Operation management aid decision-making method and device based on artificial intelligence, equipment and medium

Abstract

The embodiment of the invention provides an operation management aid decision-making method and device based on artificial intelligence, equipment and a medium. The method comprises the steps of obtaining related user data corresponding to a target object, and performing data analysis according to the related user data to obtain initial portrait information; obtaining historical score data and historical comment text corresponding to the target object under the initial portrait information; performing score correction on the historical score data to obtain corresponding initial score data of the target object under the initial portrait information; performing abnormal function identification on the target object according to the historical comment text to obtain an abnormal function type; determining an initial correction strategy corresponding to the target object according to the abnormal function type and the initial score data; obtaining the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy; performing validity evaluation on the initial correction strategy according to the latest user data to obtain a target evaluation value; and performing strategy correction on the initial correction strategy according to the target evaluation value to obtain a target correction strategy.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular, to a method, device, equipment and medium for assisting decision-making in business management based on artificial intelligence. Background Art

[0002] In today's highly competitive market environment, for a product to succeed in the market on a wide scale and achieve good sales performance, continuously revising the product based on user feedback has become an essential strategy. After a product is launched, only by collecting and analyzing user feedback can it accurately grasp users' pain points, needs and preferences, and then targetedly optimize and improve the product to maintain its competitiveness and attractiveness. However, when the existing technologies currently adopted conduct product analysis based on user feedback, they lack in-depth mining of the correlation relationships between data, and thus often have difficulty accurately extracting valuable information therefrom, unable to provide good support for subsequent decision-making assistance, and thereby reducing the promotion effect of the product. Summary of the Invention

[0003] The main objective of the embodiments of the present invention is to provide a method, device, equipment and medium for assisting decision-making in business management based on artificial intelligence, aiming to solve the problem in the related technologies that it is difficult to accurately extract valuable information therefrom, unable to provide good support for subsequent decision-making assistance, and thereby reducing the promotion effect of the product.

[0004] In a first aspect, the embodiments of the present invention provide a method for assisting decision-making in business management based on artificial intelligence, including:

[0005] Obtaining relevant user data corresponding to a target object, and performing data analysis based on the relevant user data to obtain initial portrait information corresponding to the target object;

[0006] Obtaining historical scoring data and historical review texts corresponding to the target object under the initial portrait information;

[0007] Performing scoring correction on the historical scoring data to obtain initial scoring data corresponding to the target object under the initial portrait information;

[0008] Performing abnormal function identification on the target object according to the historical review texts to obtain the abnormal function type corresponding to the target object;

[0009] Determining an initial correction strategy corresponding to the target object according to the abnormal function type and the initial scoring data;

[0010] Obtaining the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy;

[0011] Evaluate the effectiveness of the initial correction strategy based on the latest user data to obtain the target evaluation value corresponding to the initial correction strategy;

[0012] Modify the initial correction strategy according to the target evaluation value to obtain the target correction strategy corresponding to the target object.

[0013] In a second aspect, an embodiment of the present invention provides an artificial intelligence-based business management auxiliary decision-making device, including:

[0014] A data analysis module, configured to obtain relevant user data corresponding to a target object, and perform data analysis based on the relevant user data to obtain initial portrait information corresponding to the target object;

[0015] A data acquisition module, configured to obtain historical scoring data and historical review texts corresponding to the target object under the initial portrait information;

[0016] A data correction module, configured to correct the historical scoring data to obtain initial scoring data corresponding to the target object under the initial portrait information;

[0017] An anomaly recognition module, configured to perform anomaly function recognition on the target object according to the historical review texts to obtain an anomaly function type corresponding to the target object;

[0018] A strategy determination module, configured to determine an initial correction strategy corresponding to the target object according to the anomaly function type and the initial scoring data;

[0019] A data collection module, configured to obtain the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy;

[0020] A data evaluation module, configured to evaluate the effectiveness of the initial correction strategy based on the latest user data to obtain the target evaluation value corresponding to the initial correction strategy;

[0021] A strategy modification module, configured to modify the initial correction strategy according to the target evaluation value to obtain the target correction strategy corresponding to the target object.

[0022] In a third aspect, an embodiment of the present invention further provides a terminal device, which includes a processor, a memory, a computer program stored on the memory and executable by the processor, and a data bus for realizing connection communication between the processor and the memory. When the computer program is executed by the processor, the steps of any one of the artificial intelligence-based business management auxiliary decision-making methods provided in the specification of the present invention are implemented.

[0023] Fourthly, an embodiment of the present invention further provides a storage medium for computer-readable storage, characterized in that the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the artificial intelligence-based business management auxiliary decision-making methods provided in the specification of the present invention.

[0024] An embodiment of the present invention provides an artificial intelligence-based business management auxiliary decision-making method, device, equipment and medium. The method includes: obtaining relevant user data corresponding to a target object, and performing data analysis on the relevant user data to obtain initial portrait information corresponding to the target object; obtaining historical scoring data and historical review texts corresponding to the target object under the initial portrait information; performing score correction on the historical scoring data to obtain initial scoring data corresponding to the target object under the initial portrait information, so as to accurately obtain an accurate scoring result corresponding to the target object under the initial portrait information, and then identifying abnormal functions of the target object according to the historical review texts to obtain an abnormal function type corresponding to the target object; thereby determining an initial correction strategy corresponding to the target object according to the abnormal function type and the initial scoring data; then obtaining the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy; thereby evaluating the effectiveness of the initial correction strategy according to the latest user data to obtain a target evaluation value corresponding to the initial correction strategy; performing strategy correction on the initial correction strategy according to the target evaluation value to obtain a target correction strategy corresponding to the target object, and further accurately obtaining the target correction strategy corresponding to the target object according to the abnormal function type, the initial scoring data and the target evaluation value, so as to solve the problem in the related technology that it is difficult to accurately extract valuable information therefrom, thus unable to provide good support for subsequent auxiliary decision-making, and further reducing the product promotion effect, providing solid support for subsequent auxiliary decision-making, thereby attracting more potential users and enhancing the competitiveness of the product in the market. At the same time, the reliability of the target correction strategy has been greatly improved, making the product improvement and optimization process more scientific and stable. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0026] Figure 1 It is a schematic flowchart of an artificial intelligence-based business management auxiliary decision-making method provided by an embodiment of the present invention;

[0027] Figure 2 It is a schematic module structure diagram of an artificial intelligence-based business management auxiliary decision-making device provided by an embodiment of the present invention;

[0028] Figure 3 This is a schematic block diagram of a terminal device provided by an embodiment of the present invention. Specific embodiments

[0029] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0030] The flowchart shown in the accompanying drawings is only an example, and does not necessarily include all contents and operations / steps, nor does it necessarily need to be executed in the described order. For example, some operations / steps can be decomposed, combined, or partially merged, so the actual execution order may change according to the actual situation.

