A module recommendation method and system for a business system

Abstract

The present invention relates to the technical field of business systems, and in particular to a module recommendation method and system for a business system. The present invention constructs a data knowledge graph and a module knowledge graph according to data in a business system and module data, respectively, and forms a business process multivariate knowledge graph. The present invention uses a Seq2Seq algorithm model for training according to historical system operations of business personnel to obtain an operation sequence model, and outputs a standard operation sequence through the operation sequence model to reassign the relationship between data nodes and model nodes in the business process multivariate knowledge graph, thereby optimizing the business process multivariate knowledge graph and constructing a business system module recommendation model. The model is recommended for data names and operation steps input by business personnel, so that business personnel can choose modules more clearly when they are not familiar with system operations, and can also help business personnel clarify operation steps and data input situations.

Description

Technical Field

[0001] The present invention relates to the technical field of business systems, and in particular to a module recommendation method and system for a business system. Background Art

[0002] The current system faces the challenges of huge data volume and high complexity of hierarchical structure, which leads to opacity and difficulty in operation in user experience. Users may feel confused when using such a system. For example, when it comes to cross-module interaction, the information search, association and operation process are particularly cumbersome and difficult to understand. In particular, when new users operate, they are even more uncertain about how to input data and select modules.

[0003] To overcome these obstacles and improve the usability and efficiency of the system, a solution is urgently needed to clarify the use of the system. Summary of the invention

[0004] In view of the above-mentioned deficiencies in the prior art, the present invention aims to provide a module recommendation method and system for a business system, so as to overcome the obstacles in the use of the system and improve the usability and efficiency of the system.

[0005] In order to solve the above problems, the present invention adopts the following technical solutions:

[0006] In one aspect, the present invention provides a module recommendation method for a business system, comprising:

[0007] Build a data knowledge graph for the business system based on the data of the business system.

[0008] Build a module knowledge graph of the business system based on the module data of the business system.

[0009] A multi-dimensional knowledge graph of the business process is constructed based on the data knowledge graph and the module knowledge graph.

[0010] According to the operation of the business personnel on the business system, the click stream with timestamp is collected from the data entry to the result obtained through the module, and the operation sequence including the click position or content and the name of the input data and the evaluation of whether the result obtained by the module lacks the input data is obtained, and the training data set is constructed; the operation sequence model is obtained by training the Seq2Seq algorithm model, and the relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the result obtained by the module lacks the input data is obtained through the operation sequence model, the evaluation required by the module is set, and the standard operation sequence analyzed by each module is output through the operation sequence model.

[0011] Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained by evaluating different requirements in different modules. The relationship between each data name and the model and the operation order in the standard operation sequence are determined according to the standard operation sequence under the evaluation of different requirements of different modules. Based on the relationship between each data type and the model, the relationship between each data node in the data type of the data knowledge graph in the business process multi-knowledge graph and each module node in the module knowledge graph is reassigned to obtain an optimized business process multi-knowledge graph.

[0012] Construct a business system module recommendation model based on the operation sequence model and the optimized business process multi-knowledge graph.

[0013] Based on the data names and operation steps input by business personnel, the business system module recommendation model is used to calculate the possible recommended operation sequences and matching module functions of business personnel, and the module ranking is recommended to business personnel. Different prompts are marked for the data names in the business personnel's operation page according to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes.

[0014] As an implementable method, constructing a data knowledge graph of a business system based on the data of the business system includes:

[0015] Extract data from business systems using APIs, direct database connections, or log files, and perform data cleansing.

[0016] The node H of the data knowledge graph is defined by the data name.

[0017] The relationship R in the data knowledge graph is defined by the relationship between nodes.

[0018] Define the node's attribute S with the node's type.

[0019] According to the defined data knowledge graph node H, relationship R and node attribute S, the data of the business system after data cleaning is mapped to construct a data knowledge graph (H, S, R).

[0020] As an implementable method, constructing a module knowledge graph of a business system based on module data of the business system includes:

[0021] Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data according to each module.

[0022] Define the node V of the module knowledge graph with the module name.

