Enterprise customer right presentation full-link tracking method and system based on order management

Abstract

The embodiment of the invention provides an enterprise customer right presentation full-link tracking method and system based on order management, and belongs to the technical field of computers. The method comprises the following steps: acquiring original log data of an enterprise customer order; obtaining an order state evolution diagram set and an edge weight set according to the original log data; and inputting the order state evolution diagram set and the edge weight set into a preset diagram attention network to obtain an order context vector. And screening out a right trigger type from the order state evolution diagram according to a preset rule. And according to the order context vector and a preset multi-factor comprehensive scoring function, calculating a strategy score for issuing each preset right and interest category on each right and interest triggering type, and obtaining a right and interest issuing strategy set. And issuing a corresponding right category to the enterprise customer according to the right issuing strategy set when the right triggering type occurs. The problem of mismatching of rights and interests is solved, and the utilization rate of rights and interests and customer satisfaction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of computer, in particular to an enterprise customer rights and interests sending full-link tracking method and system based on order management. BACKGROUND

[0002] How enterprise customers achieve customer stickiness improvement and customer life cycle value mining through rights and interests sending, exclusive incentives and other means has become an important development direction of marketing systems. In related technologies, enterprise customers generally set rules based on customer identity level, historical purchase behavior, total consumption quota and other static characteristics to send specific points, discounts, coupon packages or targeted services. However, enterprise customers may have test procurement orders, formal procurement orders and after-sales maintenance orders at the same time, and the processing flow, duration and approval nodes of each type of order are completely different. In this context, the mode based on "customer unified rights and interests package" for sending often cannot perceive and respond to the real business status of specific orders, resulting in problems such as rights and interests abuse, wrong sending or missing incentive window, which seriously affects rights and interests usage rate and customer satisfaction. Therefore, how to improve rights and interests usage rate and customer satisfaction has become a technical problem to be solved. SUMMARY

[0003] The main purpose of the embodiments of the present application is to provide an enterprise customer rights and interests sending full-link tracking method and system based on order management, aiming to improve rights and interests usage rate and customer satisfaction.

[0004] To achieve the above-mentioned purpose, the first aspect of the embodiments of the present application provides an enterprise customer rights and interests sending full-link tracking method based on order management, which comprises:

[0005] Obtaining original log data of enterprise customer orders;

[0006] Constructing an order state evolution graph according to the original log data and calculating the behavior edge weight of the order state evolution graph to obtain an edge weight set; wherein the order state evolution graph set comprises a plurality of order state evolution graphs, the nodes of the order state evolution graph are order state types, and the edges of the order state evolution graph are change paths between the order state types;

[0007] Inputting the order state evolution graph set and the edge weight set into a preset graph attention network to obtain an order context vector;

[0008] Filtering out rights and interests trigger types from the order state evolution graph according to a preset rule;

[0009] Calculating the strategy score of each preset rights and interests category on each rights and interests trigger type according to the order context vector and a preset multi-factor comprehensive scoring function to obtain a rights and interests sending strategy set;

[0010] According to the set of benefit issuance strategies, the corresponding benefit category is issued for the enterprise customer when the benefit trigger type occurs.

[0011] In some embodiments, the original log data includes order number, order state type and state change time, the order state evolution graph is constructed and the behavior edge weight of the order state evolution graph is calculated according to the original log data, and an edge weight set is obtained, including:

[0012] According to the original log data, all events of the same order number are sorted in ascending order of the state change time, and a plurality of order processing sequences are obtained;

[0013] According to the order processing sequence, an order state evolution graph is constructed and the behavior edge weight of the order state evolution graph is calculated, and an edge weight set is obtained; wherein the order state evolution graph set includes a plurality of order state evolution graphs, the nodes of the order state evolution graph are the order state types, and the edges of the order state evolution graph are the change paths between the order state types.

[0014] In some embodiments, after the set of benefit issuance strategies is used to issue the corresponding benefit category for the enterprise customer when the benefit trigger type occurs, the method further includes:

[0015] According to the set of benefit issuance strategies, a plurality of benefit behavior nodes are obtained;

[0016] According to a preset link edge weight function, the semantic edge weight between adjacent benefit behavior nodes is calculated; wherein the benefit behavior link graph includes the benefit behavior nodes and the semantic edge weight;

[0017] According to the semantic edge weight, the risk score of the benefit behavior link graph is calculated;

[0018] If the risk score is greater than a preset risk threshold, the benefit behavior link graph is marked as an abnormal chain.

[0019] In some embodiments, after the risk score of the benefit behavior link graph is calculated according to the semantic edge weight, the method further includes:

[0020] A semantic expectation vector of the benefit category is obtained;

[0021] According to the order context vector and the semantic expectation vector, a context semantic deviation item is obtained;

[0022] If the risk score is greater than the risk threshold, and the context semantic deviation item is greater than a preset deviation threshold, a preset first result is obtained.

[0023] If the risk score is greater than the risk threshold, and the contextual semantic deviation item is less than or equal to the deviation threshold, then it is the preset second result;

[0024] If the risk score is less than or equal to the risk threshold, then it is the preset third result.

[0025] In some embodiments, the link edge weight function is:

[0026]

[0027] Where, ω uv This represents the semantic edge weight for a jump from actor node u to actor node v. u actor v Indicates the initiator before and after the action, type v Indicates the type of behavior. Δt represents the predefined set of high-risk behavior types. uv p represents the time interval between the occurrence of the behavior. uv This represents the frequency of actual transitions from behavior node u to behavior node v, given the current customer, current benefit category, and current semantic expectation vector. μ1, μ2, μ3, and μ4 represent the historical average transition probability from behavior node u to behavior node v in the overall customer group of the enterprise. μ1, μ2, μ3, and μ4 represent the weight hyperparameters used to adjust the degree of influence of each behavior dimension.

