, and then the transaction feature with the highest importance score can be determined as the evaluation feature. The input features of the model include transaction features.

[0089] In operation S220, a rating matrix for the target user is determined based on the evaluation characteristics.

[0090] According to embodiments of this disclosure, after determining the evaluation features, historical behavior information of the target user under the evaluation features is obtained, and then the rating matrix of the target user is determined based on the historical behavior information of the target user.

[0091] For example, taking loan transaction amount as the evaluation feature. For a target user, obtain their historical loan behavior information, and determine the target user's rating matrix based on the loan transaction amount within that historical information.

[0092] It should be noted that the determination of the rating matrix does not depend on the subjective ratings of the target users, but rather on the actual behavior of the target users.

[0093] In operation S230, based on the target user's rating matrix, M similar users are determined from multiple candidate users, where M is greater than or equal to 2.

[0094] According to embodiments of this disclosure, after determining the rating matrix of the target user, the similarity between multiple candidate users and the target user can be determined based on the rating matrices of the target user and candidate users. Based on the similarity, similar users who are similar to the target user are identified from the multiple candidate users.

[0095] According to embodiments of this disclosure, similar to the method for determining the rating matrix of target users, the rating matrix of candidate users is also determined based on the historical behavior information of candidate users.

[0096] According to embodiments of this disclosure, determining M similar users from a plurality of candidate users includes: determining at least one highest similarity score, and determining the candidate users corresponding to the at least one highest similarity score as similar users. For example, if M highest similarity scores exist simultaneously, the M candidate users corresponding to the M highest similarity scores are determined as M similar users.

[0097] According to embodiments of this disclosure, determining M similar users from a plurality of candidate users further includes: determining M similarities that satisfy a first threshold, and determining the M candidate users corresponding to the aforementioned M similarities as M similar users.

[0098] In operation S240, based on the historical behavior information of M similar users, the recommended functions that match the target user are determined.

[0099] According to embodiments of this disclosure, after identifying M similar users, ratings for the multiple functions by the M similar users are determined based on their historical behavioral information regarding these functions within the application scenario. Then, the ratings from the M similar users are combined to obtain a function score for each function, which is used to determine the functions to be recommended.

[0100] According to embodiments of this disclosure, collaborative filtering algorithms can be used to determine the features to be recommended. For example, a user-based collaborative filtering method can be used to identify at least one similar user who is similar to the target user, and to recommend application software features that the similar user has used but the target user has not used.

[0101] This disclosure achieves application software function recommendation by identifying evaluation features corresponding to business scenarios, where the evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario; determining a rating matrix for target users based on the evaluation features; identifying M similar users from multiple candidate users based on the target user's rating matrix; and determining the recommended functions that match the target user based on the historical behavior information of the M similar users. Since the evaluation of user preferences in the embodiments of this disclosure is based on actual behavioral features and does not rely on the user's subjective evaluation factors, the determined rating matrix is ​​more accurate, the similar users identified based on the rating matrix are more accurate, and the recommended functions identified based on similar users are more accurate, thus improving the overall accuracy of the recommendation function and enhancing the user experience.

[0102] Figure 3 A flowchart illustrating a method for determining a scoring matrix according to an embodiment of the present disclosure is shown schematically.

[0103] like Figure 3 As shown, the method 300 for determining the scoring matrix in this embodiment includes operations S321 to S322, which can be used as a specific embodiment of operation S220.

[0104] In operation S321, N scoring intervals that match the evaluation characteristics are determined, where N is greater than or equal to 2.

[0105] In operation S322, obtain the first feature value of the target user for the evaluation feature.

[0106] In operation S323, the rating matrix of the target user is determined based on the first feature value and the rating interval.

[0107] According to embodiments of this disclosure, after determining the evaluation features corresponding to the application scenario, N rating intervals are determined based on historical behavior information corresponding to the evaluation features in the database. For example, historical behavior information of all users regarding the evaluation feature is obtained from the database, and the historical behavior information of all users is used as the feature value of the evaluation feature; the above multiple feature values ​​are filtered and segmented to determine N rating intervals.

[0108] According to embodiments of this disclosure, the number of scoring intervals can be determined based on actual circumstances.

[0109] For example, the number of rating intervals is generally 5. If the number of users exceeds the first threshold, 2 more rating intervals are added. If the number of users exceeds the second threshold, 3 more rating intervals are added.

[0110] Alternatively, the number of rating intervals can be fixed at 5, and the length of each interval can be determined based on the feature values. For example, the length of each rating interval could be one-fifth of the range of feature values ​​for all users. For instance, if the feature value range is from 1 to 100, the length of each rating interval would be 20.

[0111] According to embodiments of this disclosure, after determining the rating range, the actual behavioral information of the target user in response to the evaluation features is determined as the first feature value, so as to determine the rating matrix of the target user based on the first feature value and the rating range.

[0112] The embodiments of this disclosure determine a rating matrix based on the actual feature values ​​of the target user after determining the evaluation features. This does not rely on the user's subjective features and determines an objective rating based on the actual feature values, which helps to improve the accuracy of determining user preferences.

