Intelligent printing resource allocation recommendation method integrating group splicing mode and user preference

Abstract

The invention relates to the technical field of printing resource intelligent management, in particular to a printing resource intelligent allocation recommendation method integrating a group splicing mode and user preference, which comprises the following steps of: comprehensively acquiring data, providing a basis for subsequently analyzing user requirements and resource states, and ensuring that allocation recommendation is reasonable and effective; generating a user clustering center through soft clustering, and mapping the user clustering center into a clustering preference vector to construct a dynamic field, thereby providing a basis for resource matching; by generating a resource availability potential field, resources can be reasonably allocated; by calculating a gradient vector and analyzing an interference effect, a group splicing opportunity focus and a resource conflict saddle point are identified, and the distribution efficiency is improved; by generating the grouping decision vector, an optimal grouping scheme is provided for the user, and the grouping success rate and the user satisfaction are improved; the resource pre-allocation strategy is adaptively triggered, the path is corrected, meanwhile, the path is issued to the execution node, and the guide information is pushed, so that resource conflicts are effectively solved, rapid grouping is promoted, and the user experience is enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent management of printing resources, and in particular to a printing resource intelligent allocation recommendation method fusing a group mode and user preferences. BACKGROUND

[0002] To cope with diversified user needs, modern printing service systems generally introduce an online group mode to reduce unit cost and improve resource utilization; existing technical solutions usually adopt a rule matching strategy: the system calculates the similarity according to the user's preset price, delivery time and other static labels, aggregates orders that meet the conditions into group tasks, and allocates them to available printing resources based on the first-come-first-served principle. This mode can maintain basic operation when orders are evenly distributed and resources are sufficient.

[0003] However, when faced with large-scale, multi-channel group requests, the existing method lacks the coordinated perception of user dynamic preferences and resource global state, and its static matching mechanism will cause two aspects of chain problems: on the one hand, one-sided pursuit of order aggregation will instantaneously exhaust specific types of printing resources (such as high-speed color printing devices), triggering resource preemption and order queuing congestion; on the other hand, ignoring the implicit association of users' group willingness leads to potential high-fitting group combinations being lost due to the failure to timely recommend. This mismatch between resource allocation and user matching eventually leads the system to an inefficient equilibrium of partial resource overload and partial order flow, which seriously restricts the response speed and total transaction volume of the printing service platform during peak periods. Therefore, the present application proposes a printing resource intelligent allocation recommendation method fusing a group mode and user preferences. SUMMARY

[0004] The purpose of the present application is to solve the problems in the background art, and a printing resource intelligent allocation recommendation method fusing a group mode and user preferences is proposed.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions: The printing resource intelligent allocation recommendation method fusing a group mode and user preferences comprises: S1, real-time collection of dynamic behavior data flow of a group of users and global state data of a printing resource pool; S2, soft clustering of users in a user feature space based on the dynamic behavior data flow of the group of users through a Gaussian mixture model, to generate user clustering centers; S3, mapping of each user clustering center to a group preference vector, and construction of a group willingness dynamic field based on the group preference vector; S4, calculation and analysis of the global state data of the printing resource pool, and superimposed generation of a resource availability potential field; S5, respectively calculate the gradient vectors of the teaming willingness dynamic field and the resource availability potential field, and perform interference effect analysis to identify teaming opportunity focal points and resource conflict saddle points in the interference region; S6, based on the teaming opportunity focal points, generate a teaming decision vector for the user initiating the teaming through a space-time coordination projection algorithm; S7, according to the conflict strength of the resource conflict saddle point, adaptively trigger a resource pre-deployment strategy, and apply path correction to the teaming decision vector passing through the resource conflict saddle point; S8, the corrected teaming decision vector is sent to the resource execution node, and the teaming group guidance information is pushed to the related users synchronously.

[0006] Further, the dynamic behavior data stream includes the real-time location trajectory of the user, the interface interaction event sequence, and the exploratory operation of the teaming option; the global state data of the print resource pool includes the attribute and state information of each resource unit; wherein the resource unit is the basic unit for independent scheduling and management in the print resource pool, and the attribute and state information includes: the real-time working state of each printing device, the task queue length, the future reservation schedule and the consumable inventory information.

[0007] Further, the user is soft clustered in the user feature space by a Gaussian mixture model, and the process of generating the user clustering center includes: Extract the time sequence behavior features of the user from the dynamic behavior data stream, construct a user-behavior feature matrix, and initialize the number of clustering components of the Gaussian mixture model based on the information theory criterion; Iteratively optimize the parameters of the Gaussian mixture model based on the expectation maximization algorithm, and the iteration process includes: A1, expectation step: based on the current Gaussian mixture model parameters, calculate the posterior probability of each data point in the user-behavior feature matrix belonging to each Gaussian distribution component as the clustering membership of the user; A2, maximum step: using the calculated clustering membership as the weight, recalculate the mean vector and covariance matrix of each Gaussian distribution component; The mean vector of each Gaussian distribution component obtained after iteration convergence is output as the dynamically evolving user clustering center.

