Intelligent service resource pushing system and method based on user behaviors

Abstract

The invention discloses an intelligent service resource pushing system and method based on user behaviors, and relates to the technical field of resource data pushing, and the method comprises the following steps: dividing user behavior data into a first sample and a second sample according to the urgency dimension and the importance dimension of the user behavior data, the first sample is used for calculating an emergency coefficient of the user demand, the second sample is used for calculating an importance coefficient of the user demand, and the emergency coefficient and the importance coefficient are combined and calculated through a preset formula to obtain a demand weight representing the priority of the user demand; and calculating the supply margin of the service provider, the load rate negatively related to the supply margin and the service response time based on the service provider operation data, setting a supply constraint condition, presetting a multi-objective optimization calculation logic, and screening out an optimal service resource under the condition that the supply constraint condition is satisfied by combining the demand weight, the supply margin and the service response time. And the service resource list is generated and pushed, so that the service pushing accuracy and the service ecological operation efficiency are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource data pushing, in particular to a service resource intelligent pushing system and method based on user behavior. BACKGROUND

[0002] In recent years, with the diversification of digital service scenarios, the number and types of service resources have increased exponentially. User behavior data, such as scenario triggering and browsing interaction, has become a key basis for mining demand preferences. Service resource intelligent pushing technology has become a core link connecting user personalized needs and service provider supply capabilities. It is related to whether users can quickly obtain suitable high-value services, and directly affects the resource allocation efficiency and operating costs of service providers.

[0003] However, the existing pushing scheme has obvious limitations. On the one hand, most schemes only measure demand priority through a single indicator such as user browsing frequency, without dimensionally splitting and quantifying the urgency of demand such as time sensitivity and importance such as business relevance. The weight configuration is mostly a fixed value, which cannot dynamically adapt to the characteristic differences of different demands. On the other hand, existing technologies often only focus on the user demand side, without linking and calibrating the load rate of service providers, the running state of supply margin, and demand weight. There is also a lack of multi-objective optimization filtering logic, resulting in problems such as matching low-response resources for high-urgency demands, service provider resource supply and demand imbalance, and difficulty in balancing user experience and service supply stability. SUMMARY

[0004] The technical problem solved by the present application is that the existing pushing scheme has obvious limitations. On the one hand, most schemes only measure demand priority through a single indicator such as user browsing frequency, without dimensionally splitting and quantifying the urgency of demand such as time sensitivity and importance such as business relevance. The weight configuration is mostly a fixed value, which cannot dynamically adapt to the characteristic differences of different demands. On the other hand, existing technologies often only focus on the user demand side, without linking and calibrating the load rate of service providers, the running state of supply margin, and demand weight. There is also a lack of multi-objective optimization filtering logic, resulting in problems such as matching low-response resources for high-urgency demands, service provider resource supply and demand imbalance, and difficulty in balancing user experience and service supply stability.

[0005] To solve the above technical problems, the present application provides the following technical scheme: a service resource intelligent pushing method based on user behavior, comprising the following steps: Step S1, according to the urgency dimension and importance dimension of user behavior data, the user behavior data is divided into a first sample and a second sample, the first sample is used to calculate the urgency coefficient of user demand, and the second sample is used to calculate the importance coefficient of user demand; In step S2, the emergency coefficient and the importance coefficient are combined and calculated through a preset formula to obtain a demand weight representing a priority of a user demand; The service provider's supply margin, a load rate negatively correlated with the supply margin, and a service response time are calculated based on service provider operation data, and a supply constraint condition is set; In step S3, a multi-objective optimization calculation logic is preset, and the demand weight, the supply margin, and the service response time are combined to filter out optimal service resources under the condition of meeting the supply constraint condition, and a service resource list is generated and pushed.

[0006] As a preferred scheme of the service resource intelligent pushing method based on user behavior, the user behavior data is divided into a first sample and a second sample according to an emergency dimension and an importance dimension of the user behavior data, the first sample is used to calculate an emergency coefficient of a user demand, and the second sample is used to calculate an importance coefficient of the user demand, and the emergency dimension index includes a user demand trigger frequency and a time sensitivity; The user demand trigger frequency is a trigger frequency of a scene in which the user is located and a scene associated with a similar service in a preset period, and the scene association refers to a scene matching relationship that is strongly related to a similar service function and is recognized through objective data collected by a user terminal; The similar service is a service in an industry knowledge graph that belongs to big data extraction, and the objective data is quantitative data used to represent an environment and a behavior state of the user; The time sensitivity is determined based on an instant response behavior duration of the user to the similar service pushing, the instant response behavior duration refers to a duration from the similar service pushing to the user clicking or using the similar service, and the shorter the duration, the higher the time sensitivity; The importance dimension index includes a business association degree and a user attention degree; The business association degree is determined based on a functional matching degree of a business type to which the user belongs and a service resource, and the higher the matching degree, the higher the business association degree; The user attention degree is determined based on a browsing frequency and a collection frequency of the user to the similar service resource; The user behavior data used to calculate the emergency dimension index is taken as the first sample; The user behavior data used to calculate the importance dimension index is taken as the second sample; The emergency coefficient of the user demand is calculated based on each emergency dimension index in the first sample; The importance coefficient of the user demand is calculated based on each importance dimension index in the second sample; The expression for calculating the emergency coefficient is ; wherein, This represents the urgency coefficient. This indicates the time sensitivity of the urgency dimension in the first sample. This indicates the frequency of demand triggering in the urgency dimension of the first sample. and This represents a pre-set empirical parameter, which is a constant. The calculated urgency coefficient is normalized to the first interval, which is a double-closed interval from 0 to 1. The expression for calculating the importance coefficient is as follows: ; in, Indicates importance coefficient, This indicates the business relevance of the importance dimension in the second sample. This indicates the user attention level of the importance dimension in the second sample. and This represents a pre-set empirical parameter, which is a constant. The calculated importance coefficients are normalized to the first interval.