[0113] According to embodiments of this disclosure, determining N scoring intervals that match the evaluation features includes the following steps.

[0114] Based on the evaluation features, the second feature values ​​of multiple candidate users are obtained.

[0115] Obtain the minimum and maximum second eigenvalues ​​from a plurality of second eigenvalues.

[0116] The minimum and maximum second eigenvalues ​​are divided into N scoring intervals by equal distances.

[0117] According to embodiments of this disclosure, all existing users in the database can be used as candidate users. After determining the evaluation feature, the historical behavior information of all candidate users regarding that evaluation feature is used as the second feature value of that evaluation feature.

[0118] The minimum and maximum second eigenvalues ​​are selected from the multiple second eigenvalues ​​mentioned above. Then, using the minimum second eigenvalue as the minimum value and the maximum second eigenvalue as the maximum value, the interval between the minimum and maximum values ​​is divided into N equal parts using the equal-interval binning method, resulting in N scoring intervals.

[0119] The length of each scoring interval is: W = (Xmax - Xmin) / N, where Xmax represents the largest second feature value, Xmin represents the smallest second feature value, and N is the number of scoring intervals.

[0120] According to embodiments of this disclosure, since the N scoring intervals are divided according to the magnitude of the second feature value, there is a magnitude order among the N scoring intervals. That is, the first scoring interval has the smallest value, the minimum value of the second scoring interval is greater than or equal to the maximum value of the first scoring interval, ... the minimum value of the nth scoring interval is greater than or equal to the maximum value of the (n-1)th scoring interval, ... the maximum value of the Nth scoring interval is the largest second feature value.

[0121] For example, the number of scoring intervals is 5 for scenarios such as balance inquiry, remittance, wealth management, loan, and credit card.

[0122] Table 1. Rating intervals for users across multiple scenarios.

[0123] user1 5 4 1 1 1 user2 1 1 1 1 1 user3 1 2 1 2 1

[0124] As shown in Table 1, the rating interval table for users in multiple scenarios includes balance inquiry, remittance, credit card, loan, and wealth management scenarios. User user1 is located in the 5th, 4th, 1st, 1st, and 1st rating intervals of the evaluation features in the above five scenarios. User user2 is located in the 1st, 1st, 1st, 1st, and 1st rating intervals of the above five scenarios. User user3 is located in the 1st, 2nd, 1st, 2nd, and 1st rating intervals of the above five scenarios.

[0125] The embodiments of this disclosure target multiple users under evaluation characteristics and determine objective rating intervals based on users' historical behavioral data. Compared to equal-frequency binning, the equidistant method better reflects the differences in user behavior.

[0126] According to embodiments of this disclosure, determining a rating matrix for a target user based on a first feature value and a rating interval includes the following steps.

[0127] Generate an initial rating matrix according to the order of N rating intervals. The initial rating matrix includes N elements, and the value of each element is 0.

[0128] Determine the nth scoring interval from the N scoring intervals, including the first feature value, where n is greater than or equal to 1 and less than or equal to N.

[0129] Modify the element value corresponding to the nth rating interval in the initial rating matrix to 1 to obtain the rating matrix.

[0130] According to embodiments of this disclosure, there is a numerical relationship between the rating intervals. An initial rating matrix is ​​generated based on the order of the N rating intervals. The initial rating matrix is ​​a one-dimensional matrix, and the number of elements in the one-dimensional initial rating matrix is ​​the same as the number of rating intervals, with each element having a value of 0.

[0131] For example, the evaluation feature is the loan amount, with five evaluation ranges: 0-10,000 yuan, 10,000-100,000 yuan, 100,000-1,000,000 yuan, 1,000,000-10,000,000 yuan, and over 10,000,000 yuan. The target user's actual behavior information is: loan of 900,000 yuan, i.e., the first feature value is 90.

[0132] The generated initial rating matrix is ​​[0, 0, 0, 0, 0]. Based on the first feature value and the 5 rating intervals, it can be determined that the target user's first feature value falls into the 3rd rating interval. The value of the 3rd element in the initial rating matrix is ​​updated to 1, and the target user's rating matrix is ​​[0, 0, 1, 0, 0].

[0133] Figure 4 A flowchart illustrating a method for determining similar users according to an embodiment of this disclosure is shown schematically.

[0134] like Figure 4 As shown, the method 400 for determining similar users in this embodiment includes operations S431 to S433, which can be used as a specific embodiment of operation S230.

[0135] In operation S431, a rating matrix for multiple candidate users is determined based on the rating interval and the second feature values ​​of multiple candidate users.

[0136] In operation S432, the similarity between the target user and multiple candidate users is calculated based on the rating matrix.

[0137] In operation S433, M candidate users with a similarity greater than or equal to the first threshold are identified as M similar users.

[0138] According to embodiments of this disclosure, the method for determining the rating matrix of candidate users is similar to the method for determining the rating matrix of target users, and will not be described again here.