[0008] Further, the process of mapping each user clustering center to a teaming preference vector and constructing a teaming willingness dynamic field based on the teaming preference vector includes: Obtain the output user clustering center, wherein each user clustering center is a multi-dimensional vector; Perform semantic mapping and scale scaling on each dimension of each user clustering center vector to convert it into a teaming preference vector, wherein the dimensions of the teaming preference vector include price sensitivity, delivery time requirement, and print quality level preference; In the virtual resource space, a time-varying basic strength is given to each group preference vector, and is marked as the field source strength; According to the field source strength, and in combination with the field source attenuation model, the group willingness field strength at any point in the virtual resource space is calculated, wherein the group willingness field strength is formed by the superposition of the field strength vectors of all field sources at the point; Based on the calculated full field strength distribution, a group willingness dynamic field is constructed.

[0009] Further, the process of calculating and analyzing the global state data of the print resource pool and superimposing the resource availability potential field includes: The global state data of the print resource pool is obtained, the real-time working state, task queue length and future reservation schedule of each resource unit are analyzed, and the load pressure index of each resource unit in the future target time window is predicted; The consumable inventory information is queried, and in combination with the historical failure rate of the resource unit, the service reliability coefficient is calculated; The load pressure index and the reciprocal of the service reliability coefficient are weighted and fused to generate the load potential barrier value of each resource unit, wherein the load potential barrier value is positively correlated with the load pressure of the resource unit; In the virtual resource space, each resource unit is mapped as a potential field source, wherein the potential field height of the resource unit is equal to the load potential barrier value; Through the potential field superposition algorithm, a continuous resource availability potential field covering the entire virtual resource space is generated.

[0010] Further, the gradient vectors of the group willingness dynamic field and the resource availability potential field are calculated respectively, and the interference effect analysis is performed to identify the group opportunity focus and the resource conflict saddle point in the interference region, including: The group willingness dynamic field is obtained, and the gradient vector of the group willingness dynamic field at each point in the virtual resource space is calculated; The resource availability potential field is obtained, and the gradient vector of the resource availability potential field is calculated; Based on the gradient vectors of the group willingness dynamic field and the resource availability potential field, interference analysis is performed to identify the group opportunity focus and the resource conflict saddle point.

[0011] Further, the process of interference analysis based on the gradient vectors of the group willingness dynamic field and the resource availability potential field includes: C1、The dot product operation of the gradient vectors of the group willingness dynamic field and the resource availability potential field is performed, the region with the maximum product of the gradient direction opposite and the module length is identified, and is defined as the constructive interference region, wherein the center point of the constructive interference region is the group opportunity focus; C2, the region where the equipotential lines of the teaming willingness dynamic field and the resource availability potential field intersect and the field strengths weaken each other is defined as a conflict interference region, wherein a center point of the conflict interference region is a resource conflict saddle point.

[0012] Further, the process of generating a teaming decision vector for a user initiating teaming by a space-time coordination projection algorithm includes: mapping the teaming preference vector of the user initiating teaming to a virtual resource space to determine a position of the teaming preference vector in the space as a projection starting point; taking the teaming opportunity focus as a projection ending point, and calculating an initial projection path from the starting point to the ending point; integrating the field strength of the teaming willingness dynamic field on the initial projection path as a path benefit estimation, and simultaneously integrating the potential field height variation of the resource availability potential field as a path cost estimation; maximizing the ratio of the path benefit to the path cost as an objective, and optimizing the projection path by using a variational method; combining users corresponding to user cluster centers on the new path obtained after optimization into an optimal teaming user set, and encapsulating a resource binding scheme corresponding to the path ending point into a teaming decision vector.

[0013] Further, the process of adaptively triggering a resource pre-dispatching strategy and imposing a path correction on a teaming decision vector passing through a resource conflict saddle point includes: monitoring the conflict intensity at the resource conflict saddle point, wherein the conflict intensity is quantified by a standardized difference between the teaming willingness field strength and the resource availability potential field height; presetting a system stability threshold; when the conflict intensity of any resource conflict saddle point exceeds the system stability threshold, a resource pre-dispatching process is automatically triggered; based on the collected global state data, low-load standby resource nodes are selected and dispatched from a resource pool; for all teaming decision vectors that can pass through the resource conflict saddle point, a virtual repulsive force field proportional to the conflict intensity is imposed on the projection path at the resource conflict saddle point; under the action of the repulsive force field, the projection path of the teaming decision vector is recalculated, and then a corrected teaming decision vector is output.

[0014] Further, the process of pushing teaming completion guide information to related users includes: obtaining the corrected teaming decision vector, and analyzing the optimal teaming user set encapsulated in the corrected teaming decision vector; generating individualized guide information for each user in the optimal teaming user set; In the guide information, a one-key joining shortcut operation interface carrying a unique group forming session identifier is carried; finally, complete guide information is pushed to the terminal device of the corresponding user through a system message channel.