[0031] It should be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to include the plural forms.

[0032] An embodiment of the present invention provides an artificial intelligence-based business management auxiliary decision-making method, device, equipment, and medium. Among them, the artificial intelligence-based business management auxiliary decision-making method can be applied to a terminal device, and the terminal device can be an electronic device such as a tablet computer, a notebook computer, a desktop computer, a personal digital assistant, and a wearable device. The terminal device can be a server or a server cluster.

[0033] Next, some embodiments of the present invention will be described in detail in conjunction with the accompanying drawings. Without conflict, the following embodiments and the features in the embodiments can be combined with each other.

[0034] Please refer to Figure 1 , Figure 1 This is a schematic flowchart of an artificial intelligence-based business management auxiliary decision-making method provided by an embodiment of the present invention.

[0035] As Figure 1 shown, the artificial intelligence-based business management auxiliary decision-making method includes steps S101 to S108.

[0036] Step S101, obtain relevant user data corresponding to a target object, and perform data analysis based on the relevant user data to obtain initial portrait information corresponding to the target object.

[0037] Exemplarily, the target object is a user-oriented product that has been developed, and relevant user data corresponding to the target object is obtained from the database corresponding to the target object. The relevant user data is data collected from the users targeted by the target object during the use of the target object.

[0038] Exemplarily, for example, the relevant user data includes but is not limited to user browsing behavior, click data, usage duration, function usage frequency, etc. Data clustering is performed based on the relevant user data to better understand the characteristics of different user groups. Then, data analysis methods are used to mine information related to the portrait dimensions from the integrated data to obtain portrait information corresponding to different users using the target object. Furthermore, the portrait information corresponding to different users using the target object is determined as the initial portrait information corresponding to the target object.

[0039] In some embodiments, obtaining the initial portrait information corresponding to the target object according to the data analysis of the relevant user data includes: obtaining user portrait information corresponding to a target user according to the relevant user data, and determining the information matching degree between the target user and the sub-function corresponding to the target object according to the user portrait information in combination with the relevant user data; determining the function association degree between the target user and the sub-function according to the information matching degree; obtaining the first scoring information of the target user for the sub-function corresponding to the sub-function from the relevant user data; calculating the user similarity between any two target users according to the function association degree and the first scoring information; performing user clustering on the target users according to the user similarity to obtain a user clustering result, and obtaining the second scoring information associated with each subclass cluster in the user clustering result from the first scoring information; adjusting the scores of each sub-user in the subclass cluster for the sub-function according to the second scoring information to obtain the third scoring information corresponding to the sub-user for the sub-function; obtaining the user representation sequence corresponding to the sub-user under the target object according to the third scoring information and the sub-function; performing data clustering on the user representation sequence to obtain a target clustering result, and determining the initial portrait information corresponding to the target object according to the target clustering result and the user portrait information.

[0040] Exemplarily, relevant user data corresponding to a target user is collected during the use of the target object, and data analysis is performed based on the relevant user data to obtain user portrait information corresponding to each target user under the target object.

[0041] Exemplarily, the relevant user data also includes the user information during the use of the target object by the target user. Furthermore, the data usage frequency or data usage duration corresponding to the target user under each sub-function of the target object is obtained from the relevant user data. Then, based on the data usage frequency or data usage duration corresponding to the target user under each sub-function of the target object under this user profile information, data analysis is performed to obtain the information matching degree between the target user and the sub-function corresponding to the target object.

[0042] Exemplarily, when the information matching degree is greater than the preset value, it indicates that the functional association degree between the target user and the sub-function is a high association relationship. When the information matching degree is less than or equal to the preset value, it indicates that the functional association degree between the target user and the sub-function is a low association relationship.

[0043] Exemplarily, the first rating information of the target user for each sub-function is extracted from the relevant user data. Among them, the first rating data can come from the user's active evaluation feedback, the rating mechanism in the system, etc.

[0044] Exemplarily, the functional association degree and the first rating information are used as key features for calculating the user similarity. Furthermore, the functional association degree is used as an adjustment weight for the similarity calculation of the first rating information, so as to calculate the user similarity between any two target users by combining the functional association degree with the similarity calculation method.

[0045] Exemplarily, the target users are clustered according to the user similarity using a clustering algorithm such as K-means clustering, so as to divide the target users into different sub-clusters and analyze the characteristics of each sub-cluster. The ratings of the sub-function by the users within each sub-cluster are extracted from the first rating information, and the second rating information associated with each sub-cluster is summarized.

[0046] Exemplarily, a regulation method based on the mean or median is adopted for the sub-cluster combined with the second rating information. The rating of each sub-user within the sub-cluster is compared and adjusted with the average rating of the sub-cluster, and the ratings of each sub-user in each sub-cluster for each sub-function are adjusted to obtain the third rating information corresponding to each sub-user in each sub-cluster for each sub-function.

[0047] Exemplarily, the third rating information of each sub-user for each sub-function is arranged in a certain order to form a user representation sequence corresponding to the sub-user under the target object. Then, a clustering algorithm is used to perform data clustering on the user representation sequence to obtain the target clustering result.

[0048] Exemplarily, analyze the characteristics of each cluster in the target clustering result, and combine the user portrait information constructed previously to summarize the initial portrait information corresponding to the target object. For example, if most of the users in a certain cluster are young, high-consuming, and have a high score for a specific sub-function, then these characteristics can be highlighted in the initial portrait of the target object under this sub-function.

[0049] In some embodiments, calculating the user similarity between any two of the target users according to the function association degree and the first scoring information includes: determining a first user and a second user from the target users, and determining a first associated function corresponding to the first user and a second associated function corresponding to the second user from the sub-functions; determining a first association degree corresponding to the first associated function and a second association degree corresponding to the second associated function from the function association degree; performing an intersection process on the first associated function and the second associated function to obtain a target associated function; determining the function association relationship corresponding to the first user and the second user under the target associated function according to the first association degree and the second association degree in combination with the target associated function; statistically obtaining the first same relationship quantity and the second same relationship quantity of the first user and the second user with respect to the target associated function, and obtaining the mutually exclusive relationship quantity of the first user and the second user with respect to the target associated function according to the function association relationship; determining the user similarity between the first user and the second user according to the first same relationship quantity, the second same relationship quantity, and the mutually exclusive relationship quantity in combination with the target associated function by using the first association degree and the second association degree; wherein, the user similarity is obtained according to the following formula:

[0050]

[0051] Wherein, represents the user similarity between the i-th first user and the j-th second user, ave represents the mean function, n represents the quantity corresponding to the target associated function, represents the first association degree of the i-th first user with respect to the t-th target associated function, represents the second association degree of the j-th second user with respect to the t-th target associated function, represents the first same relationship quantity, represents the first weight of the t-th target associated function, represents the second weight of the t-th target associated function, represents the second same relationship quantity, represents the mutually exclusive relationship quantity.