[0139] For example, the evaluation feature is the loan amount, and there are 5 evaluation ranges: 0-10,000 yuan, 10,000-100,000 yuan, 100,000-1,000,000 yuan, 1,000,000-10,000,000 yuan, and over 10,000,000 yuan. If user1 is a candidate user and the loan amount of candidate user1 is 5,000 yuan, then the second feature value of candidate user1 is located in the first rating range of the loan scenario. Therefore, the rating matrix of the candidate user is [1, 0, 0, 0, 0].

[0140] According to embodiments of this disclosure, the cosine similarity between the rating matrix of a candidate user and the rating matrix of a target user can be calculated, and users similar to the target user can be identified based on the cosine similarity.

[0141] The process of calculating the similarity between the target user and the candidate user satisfies:

[0142]

[0143] Where i represents the rating matrix of the target user, j represents the rating matrix of the candidate user, and sim(i,j) represents the similarity between the target user and the candidate user. ||i|| and ||j|| represent the lengths of vectors i and j, respectively.

[0144] For example, i can be [0, 0, 1, 0, 0], and j can be [1, 0, 0, 0, 0].

[0145] According to embodiments of this disclosure, the first threshold can be 1.

[0146] For example, the target user's rating matrix is ​​[0, 0, 1, 0, 0], and the rating matrices of candidate users user4, user5, and user6 are [0, 1, 0, 0, 0], [0, 0, 1, 0, 0], and [0, 0, 1, 0, 0], respectively. Therefore, the similarity between the target user and candidate user4 is calculated to be 0, and the similarity between the target user and candidate users user5 and user6 is 1. Thus, the similar users can be identified as user5 and user6.

[0147] In the embodiments of this disclosure, the calculation of similarity relies on an objectively obtained rating matrix and does not involve subjective evaluation. The similarity determined based on the rating matrix of the target user and the rating matrix of the candidate user is more accurate, and the identified similar users are also more accurate.

[0148] According to embodiments of this disclosure, determining recommended features that match a target user based on historical behavior information of M similar users includes the following steps.

[0149] Based on historical behavior information, determine the rating values ​​for P functions within the application scenario, where P is greater than or equal to 1.

[0150] Calculate the ratings for P functions based on the similarity between M similar users and the target user, and the ratings of the M similar users for the P functions.

[0151] The feature with the highest score among the P feature ratings is identified as the feature to be recommended.

[0152] According to embodiments of this disclosure, similar users may have already used multiple functions across multiple business scenarios. For example, user5 may simultaneously use functions in balance inquiry, remittance, credit card, loan, and wealth management scenarios.

[0153] According to embodiments of this disclosure, determining a user's rating based on historical behavior information includes: determining a rating range for each function based on the feature values ​​of M similar users for each of P functions; comparing the aforementioned rating range with the feature values ​​of each similar user to determine the rating range in which each similar user's feature value falls; and then determining the rating value for each function based on the falling rating range.

[0154] According to embodiments of this disclosure, the score can be determined based on the order of the score intervals. For example, scores within earlier intervals are lower, and scores within later intervals are higher. The score falling into the last score interval is the highest.

[0155] For example, user1 is located in the 5th, 4th, 1st, 1st, and 1st rating intervals of the evaluation features for the balance inquiry, remittance, credit card, loan, and wealth management scenarios, respectively. Therefore, the corresponding rating values ​​are 5, 4, 1, 1, and 1, respectively.

[0156] According to embodiments of this disclosure, based on the historical behavior information of similar users, a rating system is determined for all functions within the application scenario that the similar users have used, resulting in P rating values. Based on the similarity between the similar user and the target user, and the similar user's rating value for a specific function, a weighted value for that function is determined. Then, based on the weighted values ​​of M similar users for that function, and the similarity between the M similar users and the target user, M similar user function ratings for that function are determined. All P function ratings are calculated using the method described above.

[0157] According to embodiments of this disclosure, after calculating P function scores, the function with the highest score among the P function scores is determined as the function to be recommended.

[0158] The embodiments disclosed herein combine the similarity and rating values ​​of multiple similar users to more accurately determine the rating of each function, which helps to improve the accuracy of recommendations.

[0159] According to embodiments of this disclosure, the ratings for P functions are calculated based on the similarity between M similar users and the target user, and the ratings of the M similar users for P functions, including:

[0160] Using similarity as a weight, calculate the weighted sum of the p-th function based on the ratings of M similar users for the p-th function, where p is greater than or equal to 1 and p is less than or equal to P;

[0161] Calculate the sum of similarities between the target user and the M similar users; and

[0162] The ratio of the weighted sum to the sum of similarities is used as the function score for the p-th function.

[0163] According to embodiments of this disclosure, the process of determining the function score of the p-th function satisfies:

[0164]

[0165] Among them, R u,p w represents the feature score of the p-th feature obtained for target user u. u,s R represents the similarity between similar user s and target user u. s,p This represents the rating of similar users s for the p-th feature.

[0166] According to embodiments of this disclosure, during the application phase, since the target user has not used the application scenario before, P function ratings are determined based on the historical behavior information of M similar users, and the function with the highest rating among the P function ratings is identified as the function to be recommended.