[0015] Compared with the prior art, the beneficial effects of the printing resource intelligent allocation recommendation method fusing the group forming mode and user preferences are that: 1) By collecting group user group dynamic behavior data streams and printing resource pool global state data in real time, user behavior characteristics and resource real-time conditions are comprehensively and timely mastered, laying a foundation for subsequent analysis of user demand and resource availability, ensuring that resource allocation recommendation can be adapted to actual scenarios and improving rationality; by using a Gaussian mixture model soft clustering to generate user clustering centers and map them to group preference vectors to construct a dynamic field, the group willingness characteristics and distribution of different user groups can be described, and the preferences of users in various aspects of group forming can be clearly presented, providing a key basis for subsequent matching resources; 2) By calculating and analyzing resource availability potential fields from printing resource pool global state data, the availability of each resource unit at different times and spaces can be intuitively reflected, so that available resources can be quickly identified and intuitive reference for reasonable allocation of resources can be provided; by calculating gradient vectors and analyzing interference effects, group opportunity focal points and resource conflict saddle points are identified, and key positions and potential problem areas of resource allocation are clarified, so that resources can be concentrated to solve key problems and improve resource allocation efficiency; 3) By generating group decision vectors based on group opportunity focal points, optimal group forming schemes are provided for users, user willingness and resource conditions are comprehensively considered, and group success rate and user satisfaction are improved; by adaptively triggering resource pre-dispatching strategies and correcting paths, resources are dynamically adjusted according to conflict strength, resource conflict problems are effectively solved, and group forming is smoothly carried out; by issuing the corrected group decision vectors to execution nodes and pushing guide information, group forming is quickly formed, resource utilization is improved, personalized guide information enhances user experience, and user recognition of the system is improved. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The flowchart of the printing resource intelligent allocation recommendation method fusing the group forming mode and user preferences proposed by the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0018] REFERENCE Figure 1, a printing resource intelligent allocation recommendation method fusing the group-buying mode and user preference, comprising: S1, collecting dynamic behavior data flow of the group-buying user group and global state data of the printing resource pool in real time; wherein the dynamic behavior data flow includes real-time location trajectory of the user, interface interaction event sequence and exploratory operation of the group-buying option; the global state data of the printing resource pool includes attribute and state information of each resource unit; the resource unit is a basic unit for independent scheduling and management in the printing resource pool, and the attribute and state information includes real-time working state, task queue length, future reservation schedule and consumable inventory information of each printing device; S2, based on the dynamic behavior data flow of the group-buying user group, soft clustering the users in the user feature space through the Gaussian mixture model to generate dynamically evolving user clustering centers; S3, mapping each user clustering center to a group-buying preference vector, and constructing a group-buying willingness dynamic field based on the group-buying preference vector in the virtual resource space, wherein the group-buying willingness dynamic field represents the group-buying willingness intensity of the user in a specific resource area; S4, performing calculation and analysis on the global state data of the printing resource pool, and superimposing and generating a resource availability potential field in the virtual resource space, wherein the potential field height of the resource availability potential field is positively correlated with the load pressure of the resource unit; S5, calculating the gradient vectors of the group-buying willingness dynamic field and the resource availability potential field respectively, and performing interference effect analysis to identify the group-buying opportunity focus and the resource conflict saddle point in the interference area; S6, based on the group-buying opportunity focus, generating a group-buying decision vector including an optimal group user set and a resource binding scheme for the user initiating the group-buying through a space-time coordination projection algorithm; S7, adaptively triggering a resource pre-dispatching strategy according to the conflict strength of the resource conflict saddle point, and applying path correction to the group-buying decision vector passing through the resource conflict saddle point; S8, issuing the corrected group-buying decision vector to a resource execution node, and synchronously pushing group-buying team formation guide information to related users; wherein the resource execution node is an entity unit in the system specifically responsible for receiving instructions and executing physical printing tasks or resource allocation operations, which is a printing device controller or a regional resource scheduling server, and its function is to analyze the issued group-buying decision vector and drive the corresponding physical printing device to complete the work, or actually lock and schedule the allocated resources.

[0019] It needs to be further explained that in the specific implementation process, the process of soft clustering the users in the user feature space through the Gaussian mixture model to generate the user clustering center includes: extracting the time sequence behavior features of the user from the dynamic behavior data flow, constructing a user-behavior feature matrix, and initializing the number of clustering components of the Gaussian mixture model based on the information theory criterion; Specifically, transforming the collected raw dynamic behavior data stream into quantifiable time-series behavior features includes: converting real-time location trajectory into average moving speed of users around the print service site and residence time length in specific areas; converting interface interaction event sequence into click number distribution on different print services (such as ordinary copying, color printing, and binding services) and total click frequency per unit time; converting exploratory operation of group purchase options into switching number between price options and delivery time options, and hesitation length before final confirmation option as preference hesitation degree feature; All extracted time-series behavior features are standardized to be in similar numerical intervals; after the processing, all feature values of each user are arranged in sequence to form a user feature vector; finally, all user feature vectors are summarized to form a two-dimensional data table with the number of rows being the total number of users and the number of columns being the total number of features, i.e., a user-behavior feature matrix; The process of initializing the number of clustering components of the Gaussian mixture model based on the information theory criterion is: setting a possible clustering number range, for each candidate number in the range, training a corresponding Gaussian mixture model using the expectation maximization algorithm; wherein the clustering number range refers to the minimum value to the maximum value of the preset possible clustering number when performing soft clustering, for example, from 2 classes to 10 classes, the optimal clustering number is determined by model training and evaluation; for each trained model, calculate its Bayesian information criterion score, which integrates the goodness of fit of the model to the data and the complexity of the model itself (determined by the number of parameters and the number of samples); automatically select the clustering number corresponding to the lowest Bayesian information criterion score as the optimal component number of the Gaussian mixture model, to ensure that the discovered user categories can effectively explain the data and will not be over-fitted due to excessive complexity; Iteratively optimize the parameters of the Gaussian mixture model based on the expectation maximization algorithm, the iteration process includes: A1, expectation step: based on the current Gaussian mixture model parameters, calculate the posterior probability of each data point in the user-behavior feature matrix belonging to each Gaussian distribution component as the clustering membership of the user; A2, maximization step: use the calculated clustering membership as weights to recalculate the mean vector and covariance matrix of each Gaussian distribution component to update the Gaussian mixture model parameters; Specifically, for each user feature vector in the user-behavior feature matrix, calculate the likelihood of the user feature vector generated by the corresponding Gaussian component according to the statistical parameters (mean vector and covariance matrix) of each Gaussian component in the current Gaussian mixture model and its mixing weight; The calculated likelihoods belonging to each Gaussian component are normalized so that the sum of the likelihoods of the user belonging to all Gaussian components is 1. The set of probability values obtained after normalization for each Gaussian component is the clustering membership, which is used to quantify the similarity of the individual user behavior characteristics to each user group; The clustering membership of all users calculated is used as a weight to perform weighted statistics on the data in the user-behavior characteristic matrix. The mean vector is updated by calculating the weighted average of all user feature vectors under each Gaussian component. The covariance matrix is updated by calculating the weighted outer product of the deviation between all user feature vectors and the new mean vector of the component under each Gaussian component. At the same time, the mixing weight is updated according to the sum of the membership of each component. The purpose is to find new model parameters so that the overall probability of observing the existing data under the current model is maximized. The mean vector of each Gaussian distribution component obtained after iteration convergence is output as the dynamically evolving user clustering center.

