Method for predicting user behavior and assisting service handling by intelligent counter

Abstract

The invention discloses an intelligent counter auxiliary business handling method for predicting user behaviors, and relates to the technical field of business processing, and the auxiliary business handling method comprises the following steps: obtaining user basic information and behavior burying point information, and carrying out the preprocessing; on the basis of a naive Bayes classifier, a user behavior prediction model is constructed in combination with the preprocessed basic user information and behavior burying point information, and training is carried out; based on the user behavior prediction model, classifying the users in combination with a neighbor collaborative filtering algorithm, and predicting the service types of the classified users; and the predicted user service types are screened, and the screened service types are pushed to the client of the customer manager. According to the invention, the communication cost with the user is reduced by prompting the customer manager of the type of the service to be handled by the user.

Description

Technical Field

[0001] The present invention relates to the technical field of business processing, and in particular to a method for predicting user behavior and assisting in handling business at an intelligent counter. Background Art

[0002] With the development of Internet technology, traditional banks are also actively using big data and artificial intelligence technology. When banks collect user portrait data, these data are mostly used to predict the level of user credit risk and large-scale retail marketing. They lack connection with offline smart counter channels and cannot improve offline user satisfaction and business processing efficiency. The current practice is usually that when users go to the bank's user hall to handle business, the user manager actively inquires about the type of business the user needs to handle. The user manager helps the user select the type of business at the smart counter, and then the user takes a number and queues up to handle the business. During this period, the user manager cannot know in advance the business that the user needs to handle.

[0003] Although there is big data user portrait data in the existing technology, the big data user portrait data is not combined with the smart counter, and the user data cannot be used to predict user behavior, which increases the communication cost between user managers and users.

[0004] Currently, no effective solution has been proposed for the problems in the related technologies. Summary of the invention

[0005] In view of the problems in the related art, the present invention proposes a method for predicting user behavior and assisting in handling business at an intelligent counter to overcome the above-mentioned technical problems existing in the existing related art.

[0006] To this end, the specific technical solution adopted by the present invention is as follows:

[0007] A method for predicting user behavior and assisting in handling business at an intelligent counter includes the following steps:

[0008] S1. Obtain basic user information and behavior tracking information, and perform preprocessing;

[0009] S2. Based on the naive Bayes classifier, a user behavior prediction model is constructed and trained by combining the pre-processed user basic information and behavior embedding information;

[0010] S3, based on the predicted user behavior model, combined with the nearest neighbor collaborative filtering algorithm to classify users and predict the business type of the classified users;

[0011] S4. Filter the predicted user business types and push the filtered business types to the client of the account manager.

[0012] Furthermore, the acquisition of basic user information and behavior tracking information and preprocessing includes the following steps:

[0013] S11. Based on the function keys and operation process steps in the mobile banking, user behavior data is buried to obtain behavior buried information;

[0014] S12, extracting basic user information based on the face recognition system in the user hall;

[0015] S13. Use the big data platform to collect basic user information and behavioral information, collect the corresponding basic user information and behavioral information within the bank, and process them.

[0016] Furthermore, the behavior tracking information includes business type, number of visits and visit time;

[0017] The user's basic information includes gender, age, native place, current permanent residence, education background, occupation, nature of the employer's company, annual income, and whether the user is a corporate legal person.

[0018] Furthermore, the use of the big data platform to collect basic user information and behavior tracking information, collect the corresponding basic user information and behavior tracking information within the bank, and process them includes the following steps:

[0019] S131. Preprocessing the user's behavior information using the big data platform;

[0020] S132: Perform encoding conversion on discrete variables in the user basic information, and process corresponding continuous variables in the user basic information in combination with a probability density function.

[0021] Furthermore, the expression of the probability density function is:

[0022] p(x i |c)~N(μ,σ)

[0023]

[0024] In the formula, p(x i |c) is expressed as the probability density value of the i-th attribute of the sample in the c-th category, x i Represents the value of the i-th attribute of the sample, c represents the category of the sample, N represents normal distribution, μ represents the mean of the i-th attribute of the c-th category samples, and σ represents the variance value of the i-th attribute of the c-th category samples.