[0007] As a preferred embodiment of the intelligent push method for service resources based on user behavior described in this invention, the urgency coefficient and importance coefficient are combined and calculated using a preset formula to obtain the demand weight representing the priority of user needs. Specifically, this includes using a weighted normalization method on the urgency dimension index and the importance dimension index to generate the corresponding dynamic weight of the urgency coefficient and the dynamic weight of the importance coefficient. The expression for the dynamic weight of the urgency coefficient is, ; The expression for the dynamic weight of the importance coefficient is as follows: ; in, The dynamic weight representing the urgency factor. The dynamic weights representing the importance coefficients These represent the preset initial weights for the urgency and importance dimensions, respectively. These initial weights are constants and are determined statistically based on the service resource type. Indicates the first Standardized values ​​of each urgency dimension indicator. The range of values ​​for is the first interval. This indicates the contribution weight of the urgency dimension indicators; This represents the standardized value of the l-th importance dimension indicator, with a value range of the first interval. Indicates the contribution weight of the indicators in the importance dimension; m represents the number of emergency dimension indicators, and n represents the number of importance dimension indicators.

[0008] As a preferred solution of the service resource intelligent pushing method based on user behavior, the emergency coefficient, the importance coefficient, the dynamic weight of the emergency coefficient, and the dynamic weight of the importance coefficient are substituted into a weighted formula to obtain an initial demand weight, and an expression of the initial demand weight is, wherein, represents the initial weight, represents the emergency coefficient, represents the importance coefficient, and the value range of the emergency coefficient is a first interval.

[0009] As a preferred solution of the service resource intelligent pushing method based on user behavior, the initial demand weight is corrected based on a historical same-period load rate of a service provider to obtain a calibrated demand weight, and an expression of the correction is, wherein, represents the calibrated demand weight, both represent correction weight distribution coefficients, and , represents a correction coefficient, which is determined according to response elasticity of the service resource, represents a preset average load rate, and the value range of the average load rate is the first interval; The value range of the calibrated demand weight is the first interval, and the calibrated demand that meets a preset threshold is taken as a candidate object to enter an optimal service resource screening pool.

[0010] As a preferred solution of the service resource intelligent pushing method based on user behavior, the supply margin of the service provider, the load rate negatively correlated with the supply margin, and the service response time are calculated based on service provider operation data, and a supply constraint condition is set, and the service provider operation data specifically includes the maximum service capacity of the service provider, the occupied capacity of the service provider, the straight-line distance between the service provider and the user, the average service response speed data of the service provider, and the queuing waiting time length of the current unprocessed demand of the service provider. The supply margin is a difference between the maximum service capacity and the occupied capacity of the service provider. The maximum service capacity is a maximum number of service demands accepted by the service provider in a preset period, and the occupied capacity refers to a number of service demands currently accepted by the service provider and not completed. The quotient of the occupied capacity and the maximum service capacity of the service provider is recorded as a first quotient. ​​The load rate is in a first interval, and the load rate and the supply margin are in a negative correlation relationship, and when the supply margin increases, the load rate decreases. The quotient of the straight-line distance between the computing service provider and the user and the average service response speed of the service provider is taken as a second quotient value, and the sum of the second quotient value and the queuing waiting time of the current unprocessed demand of the service provider is taken as a service response time.

[0011] As a preferred scheme of the service resource intelligent pushing method based on user behavior, the setting content of the supply constraint condition includes supply margin constraint, load rate constraint and service response time constraint. The supply margin constraint is that the supply margin is greater than or equal to a preset supply margin threshold, and the supply margin threshold is determined according to the type of service resource. The load rate constraint is that the load rate is less than or equal to a preset load rate threshold, and the value range of the load rate threshold is a second interval, which is a double closed interval from a first value to a second value. The service response time constraint is that the service response time is less than or equal to a preset service response time threshold, and the service response time threshold is determined according to the response requirement of the service resource.

[0012] As a preferred scheme of the service resource intelligent pushing method based on user behavior, a multi-objective optimization calculation logic is preset, the optimal service resource is selected under the condition of meeting the supply constraint condition in combination with the demand weight, the supply margin and the service response time, a service resource list is generated and pushed, and specifically, the maximum of the calibrated demand weight is taken as a main target, and the maximum of the supply margin and the minimum of the service response time are taken as secondary targets to form a target priority sequence of multi-objective optimization. First, service providers meeting the supply constraint condition are filtered to obtain a candidate service provider set. The candidate service provider set is sorted based on the target priority sequence of multi-objective optimization, and specifically, the calibrated demand weight is taken as a first sorting basis, and the candidate service provider set is preliminarily sorted according to the calibrated demand weight from high to low. For the service providers with the same calibrated demand weight or with a difference less than a preset threshold in the preliminary sorting, a secondary sorting is performed according to a coordination index as a second sorting basis, and the expression of the coordination index F is, Wherein, F represents the coordination index, S represents the supply margin, and T represents the service response time, S represents the supply margin weight coefficient, T represents the service response time weight coefficient, and ;​ when the post-calibration demand weight increases, the increases, the decreases.

[0013] As a preferred solution of the service resource intelligent pushing method based on user behavior, the sorting result of the secondary sorting is modified by the business correlation degree, and the sorting position of the service provider whose business correlation degree is lower than the preset threshold is adjusted to be after the service provider whose business correlation degree is not lower than the preset threshold in the current sorting sequence. According to the modified sorting result, the first N service providers are selected to generate a service resource list, and the service resource list is pushed to the user terminal in the sorting order.