[0052] Exemplarily, two users are randomly selected from the target users and defined as the first user and the second user respectively. Then, the first associated function corresponding to the first user is found from the sub-functions; similarly, the second associated function corresponding to the second user is found. The first associated function is used to represent the sub-functions used or browsed by the first user in the target object. The second associated function is used to represent the sub-functions used or browsed by the second user in the target object. And the degree of association value corresponding to the first associated function is found from the function association degree and defined as the first degree of association; the degree of association value corresponding to the second associated function is found and defined as the second degree of association.

[0053] Exemplarily, the first associated function and the second associated function are compared one by one to find the sub-functions they have in common, and these common sub-functions constitute the target associated function. For example, if the first associated function includes sub-functions A, B, and C, and the second associated function includes sub-functions B, C, and D, then the target associated function is sub-functions B and C.

[0054] Exemplarily, for each identical function in the target associated function, the function association relationship between the first user and the second user under this sub-function is determined by combining the first degree of association and the second degree of association. If the degrees of association of both users with a certain target associated function are relatively high, it indicates that their association relationship with respect to this function is a positive strong correlation; if the degrees of association of both users with a certain target associated function are relatively low, it indicates that their association relationship with respect to this function is a negative strong correlation. If the degrees of association of the two users with a certain target associated function are one high and one low, it indicates that their association relationship with respect to this function is a mutually exclusive correlation.

[0055] Exemplarily, for each identical function in the target associated function, the number of positive strong correlation cases of the function association relationship between the first user and the second user is judged to obtain the first number of identical relationship cases between the first user and the second user under the target associated function, and for each identical function in the target associated function, the number of negative strong correlation cases of the function association relationship between the first user and the second user is judged to obtain the second number of identical relationship cases between the first user and the second user under the target associated function. At the same time, the number of mutually exclusive relationship cases between the first user and the second user under the target associated function is counted. Mutually exclusive correlation means that the association relationships of the two users with respect to a certain target associated function show opposite situations, for example, one user has a high degree of function association while the other user has a low degree of function association.

[0056] Exemplarily, obtain the total number of users involved in each identical function in the target associated function, as well as the first quantity corresponding to the high correlation degree between the function and the users under each identical function in the target associated function and the second quantity corresponding to the low correlation degree between the function and the users under each identical function in the target associated function. Then, divide the first quantity by the total number of users to obtain the first quotient value, and further use 1 minus the first quotient value to determine the first weight corresponding to the identical function. Next, divide the second quantity by the total number of users to obtain the second quotient value, and further use 1 minus the second quotient value to determine the second weight corresponding to the identical function.

[0057] Exemplarily, according to the first identical relationship quantity, the second identical relationship quantity, and the mutually exclusive relationship quantity, in combination with the target associated function and the corresponding first correlation degree and second correlation degree, the following formula is used to determine the user similarity between the first user and the second user:

[0058]

[0059] where, represents the user similarity between the i-th first user and the j-th second user, ave represents the mean function, n represents the quantity corresponding to the target associated function, represents the first correlation degree of the i-th first user with respect to the t-th target associated function, represents the second correlation degree of the j-th second user with respect to the t-th target associated function, represents the first identical relationship quantity, represents the first weight of the t-th target associated function, represents the second weight of the t-th target associated function, represents the second identical relationship quantity, represents the mutually exclusive relationship quantity.

[0060] Exemplarily, by calculating the number of users with high correlation and low correlation under each identical function and obtaining the corresponding first weight and second weight, the importance of different functions in the user group can be clarified. When calculating the user similarity, introducing the weight information of the function can more comprehensively reflect the relationship between users. Different functions have different importance to users. Simply considering the correlation degree and the number of identical relationships may not accurately measure the user similarity. By combining the function weights, the behavioral differences of users on different functions can be captured more precisely, thereby improving the accuracy of similarity calculation. The introduction of weights makes it more focused on the common interests and usage habits of users for important functions when calculating the similarity. Therefore, according to the above formula, the user similarity between the first user and the second user can be obtained more accurately.

[0061] In some embodiments, obtaining a target clustering result by clustering data according to the user representation sequence includes: determining a minimum number of clusters and a maximum number of clusters, and determining a preset step size between the minimum number of clusters and the maximum number of clusters; determining a current number of clusters according to the preset step size and the minimum number of clusters, and clustering the user representation sequence according to the current number of clusters to obtain a first clustering result; clustering the user representation sequence according to the maximum number of clusters to obtain a second clustering result; calculating a first similarity between any two of the sub-users according to the user representation sequence; determining a first clustering center corresponding to a first cluster according to the first clustering result, and calculating a second similarity between each first sub-representation sequence in the first cluster and the first clustering center; calculating a third similarity between the first clustering centers, and determining a second clustering center corresponding to a second cluster according to the second clustering result, and calculating a fourth similarity between the second clustering centers; determining an intra-cluster similarity corresponding to the current number of clusters according to the first similarity and the second similarity; determining an inter-cluster similarity corresponding to the current number of clusters according to the third similarity and the fourth similarity; determining a cluster number reliability corresponding to the current number of clusters according to the inter-cluster similarity and the intra-cluster similarity; determining an optimal number of clusters corresponding to the user representation sequence according to the cluster number reliability in combination with the current number of clusters; clustering the user representation sequence according to the optimal number of clusters to obtain the target clustering result.

[0062] Exemplarily, the minimum and maximum numbers of clusters are set manually according to expert experience, so as to obtain the minimum number of clusters and the maximum number of clusters. The preset step size is used to control the increment amplitude of the number of clusters. For example, the step size can be set to 1, which means the number of clusters will increase sequentially at intervals of 1.

[0063] Exemplarily, starting from the minimum number of clusters, the current number of clusters is incremented according to the preset step size. For example, if the minimum number of clusters is 3 and the step size is 1, then the first current number of clusters is 3. Using this current number of clusters, the user representation sequence is clustered using K-means to obtain a first clustering result. And directly using the maximum number of clusters to perform the same clustering operation on the user representation sequence to obtain a second clustering result.

[0064] Exemplarily, select any two sub - users from the user characterization sequence and use an appropriate similarity calculation method (such as cosine similarity, Euclidean distance, etc.) to calculate the first similarity between them. For each first - type cluster in the first clustering result, calculate its clustering center. The clustering center is usually a representative value such as the mean or median of all sub - characterization sequences within the cluster. Then calculate the second similarity between each first sub - characterization sequence in the first - type cluster and the clustering center of this cluster. And calculate the third similarity between each first clustering center in the first clustering result to reflect the degree of difference between different clusters.

