Business recommendation method and related equipment based on artificial intelligence

Abstract

This application proposes an artificial intelligence-based business recommendation method, device, electronic device, and storage medium. The artificial intelligence-based business recommendation method includes: analyzing business scenarios and preset business plans to obtain rated business indicators; collecting user data from the business database corresponding to the business scenarios based on the rated business indicators to construct a target user set; classifying each piece of user data in the target user set to obtain multiple user clusters; grouping the user data in each user cluster to obtain user group confidence; evaluating the business plan corresponding to the business scenario based on the user group confidence, and recommending the business plan based on the evaluation result. This method can improve the accuracy of business recommendations by reducing the user's feature orientation by multiple classifications.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based business recommendation method, device, electronic device, and storage medium. Background Art

[0002] With the development of digitalization, more and more business fields have started online operation models. For example, the financial industry and the trade industry have emerged with a variety of mobile applications with consulting and transaction functions. Enterprises will periodically recommend updated versions of business solutions in the application to users to attract users to use mobile applications and thus increase revenue. With the growing number of online users, it has become particularly important to accurately recommend business solutions to users.

[0003] Currently, users are usually randomly selected to conduct business plan experiments to determine the calibrated business plans to be recommended. However, this business recommendation method does not take into account the characteristic orientation of randomly selected customers, which may lead to inaccurate business recommendation results. Summary of the Invention

[0004] In view of the above, it is necessary to provide an artificial intelligence-based business recommendation method and related equipment to solve the technical problem of how to improve the accuracy of business recommendations, wherein the related equipment includes an artificial intelligence-based business recommendation device, an electronic device and a storage medium.

[0005] This embodiment of the present application provides a business recommendation method based on artificial intelligence, the method comprising:

[0006] Analyze business scenarios and pre-set business solutions to obtain rated business indicators. The business scenario refers to the problem that the enterprise needs to solve, and the business solution refers to the solution designed by the enterprise for the business scenario. The rated business indicators are used to characterize the quality of the business solution.

[0007] Based on the rated business indicators, user data is collected from the business database corresponding to the business scenario to build a target user set;

[0008] Classifying each piece of user data in the target user set to obtain multiple user clusters;

[0009] Grouping user data in each user cluster to obtain user grouping confidence;

[0010] The service solution corresponding to the service scenario is evaluated based on the user group confidence, and the service solution is recommended according to the evaluation result.

[0011] In the above-mentioned business recommendation method, the target user set participating in the test is determined by analyzing the business scenario, and the target user set is divided into multiple user clusters through the clustering method. The user data in the user cluster is then divided into different categories and reorganized into multiple user groups. The business recommendation results are evaluated based on the confidence level of each group of users, so that the characteristics of each group of users are as balanced as possible, thereby improving the accuracy of subsequent business recommendation results.

[0012] In some embodiments, collecting user data from a business database corresponding to the business scenario based on the rated business indicator to construct a target user set includes:

[0013] Marking user data in a business database corresponding to the business scenario according to the business indicator; if the business indicator in the user data is less than the rated business indicator, the user data is marked as a target user;

[0014] Collect statistics on the characteristics of the target user, including user age, user gender, user online time, and user preference category;

[0015] The features corresponding to each target user are used as user data, and the user data are stored as the target user set.

[0016] In this way, by comparing the business indicators and rated business indicators of user data, target users are collected and marked from the business database and stored as a target user set, providing sample support for subsequent user grouping and business recommendations, and improving the accuracy of business recommendations.

[0017] In some embodiments, the target user's characteristics include numerical data and non-numerical data. After counting the target user's characteristics, the method further includes:

[0018] Identify whether each user feature is non-numeric data, and mark each user feature according to the identification result, wherein the mark includes "yes" and "no";

[0019] Traverse the mark of each user feature in turn. If a user feature is marked as "yes", count the number of value types of the feature, mark each type in the feature in turn according to a natural number, and use the marked natural number as the encoding value of each data in the feature to obtain a numerical feature. If a user feature is marked as "no", do nothing;

[0020] Replace each data in the user feature marked as "yes" with its corresponding coded value to update the target user set.

[0021] In this way, by encoding the non-numerical data in the features of the user data to obtain fully quantified data, sample support can be provided for subsequent service recommendations, thereby improving the accuracy of service recommendations.

[0022] In some embodiments, classifying each piece of user data in the target user set to obtain multiple user clusters includes:

[0023] Calculate the cosine distance between each two pieces of user data using the cosine distance algorithm;

[0024] Classifying the target user set according to the cosine distance and a preset clustering algorithm to obtain a plurality of user clusters, wherein each user cluster includes a plurality of user data;

[0025] Counting the number and aggregation degree of user data in each user cluster, where the aggregation degree is used to characterize the diversity of features of the user data;

[0026] The quantity and aggregation degree of the user data are input into a customized integration model to obtain an integration result, which is used as the weight of the user cluster.

[0027] In this way, multiple user clusters are obtained by classifying the user data set, and the number and aggregation degree of user data in each user cluster are counted respectively, and then the weight of each user cluster is obtained based on the number and aggregation degree. The data can be divided into multiple clusters with feature similarity, and a weight is assigned to each cluster to characterize the importance of the user data in each cluster when recommending business, which can improve the accuracy of subsequent business recommendations.

[0028] In some embodiments, the customized integration model satisfies the following relationship:

[0029]

[0030] Among them, T i represents the weight of the i-th user cluster. The higher the weight, the higher the credibility of the test result when the service solution is tested using the user data in the cluster. i Represents the number of user data in the i-th user cluster, the A i The larger the value of B, the more user data is included in the cluster, and the higher the weight of the cluster should be; i represents the aggregation degree of the i-th user cluster. The lower the value of the aggregation degree, the more discrete the characteristics of the user data in the cluster, and the higher the weight of the cluster should be.

[0031] In this way, the integration result of the quantity and aggregation degree of the user data is calculated through a customized integration model. The value of the integration result can characterize the degree to which the user data in the user cluster has characteristic directionality, and can provide data support for the subsequent calculation of user grouping confidence, thereby improving the accuracy of business recommendations.

[0032] In some embodiments, grouping the user data in each user cluster to obtain the user grouping confidence includes:

[0033] Using a preset clustering algorithm to perform binary classification on the user data in each user cluster, and labeling the user data in the user cluster according to the category, wherein the labels include "experimental" and "control";

[0034] Calculate the aggregation degree of user data in each category of each user cluster as the aggregation value;

[0035] The product of the weight corresponding to each user cluster and the aggregate value is used as the confidence corresponding to each category;

[0036] Combine user data with the same tags to form the experimental group and the control group;

[0037] The confidence level of the experimental group is calculated as the sum of the confidence levels of the user data, and the confidence level of the control group is calculated as the sum of the confidence levels of the user data.

[0038] In this way, by classifying the users in each user cluster, multiple categories of users are obtained, and the confidence of each category is calculated based on the degree of aggregation of user data in each category. The higher the confidence value, the higher the credibility of the business recommendation results obtained when business recommendations are made using this group of users as samples, thereby improving the accuracy of subsequent business recommendations.

[0039] In some embodiments, evaluating the service solution corresponding to the service scenario based on the user group confidence and recommending the service solution according to the evaluation result includes:

[0040] Calculating the mean of the business indicators of the user data in the experimental group and the control group respectively as the benchmark value for each group of users;

[0041] Randomly push a service plan to each group of users, and after a preset test period, calculate the average of the service indicators of each group of users as the test value;

[0042] Calculating the difference between the test value and the benchmark value to obtain service increments for each group of users;

[0043] Calculating the product of the confidence and the service increment of each group of users as the service recommendation result for each group of users;

[0044] The business plan corresponding to the larger business recommendation result is used as the recommended calibration business plan.

[0045] In this way, by pushing different versions of the app to two groups of users and obtaining feedback on the users' online time, the business increment of each group of users was further corrected based on the confidence of each group to obtain the final experimental results, thereby improving the accuracy of the business recommendation results.

[0046] The present application also provides an artificial intelligence-based service recommendation device, comprising:

[0047] An acquisition unit, used to analyze business scenarios and preset business plans to obtain rated business indicators;

[0048] a collection unit, configured to collect user data from a service database corresponding to the service scenario based on the rated service indicator to construct a target user set;

[0049] a classification unit, configured to classify each piece of data in the target user set to obtain a plurality of user clusters;

[0050] a grouping unit, configured to group user data in each user cluster to obtain user grouping confidence;

[0051] A recommendation unit is configured to evaluate a business solution corresponding to the business scenario based on the confidence level to obtain a business recommendation result.