[0014] The service resource intelligent pushing system based on user behavior comprises a division module, a calculation module, and a pushing module. The division module divides the user behavior data into a first sample and a second sample according to the urgency dimension and the importance dimension of the user behavior data, the first sample is used to calculate an urgency coefficient of user demand, and the second sample is used to calculate an importance coefficient of user demand. The calculation module combines the urgency coefficient and the importance coefficient by a preset formula to obtain a demand weight representing a priority of user demand. The supply margin of the service provider, the load rate negatively related to the supply margin, and the service response time are calculated based on the service provider running data, and a supply constraint condition is set. The pushing module sets a multi-objective optimization calculation logic, combines the demand weight, the supply margin, and the service response time, and filters out optimal service resources under the condition of meeting the supply constraint condition to generate and push a service resource list.

[0015] The present application has the following advantages: by splitting the urgency and importance dimensions of demand and quantifying the coefficients, different demand characteristics are matched by dynamic weight adaptation, the single-index and fixed-weight coarse-grained matching problem of the existing solution is solved, the demand weight is calibrated in combination with the load rate of the service provider, and the precision matching of high-priority demand of the user and high-response resources is realized, the supply and demand load of the service provider is balanced, and the service pushing precision and service ecological operation efficiency are significantly improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A step flowchart of a service resource intelligent pushing method based on user behavior is provided for an embodiment of the present application.

[0017] Figure 2 A basic flowchart of a service resource intelligent pushing system based on user behavior is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0018] In order to make the above objectives, characteristics and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present application, not all embodiments.

[0019] Embodiment 1, refer to Figure 1 For an embodiment of the present application, a service resource intelligent pushing method based on user behavior is provided, including the following steps: Step S1, according to the urgency dimension and importance dimension of user behavior data, the user behavior data is divided into first sample and second sample, the first sample is used to calculate the urgency coefficient of user demand, and the second sample is used to calculate the importance coefficient of user demand; Step S2, the urgency coefficient and the importance coefficient are combined calculated through a preset formula, and the demand weight representing the priority of user demand is obtained; Based on the service provider operation data, the supply margin of the service provider, the load rate negatively related to the supply margin and the service response time are calculated, and the supply constraint condition is set; Step S3, a multi-objective optimization calculation logic is preset, and the optimal service resource is selected under the condition of meeting the supply constraint condition, combined with the demand weight, the supply margin and the service response time, to generate and push a service resource list.

[0020] In one embodiment, according to the urgency dimension and importance dimension of user behavior data, the first sample and the second sample are divided to calculate the urgency coefficient and the importance coefficient of user demand, and then the urgency coefficient and the importance coefficient are combined calculated through a preset formula to obtain the demand weight representing the priority of user demand. At the same time, based on the service provider operation data, the supply margin, the load rate negatively related to the supply margin and the service response time are calculated, and the supply constraint condition is set. Finally, a multi-objective optimization calculation logic is preset, and the optimal service resource is selected under the condition of meeting the supply constraint condition, combined with the demand weight, the supply margin and the service response time, to generate and push a service resource list, which realizes the precise dynamic adaptation of user demand and service resource, effectively balances the response efficiency of user high-priority demand and the supply and demand load of service provider, and significantly improves the pertinence of service pushing and the overall operation stability of service ecology.

[0021] According to the urgency dimension and importance dimension of user behavior data, the user behavior data is divided into first sample and second sample, the first sample is used to calculate the urgency coefficient of user demand, and the second sample is used to calculate the importance coefficient of user demand, the urgency dimension index includes user demand trigger frequency and time sensitivity; The user demand trigger frequency is a trigger frequency of a scene in which the user is located and a scene associated with the same type of service within a preset period. The scene association refers to a matching relationship between the scene and the same type of service function that is recognized by collecting objective data by the user terminal; The same type of service is a service in an industry knowledge graph that belongs to the same big data extraction. The objective data is quantitative data for representing the environment and behavior state of the user; The time sensitivity is determined based on the length of the immediate response behavior of the user to the same type of service push. The length of the immediate response behavior refers to the length of time from the same type of service push to the user clicking or using the same type of service. The shorter the length of time, the higher the time sensitivity; The importance dimension indicators include business association degree and user attention degree; The business association degree is determined based on the function matching degree of the business type to which the user belongs and the service resource. The higher the matching degree, the higher the business association degree; The user attention degree is determined based on the browsing frequency and the collection frequency of the user to the same type of service resource; The user behavior data for calculating the urgency dimension indicators are taken as the first sample; The user behavior data for calculating the importance dimension indicators are taken as the second sample; The urgency coefficient of the user demand is calculated based on each urgency dimension indicator in the first sample; The importance coefficient of the user demand is calculated based on each importance dimension indicator in the second sample; The expression for calculating the urgency coefficient is, ; Wherein, The urgency coefficient is represented by, The time sensitivity of the urgency dimension in the first sample is represented by, The demand trigger frequency of the urgency dimension in the first sample is represented by, And The empirical parameter is represented by, The calculated urgency coefficient is normalized to the first interval, which is a double closed interval of 0 to 1; The expression for calculating the importance coefficient is, ; Wherein, The importance coefficient is represented by, The business association degree of the importance dimension in the second sample is represented by, The user attention degree of the importance dimension in the second sample is represented by, And The empirical parameter is represented by, The calculated importance coefficient is normalized to a first interval.

[0022] In one of the embodiments, the user demand trigger frequency calculation method is that the preset period is 1 natural day, and the setting basis is to reflect the user's short-term immediate demand based on the urgency dimension index. The scene data collected by the user terminal, such as the positioning of the mobile terminal of the smart phone and the tablet computer carried by the user daily, the device use state, the scene in which the user is located is recognized, the operation trigger times of the functions strongly related to the same type of service in the scene are counted, the strongly related operation refers to the user operation directly related to the core function of the same type of service, specifically including clicking the entrance of the same type of service, initiating the demand application of the corresponding service, viewing the details or available resource list of the service, and excluding browsing irrelevant pages, clicking the general navigation bar and other operations without clear service demand direction, for example, the user is an enterprise administrative officer, and is in the enterprise office procurement scene, the number of times of clicking the office material procurement service is the user demand trigger frequency F of the current period; The time sensitivity calculation method is to count the instant response behavior time length of the same type of service push message sent to the user, and set the corresponding rules including that when the instant response behavior time length is less than or equal to 5 minutes, η is 1, when the instant response behavior time length is greater than 5 minutes and less than or equal to 30 minutes, η is 0.7, when the instant response behavior time length is greater than 30 minutes and less than or equal to 60 minutes, η is 0.1, and when the instant response behavior time length is greater than 60 minutes, η is 0.05, which indicates that the user's urgent demand for the current pushed service is low.