[0028] In some embodiments, the graph attention network includes a graph attention layer and a semantic vector layer;

[0029] The graph attention layer is:

[0030]

[0031] in, State node s j The representation at the (l+1)th layer, Represents state node s k The representation of W at layer l (l) Let σ represent the linear transformation matrix of the l-th layer of the graph network, and let σ represent the Leaky ReLU activation function. Represents state node s j The set of neighboring nodes, w jk Indicates from state node s j to s k Behavioral edge weights, w jm Indicates from state node s j to s m Behavioral edge weights;

[0032] The semantic vector layer is:

[0033]

[0034] wherein, denotes an order context vector, Z denotes a normalization factor, and γ denotes a stage regularization weight, denotes an indicator function, denotes a final embedding of a state node s j .

[0035] In some embodiments, the multi-factor comprehensive score function is:

[0036]

[0037] wherein, denotes a policy score of an order o i to issue a benefit b j on a state s k , denotes an order context vector, φ k denotes a semantic expectation vector of a benefit category b k , denotes a set of states semantically related to the benefit category b k , η1 denotes a state semantic label item weight, Var jk denotes a behavioral volatility of a state node s j to issue the benefit category b k in the issuance history, λ1 denotes a volatility suppression item weight, ρ jk denotes a structure importance enhancement item, and λ2 denotes a structure enhancement item weight, and so on.

[0038] To achieve the above object, a second aspect of the embodiments of the present application proposes an enterprise customer benefit giving full-link tracking system based on order management, which comprises:

[0039] An acquisition module, configured to acquire original log data of enterprise customer orders;

[0040] A calculation module, configured to construct an order state evolution graph and calculate a behavioral edge weight of the order state evolution graph according to the original log data, to obtain an edge weight set; wherein, an order state evolution graph set comprises a plurality of order state evolution graphs, a node of the order state evolution graph is an order state type, and an edge of the order state evolution graph is a change path between the order state types;

[0041] An input module, configured to input the order state evolution graph set and the edge weight set into a preset graph attention network, to obtain an order context vector;

[0042] a screening module configured to screen, according to a preset rule, a benefit trigger type from the order state evolution graph;

[0043] a scoring module configured to calculate, according to the order context vector and a preset multi-factor comprehensive scoring function, a policy score of each preset benefit category on each benefit trigger type, to obtain a benefit distribution policy set;

[0044] a distribution module configured to distribute, according to the benefit distribution policy set, the corresponding benefit category to the enterprise customer when the benefit trigger type occurs.

[0045] To achieve the above object, a third aspect of the embodiments of the present application provides an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the method of the first aspect when executing the computer program.

[0046] To achieve the above object, a fourth aspect of the embodiments of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method of the first aspect.

[0047] The order management-based enterprise customer benefit giving full-link tracking method and system provided by the present application obtains the original log data of the enterprise customer order. The order state evolution graph is constructed according to the original log data, and the behavior edge weight of the order state evolution graph is calculated to obtain an edge weight set; wherein the order state evolution graph set comprises a plurality of order state evolution graphs, the node of the order state evolution graph is an order state type, and the edge of the order state evolution graph is a change path between order state types. The order state evolution graph set and the edge weight set are input into a preset graph attention network to obtain an order context vector. The benefit trigger type is screened from the order state evolution graph according to a preset rule. The policy score of each preset benefit category on each benefit trigger type is calculated according to the order context vector and a preset multi-factor comprehensive scoring function, to obtain a benefit distribution policy set. The corresponding benefit category is distributed to the enterprise customer according to the benefit distribution policy set when the benefit trigger type occurs. Thus, a system capable of perceiving the order state evolution process, dynamically adapting the benefit distribution policy, and having link-level recording and auditing capability is constructed. The problem of benefit mismatch is solved, and the benefit utilization rate and customer satisfaction are improved. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 is a flowchart of the order management-based enterprise customer benefit giving full-link tracking method provided by the embodiments of the present application;

[0049] Figure 2 isFigure 1 the flowchart of step S102 in FIG. 1;

[0050] Figure 3 is a flowchart of an enterprise customer benefit gift sending full-link tracking method based on order management provided by another embodiment of the present application;

[0051] Figure 4 is a flowchart of an enterprise customer benefit gift sending full-link tracking method based on order management provided by a third embodiment of the present application;

[0052] Figure 5 is a structural schematic diagram of an enterprise customer benefit gift sending full-link tracking system based on order management provided by an embodiment of the present application;

[0053] Figure 6 is a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0054] In order to make the objects, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not intended to limit the present application.

[0055] It should be noted that although the functional modules are divided in the system schematic diagram, and the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order from the module division in the system or the order in the flowchart. The terms "first", "second", etc. in the specification and claims and the above-described drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0056] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application, and are not intended to limit the present application.

[0057] How do enterprise customers achieve customer stickiness and customer life cycle value through equity gifts, exclusive incentives and other means, become an important development direction of the marketing system. In related technologies, enterprise customers generally set rules according to customer identity level, historical purchase behavior, total consumption quota and other static characteristics, and issue specific points, discounts, coupon packages or targeted services. However, enterprise customers may have test purchase orders, formal purchase orders and after-sales maintenance orders at the same time, and the processing flow, length and approval node of each type of order are completely different. In this context, the mode based on "customer unified equity package" for issuing often cannot perceive and respond to the real business status of specific orders, resulting in problems such as equity abuse, wrong issuance or missing incentive window, which seriously affects the equity usage rate and customer satisfaction.