[0052] An embodiment of the present application further provides an electronic device, the device comprising:

[0053] a memory storing computer-readable instructions; and

[0054] A processor executes the computer-readable instructions stored in the memory to implement the artificial intelligence-based business recommendation method.

[0055] An embodiment of the present application also provides a computer-readable storage medium, in which computer-readable instructions are stored. The computer-readable instructions are executed by a processor in an electronic device to implement the artificial intelligence-based business recommendation method.

[0056] In the above-mentioned business recommendation method, the target user set participating in the test is determined by analyzing the business scenario, and the target user set is divided into multiple user clusters through clustering method. The user data in the user cluster is then divided into different categories and reorganized into experimental and control groups. The business recommendation results are evaluated and corrected based on the confidence level of each group of users, so that the characteristics of each group of users are as balanced as possible, thereby improving the accuracy of subsequent business recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a flowchart of a preferred embodiment of an artificial intelligence-based business recommendation method involved in this application.

[0058] Figure 2 This is a functional module diagram of a preferred embodiment of the artificial intelligence-based business recommendation device involved in this application.

[0059] Figure 3 It is a structural diagram of an electronic device of a preferred embodiment of the artificial intelligence-based business recommendation method involved in this application.

[0060] Figure 4 It is a structural diagram of the target user set involved in this application. DETAILED DESCRIPTION

[0061] In order to more clearly understand the purpose, features and advantages of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the case of no conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. In the following description, many specific details are set forth to facilitate a full understanding of the present application. The embodiments described are only a part of the embodiments of the present application, rather than all of the embodiments.

[0062] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the described features. Throughout the description of this application, "plurality" means two or more, unless otherwise specifically defined.

[0063] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein in the specification of this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0064] An embodiment of the present application provides an artificial intelligence-based business recommendation method, which can be applied to one or more electronic devices. The electronic device is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0065] The electronic device can be any electronic product that can interact with a user, such as a personal computer, a tablet computer, a smart phone, a personal digital assistant (PDA), a game console, an interactive network television (IPTV), a smart wearable device, etc.

[0066] The electronic device may further include a network device and / or a user device, wherein the network device includes, but is not limited to, a single network server, a server group consisting of multiple network servers, or a cloud based on cloud computing consisting of a large number of hosts or network servers.

[0067] The network where the electronic device is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN), etc.

[0068] like Figure 1 FIG2 is a flowchart of a preferred embodiment of the business recommendation method based on artificial intelligence of the present application. According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0069] S10, analyzing the business scenario and the preset business plan to obtain rated business indicators, wherein the business scenario refers to the problem that the enterprise needs to solve, the business plan refers to the solution designed by the enterprise for the business scenario, and the rated business indicators are used to characterize the quality of the business plan.

[0070] In this optional embodiment, the business scenario can be a problem that the enterprise needs to solve during operations. For example, the business scenario in this solution can refer to a certain app under the enterprise experiencing a decrease in user activity over the past period. The full name of the app is "Application," which means mobile software. User activity refers to the length of time users spend online within a particular app.

[0071] In this optional embodiment, the business plan may be a solution designed by the enterprise for the business scenario. For example, the business plan in this plan may refer to the modification and update of the APP page by the enterprise in order to improve the user activity of the APP. The preset business plan may include plan R1 and plan R2. Plan R1 may represent the design plan of the current APP, and plan R2 may represent another design plan obtained by modifying the design of the APP. The design plan of the APP may refer to the way the image of the user page in the APP is displayed or the way the transition animation moves.

[0072] In this optional embodiment, since the business scenario is that a company's app is experiencing a decline in user activity, the rated business indicator can be the average online time TIME_total of all users in the app's historical records over the past 30 calendar days. The average online time is the average of the total online time of each user over the past 30 calendar days, and "online time" refers to the time span from each time a user opens the app to each time they close it. Generally, the longer a user's online time on an app, the higher the user's attention or recognition of the app, and the greater the likelihood that the company will profit from the app. Therefore, the online time of each user over the past 30 days can be used as the user's business indicator. The business indicator can be represented by time, which represents the total online time in the user's historical records. For example, if a user logged into the app five times in the past 30 days, and the online time after each login was 20 seconds, then the user's total online time over the past 30 days is 100 seconds, which means that the user's business indicator is 100 seconds.

[0073] In this optional embodiment, the historical average online time of all users in the APP represented by TIME_total can be used as the rated business indicator, and the online time of each user in the past 30 days represented by time can be used as the business indicator corresponding to each user.

[0074] In this way, the rated business indicators are obtained by analyzing the business scenario and the preset business plan. The rated business indicators can represent the quantitative data of concern in the business scenario and can provide data support in subsequent business recommendations to improve the accuracy of business recommendation results.

[0075] S11 , collecting user data from a business database corresponding to the business scenario based on the rated business indicator to construct a target user set.

[0076] In an optional embodiment, collecting user data from the service database corresponding to the service scenario based on the rated service metrics to construct a target user set includes:

[0077] S111. Mark the user data in the service database corresponding to the service scenario according to the service metrics. If the service metrics in the user data are less than the rated service metrics, the user data is marked as a target user.

[0078] In this optional embodiment, the database is a data set designed, stored, and managed according to a data structure. Exemplarily, the service database in this solution can be a memSQL database, which is an open-source database and its function can be to store the user data corresponding to the APP.

[0079] In this optional embodiment, the specific method of marking the user data in the service database according to the service metrics is to run a preset first program in the memSQL database to obtain the program return value. The preset program can be an SQL statement, and its form can be "select 'table' from 'database' where time < TIME_total", where 'table' represents the result returned by the preset program, 'database' represents the database where the target user set is located, and the keyword "where" in the SQL statement represents conditional screening when collecting data. The condition is "time < TIME_total", where time represents the service metrics of the user, used to represent the total online duration in the historical record of a certain user, and TIME_total represents the rated service metrics. Since the screening condition in the preset program is "time < TIME_total", all the user data in the 'table' represents users with service metrics less than the rated service metrics. Therefore, each piece of data in the 'table' can be marked as the target user.

[0080] S112. Statistically analyze the characteristics of the target users. The characteristics include user age, user gender, user online duration, and user preference category.

[0081] In this optional embodiment, the characteristics of the target user can be counted. A specific implementation method can be to run a preset second program in the business database to obtain the program return value. The second program can be in the form of "selecta,b,c,dastable2fromtable". Where a can represent the age characteristic of the target user, b can represent the gender characteristic of the target user, c can represent the online time of the target user, that is, the business indicator corresponding to each target user, d can represent the preference category of the target user, and table2 can represent a data table composed of the characteristics of the target user.

[0082] In this optional embodiment, the characteristics of the target user include numerical data and non-numerical data. After counting the characteristics of the target user, the method further includes:

[0083] Identify whether each user feature is non-numeric data, and mark each user feature according to the identification result, wherein the mark includes "yes" and "no";

[0084] Traverse the mark of each user feature in turn. If a user feature is marked as "yes", count the number of value types of the feature, mark each type in the feature in turn according to a natural number, and use the marked natural number as the encoding value of each data in the feature to obtain a numerical feature. If a user feature is marked as "no", do nothing;

[0085] Replace each data in the user feature marked as "yes" with its corresponding coded value to update the target user set.

[0086] In this optional embodiment, whether each user feature belongs to non-numeric data can be determined based on a preset Python program. Taking the age feature represented by a as an example, the preset Python program can be in the form of "print(dtype(a) == int)", which means comparing the data format of the age feature represented by a with the "integer type" data format and outputting the comparison result. The comparison result includes "YES" and "NO". If the comparison result is "YES", the age feature represented by a is marked as "No", and if the comparison result is "NO", the age feature represented by a is marked as "Yes".

[0087] In this optional embodiment, the features of the user data can be traversed in sequence to obtain their labels. If the label of a column of features is "yes", the number of value types of the data in the feature is counted, and each type is marked in sequence according to a natural number, and the natural number label is used as the corresponding code of the data in the feature.

[0088] For example, taking the gender feature represented by b as an example, the gender feature includes {female, male}, and the natural number "female" is marked as 1, and the natural number "male" is marked as 2.

[0089] In this optional embodiment, the natural number label corresponding to each data in the non-numeric feature can be used as the encoding value corresponding to the data, and each data is replaced with its corresponding encoding value.

[0090] S113: Taking the features corresponding to each target user as user data, and storing the user data as the target user set.