[0065] Exemplarily, for the second clustering result, determine the second clustering center of each second - type cluster and calculate the fourth similarity between these second clustering centers.

[0066] Exemplarily, sum up all the first similarities to obtain a first value, and then sum up all the second similarities in each first - type cluster to obtain a second value. Then sum up and average all the second values in the first clustering result to obtain a third value. Divide the first value by the third value to obtain the intra - cluster similarity.

[0067] Exemplarily, sum up and average the third similarities between any two clustering centers in the first clustering result to obtain a first mean value, and sum up and average the fourth similarities between any two clustering centers in the second clustering result to obtain a second mean value. Divide the first mean value by the second mean value to obtain a relative ratio, and then subtract the relative ratio from 1 to obtain the inter - cluster similarity corresponding to the current number of clusters.

[0068] Exemplarily, sum up the intra - cluster similarity and the inter - cluster similarity to obtain the relevant similarity corresponding to the current number of clusters, and then determine the reciprocal of the relevant similarity as the reliability of the number of clusters.

[0069] Exemplarily, obtain each current number of clusters according to a preset step size, and thus obtain the reliability of the number of clusters corresponding to each current number of clusters according to the above steps. Compare the reliabilities of the number of clusters under different current numbers of clusters, and then use the current number of clusters corresponding to the minimum reliability of the number of clusters as the optimal number of clusters corresponding to the user characterization sequence. Use the determined optimal number of clusters combined with the k - means clustering algorithm to perform data clustering on the user characterization sequence to obtain the target clustering result.

[0070] Step S102: Obtain the historical scoring data and historical review texts corresponding to the target object under the initial portrait information.

[0071] Exemplarily, historical rating data and historical review texts corresponding to relevant users of the target object under the initial portrait information are obtained from the database. That is, historical rating data and historical review texts of relevant users on the target object when the user portrait is the initial portrait information are obtained.

[0072] Step S103: Perform rating correction on the historical rating data to obtain the initial rating data corresponding to the target object under the initial portrait information.

[0073] Exemplarily, calculate the average value of the historical rating data, and then adjust the rating according to the initial portrait information. For example, if the ratings of a certain type of user group are generally too high or too low, the rating can be corrected by adjusting the mean value to obtain the initial rating data, or a regression model can be established, with the initial portrait information as the independent variable and the historical rating data as the dependent variable, and the reasonable rating under the initial portrait information can be predicted through the model, so as to realize the rating correction of the historical rating data, and then sum and average the corrected data to obtain the initial rating data of the target object under the initial portrait information. That is, the initial rating data is used to represent the basic rating of the relevant users corresponding to the target object under the initial portrait information.

[0074] Initial rating data.

[0075] In some embodiments, the performing rating correction on the historical rating data to obtain the initial rating data corresponding to the target object under the initial portrait information includes: obtaining the target users involved in the target object from the historical rating data, determining the third user from the target users, and obtaining any fourth user from the target users after excluding the third user; obtaining the first relevant function corresponding to the third user and the first relevant rating corresponding to the first relevant function from the historical rating data; obtaining the second relevant function corresponding to the fourth user and the second relevant rating corresponding to the second relevant function from the historical rating data; determining the function influence degree between the third user and the fourth user according to the first relevant function and the second relevant function; obtaining the fifth user involved in the first relevant function and the third relevant rating corresponding to the fifth user from the historical rating data; determining the function influence parameter corresponding to the first relevant function according to the fifth user and the third relevant rating; determining the information similarity between the third user and the fourth user according to the function influence degree and the function influence parameter in combination with the first relevant rating and the second relevant rating; determining the similar users corresponding to the third user according to the information similarity, and performing rating correction on the first relevant rating according to the similar users to obtain the fourth relevant rating corresponding to the third user; obtaining the initial rating data corresponding to the target object under the initial portrait information according to the fourth relevant rating.

[0076] Exemplarily, the historical scoring data is data of the target user scoring different functions in the target object under the initial portrait information, and then all target users involved in scoring the functions of the target object are obtained from the initial scoring data.

[0077] Exemplarily, any user is obtained from the target users and determined as the third user, and then any user is obtained from the remaining users among the target users after excluding the third user and determined as the fourth user. All records related to the third user are screened out from the historical scoring data, the first related functions involved by the third user are determined therefrom, and the first related scores corresponding to each first related function are extracted. Using the same method, the second related functions corresponding to the fourth user and the second related scores of these functions are found from the historical scoring data.

[0078] Exemplarily, the first related functions and the second related functions are compared to find the overlap, complementarity or relevance between them. For example, if two users have more identical related functions, it indicates that they are relatively similar in function usage; if the functions of one party can supplement the function usage of the other party, there is also a certain correlation. Thus, the function influence degree between the third user and the fourth user is determined according to the matching situation of the first related functions and the second related functions. For example, the identical function information between the first related functions and the second related functions and the number of identical functions corresponding to the identical function information are calculated, and the number of the first related functions is obtained, so that the number of identical functions is divided by the number of the first related functions to obtain a third quotient value, and then the function value of the third quotient value under the exponential function with the natural constant as the base is obtained, and then the reciprocal of the function value is obtained to obtain a reciprocal value, and then 1 minus the reciprocal value is determined as the function influence degree between the third user and the fourth user.

[0079] Exemplarily, in the historical scoring data, all fifth users interacting with the first related functions are found, and the third related scores given by these users for this function are recorded, so as to calculate the mean information corresponding to the third related scores under this function, and then the scoring information of the first user under this function is subtracted from the mean information and squared to obtain a squared value, and then the squared values respectively corresponding to all the first related functions are obtained, and then the sum of all the squared values is averaged and then square-rooted to obtain the initial influence parameter corresponding to the first related functions, and then data normalization is performed according to all the initial influence parameters to obtain the function influence parameter.

[0080] Exemplarily, the intersection processing of the first related function and the second related function is performed to obtain the same function information, and further, the first average value corresponding to the first related score of the first related function corresponding to the third user and the second average value corresponding to the second related score of the second related function corresponding to the fourth user are obtained. Thus, according to the following formula, the information similarity between the third user and the fourth user is determined by combining the first average value, the second average value, the function influence degree, and the function influence parameter:

[0081]

[0082] Wherein, represents the information similarity between the a-th third user and the b-th fourth user, represents the number of functions corresponding to the same function information, represents the function influence degree between the a-th third user and the b-th fourth user, represents the initial influence parameter corresponding to the a-th third user, represents the score value corresponding to the a-th third user for the v-th same function information under the first related score, represents the first average value corresponding to the first related score of the first related function corresponding to the a-th third user, represents the score value corresponding to the b-th fourth user for the v-th same function information under the second related score, represents the second average value corresponding to the second related score of the second related function corresponding to the b-th fourth user, represents the total number corresponding to the first related function of the a-th third user, represents the initial influence parameter corresponding to the a-th third user, represents the total number corresponding to the second related function of the b-th fourth user, represents the initial influence parameter corresponding to the b-th fourth user, represents the first related score corresponding to the a-th third user under the t-th first related score, The second related score corresponding to the b-th fourth user under the x-th second related score. It should be noted that the method for calculating the initial influence parameter corresponding to the second related function of the fourth user is the same as the method for calculating the initial influence parameter corresponding to the first related function of the third user.