[0025] Furthermore, the method of classifying users based on the predicted user behavior model and combining the nearest neighbor collaborative filtering algorithm and predicting the business type of the classified users includes the following steps:

[0026] S31. Based on the basic information of the user, search for the corresponding behavior tracking information in the bank and divide the user into new users and old users;

[0027] S32, based on the nearest neighbor collaborative filtering algorithm, combining the basic information of the user to predict the business type of the new user;

[0028] S33. By predicting the user behavior model, the business type of old users is predicted using the user's basic information and the corresponding behavior tracking information within the bank.

[0029] Furthermore, the predicting of the service type of a new user based on the nearest neighbor collaborative filtering algorithm and combining the basic information of the user comprises the following steps:

[0030] S321, constructing a feature vector of word segmentation based on the preprocessed basic information of the user, and performing Pearson similarity calculation on the user according to the feature vector to obtain a Pearson similarity result;

[0031] S322. Determine the most similar user type according to the Pearson similarity result, sort and filter the business types handled by the most similar user type in ascending order by number of times, and use the filtered business types to predict the business types of new users.

[0032] Furthermore, the Pearson similarity formula is:

[0033] S XY =cov(X,Y) / (σ(X)*σ(Y))

[0034] In the formula, S XY represents Pearson similarity, cov(X,Y) represents the covariance formula, σ(X) represents the standard deviation of X, and σ(Y) represents the standard deviation of Y.

[0035] Furthermore, the expression of the predicted user behavior model is:

[0036]

[0037] Where P(Cj) represents the prior probability of category Cj, p(Cj|X) represents the conditional probability of feature X appearing under category Cj, d represents the number of attributes, Xi represents the value of X on the i-th attribute, and P(X) represents the total probability.

[0038] Furthermore, the method of filtering the predicted user service types and pushing the filtered service types to the client of the account manager includes the following steps:

[0039] S41, selecting the service type of the predicted user from high to low in combination with the probability according to the pre-configured number;

[0040] S42: Push the selected user's business type to the account manager client.

[0041] The beneficial effects of the present invention are:

[0042] The present invention realizes automatic and refined customer classification through a big data platform in combination with multi-dimensional data such as personal information, historical transaction records, behavioral habits, etc., which not only improves the accuracy of classification, but also reduces the pre-preparation time of account managers; the bank customer hall classifies customers according to their basic information; by predicting the type of business that customers may handle based on data such as personal information, historical transaction records, behavioral habits, etc., this not only reduces the communication cost between account managers and customers, but also improves the efficiency of business handling; by generating a business demand forecast report for customers in advance and sending it to the account manager before the account manager contacts the customer, this allows the account manager to prepare in advance, better respond to the customer's business needs, and improve service quality and customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 is a flow chart of a method for predicting user behavior and assisting business processing at an intelligent counter according to an embodiment of the present invention;

[0045] Figure 2 It is a flow chart of a method for predicting user behavior and assisting business handling at an intelligent counter according to an embodiment of the present invention;

[0046] Figure 3 It is a system structure diagram of a method for predicting user behavior and assisting business processing at an intelligent counter according to an embodiment of the present invention. DETAILED DESCRIPTION

[0047] To further illustrate each embodiment, the present invention provides drawings, which are part of the disclosure of the present invention and are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, ordinary technicians in the field should be able to understand other possible implementations and advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0048] According to an embodiment of the present invention, a method for predicting user behavior and assisting in handling business at an intelligent counter is provided.

[0049] The present invention will now be further described with reference to the accompanying drawings and specific embodiments. Figure 1 and Figure 2 According to an embodiment of the present invention, a method for predicting user behavior and assisting in handling business at an intelligent counter is provided. The method for assisting in handling business includes the following steps:

[0050] S1. Obtain basic user information and behavior tracking information, and perform preprocessing.

[0051] In one embodiment, the obtaining of user basic information and behavior tracking information and preprocessing includes the following steps:

[0052] S11. Based on the function keys and operation process steps in the mobile banking, user behavior data is collected to obtain behavior collection information.

[0053] In one embodiment, the behavior tracking information includes business type, number of visits and visit time; the user basic information includes gender, age, place of origin, current permanent residence, education level, occupation, nature of the employer, annual income, and whether the user is a corporate legal person.

[0054] S12. Extracting basic user information based on the face recognition system in the user hall.