[0023] The relationship K in the module knowledge graph is defined by the relationship between the node V and the data name in the module data.

[0024] The attribute P of the node V is defined by the function of the module.

[0025] The module data is mapped according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct a data knowledge graph (V, P, K).

[0026] As an implementable method, constructing a multi-knowledge graph of a business process based on the data knowledge graph and the module knowledge graph includes:

[0027] Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

[0028] As an implementable method, the step of obtaining the operation sequence model by training the Seq2Seq algorithm model includes:

[0029] The input of the Seq2Seq algorithm model is a click stream with a timestamp, and the output is a score level of the evaluation of whether the module results in the lack of input data.

[0030] The score X for evaluating whether the result obtained by the module is missing input data includes: X≤1: there is a lot of missing data, and the overall impact is large; 1<X≤2: there is some missing data, and the overall impact is large; 2<X≤3: there is some missing data, and the overall impact is small; 3<X≤4: there is some missing data, and there is no impact; 4<X≤5: there is no missing data, and there is no impact.

[0031] The loss function of the Seq2Seq algorithm model .

[0032] in, For the evaluation of manual annotation results, is the model output, are model parameters, is the regularization coefficient, and m is the operation sequence with timestamp.

[0033] As an implementable method, the optimized business process multivariate knowledge graph includes:

[0034] Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained by evaluating the different requirements of different modules. The top three standard operation sequences with the highest scores and the least data names are selected from the standard operation sequences for the evaluation of each requirement in each different module in the standard operation sequence as the screened standard operation sequences for the evaluation of each requirement in different modules.

[0035] Based on the standard operation sequence after screening based on the evaluation of each requirement in different modules, the relationship K between the node H and the node V in the standard operation sequence after screening is re-assigned according to its evaluation level, and the relationship K between the node H and the node V that appears for the first time is re-assigned based on the following order:

[0036] (1) The necessary relationship is assigned based on the relationship K between the node H and the node V in the standard operation sequence after screening with the evaluation level of 2<X≤3.

[0037] (2) The relationship K between the node H and the node V in the standard operation sequence after screening based on the evaluation level 3<X≤4 is assigned an important relationship.

[0038] (3) A non-essential relationship is assigned to the relationship K between the node H and the node V in the standard operation sequence after screening based on the evaluation level of 4<X≤5.

[0039] (4) The relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level X≤2 is assigned an unimportant relationship.

[0040] Based on the re-assignment of the relationship K between the node H and the node V, the relationship K in the business process multi-knowledge graph is adjusted to obtain an optimized business process multi-knowledge graph.

[0041] As an implementable method, based on the data name and operation steps input by the business personnel, the possible recommended operation sequence of the business personnel and the functions of the matching modules are calculated by the business system module recommendation model, and the module ranking recommendation for the business personnel includes:

[0042] The standard operation sequence after screening with the relationship K between the node H and the node V in different modules as the necessary relationship is used as the recommended operation sequence of different modules of the business system module recommendation model.

[0043] The data names and operation steps entered by the business personnel are matched with the nodes H and the order in the recommended operation sequences of different modules. The modules corresponding to the three recommended operation sequences with the highest matching degree are recommended to the business personnel, and the functional attributes of the modules are attached to the recommended modules in the display interface.

[0044] As an implementable method, the prompts for marking different data names in the operation page of the business personnel according to the module selected by the business personnel and based on the connection relationship between the data node and the module node include:

[0045] According to the module selected by the business personnel, a recommended operation sequence corresponding to the module is obtained, and a node H in the recommended operation that has a necessary relationship with the module is marked as "recommended required".

[0046] According to the module selected by the business personnel, a screened standard operation sequence is obtained for the evaluation of each requirement in the module, and nodes H in the screened standard operation sequence that have an important relationship with the module are marked as "recommended to be filled in"; nodes H in the screened standard operation sequence that have a non-essential relationship with the module are marked as "optional"; nodes H in the screened standard operation sequence that have an unimportant relationship with the module are marked as "not necessary to be filled in unless there are other requirements".