[0083] Exemplarily, represents the score value corresponding to the a-th third user for the v-th same function information under the first related score, that is, obtaining the related score corresponding to the v-th same function information from the first related score.

[0084] Exemplarily, a suitable threshold is set according to the calculated information similarity. Users with information similarity higher than this threshold are determined as similar users of the third user.

[0085] Exemplarily, obtain the information similarity between the third user and each user among the similar users, and then calculate the average value of the first scores corresponding to the first relevant scores of the third user. Further, obtain the average value of the second scores corresponding to the relevant scores of each user among the similar users. Then, determine any one of the functions in the first relevant function, and obtain the function scores corresponding to any one of the functions to be corrected in the first relevant function for each user among the similar users. Further, calculate the difference between the function score and the average value of the second scores to obtain the score difference. Then, multiply the score difference by the information similarity between the third user and the user among the similar users to obtain the first product. Repeat the above steps to obtain the first product between the third user and each user among the similar users, and sum all the first products to obtain the second product. Sum the information similarities between the third user and each user among the similar users to obtain the similarity result. Then, divide the second product by the similarity result and sum it with the average value of the first scores to obtain the corrected score of the third user under the function to be corrected. Thus, correct all the first relevant scores in the first relevant function to correct the first relevant scores of the third user to obtain the fourth relevant score.

[0086] Exemplarily, repeat the above steps to correct the scores of each third user respectively to obtain the corresponding fourth relevant scores. Then, sum and average the fourth relevant scores of all the third users under the initial portrait information to obtain the initial score data corresponding to the target object under the initial portrait information. It is also possible to calculate the score distribution of the fourth relevant scores of all the third users under the initial portrait information to determine the score corresponding to the maximum probability of the fourth relevant score distribution as the initial score data corresponding to the target object under the initial portrait information. That is, the initial score data is the score corresponding to the most frequent score distribution of the corresponding users' scores on the target object under the initial portrait information. That is, the initial score data is used to represent the overall evaluation tendency of the corresponding users on the target object under the initial portrait information. The initial score data reflects the general views and attitudes of the user group under the initial portrait information on the target object.

[0087] Step S104: Identify the abnormal functions of the target object according to the historical review text to obtain the corresponding abnormal function types of the target object.

[0088] Exemplarily, a function name mapping table is established, and then relevant functions corresponding to the historical review text are obtained by querying in the historical review text according to the function name mapping table. Then, evaluation keywords corresponding to each relevant function in the historical review text are obtained through dependency syntactic analysis. Furthermore, all evaluation keywords corresponding to the relevant functions are obtained, and clustering analysis is performed on all the evaluation keywords to obtain keyword clusters. Then, the target cluster corresponding to the maximum number in the keyword clusters is obtained, and the word types corresponding to the keywords in the target cluster, such as positive or negative, are obtained. Thus, the evaluation type of the user for the relevant functions in the target object is determined according to the word types corresponding to the keywords in the target cluster; when the evaluation type is a negative review or dissatisfaction, the relevant function corresponding to the evaluation keyword is determined as the abnormal function type corresponding to the target object.

[0089] In some embodiments, the obtaining of the abnormal function type corresponding to the target object by performing abnormal function recognition on the target object according to the historical review text includes: determining the historical review time corresponding to the historical review text, and dividing the historical review text according to the historical review time to obtain relevant review texts corresponding to different time ranges; performing sentiment analysis on the relevant review texts to obtain the sentiment types corresponding to the relevant review texts, and determining the sentiment representation values corresponding to the target object in different time ranges according to the sentiment types; performing quantity statistics on the relevant review texts to obtain the user participation degrees corresponding to different time ranges; determining the target influence values corresponding to different time ranges according to the sentiment representation values and the user participation degrees; and performing abnormal function recognition on the target object according to the target influence values in combination with the relevant review texts to obtain the abnormal function type corresponding to the target object.

[0090] Exemplarily, the historical review time corresponding to the historical review data is obtained from the database, and the earliest time value and the latest time value are obtained according to the historical review time. Then, multiple time ranges are obtained by performing time range division according to the earliest time value and the latest time value in combination with a preset time width, so as to obtain relevant review texts corresponding to different time ranges.

[0091] Exemplarily, a dictionary-based method is adopted, that is, a pre-defined sentiment dictionary is used to judge the sentiment polarity of the words in the relevant review texts; alternatively, a machine learning or deep learning model can be used to classify the sentiment types of the review texts by training on a large amount of data with sentiment labels, and then the relevant review texts are input into the selected sentiment analysis algorithm to obtain the sentiment type of each review, which is usually divided into categories such as positive, negative, and neutral.

[0092] Exemplarily, for the relevant review texts within each time range, count the number of reviews of different sentiment types. By setting certain calculation rules, such as calculating the proportion of positive reviews in the total number of reviews, the proportion of negative reviews in the total number of reviews, etc., the difference between the proportion of positive reviews in the total number of reviews and the proportion of negative reviews in the total number of reviews is determined as the sentiment representation value corresponding to the target object for this time range. Perform a simple counting operation on the relevant review texts within each time range to obtain the total number of reviews within this time range. Use the ratio between the total number of reviews and the corresponding quantity of historical review texts as the user engagement for this time range. The user engagement reflects the degree of attention of users to the target object within a specific time period.

[0093] Exemplarily, sum and average the user engagement and the sentiment representation value to obtain the target average value, and then determine the product of the target average value and the quantity of text corresponding to the relevant review texts as the target impact value for the time range corresponding to the relevant review texts. The target impact value reflects the comprehensive impact of the users' emotional attitudes and participation levels on the target object within the time range.