[0023] The value range of the business correlation degree is the left-open right-closed interval of 0 to 1. In the actual scene, the service resource and the user business type at least exist basic function matching, so R will not be 0. If an extremely special no-matching scene occurs, the minimum threshold value of the preset R is 0.01. The calculation process of the business correlation degree is that a core function label library corresponding to each business type of the enterprise is constructed in advance. Each business type corresponds to a unique core function label set. At the same time, a unique corresponding business keyword is configured for each core function label. The function description text of the service resource is extracted, the text is segmented and stop word processing is performed, and the function keyword set of the service resource is obtained. The function keywords of the service resource and the business keywords corresponding to the core function labels are matched one by one. If the function keywords of the service resource and the business keywords are the same, the function label corresponding to the business keyword is retained. If they are not the same, the function label corresponding to the business keyword is eliminated. The matched function label set is screened. According to the business type to which the user belongs, the core function label library corresponding to the business type is called. Then the matched function label set screened is compared with the labels in the user business core label library one by one. The number of function labels that match the labels in the user business core label library is counted, that is, the number of finally matched labels. The business correlation degree is calculated. The business correlation degree is the quotient value of the number of finally matched labels and the total number of labels in the user business core label library. If the calculated business correlation degree is 0, the business correlation degree is corrected to the preset minimum threshold value 0.01, so that the value of the business correlation degree meets the left-open right-closed interval of 0 to 1. The user attention degree calculation method is that the preset statistical period is 7 natural days. The setting basis is that the importance dimension index needs to reflect the characteristics of the user's medium-term attention preference. The browsing and collection frequencies of the user to the same type of service in the period are counted. Based on the collection behavior, compared with the browsing behavior, the actual demand intention of the enterprise user to the service can be better reflected. The weighted sum is calculated according to the browsing weight 0.4 plus the collection weight 0.6. Then the weighted sum is divided by the maximum weighted sum of the same type of service. For example, the maximum weighted sum of a single week in the industry is 15 to obtain the normalized result. For example, if the user browses 4 times and collects 2 times in 1 week, the weighted sum is 4*0.4+2*0.6=2.8, and the attention degree D is approximately 0.19. If the user browses 6 times and collects 3 times in 1 week, the weighted sum is 6*0.4+3*0.6=4.2, and D=0.28.

[0024] The behavioral data corresponding to the urgency index is used as the first sample, and the data corresponding to the importance index is used as the second sample. These are then substituted into the formula to calculate the urgency coefficient E and the importance coefficient I. The empirical parameters can be selected to suit the specific scenario. For example, when adapting to enterprise services, C1 is set to 2, C2 to 1, C3 to 0.8, and C4 to 0.2; when adapting to local life services, C5 is set to 3, C6 to 0.5, C7 to 1.2, and C8 to 0.3. The calculated E and I are then normalized to a double-closed interval between 0 and 1. This implementation method clarifies the quantitative logic of each dimension of the indicators, adapts empirical parameters to different service scenarios, solves the ambiguity problem in the calculation of demand coefficients, and effectively improves the accuracy of user demand quantification.

[0025] The urgency coefficient and importance coefficient are combined and calculated using a preset formula to obtain the demand weight that represents the priority of user needs. Specifically, the urgency index and importance dimension index are weighted and normalized to generate the corresponding dynamic weights of the urgency coefficient and the importance coefficient. The expression for the dynamic weight of the urgency factor is, ; The expression for the dynamic weight of the importance coefficient is, ; in, The dynamic weight representing the urgency factor. The dynamic weights representing the importance coefficients These represent the preset initial weights for the urgency and importance dimensions, respectively. These initial weights are constants and are determined statistically based on the service resource type. Indicates the first Standardized values ​​of each urgency dimension indicator. The range of values ​​for is the first interval. This indicates the contribution weight of the urgency dimension indicators; This represents the standardized value of the l-th importance dimension indicator, with a value range of the first interval. Indicates the contribution weight of the indicators in the importance dimension; m represents the number of urgency-related indicators, and n represents the number of importance-related indicators.

[0026] In one embodiment, the urgency coefficient and importance coefficient are combined by a preset formula to obtain the demand weight. Specifically, the urgency dimension index and the importance dimension index are weighted and normalized to generate the corresponding dynamic weights of the urgency coefficient and the importance coefficient. When adapting to enterprise service scenarios, first set initial weights based on statistical data of office services. The setting is 0.4, the setting of β0 is 0.6, the setting basis is that the importance priority of demand is slightly higher than the urgency in the enterprise scene, the importance of demand corresponds to the business correlation degree, and the urgency corresponds to the demand triggering frequency. Since the number of emergency degree dimension indicators m is 2, including demand triggering frequency and time sensitivity, the number of importance dimension indicators n is 2, including business correlation degree and user attention degree, the normalized value A1 of the emergency degree indicator corresponds to the normalized value 0.4 of the demand triggering frequency F in the preceding office procurement scene, and A2 corresponds to the normalized value 1 of the time sensitivity η. Since the influence degree of the two emergency degree dimension indicators of demand triggering frequency and time sensitivity on the emergency degree is similar, the contribution weight γ1 of the demand triggering frequency corresponding to the first emergency degree indicator is set, that is, γ k in the formula, k is 1, γ1 is 0.5, the contribution weight γ2 of the time sensitivity corresponding to the second emergency degree indicator is set, that is, γ k in the formula, k is 2, γ2 is 0.5, the normalized value of the importance indicator is taken, B1 corresponds to the normalized value 0.75 of the business correlation degree R, and B2 corresponds to the normalized value 0.28 of the user attention degree D. According to the fact that the business matching degree has a higher influence in the enterprise scene, the contribution weight δ1 of the business correlation degree corresponding to the first importance indicator is set, that is, δ l in the formula, l is 1, δ1 is 0.6, the contribution weight δ2 of the user attention degree corresponding to the second importance indicator is set, that is, δ l in the formula, l is 2, and δ2 is 0.4. The initial weight α0, β0, the normalized values A1 and A2 of the emergency degree indicator, and the contribution weights γ1 and γ2 of the importance indicator are substituted into the emergency coefficient dynamic weight formula, and α=(0.4+0.5×0.4+0.5×1) / (0.4+0.6+0.5×0.4+0.5×1+0.6×0.75+0.4×0.28)≈0.47 is calculated, and β≈0.53 is obtained by β=1-α. In this way, by setting the numerical value of the initial weight, the contribution weight and the setting basis, the parameter ambiguity problem in dynamic weight calculation is solved, the weight distribution is more suitable for the characteristics of the enterprise service scene, and the rationality of demand priority evaluation is effectively improved.