[0020] It needs to be further explained that, in the specific implementation process, the process of mapping each user clustering center to a group purchase preference vector and constructing a group purchase willingness dynamic field based on the group purchase preference vector includes: The output user clustering center is obtained, wherein each user clustering center is a multi-dimensional vector, and the dimension is the same as the dimension of the user-behavior characteristic matrix. Each dimension of each user clustering center vector is semantically mapped and scaled to convert it into a group purchase preference vector, wherein the dimensions of the group purchase preference vector include price sensitivity, delivery time requirement, and print quality level preference. Specifically, semantic mapping includes: setting a feature-preference mapping table to map each original dimension (such as click frequency, residence time) of the clustering center vector to a semantic dimension of group purchase preference, for example, mapping a dimension with high click frequency of price options to high price sensitivity; mapping a dimension with long residence time of express service to high delivery time requirement. Scale includes: to make the scores of different semantic dimensions comparable, the mapped values are linearly scaled to a unified evaluation interval (for example, 0 to 10 points), thereby generating a standardized group purchase preference vector, wherein the group purchase preference vector directly describes the preference intensity of the user group in terms of price, time, quality, etc. In the virtual resource space, a time-varying basic intensity is given to each group preference vector, and is marked as field source intensity; wherein the virtual resource space is an abstract mathematical space, the coordinate axes of which represent the core attributes of the printing resources, including but not limited to unit printing cost, promised delivery time length and printing quality level (for example, X axis: unit cost; Y axis: delivery time limit; Z axis: printing quality); any coordinate point in the virtual resource space uniquely corresponds to a printing service scheme with a specific attribute combination; According to the field source intensity and in combination with a field source attenuation model, the group willingness field intensity of any point in the virtual resource space is calculated, wherein the group willingness field intensity is formed by the vector superposition of the field intensities of all field sources at the point; it can be understood that the field intensity attenuation model adopts a negative exponential function of distance, and the distance is the Euclidean distance between the point in the virtual resource space and the field source in the preference vector space, wherein the preference vector space is a mathematical space spanned by each dimension of the group preference vector; The group willingness field intensity of any point in the virtual resource space is calculated, specifically including: calculating the preference difference degree (i.e. Euclidean distance) between point P and each user cluster (the position of which is determined by its group preference vector) in the virtual resource space; according to the preference difference degree, in combination with the size (i.e. the sum of cluster membership) and influence of the user cluster, the field intensity contribution of the cluster to point P is calculated through an attenuation function (such as the greater the difference degree, the weaker the contribution); finally, the total field intensity of point P is the vector sum of the field intensity contributions of all user clusters to the point; wherein the field intensity directly reflects the attractiveness of the printing service scheme represented by the point to the overall user group willingness; Based on the calculated full field intensity distribution, a group willingness dynamic field is constructed.