[0058] Based on this, the embodiment of the application provides an enterprise customer equity gift full-link tracking method and system based on order management, which aims to obtain an order state evolution graph set and an edge weight set from original log data. The order state evolution graph set and the edge weight set are input into a graph attention network to obtain an order context vector. According to a preset rule, an equity trigger type is screened out from the order state evolution graph, and a policy score of each preset equity category on each equity trigger type is calculated according to the order context vector and a preset multi-factor comprehensive scoring function, to obtain an equity issuance strategy set. According to the equity issuance strategy set, the corresponding equity category is issued to the enterprise customer when the equity trigger type occurs. Thus, a system capable of perceiving the order state evolution process, dynamically adapting the equity issuance strategy, and having link-level record and audit capability is constructed. The problem of equity mismatch is solved, and the equity usage rate and customer satisfaction are improved.

[0059] The enterprise customer equity gift full-link tracking method and system based on order management provided by the embodiment of the application are specifically explained through the following embodiments. First, the enterprise customer equity gift full-link tracking method based on order management in the embodiment of the application is described.

[0060] The embodiment of the application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence (Artificial Intelligence, AI) is the theory, method, technology and application system of using digital computer or digital computer controlled machine to simulate, extend and expand human intelligence, perceive environment, acquire knowledge and use knowledge to obtain the best results.

[0061] The artificial intelligence basic technology generally includes technologies such as sensors, special artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction system, mechatronics, etc. The artificial intelligence software technology mainly includes computer vision technology, robot technology, biometric identification technology, speech processing technology, natural language processing technology, and machine learning / deep learning, etc.

[0062] The enterprise customer rights and interests gift full-link tracking method based on order management provided by the embodiments of the present application can be applied to a terminal, can also be applied to a server end, and can also be software running in the terminal or the server end. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, etc.; the server end can be configured as an independent physical server, can also be configured as a server cluster or a distributed system composed of multiple physical servers, can also be configured as a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN, and big data and artificial intelligence platform, etc.; and the software can be an application program for implementing the enterprise customer rights and interests gift full-link tracking method based on order management, etc., but is not limited to the above forms.

[0063] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multi-processor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected by a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0064] Please refer to Figure 1 , Figure 1 is a flowchart of the enterprise customer rights and interests gift full-link tracking method based on order management provided by the embodiments of the present application, Figure 1 The method in the flowchart can include but is not limited to steps S101 to S106.

[0065] Step S101, obtaining original log data of an enterprise customer order;

[0066] Step S102, constructing an order state evolution graph according to the original log data and calculating the behavior edge weight of the order state evolution graph to obtain an edge weight set; wherein the order state evolution graph set includes a plurality of order state evolution graphs, the node of the order state evolution graph is an order state type, and the edge of the order state evolution graph is a change path between order state types;

[0067] Step S103, inputting the order state evolution graph set and the edge weight set into a preset graph attention network to obtain an order context vector;

[0068] Step S104, screening a rights trigger type from the order state evolution graph according to a preset rule;

[0069] Step S105, calculating a policy score of each preset rights category on each rights trigger type according to the order context vector and a preset multi-factor comprehensive scoring function to obtain a rights issuance policy set;

[0070] Step S106, issuing a corresponding rights category to an enterprise customer when the rights trigger type occurs according to the rights issuance policy set.

[0071] In step S101 of some embodiments, the original log data is structured log data of an enterprise internal order management system, each record is a state change event of a certain order, and specific fields include:

[0072] order_id: order number;

[0073] timestamp: state change time (Unix timestamp, accurate to minutes);

[0074] action_type: order state type, standard enumeration value, such as "submit" (ordering), "approve" (approval), "ship" (shipping), "receive" (confirming receipt), "cancel" (canceling), etc.;

[0075] source_type: operation source, including "manual" (manual), "system" (system automatic), and "api" (interface);

[0076] operator_id: user or system number performing the operation.

[0077] The original log data can be regularly pulled through an event bus or a database API query interface of the enterprise order management system, and the format is uniformly an event list in time sequence.

[0078] Please refer to Figure 2In some embodiments, the original log data includes an order number, an order state type, and a state change time, and step S102 can include but is not limited to steps S201-S202:

[0079] In step S201, all events of the same order number are sorted in ascending order of state change time according to the original log data, and a plurality of order processing sequences are obtained.

[0080] In step S202, an order state evolution graph is constructed according to the order processing sequence, and the behavior edge weight of the order state evolution graph is calculated, and a set of edge weights is obtained; wherein the set of order state evolution graphs includes a plurality of order state evolution graphs, the nodes of the order state evolution graph are order state types, and the edges of the order state evolution graph are change paths between order state types.

[0081] In step S201 of some embodiments, the original log data is aggregated according to the order number order_id, and all events of the same order are sorted in ascending order of state change time timestamp, forming an order processing sequence. When there are multiple orders, a plurality of order processing sequences are obtained. Taking an order as an example, the order number is ORD_20240401_0001, and the order has the following events in its life cycle:

[0082] 10:01 submit order;

[0083] 10:15 approve;

[0084] 11:00 ship;

[0085] 12:35 reject.