[0047] On the other hand, the present invention provides a module recommendation system for a business system, including a data knowledge graph generation module, a module knowledge graph generation module, a business process multi-knowledge graph generation module, a standard operation sequence generation module, a business process multi-knowledge graph optimization module, a business system module recommendation model construction module and a recommendation module.

[0048] The data knowledge graph generation module is used to construct a data knowledge graph of the business system based on the data of the business system.

[0049] The module knowledge graph generation module is used to construct the module knowledge graph of the business system based on the module data of the business system.

[0050] The business process multi-knowledge graph generation module is used to construct a business process multi-knowledge graph based on the data knowledge graph and the module knowledge graph.

[0051] The standard operation sequence generation module is used to collect click streams with timestamps from data entry to the process of obtaining results through modules according to the operations of business personnel on the business system, obtain an operation sequence including click positions or contents and names with input data, and an evaluation of whether the results obtained by the modules lack input data, and construct a training data set; obtain an operation sequence model by training a Seq2Seq algorithm model, obtain the relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the results obtained by the modules lack input data through the operation sequence model, set the evaluation required by the module, and output the standard operation sequence analyzed by each module through the operation sequence model.

[0052] The business process multi-knowledge graph optimization module is used to obtain a standard operation sequence under the evaluation of different requirements of different modules based on the operation sequence model through the evaluation of different requirements in different modules, determine the relationship between each data name in the standard operation sequence and the model and the operation order according to the standard operation sequence under the evaluation of different requirements of different modules, and reassign the relationship between each data node in the data type of the data knowledge graph in the business process multi-knowledge graph and each module node in the module knowledge graph based on the relationship between each data type and the model to obtain an optimized business process multi-knowledge graph.

[0053] The business system module recommendation model construction module is used to construct a business system module recommendation model based on an operation sequence model and an optimized business process multivariate knowledge graph.

[0054] The recommendation module is used to calculate the possible recommended operation sequence of the business personnel and the functions of the matching modules through the business system module recommendation model based on the data names and operation steps input by the business personnel, and recommend the module ranking to the business personnel. According to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes, different prompts are marked for the data names in the business personnel's operation page.

[0055] As an implementable method, constructing a data knowledge graph of a business system based on the data of the business system includes:

[0056] Extract data from business systems using APIs, direct database connections, or log files, and perform data cleansing.

[0057] The node H of the data knowledge graph is defined by the data name.

[0058] The relationship R in the data knowledge graph is defined by the relationship between nodes.

[0059] Define the node's attribute S with the node's type.

[0060] According to the defined data knowledge graph node H, relationship R and node attribute S, the data of the business system after data cleaning is mapped to construct a data knowledge graph (H, S, R).

[0061] The construction of the module knowledge graph of the business system according to the module data of the business system includes:

[0062] Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data according to each module.

[0063] Define the node V of the module knowledge graph with the module name.

[0064] The relationship K in the module knowledge graph is defined by the relationship between the node V and the data name in the module data.

[0065] The attribute P of the node V is defined by the function of the module.

[0066] The module data is mapped according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct a data knowledge graph (V, P, K).

[0067] The constructing of a business process multivariate knowledge graph according to the data knowledge graph and the module knowledge graph includes:

[0068] Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

[0069] The beneficial effects of the present invention are as follows: a module recommendation method and system for a business system of the present invention respectively constructs a data knowledge graph and a module knowledge graph according to the data and module data in the business system, and forms a business process multi-knowledge graph, and according to the historical system operations of business personnel, a Seq2Seq algorithm model is used for training to obtain an operation sequence model, and a standard operation sequence is output through the operation sequence model to reassign the relationship between data nodes and model nodes in the business process multi-knowledge graph, thereby optimizing the business process multi-knowledge graph; a business system module recommendation model is constructed based on the optimized business process multi-knowledge graph, and a model can be recommended for data names and operation steps input by business personnel, so that business personnel can make clearer choices of modules when they are not familiar with system operations, and the business system module can also prompt the operation steps, and can also prompt the input data names with different priorities, which can help business personnel clarify the operation steps and data input. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 The present invention is a flowchart of a module recommendation method of a business system.