[0091] In this optional embodiment, the features corresponding to each target user may be arranged column by column to construct a target user set. The target user set may be in the form of a data table, where each row represents a piece of user data and each column represents a feature of the user data.

[0092] For example, the first column in the target user set may represent the user's age a, which is a positive integer and can be in the range of [0, +∞]; the second column may represent the user's gender b, which can be in the range of {1, 2}; the third column may represent the user's online time c in the APP, which is a positive integer and can be in the range of [0, +∞]; the fourth column may represent the user's preferred content d, which can be in the range of [1, +∞]. For example, Figure 4 Schematic diagram of the structure of the target user set.

[0093] In this way, by marking the target users, counting the characteristics of each user data and storing them as the target user set, the non-numerical data in the characteristics of the user data are encoded to obtain fully quantified data, which can provide sample support for subsequent business recommendations and thus improve the accuracy of business recommendations.

[0094] S12: Classify each piece of user data in the target user set to obtain multiple user clusters.

[0095] In an optional embodiment, classifying each piece of user data in the target user set to obtain multiple user clusters includes:

[0096] S121, calculating the cosine distance between every two pieces of user data according to a cosine distance algorithm.

[0097] In this optional embodiment, each piece of user data may be used as a coding vector based on the arrangement order of the features of the user data. For example, the arrangement order of the features of the user data may be a, b, c, d, and the coding vector of a certain piece of user data may be [20, 2, 360, 1].

[0098] In this optional embodiment, the distance between each user data can be calculated according to a preset distance measurement algorithm. The preset distance measurement algorithm can be a cosine distance algorithm. For example, if the vector A corresponding to a certain user data is A=[a A ,b A ,c A ,d A ], and the vector corresponding to the other user data is B=[a B ,b B ,c B ,d B ], then the specific calculation method of the cosine distance between vectors A and B is:

[0099]

[0100] Among them, d A,B represents the cosine similarity between vector A and vector B, the A·B part represents the inner product of vector A and vector B, ‖A‖ represents the modulus of vector A, and ‖B‖ represents the modulus of vector B.

[0101] The inner product and modulus are calculated as follows:

[0102] A·B=a A ×a B +b A ×b B +c A ×c B +d A ×d b

[0103]

[0104]

[0105] The subscripts A and B represent the vectors of feature dependencies of a certain dimension.

[0106] For example, when A=[1,2,3,4] and B=[2,3,4,5], the d A,B The calculation method is:

[0107]

[0108] Then the cosine distance between the vector A and the vector B is 0.01.

[0109] In this optional embodiment, the cosine distance between the user data is used to characterize the similarity between two users. If the distance between the two user data is larger, it indicates that the similarity between the two users is lower.

[0110] S122 , classifying the target user set according to the cosine distance and a preset clustering algorithm to obtain a plurality of user clusters, wherein each user cluster includes a plurality of user data.

[0111] In this optional embodiment, the target user set may be classified according to a preset clustering algorithm to obtain multiple user clusters. The specific implementation process of the preset clustering algorithm is as follows:

[0112] A1: Mark all user data in the target user set as unvisited;

[0113] A2: Randomly select one of the user data marked as unvisited and mark it as X, and further mark the user data represented by X as visited;

[0114] A3: Select the next step to be performed according to the preset judgment condition. The preset judgment condition is that the cosine distance between the user data represented by X is less than or equal to the preset radius d e There are n pieces of user data. If n is less than the preset threshold minpts, proceed to step A4; otherwise, proceed to step A12. For example, the d e It can be 0.5, and the minpts can be 4;

[0115] A4: Create a new cluster and record it as C, and add the user X to C;

[0116] A5: Let the user data represented by X be the center and the radius be d e The user data set in the neighborhood of is N;

[0117] A6: Randomly select a piece of user data from the N and record it as Y;

[0118] A7: If the label of Y is unvisited, mark Y as visited;

[0119] A8: If the user data represented by Y is the center and the radius is d e If there are at least minpts objects in the neighborhood of , then these objects are added to the set N;

[0120] A9: If Y does not belong to any cluster, add Y to cluster C;

[0121] A10: Repeat steps A6 to A10 until N is an empty set;

[0122] A11: Note that cluster C is the first user cluster and is marked as C i , where i represents the number of cycles and the initial value of i can be set to 1;

[0123] A12: Mark X as a noise point;

[0124] A13: Repeat steps A2 to A13 until all points in the original user data set are marked as visited. The algorithm ends and multiple user clusters are obtained. The set of multiple user clusters is recorded as C. z , and can be recorded as C z =[C 1 ,C 2 ,…,C n ], where n is the C z The number of clusters in .

[0125] S123 , counting the number and aggregation degree of user data in each user cluster respectively, where the aggregation degree is used to characterize the diversity of features of the user data.

[0126] In this optional embodiment, the amount of user data in each user cluster may be counted. For example, the amount of user data in each cluster may be recorded as A={A 1 ,A 2 ,…,A n}, where A represents the set of the number of user data, each element in A corresponds to the number of user data in a certain user cluster, and the superscript of each element represents the index of the user cluster corresponding to the element.

[0127] In this optional embodiment, the variance of the user data in each user cluster can be calculated as the aggregation degree of the user data in each user cluster. The larger the variance, the greater the difference in the user data in the user cluster. The lower the aggregation degree of the user data in the user cluster, the less obvious the feature directionality of the user data in the user cluster. Therefore, the test results obtained when using the user data in the user cluster for business recommendation are more accurate.

[0128] S124 , inputting the quantity and aggregation degree of the user data into a customized integration model to obtain an integration result as the weight of the user cluster.

[0129] In this optional embodiment, the quantity and aggregation degree of the user data may be input into a custom integration model to obtain an integration result, and the custom integration model satisfies the following relationship:

[0130]

[0131] Among them, T i A represents the integrated result of the quantity and aggregation degree of user data in the i-th user cluster. The higher the value of the integrated result, the higher the credibility of the test result when the user data in the cluster is used as a sample to test the business solution.i Represents the number of user data in the i-th user cluster, the A i The larger the value of B, the more user data is included in the cluster, and the more important the cluster should be; i represents the aggregation degree of the i-th user cluster. The lower the value of the aggregation degree, the more discrete the features of the user data in the cluster are, and the higher the importance of the cluster should be.

[0132] For example, when A i =20, B i =5, the T i The representative integration result is calculated as:

[0133]

[0134] In this optional embodiment, the integrated result T of the quantity and aggregation degree of the user data in each user cluster may be used as the weight of the corresponding user cluster.

[0135] In this way, multiple user clusters are obtained by classifying the user data set, and the number and aggregation degree of user data in each user cluster are counted respectively, and then the weight of each user cluster is obtained based on the number and aggregation degree. The data can be divided into multiple clusters with feature similarity, and a weight is assigned to each cluster to characterize the importance of the user data in each cluster when recommending business, which can improve the accuracy of subsequent business recommendations.

[0136] S13, grouping the user data in each user cluster to obtain user grouping confidence.

[0137] In an optional embodiment, grouping the user data in each user cluster to obtain the user grouping confidence includes:

[0138] S131 , using a preset clustering algorithm to perform binary classification on the user data in each user cluster, and marking the user data in the user cluster according to the category, wherein the labels include “experimental” and “control”.

[0139] In this optional embodiment, each user cluster can be divided into two categories to obtain user data of two categories. The preset clustering algorithm can be a K-means clustering algorithm, and its specific implementation process is as follows:

[0140] A1, setting the number of categories K according to the number of the business solutions. For example, in this solution, K=2;

[0141] A2, randomly select K user data from all user data of a user cluster as the centroid. For example, in this solution, the K centroids can be denoted as K1 and K2 respectively;

[0142] A3, traverse each user data in the user cluster, and calculate the Euclidean distance between each user data and each centroid. For example, the Euclidean distance between a certain user data and K1 can be recorded as D1, and the Euclidean distance between it and K2 can be recorded as D2. If D1 is less than D2, then the user data is classified into the category where K1 is located. Otherwise, the user data is classified into the category where K2 is located. After the traversal is completed, two categories are obtained, each category containing multiple user data.

[0143] A4: Calculate the mean of the user data in each category and record them as K1_mean and K2_mean. For each user data in each category, calculate the Euclidean distance between it and the mean of the corresponding category. Select the user data with the smallest Euclidean distance as the new centroid of each category and record them as K1_new and K2_new.