[0027] The emergency coefficient, the importance coefficient, the dynamic weight of the emergency coefficient and the dynamic weight of the importance coefficient are substituted into the weighted formula to obtain the initial demand weight, and the expression of the initial demand weight is, ; wherein, represents the initial weight, represents the emergency coefficient, represents the importance coefficient, and the value range of the is the first interval.

[0028] Based on the historical load rate of the service provider, the average value of the load rate of the current period is preset, the initial demand weight is corrected, the calibrated demand weight is obtained, the expression of the correction is, ; wherein, represents the calibrated demand weight, both represent the correction weight distribution coefficient, and , represents the correction coefficient, which is determined according to the response elasticity of the service resource, represents the average value of the preset load rate, and the value range is the first interval; The calibrated demand weight has a value range in the first interval, and the calibrated demand that meets the preset threshold is taken as a candidate object into the optimal service resource screening pool.

[0029] In one of the embodiments, is the initial demand weight calculated by the user demand emergency coefficient, the importance coefficient and the like, which is taken as the basic reference value of the present calibration, and the part in the square bracket is the core calibration coefficient, wherein and are the correction weight distribution coefficients for meeting the constraint, which are used to divide the proportion of the initial demand weight in the two dimensions of the load adaptation reservation and the service provider load corresponding weight adjustment, the sub-item corresponding to the first dimension , is the correction coefficient determined according to the response elasticity of the service resource, is the preset average load rate of the service provider in the current period, quantifies the weakening amplitude of the load to the initial weight, so is the reserved proportion after the initial weight is deducted by the weakening, and the sub-item corresponding to the second dimension , is the influence amplitude of the load to the initial weight, and the calibration coefficient formed after the weighted fusion of the two sub-items is multiplied by the initial demand weight to obtain the calibrated demand weight , and and both need to be in the first interval, W that meets the preset threshold will be taken as a candidate object into the optimal service resource screening pool, and The constraints also ensure the rationality of the calibration logic. For example, first, the initial demand weight is calculated. Referring to the parameters of the enterprise office procurement scenario mentioned earlier, the urgency coefficient E is 0.76 after normalization, and the importance coefficient I is 0.89 after normalization. Combining this with the obtained dynamic weights α≈0.47 and β≈0.53, the initial demand weight W0=0.47×0.76+0.53×0.89≈0.83 is obtained by substituting into the formula. The value is within the double-closed interval of 0 to 1. Then, the initial weight is calibrated. The specific correction steps are as follows: based on the historical load rate statistics of office service providers, the average load rate of the current period is preset. The value is set at 0.6, based on the average load level of the enterprise's daily service hours. Considering the responsiveness of office procurement services, load changes have a significant impact on response speed, so a correction factor θ is set at 0.8, along with a correction weighting factor. 0.6 The value is 0.4, which satisfies the condition. The setting is based on prioritizing the weight of the demand itself. Substituting the above parameters into the calibration formula, we calculate W = 0.83 × [0.6 × (1 - 0.8 × 0.6) + 0.4 × 0.8 × 0.6] ≈ 0.83 × 0.50 = 0.42, with values ​​within a double-closed interval of 0 to 1. Demands corresponding to this weight that meet the preset threshold will enter the screening pool. This method clarifies the values ​​and setting basis of each parameter in the calibration process, solves the problem of relying solely on user demand weights without considering the service provider's operational status, and effectively improves the accuracy of matching demand weights with service supply capacity.

[0030] Based on the service provider's operational data, the service provider's supply margin, load factor negatively correlated with the supply margin, and service response time are calculated, and supply constraints are set. Specifically, the service provider's operational data includes the service provider's maximum service capacity, the service provider's occupied capacity, the straight-line distance between the service provider and the user, the service provider's average service response speed data, and the queuing waiting time of the service provider's currently unprocessed demands. Supply margin is the difference between a service provider's maximum service capacity and the capacity already used; Among them, the maximum service capacity is the maximum number of service requests undertaken by the service provider within the preset period, and the occupied capacity refers to the number of service requests that the service provider has undertaken but not yet completed. The quotient between the service provider's occupied capacity and the maximum service capacity is recorded as the first quotient. The load factor ranges from the first interval. The load factor and the supply margin are negatively correlated; when the supply margin increases, the load factor decreases. Calculate the quotient between the straight-line distance between the service provider and the user and the service provider's average service response speed. Record this quotient as the second quotient. The sum of the second quotient and the queuing time of the service provider's currently unprocessed requests is recorded as the service response time.