[0094] Exemplarily, according to the magnitude of the target impact value, screen out the time ranges that have a greater impact on the target object. The relevant review texts within these time periods may contain more information about the abnormal functions of the target object. In the relevant review texts of the selected key time periods, query according to the function name mapping table to obtain the corresponding relevant functions, and then obtain the evaluation keywords corresponding to each relevant function in the relevant review texts of the selected key time periods according to dependency syntactic analysis. Furthermore, obtain all the evaluation keywords corresponding to the relevant functions, perform clustering analysis on all the evaluation keywords to obtain keyword clusters, then obtain the target cluster corresponding to the largest number in the keyword clusters, and then obtain the word types corresponding to the keywords in the target cluster, such as positive or negative, etc., so as to determine the evaluation type of users on the relevant functions in the target object according to the word types corresponding to the keywords in the target cluster; when the evaluation type is a bad review or dissatisfaction, determine the relevant function corresponding to the evaluation keyword as the abnormal function type corresponding to the target object.

[0095] In some embodiments, the abnormal function identification of the target object based on the target influence value and the relevant review text to obtain the abnormal function type corresponding to the target object includes: performing word segmentation on the relevant review text to obtain the initial keywords corresponding to the relevant review text; performing word frequency statistics on the initial keywords according to the relevant review text to obtain the relevant frequencies corresponding to the initial keywords; performing word frequency quantity analysis on the relevant frequencies to obtain the frequency maximum value and the frequency maximum value corresponding to the relevant frequencies and the frequency median corresponding to the relevant frequencies; determining the initial weight corresponding to the initial keyword according to the frequency maximum value, the frequency maximum value, and the frequency median in combination with the relevant frequency corresponding to the initial keyword; determining the target weight corresponding to the initial keyword according to the initial weight and the target influence value, and determining the target keyword corresponding to the relevant review text according to the target weight; performing data analysis on the target keyword and the relevant review text to obtain the target function corresponding to the historical review text and the target description word corresponding to the target function; and determining the abnormal function type corresponding to the target function according to the target description word.

[0096] Exemplarily, Jieba word segmentation is used to split the relevant review text into individual words, and some meaningless modal particles, function words, etc. are removed to obtain the basic keywords. Then, according to the TFIDF algorithm, the keyword scores corresponding to each basic keyword are calculated, and the corresponding initial keywords are obtained according to the keyword scores.

[0097] Exemplarily, word frequency statistics are performed on the initial keywords according to the relevant review text to obtain the relevant frequencies. After counting the number of occurrences of all initial keywords, the total number of occurrences of all keywords is calculated. Then, the number of occurrences of each keyword is divided by the total number to obtain the relevant frequency of the keyword. The relevant frequency data of all initial keywords are arranged in ascending order. The first value in the arranged data is the frequency minimum value, and the last value is the frequency maximum value. If the number of initial keywords is odd, the relevant frequency value at the exact middle position of the arranged data is the frequency median; if the number is even, the average of the two middle relevant frequency values is taken as the frequency median.

[0098] Exemplarily, the initial weight corresponding to the initial keyword is obtained according to the frequency maximum value, the frequency minimum value, and the frequency median in combination with the following formula:

[0099]

[0100] Where represents the initial weight corresponding to the i-th initial keyword, represents taking the absolute value, Represents the relevant frequency corresponding to the i-th initial keyword. Represents the median frequency. Represents the maximum frequency. Represents the minimum frequency.

[0101] Exemplarily, combine the initial weight of each initial keyword with the corresponding target influence value. The multiplication method can be used, i.e., the target weight = initial weight * target influence value. Thus, set a target weight threshold, and screen out the initial keywords whose target weights are greater than this threshold. These keywords are the target keywords corresponding to the relevant review texts.

[0102] Exemplarily, associate the target keywords with the relevant review texts, carefully analyze the text paragraphs containing the target keywords, and then identify the functional content related to the target keywords from the associated text paragraphs and determine it as the target function. At the same time, find out the words describing the specific situations of these target functions, that is, the target description words. For example, if the target keyword is "retrieval" and the review text mentions "inaccurate retrieval", then "retrieval" is the target function and "inaccurate" is the target description word. Thus, judge whether the target function is an abnormal type according to the word type corresponding to the target description word. When the target description word corresponding to the target function is of the abnormal type, then determine the target function as the abnormal function type.

[0103] Step S105: Determine the initial correction strategy corresponding to the target object according to the abnormal function type and the initial scoring data.

[0104] Exemplarily, when the abnormal function type is not empty and the initial scoring data is lower than the preset score, then obtain the target description words corresponding to the abnormal function type from the historical review texts based on the determined abnormal function type according to the dependency syntactic analysis. Conduct in-depth analysis on the screened target description words to understand the specific problems reflected by each description word. For example, "lag" may mean insufficient system resources or low efficiency of program algorithms; "slow response" may be related to network latency or high server load. Thus, according to the problems reflected by the target description words, combined with the actual business situation and technical capabilities, formulate the corresponding initial correction strategy.

[0105] Step S106: Obtain the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy.

[0106] Exemplarily, execute the initial correction strategy on the target object, and thus collect the latest user data corresponding to the initial portrait information after the initial correction strategy is executed on the target object.

[0107] Step S107: Conduct an effectiveness evaluation on the initial correction strategy according to the latest user data to obtain the target evaluation value corresponding to the initial correction strategy.

[0108] Exemplarily, the latest user data includes the latest rating data. Then, rating calculation is performed on the latest rating data to obtain the average rating value of the target object under the initial portrait information. Further, the target evaluation value corresponding to the initial correction strategy is determined based on the gap between the average rating value and the initial rating data.

[0109] Step S108: Perform strategy correction on the initial correction strategy according to the target evaluation value to obtain the target correction strategy corresponding to the target object.

[0110] Exemplarily, when the target evaluation value is greater than the first preset value and the average rating value is greater than the second preset value, the initial correction strategy is retained; when the target evaluation value is less than or equal to the first preset value, or the average rating value is less than or equal to the second preset value, a correction strategy corresponding to the function abnormality type is re - formulated to obtain the target correction strategy corresponding to the target object.

[0111] Please refer to Figure 2 , Figure 2 There is provided an artificial - intelligence - based business management auxiliary decision - making device 200 according to an embodiment of the present application. The artificial - intelligence - based business management auxiliary decision - making device 200 includes a data analysis module 201, a data acquisition module 202, a data correction module 203, an abnormality identification module 204, a strategy determination module 205, a data collection module 206, a data evaluation module 207, and a strategy correction module 208. Among them, the data analysis module 201 is configured to obtain relevant user data corresponding to a target object, and perform data analysis on the relevant user data to obtain the initial portrait information corresponding to the target object; the data acquisition module 202 is configured to obtain the historical rating data and historical review texts corresponding to the target object under the initial portrait information; the data correction module 203 is configured to perform rating correction on the historical rating data to obtain the initial rating data corresponding to the target object under the initial portrait information; the abnormality identification module 204 is configured to perform abnormal function identification on the target object according to the historical review texts to obtain the abnormal function type corresponding to the target object; the strategy determination module 205 is configured to determine the initial correction strategy corresponding to the target object according to the abnormal function type and the initial rating data; the data collection module 206 is configured to obtain the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy; the data evaluation module 207 is configured to perform effectiveness evaluation on the initial correction strategy according to the latest user data to obtain the target evaluation value corresponding to the initial correction strategy; the strategy correction module 208 is configured to perform strategy correction on the initial correction strategy according to the target evaluation value to obtain the target correction strategy corresponding to the target object.