[0167] During the testing phase, target users can be those who have already used the feature to be recommended. After identifying the feature to be recommended, the target users' historical behavior information can be used to determine their ratings for the feature. Then, based on the feature ratings and their scores, evaluation metrics are determined. These evaluation metrics are used to assess the quality of the recommendation process.

[0168] According to embodiments of this disclosure, the evaluation metrics include mean absolute error (MAE) and root mean square error (RMSE).

[0169] The process of calculating the mean absolute error satisfies:

[0170]

[0171] Where n is the total number of all target users, p ui It is a predicted rating determined based on similar users, i.e., a feature rating for the feature to be recommended, r ui These are the actual rating values.

[0172] The process of calculating the root mean square error satisfies:

[0173]

[0174] Among them, observed t This represents the predicted value for the t-th target user, i.e., the feature score. t Let represent the true value for the t-th target user, i.e., the rating value determined based on historical behavior information, and N be the total number of all target users. The root mean square error (RMSE) measures the quality of the model by calculating the error between the predicted and true values. The closer the predicted and true values ​​are, the smaller the RMS error, indicating good model accuracy and better recommendation performance.

[0175] According to embodiments of this disclosure, after determining the function to be recommended, with the user's permission, subsequent historical behavior information of the target user can be obtained, and the aforementioned evaluation indicators can be determined based on the historical behavior information.

[0176] Figure 5 The illustration shows a schematic diagram of a method for determining a feature to be recommended according to a specific embodiment of the present disclosure.

[0177] like Figure 5 As shown in the diagram 500, the first feature value 501, the evaluation feature 502, the rating interval 503, the second feature value 504, the target user's rating matrix 505, the candidate user's rating matrix 505, similar users 507, the historical behavior information of similar users 508, and the recommendation function 509 are included.

[0178] According to embodiments of this disclosure, the first feature value 501 may be determined based on the target user's historical behavior information regarding the evaluation feature, and the second feature value 504 may be determined based on the candidate user's historical behavior information regarding the evaluation feature.

[0179] After determining the evaluation feature 502, the feature values ​​of all users in the database can be obtained based on the evaluation feature to determine the rating interval 503. The rating interval 503 can also be determined based on the second feature value of the candidate users.

[0180] According to embodiments of this disclosure, after obtaining the first feature value 501, the first feature value 501 is compared with the rating interval 503 to determine the rating interval into which the first feature value 501 falls, thereby determining the rating matrix 505 of the target user. After obtaining the second feature value 504, the second feature value 504 is compared with the rating interval 503 to determine the rating interval into which the second feature value 504 falls, thereby determining the rating matrix 506 of the candidate user.

[0181] Based on the target user's rating matrix 505 and the candidate user's rating matrix 506, calculate the similarity between the target user and multiple candidate users, and identify one or more similar users 507.

[0182] After identifying similar users 507, the historical behavior information 508 of similar users is determined based on similar users 507, and the functions to be recommended are determined based on the historical behavior information 508 of similar users.

[0183] According to embodiments of this disclosure, the evaluation features corresponding to the business scenario are determined, including:

[0184] For a given business scenario, the input to a pre-trained prediction model is obtained. This model predicts whether a user will use the target function within that scenario. The input to the prediction model includes the user's transaction characteristics, which include behavioral information, user profiles, and asset information. The output of the prediction model represents whether the user uses the target function. The prediction model includes a decision tree model.

[0185] The decision tree model's preset function is invoked to calculate the importance of transaction features; and

[0186] The transaction features with the highest importance are used as evaluation features.

[0187] According to embodiments of this disclosure, the prediction model can include multiple types, each used to predict multiple business scenarios. For example, for five popular functions of a mobile banking application client: remittance, balance inquiry, credit card, loan, and wealth management, models are built using the LightGBM model, ultimately forming five binary classification prediction models. LightGBM is another efficient tree-based algorithm framework after XGBoost, also belonging to the boosting algorithm family. It has advantages such as fast training speed and high accuracy, and supports parallel computing. It is currently widely used in the processing of massive financial data in the banking industry.

[0188] According to embodiments of this disclosure, various decision tree models can also be used to construct predictive models.

[0189] For example, a prediction model for mobile banking functions can be built using logistic regression, random forest, and LightGBM algorithms. By comparing the prediction accuracy, the optimal decision tree model can be selected as the prediction model.

[0190] According to embodiments of this disclosure, the LightGBM model employs a histogram-based decision tree algorithm to discretize continuous features into K values, generating a histogram of width K. When traversing the training samples, the discretized values ​​are used as indices. After each traversal, the histogram stores the statistics for each indicator. Then, the optimal split point is found by traversing the discrete values ​​of the histogram. This traversal method simultaneously reduces memory usage and time complexity.