[0021] It needs to be further explained that in the specific implementation process, the process of calculating and analyzing the global state data of the printing resource pool and superimposing the resource availability potential field includes: The global state data of the printing resource pool is obtained, the real-time working state, task queue length and future reservation schedule of each resource unit are analyzed, and the load pressure index of each resource unit in the future target time window is predicted; specifically, according to the current task queue length, processing speed and future reservation schedule of the printing device, the total workload to be processed by the printing device in the future target time window (such as the next hour) is predicted; the total workload is compared with the maximum theoretical processing capacity of the printing device in the time window, and the ratio after normalization is the load pressure index, wherein the higher the load pressure index, the busier the printing device is; query consumable inventory information, and combine the historical failure rate of the resource unit to calculate the service reliability coefficient; specifically, comprehensively evaluate whether the current consumable inventory (such as paper, toner) of the printing device is sufficient, and combine the historical failure rate data of the printing device to calculate the service reliability coefficient through a weighted scoring model (for example, service reliability coefficient = consumable sufficiency score × weight 1 + (1-historical failure rate) × weight 2), wherein the service reliability coefficient evaluates the service reliability of the printing device, and the higher the value of the service reliability coefficient, the higher the reliability of the printing device to complete the task stably. The load pressure index is weighted and fused with the reciprocal of the service reliability coefficient to generate a load barrier value of each resource unit, wherein the load barrier value is positively correlated with the load pressure of the resource unit. In the virtual resource space, each resource unit is mapped as a potential field source, wherein the potential field height of the resource unit is equal to the load barrier value; wherein the purpose of mapping as a potential field source is to convert discrete physical printing devices into potential field sources with continuous influence in the virtual resource space, so that unified resource availability evaluation can be performed on any printing service scheme (point) in the virtual resource space, laying a foundation for subsequent global analysis of resource conflicts. A continuous resource availability potential field covering the entire virtual resource space is generated through a potential field superposition algorithm; specifically, in the virtual resource space, each printing device is positioned at a specific coordinate according to the service level (such as cost, timeliness, and quality) it can provide; the load barrier value (calculated from the load pressure index and the reliability coefficient) determines the height of the potential field source; the potential field influence of a resource will smoothly decay as the spatial distance increases; the entire resource availability potential field is formed by the potential field influence of all potential field sources corresponding to the printing devices in the virtual resource space; the higher the potential field value, the greater the resource constraints, the higher the cost, or the higher the risk of executing the corresponding printing service scheme at that location.

[0022] It should be further explained that in the specific implementation process, the gradient vectors of the group-buying willingness dynamic field and the resource availability potential field are calculated respectively, and interference effect analysis is performed to identify the group-buying opportunity focus and the resource conflict saddle point in the interference area. The group-purchasing willingness dynamic field is acquired, and a gradient vector of the group-purchasing willingness dynamic field at each point in the virtual resource space is calculated, and a high-gradient region is identified as a boundary of rapid change of the group-purchasing willingness of the user; in the group-purchasing willingness dynamic field, the high-gradient region represents that the group-purchasing willingness of the user has a sharp and rapid change in the virtual resource space; specifically, one side of the boundary represents that the user group has a strong group-purchasing willingness for a certain type of resource (such as high-speed color printing), and the other side of the boundary represents that the user willingness rapidly decays or turns to other resource types (such as turning to economic black-and-white printing); the sudden change of the willingness is caused by a critical point of a preference dimension such as price or delivery timeliness; therefore, identifying the high-gradient region is essentially locating a watershed of different user preference clusters in resource selection, and is a key to finding a potential group-purchasing combination and a competitive region; The resource availability potential field is acquired, and a gradient vector of the resource availability potential field is calculated, and a high-gradient region is identified as a boundary of sharp change of resource availability; in the resource availability potential field, the high-gradient region represents that a load state or a pre-bookability of a resource unit has an abrupt change in the virtual resource space; specifically, one side of the boundary can correspond to a resource unit set that is lightly loaded and can be immediately booked, and the other side can correspond to a resource unit set that is nearly saturated, has a sharply increased queuing time, or is about to be unavailable; the abrupt change of the potential field is usually caused by factors such as a device performance bottleneck, a dense booking period, or a consumable supply state; therefore, identifying the high-gradient region is equivalent to locating a stress concentration line of a supply-demand relationship of a system resource, and is a direct basis for predicting resource congestion and triggering a pre-dispatching strategy; The gradient vectors of the group-purchasing willingness dynamic field and the resource availability potential field are subjected to interference analysis, including: C1, point multiplication is performed on the gradient vectors of the group-purchasing willingness dynamic field and the resource availability potential field, a region with the largest product of the gradient direction and the module length and with opposite gradient directions is identified as a constructive interference region, and a center point of the constructive interference region is defined as a group-purchasing opportunity focus; C2, a region where equipotential lines of the group-purchasing willingness dynamic field and the resource availability potential field intersect and field strengths mutually weaken is identified as a conflictive interference region, and a center point of the conflictive interference region is defined as a resource conflict saddle point; Specifically, the virtual resource space is subjected to discretization processing, and is divided into a three-dimensional grid structure; each grid node stores a field strength value of the group-purchasing willingness dynamic field and a potential height value of the resource availability potential field, which are calculated in advance; For the calculation of the gradient vector of the dynamic field of the group formation willingness, each grid node in the virtual resource space is traversed, and a central difference method is used for numerical calculation, including: for any current node in the virtual resource space, the gradient component of the dynamic field of the group formation willingness of the current node in the printing cost dimension is obtained by calculating the difference between the field strength numerical value of the adjacent node of the current node in the positive direction of the printing cost axis and the field strength numerical value of the adjacent node in the negative direction, and dividing by twice the grid step, and the gradient components of the current node in the delivery time dimension and the printing quality dimension are calculated in the same way; finally, the gradient vector of the dynamic field of the group formation willingness at the current node is composed of the gradient components of the printing cost, the delivery time and the printing quality; For the calculation of the gradient vector of the resource availability potential field, the same numerical central difference method as the calculation of the gradient vector of the dynamic field of the group formation willingness is used, and the same calculation process is performed on all grid nodes in the virtual resource space, wherein the numerical basis used when calculating each component of the gradient vector of the resource availability potential field is the potential field height numerical value of the resource availability potential field; thus, the complete gradient vector of the resource availability potential field at each grid node in the virtual resource space is obtained; After the gradient field calculation is completed, the effect analysis is performed: by calculating the vector dot product numerical value of the gradient vector of the dynamic field of the group formation willingness and the gradient vector of the resource availability potential field at each grid node in the virtual resource space, the constructive interference region is identified; among all the vector dot product calculation results of the entire virtual resource space, the continuous spatial region with the largest absolute value of the negative vector dot product numerical value is screened out, which indicates the spatial range with the highest degree of opposite direction of the gradient vectors of the dynamic field of the group formation willingness and the resource availability potential field, and thus the strongest synergistic driving force, and the geometric center of the identified constructive interference region is marked as the group formation opportunity focus; For the identification of the conflictive interference region, the local high field strength equipotential surface of the dynamic field of the group formation willingness intersects with the local high potential field equipotential surface of the resource availability potential field in the virtual resource space, and the field strength of the dynamic field of the group formation willingness and the potential field height of the resource availability potential field present mutual inhibition and suppression in this spatial range; at the same time, the geometric center of the conflictive interference region is marked as the resource conflict saddle point; wherein the conflictive interference region represents the spatial position directly overlapped and conflicted by the high user group formation willingness and the high resource load pressure or low resource service reliability.