[0071] Figure 2 A schematic diagram of a module recommendation system of a business system of the present invention. DETAILED DESCRIPTION

[0072] The present invention is further described in detail below in conjunction with specific embodiments.

[0073] It should be noted that these embodiments are only used to illustrate the present invention rather than to limit the present invention. Simple improvements to the method based on the concept of the present invention are within the scope of protection claimed by the present invention.

[0074] See also Figure 1 , which is a module recommendation method for a business system, including:

[0075] S100. Construct a data knowledge graph of the business system based on the data of the business system.

[0076] Extract data from the business system and preprocess it to prepare for building a data knowledge graph.

[0077] Extract data from business systems using APIs, direct database connections, or log files, and perform data cleansing.

[0078] Assume the extracted data set is , where d i Represents the i-th data record.

[0079] Remove duplicate data.

[0080] Fill missing values: For missing attributes, use the mean, median, or mode to fill in.

[0081]

[0082] .

[0083] The node H of the data knowledge graph is defined by the data name.

[0084] Identify the key entities in the business system, such as the specific names of users, products, transactions, etc., as nodes H of the knowledge graph.

[0085] Let the node set be H = {h1, h2, ..., hm}.

[0086] The relationship R in the data knowledge graph is defined by the relationship between nodes.

[0087] Analyze data and identify relationships between nodes, such as "Xiao Wang purchases goods."

[0088] Let the relation set be R = {r1, r2, ..., rj}.

[0089] Relationship assignment: Use edges to represent the relationship between nodes, such as (hi, rj, hk) means that there is a relationship rj (purchase) between nodes hi (Xiao Wang) and hk (product).

[0090] Define the node's attribute S with the node's type.

[0091] Define attributes for each node, such as the specific name of the user, product, transaction, etc., the specific type of user, product, transaction, etc., and let the attribute set be S = {s1, s2, ..., sk}.

[0092] In the field of audit technology, nodes may include "account name", "transaction name", etc.; attributes may include "user data", "transaction data", etc.; relationships may include "transfer", "attribution", etc.

[0093] According to the defined data knowledge graph node H, relationship R and node attribute S, the data of the business system after data cleaning is mapped to construct a data knowledge graph (H, S, R).

[0094] S200. Construct a module knowledge graph of the business system based on the module data of the business system.

[0095] Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data based on each module; collect the data types involved in the functions and operation of each module of the business system.

[0096] Define the node V of the module knowledge graph with the module name. For example, the module name is profit calculation, sales calculation, etc.

[0097] The relationship K in the module knowledge graph is defined by the relationship between the node V and the data name in the module data. For example, the net profit calculation module has a necessary relationship with the data names of cost price and selling price, or a non-necessary relationship with the selling cycle, etc. The relationship here is determined based on the importance of the data involved in the module or the function of the module.

[0098] The attribute P of the node V is defined by the function of the module. A module may have multiple functions, so the data P of the module may have multiple values.

[0099] The module data is mapped according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct a data knowledge graph (V, P, K).

[0100] The module knowledge graph helps to understand the relationship and dependency between modules and data, especially in the auditing industry, which helps to ensure the accuracy and compliance of data.

[0101] S300. Construct a multi-knowledge graph of business processes based on the data knowledge graph and the module knowledge graph.

[0102] Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

[0103] S400. According to the operation of the business personnel on the business system, a click stream with a timestamp is collected from data entry to the process of obtaining results through the module, and an operation sequence including the click position or content and the name of the input data and the evaluation of whether the result obtained by the module lacks input data is obtained (the evaluation of whether the result obtained by the module lacks input data here is a manually labeled result), and a training data set is constructed; an operation sequence model is obtained by training the Seq2Seq algorithm model, and the relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the result obtained by the module lacks input data is obtained through the operation sequence model, the evaluation required by the module is set, and the standard operation sequence analyzed by each module is output through the operation sequence model.