[0055] S13. Use the big data platform to collect basic user information and behavioral information, collect the corresponding basic user information and behavioral information within the bank, and process them.

[0056] In one embodiment, the use of a big data platform to collect basic user information and behavior tracking information, collect the corresponding basic user information and behavior tracking information within the bank, and process them includes the following steps:

[0057] S131. Preprocessing the user's behavior information using the big data platform;

[0058] S132: Perform encoding conversion on discrete variables in the user basic information, and process corresponding continuous variables in the user basic information in combination with a probability density function.

[0059] In one embodiment, the probability density function is expressed as:

[0060] p(x i |c)~N(μ,σ)

[0061]

[0062] In the formula, p(xi |c) is expressed as the probability density value of the i-th attribute of the sample in the c-th category, x i Represents the value of the i-th attribute of the sample, c represents the category of the sample, N represents normal distribution, μ represents the mean of the i-th attribute of the c-th category samples, and σ represents the variance value of the i-th attribute of the c-th category samples.

[0063] It needs to be explained that the function keys and operation process steps in the mobile banking are used to collect user behavior data, including the business modules entered by users (business type, number of visits, and visit time); the face recognition system in the customer hall extracts basic information such as customer facial features, gender, age, clothing, and mental state; the big data platform is used for data collection to collect personal user information within the bank and the mobile banking user behavior data. Personal user information includes: gender, age, native place, current permanent residence, education, occupation, nature of the employer, annual income, whether or not a corporate legal person, etc.; mobile banking user behavior operation data: [operation 1, operation 2, operation 3, operation 4, ...] -> [category c ], after the above steps, the business processing is completed; the big data platform is used to pre-process the customer's buried data, encode the user information data, and perform Label coding mapping on discrete variables, such as: gender is processed into female: 0, male: 1, and others: 2; if there is no level difference between unsortable variables, that is, there is no level difference between attributes, one-hot encoding can be used, for example, male: [1, 0, 0], female: [0, 1, 0], and others: [0, 0, 1]; if there is a level difference between sortable variables, that is, there is a level difference between attributes, Label coding can be used, for example, the education level is processed into education level before high school: 1, college: 3, undergraduate: 5, master's degree is processed as 7, and doctoral degree is processed as 10.

[0064] S2. Based on the naive Bayes classifier, a user behavior prediction model is constructed and trained by combining the pre-processed user basic information and behavior tracking information.

[0065] S3. Based on the predicted user behavior model, the nearest neighbor collaborative filtering algorithm is combined to classify users and predict the business type of the classified users.

[0066] In one embodiment, the method of classifying users based on the predicted user behavior model in combination with a nearest neighbor collaborative filtering algorithm and predicting the business type of the classified users includes the following steps:

[0067] S31. Based on the basic information of the user, the corresponding behavior information is searched in the line to divide the user into new users and old users.

[0068] S32. Based on the nearest neighbor collaborative filtering algorithm, the business type of the new user is predicted in combination with the user's basic information.

[0069] In one embodiment, the predicting of the business type of a new user based on a nearest neighbor collaborative filtering algorithm and combining the basic information of the user comprises the following steps:

[0070] S321, constructing a feature vector of word segmentation based on the preprocessed basic information of the user, and performing Pearson similarity calculation on the user according to the feature vector to obtain a Pearson similarity result;

[0071] S322. Determine the most similar user type according to the Pearson similarity result, sort and filter the business types handled by the most similar user type in ascending order by number of times, and use the filtered business types to predict the business types of new users.

[0072] S33. By predicting the user behavior model, the business type of old users is predicted using the user's basic information and the corresponding behavior tracking information within the bank.

[0073] In one embodiment, the Pearson similarity formula is:

[0074] S XY =cov(X,Y) / (σ(X)*σ(Y))

[0075] In the formula, S XY represents Pearson similarity, cov(X,Y) represents the covariance formula, σ(X) represents the standard deviation of X, and σ(Y) represents the standard deviation of Y.

[0076] It should be explained that the word2Vec algorithm is used to construct a feature vector for word segmentation of user behavior data: [gender, age, native place, current permanent residence, education, occupation, nature of the employer, annual income, whether a corporate legal person, operation 1, operation 2, operation 3, operation 4, ...] -> [category c].