[0191] Furthermore, during the process of discretizing the data into K data points, the optimal leaf-wise splitting of the leaf nodes is used to determine the number of K. Specifically, before splitting, all leaves in the tree are traversed first, then the leaf with the largest splitting gain is found and split again, and this operation is repeated. Experiments show that leaf-wise splitting achieves higher accuracy with the same number of splits, and a maximum depth limit to prevent overfitting is added to the leaf-wise splitting method during the construction of the prediction model.

[0192] After multiple training and optimization sessions, the final parameter values ​​for the important hyperparameters in the LightGBM model are shown in Table 2.

[0193] Table 2. Parameter values ​​of hyperparameters

[0194]

[0195] According to embodiments of this disclosure, various decision tree models can also be used to construct predictive models.

[0196] For example, a prediction model for mobile banking functions can be built using logistic regression, random forest, and LightGBM algorithms. By comparing the prediction accuracy, the optimal decision tree model can be selected as the prediction model.

[0197] According to embodiments of this disclosure, the inputs to the prediction model include transaction features, which include behavioral information, user profiles, and asset information. The user profile includes age, occupation, gender, etc. Asset information includes time deposits, current deposits, wealth management balances, held products, and assets within the bank, etc. Behavioral information includes the number of debit card transactions, the amount of debit card transactions, the number of credit card transactions, the number of times wealth management products were purchased, and the number of times loans were applied for, etc.

[0198] The aforementioned transaction characteristics were obtained through tracking information on the page with the user's permission.

[0199] According to embodiments of this disclosure, the prediction models for five business scenarios are labeled with the following information: whether the user has logged into the remittance page, whether the user has logged into the balance inquiry page, whether the user has logged into the credit card page, whether the user has logged into the loan page, and whether the user has logged into the wealth management page.

[0200] Before a predictive model can predict whether a user will use a target feature, it needs to process the data input to the model. This data processing includes data cleaning and normalization.

[0201] According to embodiments of this disclosure, for the dataset, numerical features are padded with 0 values, and discrete features are padded with 9999. This ensures both the rationality of the numerical and discrete features and a clear distinction from the original data. For example, numerical features can be amounts, frequency, or time; discrete features can be gender.

[0202] According to embodiments of this disclosure, since the units, properties, and dimensions of the various features differ, when the levels of the features vary significantly, directly using the original numerical values ​​for analysis would emphasize the role of indicators with higher numerical values ​​and weaken the role of indicators with lower numerical values. Therefore, before modeling, max-min normalization is performed on each continuous feature variable to ensure that features with different dimensions are on the same numerical order of magnitude, reducing the influence of features with large variance. The formula used for normalization is:

[0203]

[0204] Where min is the minimum value of samples with the same feature, max is the maximum value of samples with the same feature, and X represents the sample value.

[0205] According to embodiments of this disclosure, after obtaining the input of the prediction model, a preset function of the decision tree model can be called to calculate the importance of all transaction features, and the transaction feature with the highest importance can be used as the evaluation feature.

[0206] According to embodiments of this disclosure, the preset function includes a feature_importances method. For example, LightGBM's feature_importances function can measure the importance of a feature by the number of times it is split or by the gain obtained after splitting the feature. Generally, different measurement criteria will result in different order of feature importance, and multiple measurement criteria exist for multiple business scenarios.

[0207] The embodiments of this disclosure analyze the feature importance of the model, identify the most important features for the predictive model in the business scenario, and use these features as evaluation features to determine the recommended functions for the target user. The LightGBM model compensates for the lack of a rating parameter in the collaborative filtering algorithm, making the value more objective and improving the accuracy of the recommendations.

[0208] According to embodiments of this disclosure, by Figures 6A-6E Evaluation features for five business scenarios in embodiments of this disclosure are demonstrated.

[0209] Figure 6A This illustration schematically shows a diagram illustrating the importance of transaction features in a remittance scenario according to an embodiment of the present disclosure.

[0210] According to embodiments of this disclosure, for a remittance prediction model, the inputs to the model are user profiles: age, occupation, gender, etc.; asset information: time deposits, current deposits, in-bank assets, etc.; and behavioral information: number of debit card transactions, amount of debit card transactions, number of credit card transactions, amount of credit card transactions, etc.

[0211] Remittance prediction models can be used to predict user remittance usage and to identify the features that have the greatest impact on the prediction results. For example... Figure 6A In the remittance prediction model, the "debit card transaction count" feature has the highest importance. Therefore, the "debit card transaction count" is used as an evaluation feature for performing subsequent recommendation operations in remittance scenarios. For example, the "debit card transaction count" can be used as the rating parameter for the "remittance" function in a collaborative filtering model.

[0212] Figure 6B This illustration schematically shows a diagram illustrating the importance of transaction features in a balance query scenario according to an embodiment of the present disclosure.

[0213] According to embodiments of this disclosure, for the balance query prediction model, the inputs to the model are user profiles: age, occupation, gender, etc.; asset information: time deposits, current deposits, in-bank assets, etc.; and behavioral information: number of debit card transactions, debit card transaction amount, number of credit card transactions, credit card transaction amount, number of ATM transactions, etc.