[0086] These events are converted into an order processing sequence as shown in the following formula (1):

[0087] [(submit, 10:01), (approve, 10:15), (ship, 11:00), (reject, 12:35)], (1)

[0088] In step S202 of some embodiments, an order state evolution graph is constructed for the order processing sequence Each order status type (such as submit, approve, ship, reject) is a node in the graph. The edges in the graph represent the change path between order status types, for example, the edge from "submit" to "approve" represents "after submitting the order, it enters the state of approval". Each edge not only marks the state jump, but also carries the operation time interval, the operation source and whether it belongs to the key business behavior. The order status evolution graph set includes multiple order status evolution graphs, i.e., the order status evolution graph set

[0089] In the construction process, the weight of each edge is calculated when constructing the edge between each pair of adjacent states (such as submit→approve). The calculation is performed using formula (2) as follows:

[0090]

[0091] where w j,j+1 represents the edge weight from node s j to node s j+1 , which is used to express the importance of the current state jump. Δt j,j+1 represents the time interval between node s j and node s j+1 , which is directly calculated by the difference of the timestamp field. indicates an indicator function, which is 1 when the condition is met, and 0 when it is not met. source j+1 represents the source of the state change, which comes from the source_type field. s j+1 represents the order status type, which is given by the action_type field. represents the key state set defined by the system, such as {"approve", "reject", "cancel"}.

[0092] In an example, if "approve" is a manual operation and belongs to the key state set, and the time interval between "submit" and "approve" is 14 minutes, then the weight of the edge is shown in the following formula (3):

[0093] w = 0.1 x 14 + 0.5 x 1 + 1.0 x 1 = 2.9, (3)

[0094] Through the above steps S201 to S202, the order status evolution graph can be obtained, and each edge in the order status evolution graph has the meaning of structure and behavior, not only connecting two events, but also expressing whether this step is initiated by a person, whether it is important, and how long it takes.

[0095] In step S103 of some embodiments, the input to the graph attention network is a set of order state evolution graphs. Edge weight set The output is an order context vector. The order context vector set is Each order's status evolution diagram It is a state transition diagram built based on the raw log data of actual orders, V i E is a set of state nodes. i This is a set of state edges, each edge carrying a weight calculated from time, operator type, and key state markers. These diagrams provide structured behavioral process information at the order level for subsequent models, and are a structural abstraction of the enterprise customer order processing sequence.

[0096] The purpose of this step is to visualize the order status evolution graph. The graph structure information is encoded as an order context vector with contextual representation capabilities. This is used for subsequent matching of rights triggering strategies. In traditional context modeling, order metadata (such as amount, industry, time) is often used for encoding. However, enterprise orders have complex state transition behaviors, and the order processing path contains a large amount of semantics that determine "whether rights should be issued" and "what type of rights is more appropriate". Therefore, dynamic semantic representations must be extracted from the graph structure.

[0097] It should be noted that the context vector encoding method based on the order state evolution graph employs a graph attention network (GAT) and introduces a regularization term for order process stages. This term is specifically designed for enterprise customer orders, which typically include stage-based features such as "early probing," "mid-stage fulfillment," and "late delivery," capturing the differences in performance of different stages within the context and enhancing the stage-awareness capability of semantic representation. This design is particularly important in the process of matching benefits. For example, a trial coupon might be issued after "approval," logistics guarantee benefits might be issued during "fulfillment," and a follow-up incentive would be more appropriate after "receipt."

[0098] Each state node s j ∈V i initial feature vector From a state embedding layer, by a state encoder f enc The encoder is constructed by assigning values ​​to different states based on a fixed state table and learning embedding vectors (64 dimensions). Each edge e j,k ∈E i weight w jk It has been defined and calculated, and has semantics related to real business behavior.

[0099] The graph attention network consists of a graph attention layer and a semantic vector layer. The state node representation is updated through propagation through two graph attention layers. The graph attention layer is shown in the following formula (4):

[0100]

[0101] in, State node s j The representation at the (l+1)th layer, State node s k Representation at layer l. W (l) Let represent the linear transformation matrix of the l-th layer of the graph network, with dimensions 64×64. σ represents the LeakyReLU activation function. State node s j The set of neighboring nodes, w jk Indicates from state node s j to s k Behavioral edge weights, w jm Indicates from state node s j to s m Behavioral edge weights.

[0102] The above structure can effectively aggregate behavioral path information in the graph structure and focus on key turning points (such as customer cancellation, abnormal receipt, etc.) through edge weights, thereby enhancing the contextual representation's ability to understand actual business behavior.

[0103] Instead of simple averaging, a stage-aware aggregation mechanism is introduced when aggregating semantic vectors. Specifically, a stage encoding function g(s) is defined. j Each state node is mapped to a stage label (such as "early stage", "mid stage", "late stage"), and the node aggregation weights are adjusted using stage bias terms. Specifically, the semantic vector layer is shown in the following formula (5):

[0104]

[0105] in, Let Z represent the order context vector, Z represent the normalization factor to ensure that all weights sum to 1, and γ represent the stage regularization weight, with a recommended value between 0.2 and 0.5. This indicates an indicator function, which is 1 when the node is in the "mid-term" stage and 0 otherwise. State node s j The final embedding. The stage mapping g(·) is a static mapping table from state to stage (e.g., submit for early stage, approve / ship for mid-stage, receive for late stage).

[0106] This approach has strong practicality in the enterprise order context: mid-stage states are often the best window for order equity intervention (e.g. fulfillment guarantee, customer support), so this vector representation naturally introduces this bias, providing an information base for the next decision-making step that "pays more attention to the key period".

[0107] In addition, this mechanism also naturally supports the extension of different industry orders with different life cycle stage definitions. For example, the SaaS industry may have four stages of "trial / deployment / online / renewal", and the medical equipment industry may have "order / bid / acceptance / maintenance".

[0108] In step S104 of some embodiments, there are strong contextual differences in enterprise customer orders, such as some orders being annual procurement and some orders being small batch trial. It is difficult to effectively infer which equity should be issued and when the equity should be intervened in the order process based on customer level or product amount. Therefore, the goal of this step is to accurately identify "which node should the equity be issued" and "which type of equity should be issued" based on the order state evolution graph set G o and the order context vector set I o .