[0077] In addition, if the customer is a new customer, the customer is classified using the nearest neighbor collaborative filtering algorithm in combination with external customer information data and internal customer information data. The nearest neighbor collaborative filtering algorithm uses Pearson similarity to sort and generate the types of business most frequently handled by classified users.

[0078] Among them, the Pearson similarity between two vectors is:

[0079] def pearson(v1, v2):

[0080] v1_mean = np.mean(v1)

[0081] v2_mean = np.mean(v2)

[0082] return(np.dot(v1-v1_mean, v2-v2_mean)) / (np.linalg().norm(v1-v1_mean)*np.linalg().norm(v2-v2_mean))

[0083] 1) Pearson similarity calculation is performed between the new customer feature vector New_User_V and the feature vectors of the historical user list that have been labeled in the bank one by one;

[0084] 2) Sort the Pearson similarity in ascending order;

[0085] 3) Take the user categories corresponding to the first K Pearson similarities (usually K is 10);

[0086] 4) Take the category with the largest number of categories among these K as the service category of the new user and make service type recommendations.

[0087] In addition, if the customer is an existing old customer of the bank, the user's basic information, historical data of the user's operation behavior and historical data of the corresponding business types are combined into training samples to train the naive Bayes classifier prediction model. Then, when these existing old customers appear in the service hall, their basic information and recent operation behavior data are obtained to predict the type of business the user is about to handle. These predicted business types are sorted in ascending order, and the top 10 types are taken and pushed to the account manager client.

[0088] In one embodiment, the expression of the predicted user behavior model is:

[0089]

[0090] In the formula, P(Cj) represents the prior probability of category Cj, p(Cj|X) represents the conditional probability of feature X appearing under category Cj, d represents the number of attributes, Xi represents the value of X on the i-th attribute, and P(X) represents the total probability;

[0091] Since P(X) is the same for all classes, the calculation of P(X) can be omitted when comparing the probabilities of different classes.

[0092] It should be explained that the naive Bayes classifier builds a model to predict user behavior: for a given sample to be classified, the probability of it belonging to each category is calculated, and then the sample is divided into the category with the highest probability.

[0093] In addition, P(Cj) is the prior probability of category Cj, and p(Cj|X) represents the conditional probability of the occurrence of feature X under the condition of category Cj, also known as the likelihood function. For the naive Bayes classifier model, it is a classification model based on Bayes theorem and the conditional independence assumption of features. It assumes that under a given category, the features are independent of each other, that is, the probability of occurrence of a feature will not be affected by other features, so Here d is the number of attributes, Xi is the value of X on the i-th attribute; p(X) is the total probability, also known as evidence, which compresses the occurrence of the event into the sample space of X. Since p(X) is the same for all categories, the calculation of p(X) can be omitted when comparing the probabilities of different categories.

[0094] for Calculate and run Laplace smoothing to avoid the probability of one of the items being 0, which will result in a result of 0; for multiplication of multiple probabilities, that is, There will be a case where one of the categories has a profile of 0, and the final result is also 0. In order to avoid this effect, when calculating the probability of each category, the numerator is initialized to 1 and the denominator is initialized to 2. This method is called Plaus smoothing, also known as plus 1 smoothing. Multiplying too many very small numbers will cause underflow, and the natural logarithm of the product result will be taken.

[0095] S4. Filter the predicted user business types and push the filtered business types to the client of the account manager.

[0096] In one embodiment, the method of filtering the predicted user service types and pushing the filtered service types to the client of the account manager includes the following steps:

[0097] S41, selecting the service type of the predicted user from high to low in combination with the probability according to the pre-configured number;

[0098] S42: Push the selected user's business type to the account manager client.

[0099] What needs to be explained is that when a user enters the service hall, the camera is used to identify the user and obtain user information; the user's recent mobile banking operation data is called; the user information and behavioral operation data are transmitted to the Naive Bayes model to predict user behavior, sort them by probability from large to small, and take the nearly 10 high-probability behaviors, and push them to the smart client APP. The customer manager sees the predicted customer behavior, and then communicates with the customer to assist the customer in handling business.