[0112] In some embodiments, the artificial intelligence-based business management auxiliary decision-making device 200 can be applied to a terminal device.

[0113] It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the above-described artificial intelligence-based business management auxiliary decision-making device 200 can refer to the corresponding process in the foregoing embodiments of the artificial intelligence-based business management auxiliary decision-making method, and will not be elaborated herein.

[0114] Please refer to Figure 3 , Figure 3 which is a schematic block diagram of the structure of a terminal device provided by an embodiment of the present invention.

[0115] As Figure 3 shown, the terminal device 300 includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected through a bus 303, and this bus is, for example, an I2C (Inter-integrated Circuit) bus.

[0116] Specifically, the processor 301 is used to provide computing and control capabilities to support the operation of the entire terminal device. The processor 301 can be a central processing unit (CPU), and this processor 301 can also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor can be a microprocessor or this processor can also be any conventional processor, etc.

[0117] Specifically, the memory 302 can be a Flash chip, a read-only memory (ROM), a magnetic disk, an optical disc, a USB flash drive, or a mobile hard disk, etc.

[0118] Those skilled in the art can understand that Figure 3 the structure shown in

[0119] Among them, the processor is used to run the computer program stored in the memory, and when executing the computer program, it implements any one of the artificial intelligence-based business management auxiliary decision-making methods provided by the embodiments of the present invention.

[0120] In one embodiment, the processor is used to run the computer program stored in the memory, and when executing the computer program, it implements the following steps:

[0121] Obtain the relevant user data corresponding to the target object, and perform data analysis based on the relevant user data to obtain the initial portrait information corresponding to the target object;

[0122] Obtain the historical scoring data and historical review texts corresponding to the target object under the initial portrait information;

[0123] Perform scoring correction on the historical scoring data to obtain the initial scoring data corresponding to the target object under the initial portrait information;

[0124] Perform abnormal function identification on the target object according to the historical review texts to obtain the abnormal function type corresponding to the target object;

[0125] Determine the initial correction strategy corresponding to the target object according to the abnormal function type and the initial scoring data;

[0126] Obtain the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy;

[0127] Perform effectiveness evaluation on the initial correction strategy according to the latest user data to obtain the target evaluation value corresponding to the initial correction strategy;

[0128] Perform strategy correction on the initial correction strategy according to the target evaluation value to obtain the target correction strategy corresponding to the target object.

[0129] It should be noted that those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working process of the above-described terminal device can refer to the corresponding process in the embodiments of the artificial intelligence-based business management auxiliary decision-making method described above, and will not be elaborated here.

[0130] The embodiments of the present invention also provide a storage medium for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any one of the artificial intelligence-based business management auxiliary decision-making methods provided in the specification of the embodiments of the present invention.

[0131] Among them, the storage medium may be the internal storage unit of the terminal device in the foregoing embodiments, such as the hard disk or memory of the terminal device. The storage medium may also be an external storage device of the terminal device, such as a plug-in hard disk equipped on the terminal device, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc.

[0132] Those of ordinary skill in the art can understand that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, and their appropriate combinations. In the hardware embodiment, the division between the functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, one physical component may have multiple functions, or one function or step may be executed by several physical components in cooperation. Some or all of the physical components may be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or may be implemented as hardware, or may be implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which may include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disks (DVDs) or other optical disk storage, magnetic cassettes, tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically contains computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and may include any information delivery medium.

[0133] It should be understood that the term "and / or" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. It should be noted that in this text, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or system comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or system. Without further limitation, an element defined by the statement "comprising an..." does not exclude the existence of another identical element in the process, method, article or system comprising such element.

[0134] The serial numbers of the above embodiments of the present invention are only for description and do not represent the superiority or inferiority of the embodiments. The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. An artificial intelligence-based auxiliary decision-making method for business management, characterized in that, The method includes: Obtaining relevant user data corresponding to a target object, and performing data analysis based on the relevant user data to obtain initial portrait information corresponding to the target object; Obtaining historical rating data and historical review texts corresponding to the target object under the initial portrait information; Performing rating correction on the historical rating data to obtain initial rating data corresponding to the target object under the initial portrait information; Performing abnormal function recognition on the target object according to the historical review texts to obtain the abnormal function type corresponding to the target object; Determining an initial correction strategy corresponding to the target object according to the abnormal function type and the initial rating data; Obtaining the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy; Performing effectiveness evaluation on the initial correction strategy according to the latest user data to obtain a target evaluation value corresponding to the initial correction strategy; Performing strategy correction on the initial correction strategy according to the target evaluation value to obtain a target correction strategy corresponding to the target object.

2. The method according to claim 1, wherein The obtaining initial portrait information corresponding to the target object by performing data analysis based on the relevant user data includes: Obtaining user portrait information corresponding to a target user according to the relevant user data, and determining the information matching degree between the target user and the sub - functions corresponding to the target object according to the user portrait information in combination with the relevant user data; Determining the function association degree between the target user and the sub - functions according to the information matching degree; Obtaining first rating information corresponding to the sub - functions of the target user from the relevant user data; Calculating the user similarity between any two target users according to the function association degree and the first rating information; Performing user clustering on the target users according to the user similarity to obtain a user clustering result, and obtaining second rating information associated with each subclass cluster in the user clustering result from the first rating information; Adjusting the ratings of each sub - user in the subclass cluster for the sub - functions according to the second rating information to obtain third rating information corresponding to the sub - functions of the sub - users; Obtaining a user representation sequence corresponding to the sub - users under the target object according to the third rating information and the sub - functions; Performing data clustering according to the user representation sequence to obtain a target clustering result, and determining the initial portrait information corresponding to the target object according to the target clustering result and the user portrait information.