[0031] The setting content of the supply constraint includes a supply margin constraint, a load rate constraint, and a service response time constraint; The supply margin constraint is that the supply margin is greater than or equal to a preset supply margin threshold, and the supply margin threshold is determined according to the service resource type; The load rate constraint is that the load rate is less than or equal to a preset load rate threshold, and the value range of the load rate threshold is a second interval, and the second interval is a double closed interval from a first value to a second value; The service response time constraint is that the service response time is less than or equal to a preset service response time threshold, and the service response time threshold is determined according to the response requirement of the service resource.

[0032] In one of the embodiments, the service provider running data is determined first, the maximum service capacity in a 1-day period is set to 100, the setting basis is the daily average maximum capacity of this type of service provider, the current occupied capacity is 60, the load rate corresponding to the foregoing is 0.6, the supply margin is calculated as 100-60=40, the load rate is the quotient of the occupied capacity and the maximum service capacity, that is, 60 / 100=0.6, this value is negatively correlated with the supply margin, for example, if the supply margin increases to 50, the load rate decreases to 0.5, then the service response time is calculated, the straight-line distance between the service provider and the user is 5 kilometers, the average service response speed is 2.5 kilometers / hour, the quotient of the two is 2 hours, and the queuing waiting time of the current unprocessed demand is added to obtain a service response time of 3 hours. Then, the supply constraint condition is set, the supply margin threshold is set to 20, the setting basis is to reserve at least 20% of the service redundancy, the load rate threshold is set to 0.8, the second interval is 0 to 0.8, and according to historical experience, when the load rate exceeds 80%, the response efficiency of the service will be significantly reduced, and the service response time threshold is set to 4 hours, which is based on the conventional time requirement of enterprise office procurement. It can be verified that the supply margin 40 is greater than or equal to 20, the load rate 0.6 is less than or equal to 0.8, and the service response time 3 is less than or equal to 4, all of which meet the constraint condition, and the service provider can be included in the candidate resource pool. This way clearly defines the values and setting basis of the supply side parameters and constraints, solves the problem of fuzzy supply constraints, and effectively improves the accuracy of service resource selection and demand matching.

[0033] A preset multi-objective optimization calculation logic is combined with the demand weight, the supply margin, and the service response time to select the optimal service resource under the condition of meeting the supply constraint condition, generate and push a service resource list, and specifically includes that the maximum demand weight after calibration is taken as the main target, the maximum supply margin and the minimum service response time are taken as the secondary targets, and a multi-objective optimization target priority sequence is formed; The service providers meeting the supply constraint condition are filtered first to obtain a candidate service provider set; The target priority sequence based on multi-objective optimization is used to sort the candidate service provider set, specifically including: taking the calibrated demand weight as a first sorting basis, and preliminarily sorting the candidate service provider set according to the calibrated demand weight from high to low; For the service providers with the same calibrated demand weight or a difference less than a preset threshold value in the preliminary sorting, a second sorting is performed according to a coordination index, and the expression of the coordination index F is . Wherein, represents the coordination index, S represents the supply margin, T represents the service response time, represents the supply margin weight coefficient, represents the service response time weight coefficient, and . When the calibrated demand weight increases, the supply margin weight coefficient increases, and the service response time weight coefficient decreases.

[0034] The sorting result of the second sorting is modified by the business correlation degree, and the ranking of the service provider with a business correlation degree lower than a preset threshold value is adjusted to be after the service provider with a business correlation degree not lower than the preset threshold value in the current sorting sequence; According to the modified sorting result, the first N service providers are selected to generate a service resource list, and are pushed to the user terminal in the sorting order.

[0035] In one of the embodiments, the multi-objective optimization sequence is first determined, the main target is to maximize the calibrated demand weight, the secondary target is to maximize the supply margin and minimize the service response time, the service providers satisfying the supply constraint are filtered out to obtain a candidate set including service providers A and B, the parameters of the service provider A are the calibrated demand weight 0.42, the supply margin 40, and the service response time 3 hours; the parameters of the service provider B are the calibrated demand weight 0.41, the supply margin 35, and the service response time 4 hours.

[0036] The preliminary sorting is first performed according to the calibrated demand weight, the preset weight difference threshold value is 0.02, the calibrated demand weight of A is 0.42 which is greater than the calibrated demand weight of B which is 0.41, and the difference 0.01 is less than 0.02, so the second sorting is needed, because the calibrated demand weight of A is larger, the supply margin weight coefficient is set to 0.6, and the service response time weight coefficient is set to 0.4, which satisfies , the F of the collaborative index F, A is F=0.6*40+0.4*(1 / 3)≈24.13, the F of B is F=0.6*35+0.4*(1 / 4)=21.1, after secondary sorting, A is still in the front. The matching degree of the service provider's service capability and the user's business type is measured by the correlation degree, and the business correlation degree itself is obtained based on the matching degree of the quantification of the core function label of the user's business type and the function of the service provider. The evaluation dimensions of the business correlation degree and the matching degree are completely consistent, and the business correlation degree has been pre-quantified to the double closed interval of 0 to 1. For example, the business correlation degree of A is 0.75, and the correlation degree is 0.75. The business correlation degree of B is 0.68, and the preset correlation degree threshold is 0.7, so the position of B is adjusted to after A.

[0037] Finally, the first one is selected, N is 1, and the setting basis is that the user preferentially selects one optimal resource service provider to generate a list and push under the enterprise office procurement scene.

[0038] Embodiment 2, refer to Figure 2 As another embodiment of the application, which is different from the first embodiment, a service resource intelligent push system based on user behavior is provided, comprising a division module, a calculation module and a push module; The division module divides the user behavior data into a first sample and a second sample according to the urgency dimension and the importance dimension of the user behavior data, the first sample is used to calculate the urgency coefficient of the user demand, and the second sample is used to calculate the importance coefficient of the user demand; The calculation module combines the urgency coefficient and the importance coefficient by a preset formula to obtain a demand weight representing the priority of the user demand; Based on the service provider running data, the supply margin of the service provider, the load rate negatively related to the supply margin and the service response time are calculated, and the supply constraint condition is set; The push module sets a multi-objective optimization calculation logic, combines the demand weight, the supply margin and the service response time, filters out the optimal service resource under the condition of meeting the supply constraint condition, generates and pushes a service resource list.