[0100] In addition, Bayesian algorithm is a general term for a class of algorithms based on Bayesian theorem, which are widely used in statistics, machine learning, data mining and other fields. Bayesian theorem is a theorem about the conditional probability of random events A and B, and its formula is:

[0101] P(A|B)=P(B|A)*P(A) / P(B)

[0102] In the formula, P(A|B) represents the probability of event A occurring under the condition that event B occurs; P(B|A) represents the probability of event B occurring under the condition that event A occurs; P(A) represents the probability of event A occurring; P(B) represents the probability of event B occurring.

[0103] The core idea of ​​the Bayesian algorithm is to use known probabilities to predict unknown probabilities. In practical applications, the Bayesian algorithm is often used for classification problems, especially when the amount of data is insufficient. It can make predictions through prior probabilities and posterior probabilities.

[0104] In addition, the normal distribution, also known as the normal distribution or Gaussian distribution, is usually recorded as X~N(μ,σ). Among them, μ represents the mathematical expectation (mean) of the normal distribution, and σ represents the variance of the normal distribution. The normal distribution with μ=0 and σ=1 is called the standard normal distribution.

[0105] Covariance is used in probability theory and statistics to measure the overall error of two variables. Variance is a special case of covariance, that is, when the two variables are the same.

[0106] Covariance represents the overall error of two variables, which is different from the variance of the error of only one variable. If the two variables have the same trend, that is, if one is greater than its expected value and the other is also greater than its expected value, then the covariance between the two variables is positive. If the two variables have opposite trends, that is, one is greater than its expected value and the other is less than its expected value, then the covariance between the two variables is negative.

[0107] The covariance Cov(X, Y) between two real random variables X and Y with expected values ​​E[X] and E[Y] is defined as:

[0108] Coυ(X,Y)=E[(XE[X])(YE[Y])]

[0109] =E[XY]-2E[Y]E[X]+E[X]E[Y]

[0110] =E[XY]-E[X]E[Y]

[0111] Pearson similarity, also known as Pearson correlation coefficient, is used to measure the degree of correlation between two variables. The value range is [-1, 1]; 1 represents perfect correlation, 0 represents no correlation, and -1 represents perfect negative correlation.

[0112] In order to facilitate understanding of the above technical solutions of the present invention, the above technical solutions of the present invention are further described below from the perspective of architecture and principle, as follows: Figure 3 As shown:

[0113] The client sends a request to the access scheduling layer. At the same time, the protocol access module communicates with the distribution scheduling module. The access scheduling layer sends a request to the recommendation algorithm layer. At the same time, the recall module in the recommendation algorithm layer communicates with the filtering module, the fine sorting module and the mixed sorting module in turn. The recall module communicates with the recall service module in the public component, and the fine sorting module communicates with the sorting service module in the public component.

[0114] The client reports logs to the reporting center, and the reporting center sends a message queue to the real-time processing module. The real-time processing module includes a log processing module, a data statistics module, a click rate update module and a portrait update module. The real-time processing module sends landing data to the data warehouse in the storage unit. At the same time, the real-time processing module sends log statistics to the indicator object data module in the storage unit. The real-time processing module sends feature updates to the high-speed memory buffer database in the storage unit. The data warehouse in the storage unit sends offline data indicators to the indicator object data module in the storage unit. The indicator object data module in the storage unit communicates with the reporting system in the storage unit. At the same time, the indicator object data module in the storage unit sends a model to the sorting service module in the public component, the high-speed memory buffer database in the storage unit sends user features to the sorting service module in the public component, and the memory index module in the storage unit sends an index table to the recall service module in the public component. The public component communicates with the recommendation algorithm layer, the recommendation algorithm layer returns the result to the access scheduling layer, and the access scheduling layer returns the result to the client.

[0115] To sum up, with the help of the above-mentioned technical scheme of the present invention, the present invention realizes automatic and refined customer classification through a big data platform, combined with multi-dimensional data such as customer's personal information, historical transaction records, behavioral habits, etc., which not only improves the accuracy of classification, but also reduces the pre-preparation time of account managers; the bank customer hall classifies customers according to their basic information; by predicting the type of business that customers may handle based on their personal information, historical transaction records, behavioral habits and other data, this not only reduces the communication cost between account managers and customers, but also improves the efficiency of business handling; by generating a customer's business demand forecast report in advance and sending it to the account manager before the account manager contacts the customer, this allows the account manager to prepare in advance, better respond to the customer's business needs, and improve service quality and customer satisfaction.