[0104] Among them, the operation sequence model obtained by training the Seq2Seq algorithm model includes:

[0105] The Seq2Seq algorithm model is an important RNN model, also known as the Encoder-Decoder model, which can be understood as an N×M model. The model consists of two parts: Encoder is used to encode sequence information, encoding sequence information of any length into a vector c. The Decoder is a decoder. After the decoder obtains the context information vector c, it can decode the information and output it as a sequence. The present invention utilizes the ability of the Seq2Seq algorithm model to process sequence information of any length to realize the processing of operation sequences of different lengths formed by data input due to different modules or their functions and different operation complexities in the present invention.

[0106] The input of the Seq2Seq algorithm model is a clickstream / operation sequence with a timestamp, and the output is a score level of the evaluation of whether the module's results are missing input data;

[0107] The scores X for evaluating whether the module results are missing input data include: X≤1: there are many missing data, and the overall impact is large; 1<X≤2: there are some missing data, and the overall impact is large; 2<X≤3: there are some missing data, and the overall impact is small; 3<X≤4: there are some missing data, and there is no impact; 4<X≤5: there is no missing data, and there is no impact;

[0108] Loss function of Seq2Seq algorithm model ;

[0109] in, For the evaluation of manual annotation results, is the model output, are model parameters, is the regularization coefficient, and m is the click stream / operation sequence with timestamp.

[0110] S500. Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained by evaluating different requirements in different modules. The relationship between each data name in the standard operation sequence and the model and the operation order are determined according to the standard operation sequence under the evaluation of different requirements of different modules. Based on the relationship between each data type and the model, the relationship between each data node in the data type of the data knowledge graph in the business process multi-knowledge graph and each module node in the module knowledge graph is reassigned to obtain an optimized business process multi-knowledge graph.

[0111] Among them, the optimized business process multi-knowledge graph includes:

[0112] Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained by evaluating the different requirements of different modules. The top three standard operation sequences with the highest scores and the least data names are selected from the standard operation sequences for the evaluation of each requirement in each different module in the standard operation sequence as the screened standard operation sequences for the evaluation of each requirement in different modules.

[0113] Based on the standard operation sequence after screening based on the evaluation of each requirement in different modules, the relationship K between the node H and the node V in the standard operation sequence after screening is re-assigned according to its evaluation level, and the relationship K between the node H and the node V that appears for the first time is re-assigned based on the following order:

[0114] (1) Assign the necessary relationship based on the relationship K between node H and node V in the standard operation sequence after screening with an evaluation level of 2<X≤3;

[0115] (2) assigning an important relationship to the relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level of 3<X≤4;

[0116] (3) assigning a non-essential relationship to the relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level of 4<X≤5;

[0117] (4) The relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level X≤2 is assigned an unimportant relationship;

[0118] Based on the re-assignment of the relationship K between the node H and the node V, the relationship K in the business process multi-knowledge graph is adjusted to obtain an optimized business process multi-knowledge graph.

[0119] S600. Construct a business system module recommendation model based on the operation sequence model and the optimized business process multi-knowledge graph.

[0120] The business system module recommendation model is equivalent to including an operation sequence model, which can output standard operation sequences under the evaluation of different requirements of different modules, as well as the standard operation sequences after screening of the evaluation of each requirement in different modules after processing.

[0121] The optimized business process multivariate knowledge graph redefines the relationship K between a model node V and its related data node H based on the screened standard operation sequence.

[0122] S700. Based on the data names and operation steps input by the business personnel, the possible recommended operation sequences and matching module functions of the business personnel are calculated through the business system module recommendation model, and the module ranking is recommended to the business personnel. Different prompts are marked for the data names in the business personnel operation page according to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes.

[0123] Among them, based on the data name and operation steps input by the business personnel, the possible recommended operation sequence of the business personnel and the functions of the matching modules are calculated through the business system module recommendation model, and the module ranking recommendation for the business personnel includes:

[0124] The standard operation sequence after screening with the relationship K between the node H and the node V in different modules as the necessary relationship is used as the recommended operation sequence of different modules of the business system module recommendation model;

[0125] The data names and operation steps entered by the business personnel are matched with the nodes H and the order in the recommended operation sequences of different modules. The modules corresponding to the three recommended operation sequences with the highest matching degree are recommended to the business personnel, and the functional attributes of the modules are attached to the recommended modules in the display interface.