[0214] Balance query prediction models can be used to predict user usage of the balance query function, and the model can also identify the features that have the greatest impact on the results of balance query prediction. For example... Figure 6B In the balance query prediction model, the "ATM transaction count" feature has the highest importance. Therefore, the "ATM transaction count" is used as an evaluation feature for executing subsequent recommendation operations in the balance query scenario. For example, the "ATM transaction count" can be used as the rating parameter for the "balance query" function in the collaborative filtering model.

[0215] Figure 6C This illustration schematically shows the importance of transaction features in a financial management scenario according to an embodiment of the present disclosure.

[0216] According to embodiments of this disclosure, for a wealth management prediction model, the inputs to the model are: user profile (age, occupation, gender, etc.); asset information (time deposits, current deposits, in-bank assets, wealth management scale, etc.); and behavioral information (number of debit card transactions, amount of debit card transactions, number of credit card transactions, amount of credit card transactions, amount of wealth management products purchased, etc.).

[0217] Financial prediction models can be used to predict user spending on financial management features, and the models can also identify the features that have the greatest impact on the prediction results. For example... Figure 6CIn wealth management prediction models, the "scale of wealth management" feature has the highest importance. Therefore, "scale of wealth management" is used as an evaluation feature for subsequent recommendation operations in wealth management scenarios. For example, "scale of wealth management" can be used as the rating parameter for the "wealth management" function in collaborative filtering models.

[0218] Figure 6D This illustration schematically shows the importance of transaction features in a loan scenario according to an embodiment of the present disclosure.

[0219] According to embodiments of this disclosure, for a loan prediction model, the inputs to the model are: user profile (age, occupation, gender, etc.); asset information (time deposits, current deposits, bank assets, etc.); and behavioral information (number of debit card transactions, amount of debit card transactions, number of credit card transactions, amount of credit card transactions, amount of loan transactions, etc.).

[0220] Loan models can be used to predict user loan usage and to identify the features that have the greatest impact on the loan prediction model's results. For example... Figure 6D In the loan prediction model, the "loan transaction amount" feature has the highest importance; therefore, it is used as the evaluation feature for subsequent recommendation operations in loan scenarios. For example, the "loan transaction amount" can be used as the rating parameter for the "loan" function in a collaborative filtering model.

[0221] Figure 6E This illustration schematically shows the importance of transaction features in a credit card scenario according to an embodiment of the present disclosure.

[0222] According to embodiments of this disclosure, for a credit card prediction model, the inputs to the model are user profiles: age, occupation, gender, etc.; asset information: time deposits, current deposits, in-bank assets, credit card holdings, etc.; and behavioral information: number of debit card transactions, debit card transaction amount, number of credit card transactions, credit card transaction amount, etc.

[0223] Credit card prediction models can be used to predict user credit card usage and identify the features that have the greatest impact on the prediction results. For example... Figure 6E In credit card prediction models, the "number of credit cards held" feature has the highest importance. Therefore, the "number of credit cards held" is used as an evaluation feature for subsequent recommendation operations in credit card scenarios. For example, the "number of credit cards held" can be used as the rating parameter for the "credit card" function in a collaborative filtering model.

[0224] Figure 7 A schematic block diagram of a function recommendation device for application software according to an embodiment of the present disclosure is shown.

[0225] like Figure 7As shown, the application software function recommendation device 700 of this embodiment includes a feature determination module 710, a matrix determination module 720, a user determination module 730, and a function determination module 740.

[0226] The feature determination module 710 is used to determine the evaluation features corresponding to the business scenario. The evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario. In one embodiment, the feature determination module 710 can be used to perform the operation S210 described above, which will not be repeated here.

[0227] The matrix determination module 720 is used to determine a rating matrix for the target user based on the evaluation features. In one embodiment, the matrix determination module 720 can be used to perform the operation S220 described above, which will not be repeated here.

[0228] User determination module 730 is used to determine M similar users from multiple candidate users based on the target user's rating matrix, where M is greater than or equal to 2. In one embodiment, user determination module 730 can be used to perform the operation S230 described above, which will not be repeated here.

[0229] The function determination module 740 is used to determine the recommended functions that match the target user based on the historical behavior information of M similar users. In one embodiment, the function determination module 740 can be used to perform the operation S240 described above, which will not be repeated here.

[0230] According to embodiments of this disclosure, the matrix determination module 720 includes a first determination unit, a second determination unit, and a third determination unit.

[0231] The first determining unit is used to determine N scoring intervals that match the evaluation features, where N is greater than or equal to 2. In one embodiment, the first determining unit can be used to perform the operation S321 described above, which will not be repeated here.

[0232] The second determining unit is used to obtain the first feature value of the target user for the evaluation features. In one embodiment, the second determining unit can be used to perform the operation S322 described above, which will not be repeated here.

[0233] The third determining unit is used to determine the rating matrix of the target user based on the first feature value and the rating interval. In one embodiment, the third determining unit can be used to perform the operation S323 described above, which will not be repeated here.

[0234] According to embodiments of this disclosure, the third determining unit includes a first determining subunit, a second determining subunit, and a third determining subunit.