[0023] It should be further explained that, in the specific implementation process, the process of generating the group formation decision vector for the user initiating the group formation includes: Mapping the group formation preference vector of the user initiating the group formation to the virtual resource space to determine the position of the group formation preference vector in the space as the projection starting point; Taking the group-buying opportunity focus as the projection end point, an initial projection path from the start point to the end point is calculated; specifically, a target-oriented gradient ascent algorithm is adopted, which starts from the mapping point of the group-buying initiator in the virtual resource space, at each path extension step, the group-buying willingness field strength (benefit) of all adjacent grid points and the directional proximity (directivity) between the adjacent points and the target group-buying opportunity focus are comprehensively evaluated through a weighted sum function, and the adjacent point with the highest weighted sum is selected as the next path point; this process is iterated to ensure that the path is effectively directed to the final target focus while tending to the high group-buying willingness area, thereby forming a connected initial path from the start point to the end point; On the initial projection path, the field strength of the group-buying willingness dynamic field is integrated to calculate the path benefit estimate; at the same time, the potential field height variation of the resource availability potential field is integrated to calculate the path cost estimate; specifically, the planned path is segmented and accumulated for evaluation; the path is divided into a plurality of small line segments, for each small line segment, the average of the field strengths (or potential fields) of its start point and end point is taken, multiplied by the length of the line segment to obtain a differential quantity; the path benefit estimate is the cumulative sum of the differentials of the group-buying willingness field strengths on all line segments; the path cost estimate is the cumulative sum of the differentials of the potential field heights of the resource availability potential field on all line segments; the two integral values respectively represent the total willingness benefit and the total resource cost that can be obtained by selecting the path; To maximize the ratio of path benefit to path cost, the projection path is optimized using the variational method; specifically, the path is represented as a series of control points; by fine-tuning the positions of the control points, the benefit-cost ratio of the new path is repeatedly calculated, where the benefit-cost ratio is obtained as follows: for any path to be evaluated, the path is discretized into a dense sequence of points; by numerical integration method, the integral sum of the group-buying willingness field strength values of all sequence points on the path along the path is calculated as the total benefit estimate value of the path; simultaneously, the integral sum of the potential field height values of the resource availability potential field of all sequence points on the path along the path is calculated as the total cost estimate value of the path; finally, the total benefit estimate value is divided by the total cost estimate value, and the quotient obtained is the benefit-cost ratio of the path; by comparing the changes in the benefit-cost ratio before and after adjustment, the path is guided to evolve in a more optimal direction (e.g., more through high group-buying willingness dynamic fields while avoiding high resource availability potential fields), and after multiple iterations, it finally converges to an optimized path that achieves the best balance between benefit and cost; The users corresponding to the user cluster centers passed through by the new path obtained after optimization are combined into an optimal group-buying user set, and the resource binding scheme corresponding to the path end point is encapsulated into a group-buying decision vector.

[0024] It needs to be further explained that in the specific implementation process, the adaptive triggering of the resource pre-dispatching strategy and the application of path correction to the group-buying decision vector passing through the resource conflict saddle point include: monitoring the conflict intensity at the resource conflict saddle point, wherein the conflict intensity is quantified by the normalized difference between the field strength of the group formation willingness field and the potential height of the resource availability potential field, and the formula for calculating the normalized difference is ; in the formula, is the conflict intensity; is the field strength of the group formation willingness field; is the potential height of the resource availability potential field; is the field strength-potential coupling coefficient, used to unify the dimensions; is a very small constant to prevent division by zero errors; a preset system stability threshold, wherein the system stability threshold is determined by analyzing historical system operation data, and is used as a conflict intensity threshold value for distinguishing between a system stable state and a critical overload state; when the conflict intensity of any resource conflict saddle point exceeds the system stability threshold, a resource pre-dispatching process is automatically triggered; based on the collected global state data, low-load backup resource nodes are screened and dispatched from the resource pool to reduce the potential height of the resource availability potential field at the resource conflict saddle point; by introducing the backup resource nodes into the virtual resource space in a virtual injection manner, a resource depression with a lower resource availability potential field is formed at the corresponding position, thereby directly weakening the potential height of the resource availability potential field at the resource conflict saddle point; for all group decision vectors that may pass through the resource conflict saddle point, a virtual repulsive force field proportional to the conflict intensity is applied on the projection path of the resource conflict saddle point; wherein the virtual repulsive force field is centered on the resource conflict saddle point, and the influence range of the virtual repulsive force field expands as the conflict intensity increases; under the action of the repulsive force field, the projection path of the group decision vector is recalculated to bypass or quickly pass through the resource conflict saddle point, and then the corrected group decision vector is output.