[0126] Among them, the tips for marking different data names in the operation page of the business personnel according to the module selected by the business personnel and based on the connection relationship between the data node and the module node include:

[0127] According to the module selected by the business personnel, the recommended operation sequence corresponding to the module is obtained, and the node H in the recommended operation that has a necessary relationship with the module is marked as "recommended required";

[0128] According to the module selected by the business personnel, a screened standard operation sequence is obtained for the evaluation of each requirement in the module, and nodes H in the screened standard operation sequence that have an important relationship with the module are marked as "recommended to be filled in"; nodes H in the screened standard operation sequence that have a non-essential relationship with the module are marked as "optional"; nodes H in the screened standard operation sequence that have an unimportant relationship with the module are marked as "not necessary to be filled in unless there are other requirements".

[0129] join Figure 2 , which is a module recommendation system for a business system, including a data knowledge graph generation module 100, a module knowledge graph generation module 200, a business process multi-knowledge graph generation module 300, a standard operation sequence generation module 400, a business process multi-knowledge graph optimization module 500, a business system module recommendation model construction module 600 and a recommendation module 700.

[0130] The data knowledge graph generation module 100 is used to construct a data knowledge graph of a business system based on the data of the business system.

[0131] The module knowledge graph generation module 200 is used to construct a module knowledge graph of a business system based on the module data of the business system.

[0132] The business process multi-knowledge graph generation module 300 is used to construct a business process multi-knowledge graph based on the data knowledge graph and the module knowledge graph.

[0133] The standard operation sequence generation module 400 is used to collect click streams with timestamps from data entry to the result obtained by the module according to the operation of the business personnel on the business system, and obtain an operation sequence including the click position or content and the name of the input data and the evaluation of whether the result obtained by the module lacks input data, so as to construct a training data set; obtain an operation sequence model by training the Seq2Seq algorithm model, obtain the relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the result obtained by the module lacks input data through the operation sequence model, set the evaluation required by the module, and output the standard operation sequence analyzed by each module through the operation sequence model.

[0134] The business process multi-knowledge graph optimization module 500 is used to obtain the standard operation sequence under the evaluation of different requirements of different modules based on the operation sequence model through the evaluation of different requirements in different modules, determine the relationship between each data name and the model and the operation order in the standard operation sequence according to the standard operation sequence under the evaluation of different requirements of different modules, and reassign the relationship between each data node in the data type of the data knowledge graph in the business process multi-knowledge graph and each module node in the module knowledge graph based on the relationship between each data type and the model to obtain an optimized business process multi-knowledge graph.

[0135] The business system module recommendation model construction module 600 is used to construct a business system module recommendation model based on the operation sequence model and the optimized business process multivariate knowledge graph.

[0136] The recommendation module 700 is used to calculate the possible recommended operation sequence of the business personnel and the functions of the matching modules through the business system module recommendation model based on the data names and operation steps input by the business personnel, and recommend the module ranking to the business personnel. According to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes, different prompts are marked for the data names in the business personnel's operation page.

[0137] Among them, building a data knowledge graph of the business system based on the data of the business system includes:

[0138] Extract data from business systems using APIs, direct database connections, or log files, and perform data cleansing.

[0139] The node H of the data knowledge graph is defined by the data name.

[0140] The relationship R in the data knowledge graph is defined by the relationship between nodes.

[0141] Define the node's attribute S with the node's type.

[0142] According to the defined data knowledge graph node H, relationship R and node attribute S, the data of the business system after data cleaning is mapped to construct a data knowledge graph (H, S, R).

[0143] Building a module knowledge graph of a business system based on the module data of the business system includes:

[0144] Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data according to each module.

[0145] Define the node V of the module knowledge graph with the module name.

[0146] Define the relationship K in the module knowledge graph based on the relationship between the node V and the data name in the module data;

[0147] The attribute P of the node V is defined by the function of the module.

[0148] The module data is mapped according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct a data knowledge graph (V, P, K).