[0235] The first determining subunit is used to generate an initial scoring matrix according to the order of N scoring intervals. The initial scoring matrix includes N elements, and the element value of each element is 0.

[0236] The second determining subunit is used to determine the nth scoring interval, which includes the first feature value, from N scoring intervals, where n is greater than or equal to 1 and n is less than or equal to N.

[0237] The third determining sub-unit is used to modify the element value corresponding to the nth scoring interval in the initial scoring matrix to 1, thus obtaining the scoring matrix.

[0238] According to embodiments of this disclosure, the first determining unit includes a fourth determining subunit, a fifth determining subunit, and a sixth determining subunit.

[0239] The fourth determining subunit is used to obtain the second feature values ​​of multiple candidate users based on the evaluation features.

[0240] The fifth determining subunit is used to obtain the minimum and maximum second eigenvalues ​​from a plurality of second eigenvalues.

[0241] The sixth sub-unit is used to divide the smallest and largest second eigenvalues ​​at equal intervals, resulting in N scoring intervals.

[0242] According to embodiments of this disclosure, the user determination module 730 includes a fourth determination unit, a fifth determination unit, and a sixth determination unit.

[0243] The fourth determining unit is used to determine a rating matrix for multiple candidate users based on the rating interval and the second feature values ​​of multiple candidate users. In one embodiment, the fourth determining unit can be used to perform the operation S431 described above, which will not be repeated here.

[0244] The fifth determining unit is used to calculate the similarity between the target user and multiple candidate users based on the rating matrix. In one embodiment, the fifth determining unit can be used to perform the operation S432 described above, which will not be repeated here.

[0245] The sixth determining unit is used to determine the M candidate users whose similarity is greater than or equal to the first threshold as M similar users. In one embodiment, the sixth determining unit can be used to perform the operation S433 described above, which will not be repeated here.

[0246] According to embodiments of this disclosure, the function determination module 740 includes a seventh determination unit, an eighth determination unit, and a ninth determination unit.

[0247] The seventh determining unit is used to determine the score value of P functions in the application scenario based on historical behavior information, where P is greater than or equal to 1.

[0248] The eighth determining unit is used to calculate the scores of P functions based on the similarity between M similar users and the target user, and the ratings of the M similar users for the P functions.

[0249] The ninth determination unit is used to determine the function with the highest score among the P function scores as the function to be recommended.

[0250] According to embodiments of this disclosure, the eighth determining unit includes a seventh determining subunit, an eighth determining subunit, and a ninth determining subunit.

[0251] The seventh determination subunit is used to use similarity as a weight to calculate the weighted sum of the p-th function based on the ratings of M similar users for the p-th function, where p is greater than or equal to 1 and p is less than or equal to P.

[0252] The eighth determination subunit is used to calculate the sum of similarities between the M similar users and the target user.

[0253] The ninth determining subunit is used to take the ratio of the weighted sum to the sum of similarities as the function score of the p-th function.

[0254] According to embodiments of this disclosure, the feature determination module 710 includes an input feature acquisition unit, an importance calculation unit, and an evaluation feature determination unit.

[0255] The input feature acquisition unit is used to acquire the input of the pre-trained prediction model for the business scenario. The prediction model is used to predict whether the user will use the target function in the business scenario. The input of the prediction model includes the user's transaction features, which include behavioral information, user profile and asset information. The output of the prediction model represents whether the user uses the target function. The prediction model includes a decision tree model.

[0256] The importance calculation unit is used to call the preset function of the decision tree model to calculate the importance of transaction features.

[0257] The evaluation feature determination unit is used to select the transaction features with the highest importance as evaluation features.

[0258] According to embodiments of this disclosure, any multiple modules among the feature determination module 710, matrix determination module 720, user determination module 730, and function determination module 740 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module.

[0259] According to embodiments of this disclosure, at least one of the feature determination module 710, matrix determination module 720, user determination module 730, and function determination module 740 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the feature determination module 710, matrix determination module 720, user determination module 730, and function determination module 740 can be at least partially implemented as a computer program module, which, when run, can perform corresponding functions.

[0260] Figure 8 A block diagram of an electronic device for a function recommendation method suitable for application software according to an embodiment of the present disclosure is illustrated schematically.

[0261] like Figure 8 As shown, an electronic device 800 according to an embodiment of this disclosure includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage portion 808 into a random access memory (RAM) 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this disclosure.

[0262] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 802 and / or RAM 803. It should be noted that the programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0263] According to embodiments of this disclosure, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the I / O interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the I / O interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0264] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0265] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 802 and / or RAM 803 and / or one or more memories other than ROM 802 and RAM 803 described above.

[0266] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this disclosure.

[0267] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0268] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0269] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of this disclosure embodiment. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0270] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0271] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0272] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0273] The specific embodiments described above further illustrate the purpose, technical solutions, and beneficial effects of this disclosure. It should be understood that the above descriptions are merely specific embodiments of this disclosure and are not intended to limit this disclosure. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the protection scope of this disclosure.