[0144] A5, calculate the Euclidean distance between K1_new and K1 and record it as D1_new, calculate the Euclidean distance between K2_new and K2 and record it as D2_new, and record the preset distance threshold as Dt;

[0145] A6: If both D1_new and D2_new are less than Dt, the algorithm terminates and the category where K1 and K2 are located are used as the final classification results. Otherwise, steps A3 to A6 are repeated until the algorithm terminates.

[0146] In this optional embodiment, the classification result of each user cluster can be obtained based on the execution process of the above Kmeans algorithm. Each user cluster can be divided into two categories. Each category in each user cluster can be recorded as G = [(G 1实验 ,G 1对照 ),(G 2实验 ,G 2对照 ),…,(G n实验 ,G n对照 )], exemplary, G 1实验 represents the category marked as “experiment” in the user cluster with index 1, G 1 对照 Represents the category labeled as “control” in the user cluster with index 1, where each category contains multiple user data.

[0147] S132 , respectively calculating the aggregation degree of the user data in each category of each user cluster as an aggregation value.

[0148] In this optional embodiment, the variance of the user data of each category in each user cluster may be calculated. The variance is used to characterize the degree of aggregation of users in each category. The calculation formula for the variance is:

[0149]

[0150] Where S represents the variance of user data in a certain category, which is used to characterize the degree of aggregation of user data in this category. A higher value of S indicates a lower degree of aggregation of user data in this category, and the user data in this category has stronger characteristic directionality; u represents the number of user data in this category; j represents the index of user data in this category; x represents the value of a certain user data in this category; Represents the mean value of user data in this category.

[0151] For example, when u=3, x1=[1, 2, 3, 4], x2=[2, 3, 4, 5], and x3=[3, 4, 5, 6], S is calculated as follows:

[0152]

[0153] In this optional embodiment, the inverse of the variance S can be used as the aggregation value corresponding to each category in each user cluster. The higher the aggregation value, the lower the characteristic directionality of the user data in the corresponding category, and the more accurate the test results obtained when the user data in the category is used as a sample for business recommendation.

[0154] S133: The product of the weight corresponding to each user cluster and the aggregation value is used as the confidence corresponding to each category.

[0155] In this optional embodiment, the product of the weight corresponding to each user cluster and the aggregate value can be used as the confidence corresponding to each category, and the confidence satisfies the following relationship:

[0156]

[0157] Among them, t we Represents the confidence level of the category marked as e in the user cluster with index w. The higher the confidence level, the higher the confidence level of the user data in this category. we Represents the variance corresponding to the user data in this category, T w Represents the confidence of the user cluster with index w.

[0158] For example, when S 1实验 =3.622, T 1 =0.982, the t we The confidence of the representative is calculated as:

[0159]

[0160] In this optional embodiment, the t we The value of t is used as the confidence corresponding to each category. weThe higher the value, the higher the credibility of the test results obtained when using user data in this category as samples for business recommendation.

[0161] S134: Combine user data with the same tag to form an experimental group and a control group.

[0162] In this optional embodiment, categories with the same label in all user clusters can be combined into an experimental group and a control group. Exemplarily, this solution can be expressed as:

[0163] Experimental group = [G 1实验 ,G 2实验 ,…,G n实验 ];

[0164] Control group = [G 1对照 ,G 2对照 ,…,G n对照 ].

[0165] In this optional embodiment, both the experimental group and the control group contain multiple user data.

[0166] S135 , calculating the sum of the confidences of the user data in the experimental group as the confidence of the experimental group, and calculating the sum of the confidences of the data in the control group as the confidence of the control group.

[0167] In this optional embodiment, the sum of the confidences of the user data in the experimental group can be calculated as the confidence of the experimental group. For example, if there are 3 user clusters in this solution and t 1实验 =0.271, t 2实验 =0.28, t 3实验 =0.29, then the confidence level of the experimental group is calculated as:

[0168] Texperiment = 0.271 + 0.28 + 0.29 = 0.841

[0169] In this optional embodiment, the sum of the confidences of the user data in the control group can be calculated as the confidence of the control group. For example, if there are 3 user clusters in this solution, and t 1对照 =0.29, t 2对照 =0.289, t 3对照 =0.35, then the confidence level of the control group is calculated as:

[0170] T control = 0.29 + 0.289 + 0.35 = 0.929

[0171] In this optional embodiment, the normalized confidence of the experimental group and the control group can be calculated according to the maximum normalization algorithm. For example, when Texperimental = 0.841 and Tcontrol = 0.929, the normalized confidence of the experimental group is calculated as follows:

[0172]

[0173] The normalized confidence of the control group is calculated as follows:

[0174]

[0175] In this optional embodiment, the normalized confidence can be used as the user grouping confidence of the experimental group and the control group respectively.

[0176] In this way, by classifying the users in each user cluster, multiple categories of users are obtained, and the confidence of each category is calculated based on the degree of aggregation of user data in each category. The higher the confidence value, the higher the credibility of the business recommendation results obtained when business recommendations are made using this group of users as samples, thereby improving the accuracy of subsequent business recommendations.

[0177] S14: Evaluate a service solution corresponding to the service scenario based on the user group confidence, and recommend the service solution according to the evaluation result.

[0178] In this optional embodiment, evaluating the service solution corresponding to the service scenario based on the user group confidence level and recommending the service solution according to the evaluation result includes:

[0179] S141, respectively calculating the mean of the service indicators of the user data in the experimental group and the control group to serve as the benchmark value of each group of users.

[0180] In this optional embodiment, the mean of the service indicators of the user data in the experimental group and the control group can be calculated respectively. The service indicator is the online time in the user data. For example, when the experimental group includes 10 users and the mean online time in the user data in the experimental group is 300 seconds, it can be recorded as When the control group contains 20 users and the average online time in the user data of the control group is 360 seconds, it can be recorded as

[0181] In this optional embodiment, the The value of is used as the benchmark value of the users in the experimental group, and the The value of is used as the benchmark value for users in the control group.

[0182] S142: randomly push a service plan to each group of users, and after a preset test period, calculate the average value of the service indicators of each group of users as a test value.

[0183] In this optional embodiment, the enterprise can push the app with the R1 service plan to the users in the experimental group and push the app with the R2 service plan to the users in the control group, and can count the average online time of the users in the experimental group after the preset test period and record it as The average online time of users in the control group is counted and recorded as

[0184] In this optional embodiment, the The value of is taken as the test value of the experimental group, and the The value of was taken as the test value of the control group.

[0185] S143: Calculate the difference between the test value and the reference value to obtain the service increment of each group of users.

[0186] In this optional embodiment, the service increment of the experimental group can be recorded as V 实验 , the business increment of the control group can be recorded as V 对照 , the calculation method of the business increment of the experimental group is:

[0187]

[0188] In this optional embodiment, the business increment of the control group is calculated as follows:

[0189]

[0190] In this optional embodiment, the V 实验 The value of V 对照 The value of is used as the business increment of the control group.

[0191] S144: Calculate the product of the confidence and the service increment of each group of users as the service recommendation result for each group of users.

[0192] In this optional embodiment, the confidence level Tg 实验 and Tg 对照 The service increment of each user group is corrected to obtain a service recommendation result for each user group. The service recommendation result is calculated as follows:

[0193] R 实验 =Tg 实验 ×V 实验

[0194] R 对照 =Tg 对照 ×V 对照

[0195] Among them, Tg 实验 Represents the confidence level of the experimental group, Tg 对照 Represents the confidence of the control group, V 实验 represents the business increment of the experimental group, V 对照 represents the business increment of the control group, R 实验 represents the business recommendation results of the experimental group, R 对照 Represents the business recommendation results of the control group.

[0196] S145 , taking the business solution corresponding to the larger business recommendation result as the recommended calibration business solution.

[0197] In this optional embodiment, the R 实验 The value of R 对照 The business plan corresponding to the larger value is used as the calibration business plan for enterprise promotion.

[0198] For example, when R 实验 =10 and R 对照 =20, since the business plan pushed by the enterprise to the users in the control group is the R2 business plan, the R2 business plan can be used as the calibration business plan for enterprise promotion.

[0199] In this way, by pushing different versions of the app to two groups of users and obtaining feedback on the users' online time, the business increment of each group of users was further corrected based on the confidence of each group to obtain the final experimental results, thereby improving the accuracy of the business recommendation results.

[0200] In the above-mentioned business recommendation method, the target user set participating in the test is determined by analyzing the business scenario, and the target user set is divided into multiple user clusters through clustering method. The user data in the user cluster is then divided into different categories and reorganized into experimental and control groups. The business recommendation results are evaluated and corrected based on the confidence level of each group of users, so that the characteristics of each group of users are as balanced as possible, thereby improving the accuracy of subsequent business recommendations.