[0109] To achieve this goal, potential equity trigger types are extracted from the order state evolution graph and a strategy score is calculated for each equity trigger type Combining semantic similarity, node semantic category matching, and behavior risk penalty to make a combined judgment. Finally, the most suitable equity type is selected, and the corresponding equity issuance strategy set (t j ,b k ) is output.

[0110] Specifically, the preset rules define potential equity trigger types as nodes that meet one of the following conditions:

[0111] The state s j corresponding to the node is in the set of system-defined strategy-sensitive nodes {“approve”, “ship”, “receive”, “reject”};

[0112] The weight w j,j+1 of the forward edge of the node is greater than τ (e.g. τ = 2.5), indicating that this jump may involve critical operations such as manual intervention or extraordinary delay;

[0113] The node has multiple incoming edges (in-degree ≥ 2 in the graph), indicating that the state may be a fusion node of multiple process paths, with behavior intersection.

[0114] In step S105 of some embodiments, a strategy score of each preset benefit category is calculated on each benefit trigger type according to the order context vector and a preset multi-factor comprehensive scoring function, to obtain a set of benefit issuance strategies.

[0115] Specifically, for each benefit trigger type t j and each benefit category b k (total K categories), a multi-factor comprehensive scoring function is constructed as shown in the following formula (6) and formula (7):

[0116]

[0117] Wherein, represents the strategy score of the order o i issuing the benefit b j in the state s k , represents the order context vector, which is the behavior flow information of the entire order. φ k represents the semantic expectation vector of the benefit category b k , which can be obtained through historical training. represents the state set related to the semantic of the benefit category b k , for example, the warranty benefit usually matches the states of “receive”, “complete” and the like. η1 represents the state semantic label item weight (such as 0.5-1.0). Var jk represents the behavior volatility of the state node s j issuing the benefit category b k in the issuance history (such as the variance of the return rate / usage rate), which is used to punish unstable strategies. λ1 represents the volatility suppression item weight, ρ jk represents the structure importance enhancement item, which represents the structural similarity of the state s j to the high-risk / high-value state node in the order state evolution graph, and the higher the structural similarity, the greater the possible value of the node strategy. λ2 represents the structure enhancement item weight (such as 0.2-0.4). The high-risk state set represents the shortest path set of the state s j to the high-risk state, and len(p) represents the hop number of the path p. The shorter the path, the closer s j is to the risk node, and the greater ρ jk is, indicating that the benefit is issued in advance at the node has “buffering” potential. This design makes the benefit strategy defensive, that is, even if the customer does not explicitly cancel or return, if the customer is close to the risk node, the customer service follow-up, delayed performance and other benefits can be issued in advance.

[0118] For each benefit trigger type t j, select the benefit category with the maximum strategy score As shown in the following formula (8):

[0119]

[0120] wherein, denotes the benefit category corresponding to the maximum strategy score, denotes the order o i The strategy score of the benefit b j issued in the state s k , and K denotes the total K benefit categories.

[0121] The final benefit issuance strategy of the order o i , as shown in the following formula (9):

[0122]

[0123] wherein, denotes the benefit issuance strategy of the order o i , t j denotes the benefit trigger type, denotes the benefit category with the maximum strategy score, denotes the set of benefit trigger types. The set of benefit issuance strategies

[0124] This step realizes intelligent benefit judgment by combining the order context vector I o and the state graph structure G o , and introducing a comprehensive scoring mechanism . In particular, a structure importance enhancement term p jk is designed in the scoring function, which is based on the risk distance calculation in the order state evolution graph, and injects a structural early warning capability into the benefit issuance strategy. At the same time, the historical behavior volatility is introduced as a penalty term, realizing dynamic adjustment based on experience feedback. This benefit decision mechanism combining behavior path graph, semantic intention vector and structural risk indicator significantly improves the pertinence and timeliness of benefit allocation in enterprise customer scenarios.

[0125] In step S106 of some embodiments, the corresponding benefit category is issued to the enterprise customer according to the benefit issuance strategy set when the benefit trigger type occurs. Specifically, for the benefit issuance strategy set wherein each denotes that the state node t i in the order o j is identified as a benefit trigger type, and the benefit category b

[0126] The steps S101 to S106 shown in the embodiments of the present application obtain the original log data of the enterprise customer order. An order state evolution graph is constructed according to the original log data, and the behavior edge weight of the order state evolution graph is calculated to obtain an edge weight set; wherein the order state evolution graph set includes a plurality of order state evolution graphs, the node of the order state evolution graph is an order state type, and the edge of the order state evolution graph is a change path between order state types. The order state evolution graph set and the edge weight set are input into a preset graph attention network to obtain an order context vector. According to a preset rule, a right benefit trigger type is screened out from the order state evolution graph. According to the order context vector and a preset multi-factor comprehensive scoring function, a policy score of each preset right benefit category on each right benefit trigger type is calculated to obtain a right benefit policy set. According to the right benefit policy set, the corresponding right benefit category is issued to the enterprise customer when the right benefit trigger type occurs. Thus, a system capable of perceiving the order state evolution process, dynamically adapting the right benefit issuing strategy, and having link-level recording and auditing capability is constructed. The problem of right benefit mismatch is solved, and the right benefit utilization rate and customer satisfaction are improved.