[0039] In one of the embodiments, the dividing module divides the user behavior data according to the urgency dimension and the importance dimension to calculate the urgency coefficient and the importance coefficient of the user demand respectively, the calculating module combines the urgency coefficient and the importance coefficient through a preset formula to obtain a demand weight representing the priority of the user demand, meanwhile, the supply margin, the load rate negatively correlated with the supply margin and the service response time are calculated based on the service provider operation data, and the supply constraint condition is set, the pushing module sets a multi-objective optimization calculation logic, under the condition of meeting the supply constraint condition, the optimal service resource is selected in combination with the demand weight, the supply margin and the service response time, a service resource list is generated and pushed, and the accurate dynamic adaptation of the user demand and the service resource is realized, the response efficiency of the user high-priority demand and the supply and demand load of the service provider are effectively balanced, and the pertinence of the service pushing and the overall operation stability of the service ecology are significantly improved.

[0040] The present application solves the problem of coarse-grained matching of the single index and fixed weight of the existing scheme by splitting the urgency and importance dimensions of the demand and quantifying the coefficients, and matching different demand characteristics by matching dynamic weight, and balances the supply and demand load of the service provider by combining the demand weight with the service provider load rate, and corrects the correlation degree by assisting multi-objective optimization sorting, which realizes the accurate matching of the user high-priority demand and the high-response resource, balances the supply and demand load of the service provider, and significantly improves the service pushing accuracy and the service ecological operation efficiency.

[0041] Those skilled in the art will appreciate that embodiments of the present application can be readily used as a method, a system or a computer program product. Accordingly, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (or computer- readable storage media) having computer-usable program code embodied in the medium. The medium can be any available storage media that can be accessed by a computer. By way of example, and not limitation, such computer-usable storage media can include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other storage medium(s) that can be used to carry or store desired computer program code in the form of instructions or data structures and that can be accessed by a computer. Also, the present application can be embodied in a computer program product that can be traded as goods or merchandise, through a computer-based platform or via a network. Figure 1 one or more functions specified in the flow or flows and / or blocks Figure 1 one or more functions specified in the flow or flows and / or blocks

[0042] It should be noted that the above-mentioned embodiments are only used to illustrate but not to limit the technical solutions of the present application. Although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the present application, and they should be covered in the scope of the claims of the present application.

Claims

1. A service resource intelligent pushing method based on user behavior, characterized in that, The method comprises the following steps: Step S1, dividing user behavior data into a first sample and a second sample according to an urgency dimension and an importance dimension of the user behavior data, the first sample being used to calculate an urgency coefficient of user demand, and the second sample being used to calculate an importance coefficient of user demand; Step S2, combining and calculating the urgency coefficient and the importance coefficient through a preset formula to obtain a demand weight representing a priority of user demand; Based on service provider operation data, the service provider's supply margin, a load rate negatively correlated with the supply margin, and a service response time are calculated, and a supply constraint condition is set; Step S3, a multi-objective optimization calculation logic is preset, and the demand weight, the supply margin, and the service response time are combined to filter out optimal service resources under the condition of meeting the supply constraint condition, and a service resource list is generated and pushed.

2. The user behavior based service resource intelligent push method of claim 1, wherein, According to the urgency dimension and the importance dimension of the user behavior data, the user behavior data is divided into a first sample and a second sample, the first sample is used to calculate an urgency coefficient of user demand, and the second sample is used to calculate an importance coefficient of user demand, and the urgency dimension index includes a user demand trigger frequency and a time sensitivity; The user demand trigger frequency is a trigger frequency of a scene in which a user is located and a scene associated with a similar service within a preset period, and the scene association refers to a scene matching relationship that is strongly related to a similar service function and is recognized through objective data collected by a user terminal; Wherein, the similar service refers to a service in an industry knowledge graph that belongs to big data extraction, and the objective data is quantitative data used to represent the environment and behavior state of the user; The time sensitivity is determined based on a length of an instant response behavior of the user to a similar service push, and the length of the instant response behavior refers to a length of time from when a similar service push is sent to when the user clicks or uses the similar service, and the shorter the length of time, the higher the time sensitivity; The importance dimension index includes a business association degree and a user attention degree; The business association degree is determined based on a functional matching degree of a business type to which the user belongs and a service resource, and the higher the matching degree, the higher the business association degree; The user attention degree is determined based on a browsing frequency and a collection frequency of the user to a similar service resource; The user behavior data used to calculate the urgency dimension index is taken as the first sample; The user behavior data used to calculate the importance dimension index is taken as the second sample; The urgency coefficient of user demand is calculated based on each urgency dimension index in the first sample; The importance coefficient of user demand is calculated based on each importance dimension index in the second sample; The expression for calculating the emergency coefficient is ; wherein, denotes the urgency coefficient, denotes the time sensitivity of the urgency dimension in the first sample, denotes the demand trigger frequency of the urgency dimension in the first sample, and denotes a pre-set empirical parameter, which is a constant; The calculated urgency coefficient is normalized to a first interval, and the first interval is a double closed interval of 0 to 1; The expression for calculating the importance coefficient is ; wherein, represents an importance coefficient, represents a service relevance of the importance dimension in the second sample, represents a user attention of the importance dimension in the second sample, and represents a pre-set empirical parameter, which is a constant; The calculated importance coefficient is normalized to the first interval.