[0201] like Figure 2, which is a functional module diagram of a preferred embodiment of an artificial intelligence-based business recommendation device provided in an embodiment of the present application. The artificial intelligence-based business recommendation device 11 includes an acquisition unit 110, a collection unit 111, a classification unit 112, a grouping unit 113, and a recommendation unit 114. The modules / units referred to herein refer to a series of computer program segments that can be executed by the processor 13 and perform fixed functions, and are stored in the memory 12. In this embodiment, the functions of each module / unit will be described in detail in subsequent embodiments.

[0202] In an optional embodiment, the acquisition unit 110 is used to analyze business scenarios and preset business plans to obtain rated business indicators, where the business scenario refers to a problem that the enterprise needs to solve, and the business plan refers to a solution designed by the enterprise for the business scenario. The rated business indicators are used to characterize the quality of the business plan.

[0203] In this optional embodiment, the business scenario can be a problem that the enterprise needs to solve during operations. For example, the business scenario in this solution can refer to a certain app under the enterprise experiencing a decrease in user activity over the past period. The full name of the app is "Application," which means mobile software. User activity refers to the length of time users spend online within a particular app.

[0204] In this optional embodiment, the business plan may be a solution designed by the enterprise for the business scenario. For example, the business plan in this plan may refer to the modification and update of the APP page by the enterprise in order to improve the user activity of the APP. The preset business plan may include plan R1 and plan R2. Plan R1 may represent the design plan of the current APP, and plan R2 may represent another design plan obtained by modifying the design of the APP. The design plan of the APP may refer to the way the image of the user page in the APP is displayed or the way the transition animation moves.

[0205] In this optional embodiment, since the business scenario is that a company's app is experiencing a decline in user activity, the rated business indicator can be the average online time TIME_total of all users in the app's historical records over the past 30 calendar days. The average online time is the average of the total online time of each user over the past 30 calendar days, and "online time" refers to the time span from each time a user opens the app to each time they close it. Generally, the longer a user's online time on an app, the higher the user's attention or recognition of the app, and the greater the likelihood that the company will profit from the app. Therefore, the online time of each user over the past 30 days can be used as the user's business indicator. The business indicator can be represented by time, which represents the total online time in the user's historical records. For example, if a user logged into the app five times in the past 30 days, and the online time after each login was 20 seconds, then the user's total online time over the past 30 days is 100 seconds, which means that the user's business indicator is 100 seconds.

[0206] In this optional embodiment, the historical average online time of all users in the APP represented by TIME_total can be used as the rated business indicator, and the online time of each user in the past 30 days represented by time can be used as the business indicator corresponding to each user.

[0207] In an optional embodiment, the collecting unit 111 is configured to collect user data from a service database corresponding to the service scenario based on the rated service indicator to construct a target user set.

[0208] In an optional embodiment, collecting user data from a business database corresponding to the business scenario based on the rated business indicator to construct a target user set includes:

[0209] Marking user data in a business database corresponding to the business scenario according to the business indicator; if the business indicator in the user data is less than the rated business indicator, the user data is marked as a target user;

[0210] Collect statistics on the characteristics of the target user, including user age, user gender, user online time, and user preference category;

[0211] The features corresponding to each target user are used as user data, and the user data are stored as the target user set.

[0212] In this optional embodiment, the database is a data set that designs, stores, and manages data information according to a data structure. Exemplarily, the business database in this solution can be a memSQL database, which is an open-source database, and its function can be to store user data corresponding to the APP.

[0213] In this optional embodiment, the specific method of marking user data in the business database according to business indicators is to run a preset first program in the memSQL database to obtain a program return value. The preset program can be an SQL statement, and its form can be "select 'table' from 'database' where time < TIME_total", where 'table' represents the result returned by the preset program, 'database' represents the database where the target user set is located, and the keyword "where" in the SQL statement represents conditional filtering when collecting data. The condition is "time < TIME_total", where time represents the business indicator of the user, used to represent the total online duration in the historical record of a certain user, and TIME_total represents the rated business indicator. Since the filtering condition in the preset program is "time < TIME_total", all user data in the 'table' represents users with business indicators less than the rated business indicator. Therefore, each piece of data in the 'table' can be marked as the target user.

[0214] In this optional embodiment, the characteristics of the target user can be statistically analyzed. The specific implementation method can be to run a preset second program in the business database to obtain a program return value. The form of the second program can be "select a, b, c, d as table2 from table". Among them, a can represent the age characteristic of the target user, b can represent the gender characteristic of the target user, c can represent the online duration of the target user, that is, the business indicator corresponding to each target user, d can represent the preference category of the target user, and table2 can represent the data table composed of the characteristics of the target user.

[0215] In this optional embodiment, the characteristics of the target user include numerical data and non-numerical data. After statistically analyzing the characteristics of the target user, the method further includes:

[0216] Identify whether each user characteristic belongs to non-numerical data, and mark each user characteristic according to the identification result. The marks include "yes" and "no";

[0217] Traverse the mark of each user feature in turn. If a user feature is marked as "yes", count the number of value types of the feature, mark each type in the feature in turn according to a natural number, and use the marked natural number as the encoding value of each data in the feature to obtain a numerical feature. If a user feature is marked as "no", do nothing;

[0218] Replace each data in the user feature marked as "yes" with its corresponding coded value to update the target user set.

[0219] In this optional embodiment, whether each user feature belongs to non-numeric data can be determined based on a preset Python program. Taking the age feature represented by a as an example, the preset Python program can be in the form of "print(dtype(a) == int)", which means comparing the data format of the age feature represented by a with the "integer type" data format and outputting the comparison result. The comparison result includes "YES" and "NO". If the comparison result is "YES", the age feature represented by a is marked as "No", and if the comparison result is "NO", the age feature represented by a is marked as "Yes".

[0220] In this optional embodiment, the features of the user data can be traversed in sequence to obtain their labels. If the label of a column of features is "yes", the number of value types of the data in the feature is counted, and each type is marked in sequence according to a natural number, and the natural number label is used as the corresponding code of the data in the feature.

[0221] For example, taking the gender feature represented by b as an example, the gender feature includes {female, male}, and the natural number "female" is marked as 1, and the natural number "male" is marked as 2.

[0222] In this optional embodiment, the natural number label corresponding to each data in the non-numeric feature can be used as the encoding value corresponding to the data, and each data is replaced with its corresponding encoding value.

[0223] In this optional embodiment, the features corresponding to each target user may be arranged column by column to construct a target user set. The target user set may be in the form of a data table, where each row represents a piece of user data and each column represents a feature of the user data.

[0224] For example, the first column in the target user set may represent the user's age a, which is a positive integer and can be in the range of [0, +∞]; the second column may represent the user's gender b, which can be in the range of {1, 2}; the third column may represent the user's online time c in the APP, which is a positive integer and can be in the range of [0, +∞]; the fourth column may represent the user's preferred content d, which can be in the range of [1, +∞]. For example, Figure 4 is a schematic diagram of the target user set.

[0225] In an optional embodiment, the classification unit 112 is configured to classify each piece of user data in the target user set to obtain multiple user clusters.

[0226] In this optional embodiment, classifying each piece of user data in the target user set to obtain multiple user clusters includes:

[0227] Calculate the cosine distance between each two pieces of user data using the cosine distance algorithm;

[0228] Classifying the target user set according to the cosine distance and a preset clustering algorithm to obtain a plurality of user clusters, wherein each user cluster includes a plurality of user data;

[0229] Counting the number and aggregation degree of user data in each user cluster, where the aggregation degree is used to characterize the diversity of features of the user data;

[0230] The quantity and aggregation degree of the user data are input into a customized integration model to obtain an integration result, which is used as the weight of the user cluster.

[0231] In this optional embodiment, each piece of user data may be used as a coding vector based on the arrangement order of the features of the user data. For example, the arrangement order of the features of the user data may be a, b, c, d, and the coding vector of a certain piece of user data may be [20, 2, 360, 1].

[0232] In this optional embodiment, the distance between each user data can be calculated according to a preset distance measurement algorithm. The preset distance measurement algorithm can be a cosine distance algorithm. For example, if the vector A corresponding to a certain user data is A=[a A ,b A ,c A ,d A ], and the vector corresponding to the other user data is B=[a B ,b B ,c B ,d B ], then the specific calculation method of the cosine distance between vectors A and B is:

[0233]

[0234] Among them, d A,B represents the cosine similarity between vector A and vector B, the A·B part represents the inner product of vector A and vector B, ‖A‖ represents the modulus of vector A, and ‖B‖ represents the modulus of vector B.