[0025] It should be further explained that, in the specific implementation process, the process of pushing the group formation guidance information to the relevant users includes: obtaining the corrected group decision vector, and analyzing the optimal group participant set encapsulated in the corrected group decision vector; For each user in the optimal group user set, personalized guide information is generated; wherein, by querying the historical dynamic behavior data stream of the user, and using the pre-trained feature encoding network, the real-time position trajectory, interface interaction event sequence and group forming option exploratory operation of the user are mapped into a low-dimensional dense user individual preference vector; by calculating the similarity of the user individual preference vector and the group forming preference vector corresponding to the group forming scheme in the shared semantic space, the preference fit degree of the user and the group forming scheme is quantified, and the scheme advantage most associated with the quantified preference of the user is highlighted in the information content, for example, for price-sensitive users, the saving proportion is highlighted, and for users with high time efficiency requirements, the pre-locked delivery time commitment is explicitly provided; In the guide information, a one-key joining shortcut operation interface carrying a unique group forming session identifier; finally, the complete guide information is pushed to the terminal device of the corresponding user through the system message channel.

[0026] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments focuses on the differences from other embodiments. Especially, for the device embodiments, since they are basically based on the method embodiments, the description is relatively simple, and the related parts can be referred to the part of the method embodiments.

[0027] For the convenience of description, the above device is described as various units divided by functions. Of course, in the implementation of the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0028] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0029] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams can be implemented by computer program instructions, and the combination of flows and / or blocks in the flowcharts and / or block diagrams. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a machine that implements the functions described in the flowcharts and / or block diagrams. Figure 1 one flow or multiple flows and / or blocksFigure 1 means for performing the function specified in the block or blocks.

[0030] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flow Figure 1 flow or flows and / or blocks Figure 1 means for performing the function specified in the block or blocks.

[0031] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flow Figure 1 flow or flows and / or blocks Figure 1 means for performing the function specified in the block or blocks.

[0032] Secondly, the drawings of the embodiments disclosed in the present application only involve the structures related to the embodiments disclosed in the present application, and other structures can refer to the general design, and the same embodiments and different embodiments of the present application can be combined with each other without conflict; Finally, the above description is only the preferred specific implementation of the present application, but the protection scope of the present application is not limited to this, any person skilled in the art can make equivalent replacement or change according to the technical scheme and the inventive concept of the present application within the technical range disclosed in the present application, which should be covered in the protection scope of the present application.

Claims

1. A method for intelligent allocation recommendation of printing resources based on fusion of group-buying mode and user preferences, characterized by: S1, real-time collection of dynamic behavior data stream of group-buying user group and global state data of printing resource pool; S2, soft clustering of users in user feature space based on dynamic behavior data stream of group-buying user group through Gaussian mixture model, to generate user clustering centers; S3, mapping of each user clustering center to a group-buying preference vector, and construction of a group-buying willingness dynamic field based on the group-buying preference vector; S4, calculation and analysis of global state data of the printing resource pool, and superimposed generation of resource availability potential field; S5, calculation of gradient vectors of the group-buying willingness dynamic field and the resource availability potential field respectively, and interference effect analysis to identify group-buying opportunity focal points and resource conflict saddle points in the interference area; S6, generation of group-buying decision vectors for users initiating group buying based on group-buying opportunity focal points through space-time collaborative projection algorithm; S7, adaptive triggering of resource pre-dispatching strategy according to the conflict strength of the resource conflict saddle point, and imposing path correction on the group-buying decision vector passing through the resource conflict saddle point; S8, downward transmission of the corrected group-buying decision vector to resource execution nodes, and synchronous push of group-buying team formation guidance information to related users. 2.The method of claim 1, wherein, The dynamic behavior data stream includes real-time location trajectory, interface interaction event sequence and exploratory operation of group-buying options of the user; the global state data of the printing resource pool includes attribute and state information of each resource unit; wherein the resource unit is the basic unit for independent scheduling and management in the printing resource pool, and the attribute and state information includes real-time working state, task queue length, future reservation schedule and consumable inventory information of each printing device. 3.The method of claim 2, wherein, The process of soft clustering of users in user feature space through Gaussian mixture model to generate user clustering centers includes: extracting time series behavior features of users from dynamic behavior data stream, constructing user-behavior feature matrix, and initializing the number of clustering components of Gaussian mixture model based on information theory criteria; iterative optimization of parameters of Gaussian mixture model based on expectation maximization algorithm, the iterative process including: A1, expectation step: based on the current Gaussian mixture model parameters, calculating the posterior probability of each data point in the user-behavior feature matrix belonging to each Gaussian distribution component as the clustering membership of the user; A2, maximization step: using the calculated clustering membership as weights to recalculate the mean vector and covariance matrix of each Gaussian distribution component; outputting the mean vector of each Gaussian distribution component obtained after iterative convergence as the dynamically evolving user clustering center.