[0127] Please refer to Figure 3 In some embodiments, after step S106, the enterprise customer right benefit giving full-link tracking method based on order management can further include but is not limited to steps S301 to S304:

[0128] Step S301, obtaining a plurality of right benefit behavior nodes according to the right benefit policy set;

[0129] Step S302, calculating the semantic edge weight between adjacent right benefit behavior nodes according to a preset link edge weight function; wherein the right benefit behavior link graph includes the right benefit behavior node and the semantic edge weight;

[0130] Step S303, calculating the risk score of the right benefit behavior link graph according to the semantic edge weight;

[0131] Step S304, if the risk score is greater than a preset risk threshold, the right benefit behavior link graph is marked as an abnormal chain.

[0132] In steps S301 to S302 of some embodiments, the right benefit trigger type in the right benefit policy set is the right benefit behavior node. A plurality of right benefit behavior nodes are obtained according to the right benefit policy set. The semantic edge weight between adjacent right benefit behavior nodes is calculated according to a preset link edge weight function. Wherein the right benefit behavior link graph includes the right benefit behavior node and the semantic edge weight, and the right benefit behavior link graph G b =(V b ,E b ), V b represents the right benefit behavior node, and E brepresenting the transition path between behavior events, each edge with a calculated semantic edge weight ω uv , and can be accompanied by behavior metadata, such as log ID, operating device, behavior channel (such as mobile terminal, interface), etc.

[0133] It should be noted that, in order to track the use and transfer of rights in the actual operation process, the right event flow log data provided by the enterprise marketing system is introduced in this step, including but not limited to the following event types: right issuance, customer collection, use, return, freezing, transfer, expiration, etc. These events are automatically synchronized through the system interface, and the fields include right_id (right instance number), event_type (behavior type), timestamp (time), actor_id (behavior initiator), etc.

[0134] In the traditional right management system, the state change of the right is often recorded in multiple fields in the database, and is not modeled as a "behavior path", which makes it difficult for the system to perform structured backtracking and semantic analysis on the right flow process when there is a dispute over customer rights or the right strategy needs to be reviewed. The embodiment aims to build a right behavior link graph G b , which organizes the entire behavior path of each right from being triggered, used, returned, etc. into a queryable, auditable, and featureable graph structure.

[0135] The construction process constructs a right behavior link graph G for each right instance , and binds order number, state node, right type, trigger time, etc. meta information. The system then scans the right event flow log to retrieve all related behavior events, and constructs a behavior sequence in ascending order of time.

[0136] These behavior nodes include: issuance, customer viewing, customer collection, use success, return application, transfer success, expiration, system freezing, etc. The nodes are organized as a directed chain structure, where each edge represents a behavior transition. To achieve subsequent explainable analysis, a link edge weight function ω uv is designed for each edge, which not only reflects the time relationship, but also considers behavior risk, behavior subject, and deviation from issuance logic.

[0137] Specifically, in order to better measure "right behavior chain complexity" and "behavior chain breakage warning" from the perspective of enterprise scenarios, the designed link edge weight function is as shown in the following formula (10):

[0138]

[0139] where ω uv ​actor u , actor v type v type Δt uv p uv , actor μ1,μ2,μ3,μ4

[0140]

[0141] In some embodiments, in steps S303 to S304, all equity behavior chains P b = {r i →a1→...→a i} are extracted from the equity behavior link graph G m , and for each chain, the risk score of the equity behavior link graph is calculated based on the semantic edge weight ω uv . As shown in the following formula (11):

[0142]

[0143] wherein, ω uv P i G b τ r When the risk threshold is τ , the equity behavior link graph is marked as an abnormal chain when

[0144] ​​In steps S301 to S304 shown in the embodiment, by comparing the risk score and the risk threshold, the equity behavior link icon is marked as an abnormal chain when the risk score is greater than the risk threshold. Thus, the system can record each triggered equity instance in the enterprise order as a structured node, and construct a complete equity behavior link based on the customer's behavior log such as taking, using and returning; by calculating the semantic edge weight of each jump behavior in the link, the system can identify the chain segment with high operational risk and abnormal behavior deviating from the expectation, and further judge whether the whole link is abnormal through aggregation score; finally, the system can automatically mark the high-risk equity chain, provide clear basis for timely adjustment of strategies (such as suspension, reduction, manual review) for the operation side, and ensure dynamic monitoring and control of the whole process of customer equity issuance by enterprises in complex order scenarios.

[0145] Referring to Figure 4 In some embodiments, after step S304, the enterprise customer equity gift full-link tracking method based on order management can further include but is not limited to steps S401 to S405:

[0146] Step S401, obtaining a semantic expectation vector of the equity category;

[0147] Step S402, obtaining a context semantic deviation item according to the order context vector and the semantic expectation vector;

[0148] Step S403, if the risk score is greater than the risk threshold, and the context semantic deviation item is greater than a preset deviation threshold, a preset first result is obtained;

[0149] Step S404, if the risk score is greater than the risk threshold, and the context semantic deviation item is less than or equal to the deviation threshold, a preset second result is obtained;

[0150] Step S405, if the risk score is less than or equal to the risk threshold, a preset third result is obtained.

[0151] In steps S401 to S402 of some embodiments, in order to further avoid misjudgment of the equity behavior link graph, a context semantic deviation item is introduced to measure the expected difference between the actual behavior and the order context vector. Therefore, the semantic expectation vector of the equity category is obtained, and the context semantic deviation item is obtained according to the order context vector and the semantic expectation vector. As shown in the following formula (12):

[0152]

[0153] Wherein, the context semantic deviation item is denoted as the order context vector is denoted as φk representing the equity category b k corresponding semantic expectation vector. That is, the matching degree of the order context and the equity if significantly deviates (for example, it is actually a "compensatory equity" but used for "active marketing"), it means that the strategy itself has a problem, or the node selection is inaccurate.