[0235] The inner product and modulus are calculated as follows:

[0236] A·B=a A ×a B +b A ×b B +c A ×c B +d A ×d b

[0237]

[0238]

[0239] The subscripts A and B represent the vectors of feature dependencies of a certain dimension.

[0240] For example, when A=[1,2,3,4] and B=[2,3,4,5], the d A,B The calculation method is:

[0241]

[0242] Then the cosine distance between the vector A and the vector B is 0.01.

[0243] In this optional embodiment, the cosine distance between the user data is used to characterize the similarity between two users. If the distance between the two user data is larger, it indicates that the similarity between the two users is lower.

[0244] In this optional embodiment, the target user set may be classified according to a preset clustering algorithm to obtain multiple user clusters. The specific implementation process of the preset clustering algorithm is as follows:

[0245] A1: Mark all user data in the target user set as unvisited;

[0246] A2: Randomly select one of the user data marked as unvisited and mark it as X, and further mark the user data represented by X as visited;

[0247] A3: Select the next step to be performed according to the preset judgment condition. The preset judgment condition is that the cosine distance between the user data represented by X is less than or equal to the preset radius d e There are n pieces of user data. If n is less than the preset threshold minpts, proceed to step A4; otherwise, proceed to step A12. For example, the d e It can be 0.5, and the minpts can be 4;

[0248] A4: Create a new cluster and record it as C, and add the user X to C;

[0249] A5: Let the user data represented by X be the center and the radius be d e The user data set in the neighborhood of is N;

[0250] A6: Randomly select a piece of user data from the N and record it as Y;

[0251] A7: If the label of Y is unvisited, mark Y as visited;

[0252] A8: If the user data represented by Y is the center and the radius is d e If there are at least minpts objects in the neighborhood of , then these objects are added to the set N;

[0253] A9: If Y does not belong to any cluster, add Y to cluster C;

[0254] A10: Repeat steps A6 to A10 until N is an empty set;

[0255] A11: Note that cluster C is the first user cluster and is marked as C i , where i represents the number of cycles and the initial value of i can be set to 1;

[0256] A12: Mark X as a noise point;

[0257] A13: Repeat steps A2 to A13 until all points in the original user data set are marked as visited. The algorithm ends and multiple user clusters are obtained. The set of multiple user clusters is recorded as C. z , and can be recorded as C z =[C 1 ,C 2 ,…,C n ], where n is the C z The number of clusters in .

[0258] In this optional embodiment, the amount of user data in each user cluster may be counted. For example, the amount of user data in each cluster may be recorded as A={A 1 ,A 2 ,…,A n}, where A represents the set of the number of user data, each element in A corresponds to the number of user data in a certain user cluster, and the superscript of each element represents the index of the user cluster corresponding to the element.

[0259] In this optional embodiment, the variance of the user data in each user cluster can be calculated as the aggregation degree of the user data in each user cluster. The larger the variance, the greater the difference in the user data in the user cluster. The lower the aggregation degree of the user data in the user cluster, the less obvious the feature directionality of the user data in the user cluster. Therefore, the test results obtained when using the user data in the user cluster for business recommendation are more accurate.

[0260] In this optional embodiment, the quantity and aggregation degree of the user data may be input into a custom integration model to obtain an integration result, and the custom integration model satisfies the following relationship:

[0261]

[0262] Among them, T i A represents the integrated result of the quantity and aggregation degree of user data in the i-th user cluster. The higher the value of the integrated result, the higher the credibility of the test result when the user data in the cluster is used as a sample to test the business solution. i Represents the number of user data in the i-th user cluster, the A i The larger the value of B, the more user data is included in the cluster, and the more important the cluster should be; i represents the aggregation degree of the i-th user cluster. The lower the value of the aggregation degree, the more discrete the features of the user data in the cluster are, and the higher the importance of the cluster should be.

[0263] For example, when A i =20, B i =5, the T i The representative integration result is calculated as:

[0264]

[0265] In this optional embodiment, the integrated result T of the quantity and aggregation degree of user data in each user cluster may be used as the weight of the corresponding user cluster.

[0266] In an optional embodiment, the grouping unit 113 is configured to group the user data in each user cluster to obtain user grouping confidence.

[0267] In this optional embodiment, grouping the user data in each user cluster to obtain the user grouping confidence includes:

[0268] Using a preset clustering algorithm to perform binary classification on the user data in each user cluster, and labeling the user data in the user cluster according to the category, wherein the labels include "experimental" and "control";

[0269] Calculate the aggregation degree of user data in each category of each user cluster as the aggregation value;

[0270] The product of the weight corresponding to each user cluster and the aggregate value is used as the confidence corresponding to each category;

[0271] Combine user data with the same tags to form the experimental group and the control group;

[0272] The confidence level of the experimental group is calculated as the sum of the confidence levels of the user data, and the confidence level of the control group is calculated as the sum of the confidence levels of the user data.

[0273] In this optional embodiment, each user cluster can be divided into two categories to obtain user data of two categories. The preset clustering algorithm can be a K-means clustering algorithm, and its specific implementation process is as follows:

[0274] A1, setting the number of categories K according to the number of the business solutions. For example, in this solution, K=2;

[0275] A2, randomly select K user data from all user data of a user cluster as the centroid. For example, in this solution, the K centroids can be denoted as K1 and K2 respectively;

[0276] A3, traverse each user data in the user cluster, and calculate the Euclidean distance between each user data and each centroid. For example, the Euclidean distance between a certain user data and K1 can be recorded as D1, and the Euclidean distance between it and K2 can be recorded as D2. If D1 is less than D2, then the user data is classified into the category where K1 is located. Otherwise, the user data is classified into the category where K2 is located. After the traversal is completed, two categories are obtained, each category containing multiple user data.

[0277] A4: Calculate the mean of the user data in each category and record them as K1_mean and K2_mean. For each user data in each category, calculate the Euclidean distance between it and the mean of the corresponding category. Select the user data with the smallest Euclidean distance as the new centroid of each category and record them as K1_new and K2_new.

[0278] A5, calculate the Euclidean distance between K1_new and K1 and record it as D1_new, calculate the Euclidean distance between K2_new and K2 and record it as D2_new, and record the preset distance threshold as Dt;

[0279] A6: If both D1_new and D2_new are less than Dt, the algorithm terminates and the category where K1 and K2 are located are used as the final classification results. Otherwise, steps A3 to A6 are repeated until the algorithm terminates.

[0280] In this optional embodiment, the classification result of each user cluster can be obtained based on the execution process of the above Kmeans algorithm. Each user cluster can be divided into two categories. Each category in each user cluster can be recorded as G = [(G 1实验 ,G 1对照 ),(G 2实验 ,G 2对照 ),…,(G n实验 ,G n对照 )], exemplary, G 1实验 represents the category marked as “experiment” in the user cluster with index 1, G 1 对照 Represents the category labeled as “control” in the user cluster with index 1, where each category contains multiple user data.

[0281] In this optional embodiment, the variance of the user data of each category in each user cluster may be calculated. The variance is used to characterize the degree of aggregation of users in each category. The calculation formula for the variance is:

[0282]

[0283] Where S represents the variance of user data in a certain category, which is used to characterize the degree of aggregation of user data in this category. A higher value of S indicates a lower degree of aggregation of user data in this category, and the user data in this category has stronger characteristic directionality; u represents the number of user data in this category; j represents the index of user data in this category; x represents the value of a certain user data in this category; Represents the mean value of user data in this category.

[0284] For example, when u=3, x1=[1, 2, 3, 4], x2=[2, 3, 4, 5], and x3=[3, 4, 5, 6], S is calculated as follows:

[0285]

[0286] In this optional embodiment, the inverse of the variance S can be used as the aggregation value corresponding to each category in each user cluster. The higher the aggregation value, the lower the characteristic directionality of the user data in the corresponding category, and the more accurate the test results obtained when the user data in the category is used as a sample for business recommendation.

[0287] In this optional embodiment, the product of the weight corresponding to each user cluster and the aggregate value can be used as the confidence corresponding to each category, and the confidence satisfies the following relationship:

[0288]

[0289] Among them, t we Represents the confidence level of the category marked as e in the user cluster with index w. The higher the confidence level, the higher the confidence level of the user data in this category. we Represents the variance corresponding to the user data in this category, T w Represents the confidence of the user cluster with index w.