4. The method of claim 3, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation. The process of mapping each user clustering center to a group-buying preference vector and constructing a group-buying willingness dynamic field based on the group-buying preference vector includes: obtaining the output user clustering centers, each of which is a multi-dimensional vector; performing semantic mapping and scale scaling on each dimension of each user clustering center vector to convert it to a group-buying preference vector, wherein the dimensions of the group-buying preference vector include price sensitivity, delivery time requirement and printing quality level preference; In the virtual resource space, a time-varying basic strength is given to each group preference vector, and is marked as the field source strength; According to the field source strength, and combined with the field source attenuation model, the group willingness field strength of any point in the virtual resource space is calculated, wherein the group willingness field strength is formed by the superposition of the field strength vectors of all field sources at the point; Based on the calculated full field strength distribution, a group willingness dynamic field is constructed.

5. The method of claim 2, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation, characterized in that, The process of calculating and analyzing the global state data of the print resource pool and superimposing the resource availability potential field includes: Obtain the global state data of the print resource pool, analyze the real-time working state, task queue length and future reservation schedule of each resource unit, and predict the load pressure index of each resource unit in the future target time window; Inquire the consumable inventory information, and combine the historical failure rate of the resource unit to calculate its service reliability coefficient; Weight and fuse the load pressure index and the reciprocal of the service reliability coefficient to generate the load potential barrier value of each resource unit, wherein the load potential barrier value is positively correlated with the load pressure of the resource unit; In the virtual resource space, each resource unit is mapped as a potential field source, wherein the potential field height of the resource unit is equal to the load potential barrier value; Through the potential field superposition algorithm, a continuous resource availability potential field covering the entire virtual resource space is generated.

6. The method of claim 1, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation, characterized in that, The process of calculating the gradient vectors of the group willingness dynamic field and the resource availability potential field respectively, and performing interference effect analysis to identify the group opportunity focus and resource conflict saddle point in the interference area includes: Obtain the group willingness dynamic field, and calculate the gradient vector of the group willingness dynamic field at each point in the virtual resource space; Obtain the resource availability potential field, and calculate the gradient vector of the resource availability potential field; Based on the gradient vectors of the group willingness dynamic field and the resource availability potential field, interference analysis is performed to identify the group opportunity focus and resource conflict saddle point.

7. The method of claim 6, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation. The process of interference analysis based on the gradient vectors of the group willingness dynamic field and the resource availability potential field includes: C1, perform dot product operation on the gradient vectors of the group willingness dynamic field and the resource availability potential field, identify the area with the maximum product of the gradient direction opposite and the module length, and define it as the constructive interference area, wherein the center point of the constructive interference area is the group opportunity focus; C2, identify the area where the equipotential lines of the group willingness dynamic field and the resource availability potential field intersect and the field strengths weaken each other, and define it as the conflict interference area, wherein the center point of the conflict interference area is the resource conflict saddle point.

8. The method of claim 7, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation. The process of generating a group decision vector for a user initiating a group through a space-time collaborative projection algorithm includes: Map the group preference vector of the user initiating the group to the virtual resource space, and determine the position of the group preference vector in the space as the projection starting point; Take the group opportunity focus as the projection endpoint, and calculate the initial projection path from the starting point to the endpoint; On the initial projection path, integrate the field strength of the group willingness dynamic field as the path benefit estimation, and at the same time, integrate the potential field height change of the resource availability potential field as the path cost estimation; Maximize the ratio of path benefit to path cost as the target, and optimize the projection path by using the variational method; The users corresponding to the user cluster centers passed by the new path obtained after optimization are combined into an optimal group user set, and the resource binding scheme corresponding to the path end point is encapsulated into a group decision vector.

9. The method of claim 8, wherein the fusion of the cluster pattern and the user preference is recommended to the printing resource intelligent allocation. The adaptive triggering resource pre-disposition strategy and the process of imposing path correction on the group decision vector passing through the resource conflict saddle point include: Monitoring the conflict intensity at the resource conflict saddle point, wherein the conflict intensity is quantified by the standardized difference between the group willingness field strength and the potential field height of the resource availability potential field; A system stability threshold is preset; when the conflict intensity of any resource conflict saddle point exceeds the system stability threshold, the resource pre-disposition process is automatically triggered; based on the collected global state data, low-load standby resource nodes are screened and scheduled from the resource pool; For all group decision vectors that may pass through the resource conflict saddle point, a virtual repulsive force field proportional to the conflict intensity is imposed on the projected path of the resource conflict saddle point; Under the action of the repulsive force field, the projected path of the group decision vector is recalculated, and then the corrected group decision vector is output.

10. The method of claim 1, wherein the fusion of the pattern of the user's preference and the print resource is determined by a fusion algorithm. The process of pushing group formation guide information to related users includes: Obtaining the corrected group decision vector, analyzing the optimal group user set encapsulated in the corrected group decision vector; For each user in the optimal group user set, personalized guide information is generated; In the guide information, a one-key join shortcut operation interface carrying a unique group session identifier is carried; finally, the complete guide information is pushed to the terminal device of the corresponding user through the system message channel.

Images