Claims

1. A method for recommending features in application software, comprising: The evaluation features corresponding to the business scenario are determined. The evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario. The evaluation features are determined by a trained prediction model. Based on the evaluation characteristics, determine the rating matrix for the target user; Based on the target user's rating matrix, determine M similar users from multiple candidate users, where M is greater than or equal to 2; as well as Based on the historical behavior information of the M similar users, determine the recommended functions that match the target user; Determining the rating matrix for the target user based on the evaluation features includes: Based on the evaluation features, second feature values ​​of multiple candidate users are obtained; the second feature values ​​are obtained by using historical behavioral information related to the evaluation features as feature values. Obtain the minimum and maximum second eigenvalues ​​from multiple second eigenvalues; using the minimum second eigenvalue as the minimum value and the maximum second eigenvalue as the maximum value, divide the minimum and maximum values ​​into N equal parts using the equal-interval binning method to obtain N scoring intervals. The N scoring intervals have equal lengths and exist in a size order, and N is greater than or equal to 2. Obtain the first feature value of the target user for the evaluation feature; and An initial rating matrix is ​​generated according to the order of the N rating intervals. The initial rating matrix includes N elements, and the value of each element is 0. From the N scoring intervals, determine the nth scoring interval that includes the first feature value, where n is greater than or equal to 1 and n is less than or equal to N; and The element value corresponding to the nth scoring interval in the initial scoring matrix is ​​modified to 1 to obtain the scoring matrix.

2. The method according to claim 1, wherein, The step of determining M similar users from multiple candidate users based on the target user's rating matrix includes: Based on the rating interval and the second feature values ​​of the multiple candidate users, a rating matrix for the multiple candidate users is determined; Based on the rating matrix, the similarity between the target user and the plurality of candidate users is calculated; and The M candidate users whose similarity is greater than or equal to the first threshold are identified as the M similar users.

3. The method according to claim 2, wherein, The step of determining the recommended function that matches the target user based on the historical behavior information of the M similar users includes: Based on the historical behavior information of the M similar users, determine the rating values ​​of the M similar users for P functions in the business scenario, where P is greater than or equal to 1; Calculate the ratings for P functions based on the similarity between the M similar users and the target user, and the ratings of the M similar users for the P functions. The function with the highest score among the P function ratings is identified as the function to be recommended.

4. The method according to claim 3, wherein, The step of calculating the ratings for P functions based on the similarity between the M similar users and the target user, and the ratings of the M similar users for the P functions, includes: Using the similarity as a weight, a weighted sum of the p-th function is calculated based on the ratings of the M similar users for the p-th function, where p is greater than or equal to 1 and p is less than or equal to P; Calculate the sum of similarities between the M similar users and the target user; and The ratio of the weighted sum to the sum of the similarities is used as the function score for the p-th function.

5. The method according to claim 1, wherein, The determination of evaluation features corresponding to the business scenario includes: For the aforementioned business scenario, the input of a pre-trained prediction model is obtained. The prediction model is used to predict whether a user will use the target function in the business scenario. The input of the prediction model includes the user's transaction characteristics, which include behavioral information, user profile, and asset information. The output of the prediction model represents whether the user uses the target function. The prediction model includes a decision tree model. The importance of the transaction features is calculated by calling a preset function of the decision tree model; and The transaction feature with the highest importance is used as the evaluation feature.

6. A function recommendation device for application software, comprising: The feature determination module is used to determine the evaluation features corresponding to the business scenario. The evaluation features are the transaction features with the highest importance in the business scenario, and the importance indicates the degree of influence of the transaction features in the business scenario. The evaluation features are determined by a trained prediction model. The matrix determination module is used to determine a rating matrix for the target user based on the evaluation features. The user identification module is used to identify M similar users from multiple candidate users based on the target user's rating matrix, where M is greater than or equal to 2; as well as The function determination module is used to determine the recommended functions that match the target user based on the historical behavior information of the M similar users. The matrix determination module is further configured to: obtain second feature values ​​of multiple candidate users based on the evaluation features; The second feature value is obtained by using historical behavioral information for the evaluation feature as the feature value; Obtain the minimum and maximum second eigenvalues ​​from a plurality of second eigenvalues; Using the smallest second eigenvalue as the minimum value and the largest second eigenvalue as the maximum value, the area between the minimum and maximum values ​​is divided into N equal parts using the equidistant binning method, resulting in N rating intervals. The N rating intervals have equal lengths and are ordered in size, with N greater than or equal to 2. The first eigenvalue of the target user for the evaluation feature is obtained. An initial rating matrix is ​​generated according to the order of the N rating intervals, containing N elements, each with a value of 0. The nth rating interval, containing the first eigenvalue, is determined from the N rating intervals, where n is greater than or equal to 1 and less than or equal to N. The element value corresponding to the nth rating interval in the initial rating matrix is ​​modified to 1 to obtain the rating matrix.

7. An electronic device, comprising: One or more processors; Storage device for storing one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors perform the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the method according to any one of claims 1 to 5.

9. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images