3. The method according to claim 2, wherein The calculating the user similarity between any two target users according to the function association degree and the first rating information includes: Determining a first user and a second user from the target users, and determining a first associated function corresponding to the first user and a second associated function corresponding to the second user from the sub - functions; Determining a first association degree corresponding to the first associated function and a second association degree corresponding to the second associated function from the function association degree; Performing an intersection process on the first associated function and the second associated function to obtain a target associated function; Determine the functional association relationship corresponding to the first user and the second user under the target association function according to the first association degree, the second association degree, and the target association function; Statistically analyze the relationship between the first user and the second user according to the functional association relationship to obtain the first number of identical relationships and the second number of identical relationships of the first user and the second user with respect to the target association function, and obtain the number of mutually exclusive relationships of the first user and the second user with respect to the target association function; Determine the user similarity between the first user and the second user according to the first number of identical relationships, the second number of identical relationships, the number of mutually exclusive relationships, the target association function, the first association degree, and the second association degree; Wherein, the user similarity is obtained according to the following formula: ； Among them, represents the user similarity between the i-th first user and the j-th second user, ave represents the mean function, and n represents the quantity corresponding to the target association function. represents the first association degree of the i-th first user to the t-th target association function. represents the second association degree of the j-th second user to the t-th target association function. represents the number of first identical relationships. represents the first weight of the t-th target association function. represents the second weight of the t-th target association function. represents the number of second identical relationships. represents the number of exclusive relationships.

4. The method according to claim 2, characterized in that, The data clustering of the user representation sequence to obtain the target clustering result includes: Determine the minimum number of clusters and the maximum number of clusters, and determine a preset step between the minimum number of clusters and the maximum number of clusters; Determine the current number of clusters according to the preset step and the minimum number of clusters, and perform data clustering on the user representation sequence according to the current number of clusters to obtain a first clustering result; Perform data clustering on the user representation sequence according to the maximum number of clusters to obtain a second clustering result; Calculate the first similarity between any two of the sub-users according to the user representation sequence; Determine the first clustering center corresponding to the first cluster according to the first clustering result, and calculate the second similarity between each first sub-representation sequence in the first cluster and the first clustering center; Calculate the third similarity between the first clustering centers, and determine the second clustering center corresponding to the second cluster according to the second clustering result, and calculate the fourth similarity between the second clustering centers; Determine the intra-cluster similarity corresponding to the current number of clusters according to the first similarity and the second similarity; Determine the inter-cluster similarity corresponding to the current number of clusters according to the third similarity and the fourth similarity; Determine the cluster number reliability corresponding to the current number of clusters according to the inter-cluster similarity and the intra-cluster similarity; Determine the optimal number of clusters corresponding to the user representation sequence according to the cluster number reliability and the current number of clusters; Perform data clustering on the user representation sequence according to the optimal number of clusters to obtain the target clustering result.

5. The method according to claim 1, wherein The scoring correction of the historical scoring data to obtain the initial scoring data corresponding to the target object under the initial portrait information includes: Obtain the target users involved in the target object from the historical scoring data, determine the third user from the target users, and obtain any fourth user from the target users excluding the third user; Obtain the first related function corresponding to the third user and the first related score corresponding to the first related function from the historical scoring data; Obtain the second related function corresponding to the fourth user and the second related score corresponding to the second related function from the historical score data; Determine the function influence degree between the third user and the fourth user according to the first related function and the second related function; Obtain the fifth user involved in the first related function and the third related score corresponding to the fifth user from the historical score data; Determine the function influence parameter corresponding to the first related function according to the fifth user and the third related score; Determine the information similarity between the third user and the fourth user according to the function influence degree and the function influence parameter, combining the first related score and the second related score; Determine the similar users corresponding to the third user according to the information similarity, and correct the first related score according to the similar users to obtain the fourth related score corresponding to the third user; Obtain the initial score data corresponding to the target object under the initial portrait information according to the fourth related score; 6. The method according to claim 1, wherein The abnormal function identification of the target object according to the historical review text to obtain the abnormal function type corresponding to the target object includes: Determine the historical review time corresponding to the historical review text, and divide the historical review text according to the historical review time to obtain the relevant review texts corresponding to different time ranges; Perform sentiment analysis on the relevant review texts to obtain the sentiment types corresponding to the relevant review texts, and determine the sentiment representation values corresponding to the target object in different time ranges according to the sentiment types; Perform quantity statistics on the relevant review texts to obtain the user participation degrees corresponding to different time ranges; Determine the target influence values corresponding to different time ranges according to the sentiment representation values and the user participation degrees; Perform abnormal function identification on the target object according to the target influence values in combination with the relevant review texts to obtain the abnormal function type corresponding to the target object; 7. The method according to claim 6, characterized in that, The abnormal function identification of the target object according to the target influence values in combination with the relevant review texts to obtain the abnormal function type corresponding to the target object includes: Perform word segmentation on the relevant review texts to obtain the initial keywords corresponding to the relevant review texts; Perform word frequency statistics on the initial keywords according to the relevant review texts to obtain the relevant frequencies corresponding to the initial keywords; Perform word frequency quantity analysis according to the relevant frequencies to obtain the frequency maximum value and the frequency maximum value corresponding to the relevant frequencies as well as the frequency median corresponding to the relevant frequencies; Determine the initial weights corresponding to the initial keywords according to the frequency maximum value, the frequency maximum value and the frequency median in combination with the relevant frequencies corresponding to the initial keywords; Determine the target weights corresponding to the initial keywords according to the initial weights and the target influence values, and determine the target keywords corresponding to the relevant review texts according to the target weights; Performing data analysis based on the target keyword and the relevant review text to obtain the corresponding target function in the historical review text and the target description words corresponding to the target function; Determining the abnormal function type corresponding to the target function according to the target description words.

8. An operation and management auxiliary decision-making device based on artificial intelligence, characterized in that Including: A data analysis module, configured to obtain relevant user data corresponding to a target object, and perform data analysis based on the relevant user data to obtain initial portrait information corresponding to the target object; A data acquisition module, configured to obtain historical rating data and historical review text corresponding to the target object under the initial portrait information; A data correction module, configured to perform rating correction on the historical rating data to obtain initial rating data corresponding to the target object under the initial portrait information; An abnormal recognition module, configured to perform abnormal function recognition on the target object according to the historical review text to obtain the abnormal function type corresponding to the target object; A strategy determination module, configured to determine an initial correction strategy corresponding to the target object according to the abnormal function type and the initial rating data; A data collection module, configured to obtain the latest user data corresponding to the target object under the initial portrait information under the initial correction strategy; A data evaluation module, configured to perform effectiveness evaluation on the initial correction strategy according to the latest user data to obtain a target evaluation value corresponding to the initial correction strategy; A strategy correction module, configured to perform strategy correction on the initial correction strategy according to the target evaluation value to obtain a target correction strategy corresponding to the target object.

9. A terminal device, characterized in that, The terminal device includes a processor and a memory; The memory is used to store a computer program; The processor is configured to execute the computer program and implement the artificial intelligence-based business management auxiliary decision-making method according to any one of claims 1 to 7 when executing the computer program.

10. A computer storage medium for computer storage, characterized in that, The computer storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the artificial intelligence-based business management auxiliary decision-making method according to any one of claims 1 to 7.

Images