[0290] For example, when S 1实验 =3.622, T 1 =0.982, the t we The confidence of the representative is calculated as:

[0291]

[0292] In this optional embodiment, the t we The value of t is used as the confidence corresponding to each category. we The higher the value, the higher the credibility of the test results obtained when using user data in this category as samples for business recommendation.

[0293] In this optional embodiment, categories with the same label in all user clusters can be combined into an experimental group and a control group. Exemplarily, this solution can be expressed as:

[0294] Experimental group = [G 1实验 ,G 2实验 ,…,G n实验 ];

[0295] Control group = [G 1对照 ,G 2对照 ,…,G n对照 ].

[0296] In this optional embodiment, both the experimental group and the control group contain multiple user data.

[0297] In this optional embodiment, the sum of the confidences of the user data in the experimental group can be calculated as the confidence of the experimental group. For example, if there are 3 user clusters in this solution and t 1实验 =0.271, t 2实验 =0.28, t 3实验 =0.29, then the confidence level of the experimental group is calculated as:

[0298] Texperiment = 0.271 + 0.28 + 0.29 = 0.841

[0299] In this optional embodiment, the sum of the confidences of the user data in the control group can be calculated as the confidence of the control group. For example, if there are 3 user clusters in this solution, and t 1对照 =0.29, t2对照 =0.289, t 3对照 =0.35, then the confidence level of the control group is calculated as:

[0300] T control = 0.29 + 0.289 + 0.35 = 0.929

[0301] In this optional embodiment, the normalized confidence of the experimental group and the control group can be calculated according to the maximum normalization algorithm. For example, when Texperimental = 0.841 and Tcontrol = 0.929, the normalized confidence of the experimental group is calculated as follows:

[0302]

[0303] The normalized confidence of the control group is calculated as follows:

[0304]

[0305] In this optional embodiment, the normalized confidence can be used as the user grouping confidence of the experimental group and the control group respectively.

[0306] In an optional embodiment, the recommendation unit 114 is configured to evaluate the service solution corresponding to the service scenario based on the user group confidence level, and recommend the service solution according to the evaluation result.

[0307] In this optional embodiment, evaluating the service solution corresponding to the service scenario based on the user group confidence, and recommending the service solution according to the evaluation result includes:

[0308] Calculating the mean of the business indicators of the user data in the experimental group and the control group respectively as the benchmark value for each group of users;

[0309] Randomly push a service plan to each group of users, and after a preset test period, calculate the average of the service indicators of each group of users as the test value;

[0310] Calculating the difference between the test value and the benchmark value to obtain service increments for each group of users;

[0311] Calculating the product of the confidence and the service increment of each group of users as the service recommendation result for each group of users;

[0312] The business plan corresponding to the larger business recommendation result is used as the recommended calibration business plan.

[0313] In this optional embodiment, the mean of the service indicators of the user data in the experimental group and the control group can be calculated respectively. The service indicator is the online time in the user data. For example, when the experimental group includes 10 users and the mean online time in the user data in the experimental group is 300 seconds, it can be recorded as When the control group contains 20 users and the average online time in the user data of the control group is 360 seconds, it can be recorded as

[0314] In this optional embodiment, the The value of is used as the benchmark value of the users in the experimental group, and the The value of is used as the benchmark value for users in the control group.

[0315] In this optional embodiment, the enterprise can push the app with the R1 service plan to the users in the experimental group and push the app with the R2 service plan to the users in the control group, and can count the average online time of the users in the experimental group after the preset test period and record it as The average online time of users in the control group is counted and recorded as

[0316] In this optional embodiment, the The value of is taken as the test value of the experimental group, and the The value of was taken as the test value of the control group.

[0317] In this optional embodiment, the service increment of the experimental group can be recorded as V 实验 , the business increment of the control group can be recorded as V 对照 , the calculation method of the business increment of the experimental group is:

[0318]

[0319] In this optional embodiment, the business increment of the control group is calculated as follows:

[0320]

[0321] In this optional embodiment, the V 实验 The value of V control can be used as the business increment of the experimental group, and the value of V control can be used as the business increment of the control group.

[0322] In this optional embodiment, the confidence level Tg 实验 and Tg 对照 The service increment of each user group is corrected to obtain a service recommendation result for each user group. The service recommendation result is calculated as follows:

[0323] R 实验 =Tg 实验 ×V 实验

[0324] R 对照 =Tg 对照 ×V 对照

[0325] Among them, Tg 实验 Represents the confidence level of the experimental group, Tg 对照 Represents the confidence of the control group, V 实验 represents the business increment of the experimental group, V 对照 represents the business increment of the control group, R 实验 represents the business recommendation results of the experimental group, R 对照 Represents the business recommendation results of the control group.

[0326] In this optional embodiment, the R 实验 The value of R 对照 The business plan corresponding to the larger value is used as the calibration business plan for enterprise promotion.

[0327] For example, when R 实验 =10 and R 对照 =20, since the business plan pushed by the enterprise to the users in the control group is the R2 business plan, the R2 business plan can be used as the calibration business plan for enterprise promotion.

[0328] like Figure 3 FIG2 is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Electronic device 1 includes memory 12 and processor 13. Memory 12 is used to store computer-readable instructions, and processor 13 executes the computer-readable instructions stored in memory to implement the artificial intelligence-based business recommendation method of any of the above embodiments.

[0329] In an optional embodiment, the electronic device 1 further includes a bus, a computer program stored in the memory 12 and executable on the processor 13, such as a business recommendation program based on artificial intelligence.

[0330] Figure 3 Only the electronic device 1 having components 12-13 is shown, and it can be understood by those skilled in the art that Figure 3 The structure shown does not constitute a limitation on the electronic device 1 , and the electronic device 1 may include fewer or more components than shown in the figure, or combine certain components, or arrange the components differently.

[0331] Combine Figure 1The memory 12 in the electronic device 1 stores a plurality of computer-readable instructions to implement a business recommendation method based on artificial intelligence, and the processor 13 can execute the plurality of instructions to implement:

[0332] Analyze business scenarios and pre-set business solutions to obtain rated business indicators. The business scenario refers to the problem that the enterprise needs to solve, and the business solution refers to the solution designed by the enterprise for the business scenario. The rated business indicators are used to characterize the quality of the business solution.

[0333] Based on the rated business indicators, user data is collected from the business database corresponding to the business scenario to build a target user set;

[0334] Classifying each piece of user data in the target user set to obtain multiple user clusters;

[0335] Grouping user data in each user cluster to obtain user grouping confidence;

[0336] The service solution corresponding to the service scenario is evaluated based on the user group confidence, and the service solution is recommended according to the evaluation result.

[0337] Specifically, the specific implementation method of the processor 13 for the above instructions can refer to Figure 1 The description of the relevant steps in the corresponding embodiments will not be repeated here.

[0338] Those skilled in the art will understand that the schematic diagram is merely an example of the electronic device 1 and does not constitute a limitation on the electronic device 1. The electronic device 1 may have either a bus structure or a star structure. The electronic device 1 may also include more or less other hardware or software than shown in the figure, or a different arrangement of components. For example, the electronic device 1 may also include input and output devices, network access devices, etc.

[0339] It should be noted that the electronic device 1 is only an example, and other existing or future electronic products that are suitable for this application should also be included in the scope of protection of this application and incorporated herein by reference.

[0340] Among them, the memory 12 includes at least one type of readable storage medium, and the readable storage medium can be non-volatile or volatile. The readable storage medium includes flash memory, mobile hard disk, multimedia card, card-type memory (for example: SD or DX memory, etc.), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 12 can be an internal storage unit of the electronic device 1, such as a mobile hard disk of the electronic device 1. In other embodiments, the memory 12 can also be an external storage device of the electronic device 1, such as a plug-in mobile hard disk, smart memory card (SmartMediaCard, SMC), secure digital (SecureDigital, SD) card, flash card (FlashCard), etc. equipped on the electronic device 1. Furthermore, the memory 12 can also include both an internal storage unit of the electronic device 1 and an external storage device. The memory 12 can not only be used to store application software and various types of data installed on the electronic device 1, such as the code of a business recommendation program based on artificial intelligence, but can also be used to temporarily store data that has been output or is to be output.

[0341] In some embodiments, the processor 13 may be composed of an integrated circuit, such as a single packaged integrated circuit or a plurality of packaged integrated circuits with the same or different functions, including one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and a combination of various control chips. The processor 13 is the control core (Control Unit) of the electronic device 1, connecting the various components of the entire electronic device 1 using various interfaces and circuits. It executes or executes programs or modules stored in the memory 12 (such as executing an artificial intelligence-based business recommendation program) and calls data stored in the memory 12 to perform various functions of the electronic device 1 and process data.