[0154] In steps S403 to S405 of some embodiments, the first result is strategy mismatch, the second result is customer abnormal behavior, and the third result is normal behavior without intervention. Therefore, when the risk threshold is τ r , the deviation threshold τ s , if and , it is attributed to "strategy mismatch"; if and , it is attributed to "customer abnormal behavior"; if , it is considered normal behavior without intervention.

[0155] In steps S401 to S405 shown in the embodiment, by comparing the risk score with the risk threshold and comparing the context semantic deviation item with the deviation threshold, a more accurate judgment result of the equity behavior link graph is obtained, and misjudgment of the equity behavior link graph is avoided.

[0156] Referring to Figure 5 , the embodiment of the application further provides an enterprise customer equity gift full-link tracking system based on order management, which can implement the above-mentioned enterprise customer equity gift full-link tracking method based on order management. The system comprises:

[0157] An acquisition module 501 is configured to acquire original log data of an enterprise customer order;

[0158] A calculation module 502 is configured to construct an order state evolution graph and calculate a behavior edge weight of the order state evolution graph according to the original log data, to obtain an edge weight set. The order state evolution graph set comprises a plurality of order state evolution graphs, the node of the order state evolution graph is an order state type, and the edge of the order state evolution graph is a change path between order state types.

[0159] An input module 503 is configured to input the order state evolution graph set and the edge weight set into a preset graph attention network, to obtain an order context vector;

[0160] A screening module 504 is configured to screen an equity trigger type from the order state evolution graph according to a preset rule;

[0161] The scoring module 505 is configured to calculate a strategy score of each preset benefit category on each benefit trigger type according to the order context vector and a preset multi-factor comprehensive scoring function, to obtain a benefit distribution strategy set.

[0162] The distribution module 506 is configured to distribute a corresponding benefit category for the enterprise customer according to the benefit distribution strategy set when the benefit trigger type occurs.

[0163] The specific implementation of the order management-based enterprise customer benefit giving full-link tracking system is basically the same as that of the order management-based enterprise customer benefit giving full-link tracking method, and will not be repeated here.

[0164] The embodiments of the present application also provide an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the order management-based enterprise customer benefit giving full-link tracking method. The electronic device can be any smart terminal, such as a tablet computer or a vehicle-mounted computer.

[0165] Please refer to Figure 6 , Figure 6 The hardware structure of the electronic device of another embodiment is illustrated, which includes:

[0166] The processor 601 can be implemented in a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, and is used to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0167] The memory 602 can be implemented in the form of a ROM (Read Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory). The memory 602 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 602 and called and executed by the processor 601 to implement the order management-based enterprise customer benefit giving full-link tracking method.

[0168] The input / output interface 603 is used to realize information input and output.

[0169] The communication interface 604 is configured to realize the communication interaction between the device and other devices, and can realize the communication through a wired manner (for example, a USB, a network cable, or the like) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, or the like).

[0170] The bus 605 is configured to transmit information between various components (for example, the processor 601, the memory 602, the input / output interface 603, and the communication interface 604) of the device.

[0171] The processor 601, the memory 602, the input / output interface 603, and the communication interface 604 are connected to each other through the bus 605 to realize the communication connection between the device.

[0172] The embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the order management-based enterprise customer rights and interests gift full-link tracking method.

[0173] The memory is a non-transitory computer readable storage medium, and can be used to store a non-transitory software program and a non-transitory computer executable program. In addition, the memory can include a high-speed random access memory, and can further include a non-transitory memory, for example, at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include a memory remotely arranged relative to the processor, and the remote memory can be connected to the processor through a network. Examples of the network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0174] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0175] Those skilled in the art can understand that the technical solutions shown in the figure do not constitute a limitation on the embodiments of the present application, and can include more or fewer steps than the figure, or combine certain steps, or different steps.

[0176] The system embodiments described above are only schematic, and the units described as separate components can be or can not be physically separated, that is, can be located in one place, or can be distributed on multiple network units. According to actual needs, part or all of the modules can be selected to realize the purpose of the embodiments.

[0177] Those skilled in the art can understand that all or some of the steps in the method disclosed above, the function modules / units in the system and the device can be implemented as software, firmware, hardware or appropriate combination thereof.

[0178] The terms "first", "second", "third", "fourth" and the like in the description of the application and in the claims, if any, are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the use of these terms herein is to be construed to cover a general order and / or structure unless otherwise indicated. Furthermore, the terms "comprise", "comprising", "has", "having", "includes", "including", "contain", "containing" or any other similar forms are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises, has, includes, contains items or components does not include items or components not explicitly recited. The terms "a" or "an", as used herein in the detailed description and in the claims, mean "one or more" or "at least one", unless otherwise indicated.

[0179] It should be understood that, in the present application, "at least one" means one or more, and "multiple" means two or more. "And / or" is used to describe the relationship between associated objects, which means that there can be three relationships, for example, "A and / or B" can mean that there are three cases: only A, only B, and A and B at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can mean a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be singular or plural.

[0180] In several embodiments provided in the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the above-described system embodiments are only illustrative, for example, the division of the above-mentioned units is only a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed objects can be indirect coupling or communication connection between the systems or units through some interfaces, and can be electrical, mechanical or other forms.

[0181] The units described as separate components above can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0182] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0183] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical scheme of the present application or the part that contributes to the prior art or the whole or part of the technical scheme can be embodied in the form of a software product. The computer software product is stored in a storage medium, including multiple instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0184] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, but this does not limit the scope of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the embodiments of the present application.