[0116] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting user behavior and assisting in handling business at an intelligent counter, characterized in that: The auxiliary business handling method comprises the following steps: S1. Obtain basic user information and behavior tracking information, and perform preprocessing; S2. Based on the naive Bayes classifier, a user behavior prediction model is constructed and trained by combining the pre-processed user basic information and behavior embedding information; S3, based on the predicted user behavior model, combined with the nearest neighbor collaborative filtering algorithm to classify users and predict the business type of the classified users; S4. Filter the predicted user business types and push the filtered business types to the client of the account manager.

2. According to claim 1, a method for predicting user behavior and assisting in handling business at an intelligent counter is characterized in that: The acquisition of user basic information and behavior tracking information and preprocessing includes the following steps: S11. Based on the function keys and operation process steps in the mobile banking, user behavior data is buried to obtain behavior buried information; S12, extracting basic user information based on the face recognition system in the user hall; S13. Use the big data platform to collect basic user information and behavioral information, collect the corresponding basic user information and behavioral information within the bank, and process them.

3. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 2 is characterized in that: The behavior tracking information includes business type, number of visits and visit time; The user's basic information includes gender, age, native place, current permanent residence, education background, occupation, nature of the employer's company, annual income, and whether the user is a corporate legal person.

4. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 2 is characterized in that: The method of collecting basic user information and behavior information using a big data platform, collecting the corresponding basic user information and behavior information within the bank, and processing the information includes the following steps: S131. Preprocessing the user's behavior information using the big data platform; S132: Perform encoding conversion on discrete variables in the user basic information, and process corresponding continuous variables in the user basic information in combination with a probability density function.

5. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 4 is characterized in that: The expression of the probability density function is: p(x i |c)~N(μ,σ) In the formula, p(x i |c) is expressed as the probability density value of the i-th attribute of the sample in the c-th category, x i Represents the value of the i-th attribute of the sample, c represents the category of the sample, N represents normal distribution, μ represents the mean of the i-th attribute of the c-th category samples, and σ represents the variance value of the i-th attribute of the c-th category samples.

6. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 1 is characterized in that: The method of classifying users based on the predicted user behavior model and combining the nearest neighbor collaborative filtering algorithm and predicting the business type of the classified users includes the following steps: S31. Based on the basic information of the user, search for the corresponding behavior tracking information in the bank and divide the user into new users and old users; S32, based on the nearest neighbor collaborative filtering algorithm, combining the basic information of the user to predict the business type of the new user; S33. By predicting the user behavior model, the business type of old users is predicted using the user's basic information and the corresponding behavior tracking information within the bank.

7. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 6, characterized in that: The method of predicting the service type of a new user based on the nearest neighbor collaborative filtering algorithm and combining the basic information of the user comprises the following steps: S321, constructing a feature vector of word segmentation based on the preprocessed basic information of the user, and performing Pearson similarity calculation on the user according to the feature vector to obtain a Pearson similarity result; S322. Determine the most similar user type according to the Pearson similarity result, sort and filter the business types handled by the most similar user type in ascending order by number of times, and use the filtered business types to predict the business types of new users.

8. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 7 is characterized in that: The Pearson similarity formula is: S XY =cov(X,Y) / (σ(X)*σ(Y)) In the formula, S XY represents Pearson similarity, cov(X,Y) represents the covariance formula, σ(X) represents the standard deviation of X, and σ(Y) represents the standard deviation of Y.

9. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 6, characterized in that: The expression of the predicted user behavior model is: Where P(Cj) represents the prior probability of category Cj, p(Cj|X) represents the conditional probability of feature X appearing under category Cj, d represents the number of attributes, Xi represents the value of X on the i-th attribute, and P(X) represents the total probability.

10. The method for predicting user behavior and assisting business handling at an intelligent counter according to claim 1, characterized in that: The method of filtering the predicted user business types and pushing the filtered business types to the client of the account manager includes the following steps: S41, selecting the service type of the predicted user from high to low in combination with the probability according to the pre-configured number; S42: Push the selected user's business type to the account manager client.

Images