[0342] The processor 13 executes the operating system of the electronic device 1 and various installed applications. The processor 13 executes the applications to implement the steps in the above-mentioned embodiments of the business recommendation method based on artificial intelligence, such as Figure 1 Steps shown.

[0343] Exemplarily, the computer program may be divided into one or more modules / units, which are stored in the memory 12 and executed by the processor 13 to complete the present application. The one or more modules / units may be a series of computer-readable instruction segments capable of performing specific functions, which are used to describe the execution process of the computer program in the electronic device 1. For example, the computer program may be divided into an acquisition unit 110, a collection unit 111, a classification unit 112, a grouping unit 113, and a recommendation unit 114.

[0344] The above-mentioned integrated unit implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, computer device, or network device, etc.) or a processor to execute part of the artificial intelligence-based business recommendation method described in various embodiments of the present application.

[0345] If the modules / units integrated in the electronic device 1 are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the processes in the above-mentioned embodiment methods, and can also instruct the relevant hardware devices to complete them through a computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by a processor, it can implement the steps of each of the above-mentioned method embodiments.

[0346] The computer program includes computer program code, which may be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium may include any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory, or other memory.

[0347] Furthermore, the computer-readable storage medium may mainly 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, etc.; the data storage area may store data created according to the use of the blockchain node, etc.

[0348] The blockchain referred to in this application is a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0349] The bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 The diagram is represented by only one arrow, but it does not mean that there is only one bus or one type of bus. The bus is configured to implement connection and communication between the memory 12 and at least one processor 13, etc.

[0350] Although not shown, the electronic device 1 may also include a power source (such as a battery) to power various components. Preferably, the power source may be logically connected to the at least one processor 13 via a power management device, thereby implementing functions such as charging management, discharging management, and power consumption management through the power management device. The power source may also include one or more DC or AC power sources, a recharging device, a power failure detection circuit, a power converter or inverter, a power status indicator, and other arbitrary components. The electronic device 1 may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be detailed here.

[0351] Furthermore, the electronic device 1 may also include a network interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a WI-FI interface, a Bluetooth interface, etc.), which is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0352] Optionally, the electronic device 1 may further include a user interface, which may be a display, an input unit (such as a keyboard), or a standard wired interface or a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0353] An embodiment of the present application also provides a computer-readable storage medium (not shown), in which computer-readable instructions are stored. The computer-readable instructions are executed by a processor in an electronic device to implement the artificial intelligence-based business recommendation method described in any of the above embodiments.

[0354] It should be understood that the embodiment is for illustration only and the scope of the patent application is not limited to this structure.

[0355] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the module division is merely a logical function division, and other division methods may be used in actual implementation.

[0356] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed across multiple network elements. Some or all of the modules may be selected to achieve the purpose of the solution of this embodiment according to actual needs.

[0357] In addition, the functional modules in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0358] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices listed in the specification may also be implemented by a single unit or device through software or hardware. Terms such as first and second are used to indicate names and do not imply any particular order.

[0359] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. A business recommendation method based on artificial intelligence, characterized in that: The method comprises: Analyze business scenarios and pre-set business solutions to obtain rated business indicators. The business scenario refers to the problem that the enterprise needs to solve, and the business solution refers to the solution designed by the enterprise for the business scenario. The rated business indicators are used to characterize the quality of the business solution. Based on the rated business indicators, user data is collected from the business database corresponding to the business scenario to build a target user set; Classifying each piece of user data in the target user set to obtain multiple user clusters; The user data in each user cluster is grouped to obtain a user grouping confidence; the quantity and aggregation degree of the user data in each user cluster are counted, where the aggregation degree is used to characterize the diversity of the characteristics of the user data; the quantity and aggregation degree of the user data are input into a custom integration model to obtain an integration result, which is used as the weight of the user cluster; wherein determining the user grouping confidence includes: calculating the aggregation degree of the user data in each category of each user cluster as an aggregation value; and determining the user grouping confidence based on the product of the weight corresponding to each user cluster and the aggregation value; wherein the custom integration model satisfies the following relationship: ;in, represents the weight of the i-th user cluster. The higher the weight, the higher the credibility of the test results when the service solution is tested using the user data in the cluster. represents the number of user data in the i-th user cluster, The larger the value of , the more user data the cluster contains, and the higher the weight of the cluster should be; represents the aggregation degree of the i-th user cluster. The lower the value of the aggregation degree, the more discrete the characteristics of the user data in the cluster, and the higher the weight of the cluster should be. The service solution corresponding to the service scenario is evaluated based on the user group confidence, and the service solution is recommended according to the evaluation result.

2. The business recommendation method based on artificial intelligence according to claim 1, characterized in that: The collecting user data from the business database corresponding to the business scenario based on the rated business indicator to construct a target user set includes: Marking user data in a business database corresponding to the business scenario according to the business indicator; if the business indicator in the user data is less than the rated business indicator, the user data is marked as a target user; Collect statistics on the characteristics of the target user, including user age, user gender, user online time, and user preference category; The features corresponding to each target user are used as user data, and the user data are stored as the target user set.

3. The business recommendation method based on artificial intelligence according to claim 2, characterized in that: The characteristics of the target user include numerical data and non-numerical data. After counting the characteristics of the target user, the method further includes: Identify whether each user feature is non-numeric data, and mark each user feature according to the identification result, wherein the mark includes "yes" and "no"; Traverse the flags of each user feature in turn. If a user feature is marked as "yes", count the number of value types of the feature, mark each type in the feature in turn according to a natural number, and use the marked natural number as the encoding value of each data in the feature to obtain a numerical feature. If a user feature is marked as "no", do nothing; Replace each data in the user feature marked "yes" with its corresponding coded value to update the target user set.

4. The business recommendation method based on artificial intelligence according to claim 1, characterized in that: Classifying each piece of user data in the target user set to obtain multiple user clusters includes: Calculate the cosine distance between each two pieces of user data using the cosine distance algorithm; The target user set is classified according to the cosine distance and a preset clustering algorithm to obtain a plurality of user clusters, wherein each user cluster contains a plurality of user data.

5. The business recommendation method based on artificial intelligence according to claim 1, characterized in that: The step of grouping the user data in each user cluster to obtain the user grouping confidence comprises: Using a preset clustering algorithm to perform binary classification on the user data in each user cluster, and labeling the user data in the user cluster according to the category, the labels include "experimental" and "control"; Calculate the aggregation degree of user data in each category of each user cluster as the aggregation value; The product of the weight corresponding to each user cluster and the aggregate value is used as the confidence corresponding to each category; Combine user data with the same tags to form the experimental group and the control group; The confidence level of the experimental group is calculated as the sum of the confidence levels of the user data, and the confidence level of the control group is calculated as the sum of the confidence levels of the user data.

6. The artificial intelligence-based business recommendation method according to claim 5, characterized in that: The evaluating the service solution corresponding to the service scenario based on the user group confidence and recommending the service solution according to the evaluation result includes: Calculating the mean of the business indicators of the user data in the experimental group and the control group respectively as the benchmark value for each group of users; Randomly push a service plan to each group of users, and after a preset test period, calculate the average of the service indicators of each group of users as the test value; Calculating the difference between the test value and the benchmark value to obtain service increments for each group of users; Calculating the product of the confidence and the service increment of each group of users as the service recommendation result for each group of users; The business solution corresponding to the business recommendation result with a large value is used as the recommended calibration business solution.

7. A business recommendation device based on artificial intelligence, characterized in that: The apparatus comprises a unit for implementing the method according to any one of claims 1 to 6, the apparatus comprising: An acquisition unit, used to analyze business scenarios and preset business plans to obtain rated business indicators; a collection unit, configured to collect user data from a service database corresponding to the service scenario based on the rated service indicator to construct a target user set; a classification unit, configured to classify each piece of data in the target user set to obtain a plurality of user clusters; a grouping unit, configured to group user data in each user cluster to obtain user grouping confidence; A recommendation unit is used to evaluate the service solution corresponding to the service scenario based on the user group confidence level, and recommend the service solution according to the evaluation result.

8. An electronic device, characterized in that: The electronic device comprises: a memory storing computer-readable instructions; and A processor executes computer-readable instructions stored in the memory to implement the artificial intelligence-based business recommendation method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions are executed by a processor in an electronic device to implement the artificial intelligence-based business recommendation method according to any one of claims 1 to 6.

Images