Claims

1. An order management-based enterprise customer equity gift giving full-link tracking method, characterized in that, The method comprises: obtaining original log data of enterprise customer orders; constructing an order state evolution graph according to the original log data and calculating the behavior edge weight of the order state evolution graph to obtain an edge weight set; wherein the order state evolution graph set comprises a plurality of order state evolution graphs, the nodes of the order state evolution graph are order state types, and the edges of the order state evolution graph are change paths between the order state types; inputting the order state evolution graph set and the edge weight set into a preset graph attention network to obtain an order context vector; screening a benefit trigger type from the order state evolution graph according to a preset rule; calculating a policy score of each preset benefit category on each benefit trigger type according to the order context vector and a preset multi-factor comprehensive scoring function to obtain a benefit distribution strategy set; distributing the corresponding benefit category to the enterprise customer when the benefit trigger type occurs according to the benefit distribution strategy set.

2. The method of claim 1, wherein, The original log data comprises an order number, an order state type and a state change time, and the order state evolution graph is constructed according to the original log data and the behavior edge weight of the order state evolution graph is calculated to obtain an edge weight set, which comprises: sorting all events of the same order number in ascending order according to the state change time to obtain a plurality of order processing sequences according to the original log data; constructing an order state evolution graph according to the order processing sequence and calculating the behavior edge weight of the order state evolution graph to obtain an edge weight set; wherein the order state evolution graph set comprises a plurality of order state evolution graphs, the nodes of the order state evolution graph are the order state types, and the edges of the order state evolution graph are change paths between the order state types.

3. The method of claim 1, wherein, After the benefit distribution strategy set is used to distribute the corresponding benefit category to the enterprise customer when the benefit trigger type occurs, the method further comprises: obtaining a plurality of benefit behavior nodes according to the benefit distribution strategy set; calculating the semantic edge weight between adjacent benefit behavior nodes according to a preset link edge weight function; wherein the benefit behavior link graph comprises the benefit behavior nodes and the semantic edge weight; calculating the risk score of the benefit behavior link graph according to the semantic edge weight; if the risk score is greater than a preset risk threshold, the benefit behavior link graph is marked as an abnormal chain.

4. The method of claim 3, wherein, After the risk score of the benefit behavior link graph is calculated according to the semantic edge weight, the method further comprises: obtaining a semantic expectation vector of the benefit category; obtaining a context semantic deviation item according to the order context vector and the semantic expectation vector; if the risk score is greater than the risk threshold and the context semantic deviation item is greater than a preset deviation threshold, a preset first result is obtained; if the risk score is greater than the risk threshold and the context semantic deviation item is less than or equal to the deviation threshold, a preset second result is obtained; if the risk score is less than or equal to the risk threshold, a preset third result is obtained.

5. The method of claim 3, wherein, The link edge weight function is: wherein ω uv denotes the semantic edge weight of jumping from behavior node u to behavior node v, actor u , actor v denotes the initiator before and after the behavior, type v denotes the behavior type, denotes the preset high-risk behavior type set, Δt uv denotes the behavior occurrence interval time, p uv denotes the transition frequency of actually occurring from behavior node u to behavior node v under the current customer, the current equity category, and the current semantic expectation vector, denotes the historical average transition probability of jumping from behavior node u to behavior node v in the overall customer group of the enterprise, μ1, μ2, μ3, μ4 denote weight hyperparameters for adjusting the influence degree of each behavior dimension.

6. The method of claim 1, wherein, The graph attention network comprises a graph attention layer and a semantic vector layer; The graph attention layer is: in, Represents state node s j The representation at the (l+1)th layer, State node s k The representation of W at layer l (l) Let σ represent the linear transformation matrix of the l-th layer of the graph network, and let σ represent the Leaky ReLU activation function. Represents state node s j The set of neighboring nodes, w jk Indicates from state node s j to s k Behavioral edge weights, w jm Indicates from state node s j to s m Behavioral edge weights; The semantic vector layer is: wherein, denotes the order context vector, Z denotes a normalization factor, γ denotes a stage regularization weight, denotes an indicator function, denotes the final embedding of the state node s j .

7. The method of claim 1, wherein, The multi-factor comprehensive scoring function is: in, Indicates order o i In state s j Rights and benefits issued above k Strategy score, Represents the order context vector, φ k Indicates equity category b k The semantic expectation vector, Indicates the relationship with rights category b k A semantically related set of states, where η1 represents the weight of the state semantic label item, and Var jk This indicates the state node s in the issuance history. j Issuance of rights category b k The behavioral volatility, where λ1 represents the weight of the volatility suppression term, ρ jk λ² represents the structural importance enhancement term, and λ² represents the weight of the structural enhancement term. Represents state s j The set of shortest paths to high-risk states, where len(p) represents the number of hops for path p.

8. An order management based enterprise customer equity gift program end-to-end tracking system, comprising: The system comprises: An acquisition module configured to acquire original log data of enterprise customer orders; A calculation module configured to construct an order state evolution graph and calculate behavior edge weights of the order state evolution graph according to the original log data, to obtain an edge weight set; wherein the order state evolution graph set comprises a plurality of order state evolution graphs, a node of the order state evolution graph is an order state type, and an edge of the order state evolution graph is a change path between the order state types; An input module configured to input the order state evolution graph set and the edge weight set into a preset graph attention network, to obtain an order context vector; A screening module configured to screen out a right trigger type from the order state evolution graph according to a preset rule; A scoring module configured to calculate a policy score of each preset right category on each right trigger type according to the order context vector and a preset multi-factor comprehensive scoring function, to obtain a right issuance policy set; An issuance module configured to issue a corresponding right category to the enterprise customer when the right trigger type occurs according to the right issuance policy set.

9. An electronic device, comprising: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the order management-based enterprise customer right gift full-link tracking method of any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1-9. The computer program is executed by the processor to implement the order management-based enterprise customer right gift full-link tracking method of any one of claims 1 to 7.

Images