3. The user behavior based service resource intelligent push method of claim 2, wherein, The urgency coefficient and the importance coefficient are combined and calculated through a preset formula to obtain a demand weight representing a priority of user demand, and specifically, a dynamic weight of the urgency coefficient and a dynamic weight of the importance coefficient are generated by using a weighted normalization method on the urgency dimension index and the importance dimension index; The expression of the dynamic weight of the emergency coefficient is ; The expression of the dynamic weight of the importance coefficient is ; in, The dynamic weight representing the urgency factor. The dynamic weights representing the importance coefficients These represent the preset initial weights for the urgency and importance dimensions, respectively. These initial weights are constants and are determined statistically based on the service resource type. Indicates the first Standardized values ​​of each urgency dimension indicator. The range of values ​​for is the first interval. This indicates the contribution weight of the urgency dimension indicators; a standardized value of the lth importance dimension indicator, taking a value in a first interval, a contribution weight of the importance dimension indicator; m represents the number of the urgency dimension indexes, and n represents the number of the importance dimension indexes.

4. The user behavior based service resource intelligent push method of claim 3, wherein, The emergency coefficient, the importance coefficient, the dynamic weight of the emergency coefficient and the dynamic weight of the importance coefficient are substituted into a weighted formula to obtain an initial demand weight, and an expression of the initial demand weight is, ; wherein denotes an initial weight, denotes an urgency coefficient, denotes an importance coefficient, the has a value in a first interval.

5. The user behavior based intelligent push of service resources method of claim 4, wherein, Based on the service provider historical same period load rate, the preset current period load rate average value is used to correct the initial demand weight to obtain the calibrated demand weight, and the correction expression is, ; wherein, denotes the calibrated demand weight, both denote the modified weight distribution coefficient, and , denotes the modified coefficient, determined according to the response elasticity of the service resource, denotes the mean value of the preset load rate, and the value range is the first interval; The calibrated demand weight is in a first interval, and the calibrated demand meeting a preset threshold is taken as a candidate object to enter an optimal service resource screening pool.

6. The user behavior based intelligent push of service resources method of claim 5, wherein, The service provider running data is used to calculate a supply margin of the service provider, a load rate negatively correlated with the supply margin, and a service response time, and a supply constraint condition is set, specifically including that the service provider running data includes a maximum service capacity of the service provider, an occupied capacity of the service provider, a straight-line distance between the service provider and the user, average service response speed data of the service provider, and a queuing waiting time length of a current unprocessed demand of the service provider; The supply margin is a difference between the maximum service capacity and the occupied capacity of the service provider; The maximum service capacity is a maximum number of service demands accepted by the service provider in a preset period, and the occupied capacity refers to a number of service demands currently accepted by the service provider and not completed. A quotient of the occupied capacity and the maximum service capacity of the service provider is taken as a first quotient value; The load rate is in a first interval, and the load rate and the supply margin are in a negative correlation, and when the supply margin increases, the load rate decreases. A quotient of the straight-line distance between the service provider and the user and the average service response speed of the service provider is taken as a second quotient value, and a sum of the second quotient value and the queuing waiting time length of the current unprocessed demand of the service provider is taken as the service response time.

7. The user behavior based intelligent push of service resources method of claim 6, wherein, The supply constraint condition includes a supply margin constraint, a load rate constraint, and a service response time constraint; The supply margin constraint is that the supply margin is greater than or equal to a preset supply margin threshold, and the supply margin threshold is determined according to the type of the service resource; The load rate constraint is that the load rate is less than or equal to a preset load rate threshold, and the load rate threshold is in a second interval, and the second interval is a double closed interval from a first value to a second value; The service response time constraint is that the service response time is less than or equal to a preset service response time threshold, and the service response time threshold is determined according to the response requirement of the service resource.

8. The user behavior based service resource intelligent push method of claim 7, wherein, A preset multi-objective optimization calculation logic is used to screen out optimal service resources under the condition of meeting the supply constraint condition, generate and push a service resource list in combination with the demand weight, the supply margin, and the service response time, specifically including taking the maximization of the calibrated demand weight as a main target, and taking the maximization of the supply margin and the minimization of the service response time as secondary targets to form a target priority sequence of multi-objective optimization; First, service providers meeting the supply constraint condition are filtered to obtain a candidate service provider set; The candidate service provider set is sorted based on the target priority sequence of multi-objective optimization, specifically including taking the calibrated demand weight as a first sorting basis, and preliminarily sorting the candidate service provider set according to the calibrated demand weight from high to low; The service providers with the same demand weight or the difference less than the preset threshold after calibration in the preliminary sorting are subjected to secondary sorting with the coordination index as the second sorting basis, and the expression of the coordination index F is, ; wherein, represents a synergy index, S represents the supply margin, T represents the service response time, represents a supply margin weight coefficient, represents a service response time weight coefficient, and ; when an increase in the post-calibration demand weight is detected, the is increased, the is decreased.

9. The user behavior-based service resource intelligent pushing method of claim 8, characterized in that The sorting result of the secondary sorting is modified through the business association degree, and the sorting position of the service provider with a business association degree lower than a preset threshold is adjusted to be after the service provider with a business association degree not lower than the preset threshold in the current sorting sequence. Based on the revised sorting results, the top N service providers are selected to generate a service resource list, which is then pushed to the user's terminal in sorted order.

10. A service resource intelligent pushing system based on user behavior, applied to the service resource intelligent pushing method based on user behavior in any one of claims 1-9, characterized in that, It includes a partitioning module, a calculation module, and a push module; The segmentation module divides user behavior data into a first sample and a second sample based on the urgency and importance dimensions of the user behavior data. The first sample is used to calculate the urgency coefficient of the user's needs, and the second sample is used to calculate the importance coefficient of the user's needs. The calculation module combines the urgency coefficient and importance coefficient using a preset formula to calculate the demand weight that represents the priority of user needs. Based on the service provider's operational data, calculate the service provider's supply margin, load factor negatively correlated with the supply margin, and service response time, and set supply constraints. The push module has a preset multi-objective optimization calculation logic. Combining the demand weight, supply margin and service response time, it selects the optimal service resources under the condition of satisfying the supply constraints, and generates and pushes a list of service resources.

Images