[0149] Building a multi-knowledge graph of business processes based on the data knowledge graph and the module knowledge graph includes:

[0150] Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

[0151] Among them, the operation sequence model obtained by training the Seq2Seq algorithm model includes:

[0152] The input of the Seq2Seq algorithm model is a click stream with a timestamp, and the output is a score level of the evaluation of whether the module's results are missing input data.

[0153] The score X for evaluating whether the module results are missing input data includes: X≤1: there is a lot of missing data, and the overall impact is large; 1<X≤2: there is some missing data, and the overall impact is large; 2<X≤3: there is some missing data, and the overall impact is small; 3<X≤4: there is some missing data, and there is no impact; 4<X≤5: there is no missing data, and there is no impact.

[0154] Loss function of Seq2Seq algorithm model .

[0155] in, For the evaluation of manual annotation results, is the model output, are model parameters, is the regularization coefficient, and m is the operation sequence with timestamp.

[0156] Among them, the optimized business process multi-knowledge graph includes:

[0157] Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained by evaluating the different requirements of different modules. The top three standard operation sequences with the highest scores and the least data names are selected from the standard operation sequences for the evaluation of each requirement in each different module in the standard operation sequence as the screened standard operation sequences for the evaluation of each requirement in different modules.

[0158] Based on the screened standard operation sequence of the evaluation of each requirement in different modules, the relationship K between the node H and the node V in the screened standard operation sequence is reassigned according to its evaluation level, and the relationship K between the node H and the node V that appears for the first time is reassigned based on the following order.

[0159] (1) The necessary relationship is assigned based on the relationship K between the node H and the node V in the standard operation sequence after screening with the evaluation level of 2<X≤3.

[0160] (2) The relationship K between the node H and the node V in the standard operation sequence after screening based on the evaluation level 3<X≤4 is assigned an important relationship.

[0161] (3) A non-essential relationship is assigned to the relationship K between the node H and the node V in the standard operation sequence after screening based on the evaluation level of 4<X≤5.

[0162] (4) The relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level X≤2 is assigned an unimportant relationship.

[0163] Based on the re-assignment of the relationship K between the node H and the node V, the relationship K in the business process multi-knowledge graph is adjusted to obtain an optimized business process multi-knowledge graph.

[0164] Among them, based on the data name and operation steps input by the business personnel, the possible recommended operation sequence of the business personnel and the functions of the matching modules are calculated through the business system module recommendation model, and the module ranking recommendation for the business personnel includes:

[0165] The standard operation sequence after screening with the relationship K between the node H and the node V in different modules as the necessary relationship is used as the recommended operation sequence of different modules of the business system module recommendation model.

[0166] The data names and operation steps entered by the business personnel are matched with the nodes H and the order in the recommended operation sequences of different modules. The modules corresponding to the three recommended operation sequences with the highest matching degree are recommended to the business personnel, and the functional attributes of the modules are attached to the recommended modules in the display interface.

[0167] Among them, the tips for marking different data names in the operation page of the business personnel according to the module selected by the business personnel and based on the connection relationship between the data node and the module node include:

[0168] According to the module selected by the business personnel, a recommended operation sequence corresponding to the module is obtained, and a node H in the recommended operation that has a necessary relationship with the module is marked as "recommended required".

[0169] According to the module selected by the business personnel, a screened standard operation sequence is obtained for the evaluation of each requirement in the module, and nodes H in the screened standard operation sequence that have an important relationship with the module are marked as "recommended to be filled in"; nodes H in the screened standard operation sequence that have a non-essential relationship with the module are marked as "optional"; nodes H in the screened standard operation sequence that have an unimportant relationship with the module are marked as "not necessary to be filled in unless there are other requirements".

[0170] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described with reference to the preferred embodiments of the present invention, it should be understood by those skilled in the art that various changes may be made in form and details without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. A module recommendation method for a business system, characterized in that: include: Build a data knowledge graph of the business system based on the data of the business system; Build a module knowledge graph of the business system based on the module data of the business system; Construct a multi-knowledge graph of business processes according to the data knowledge graph and the module knowledge graph; According to the operation of the business personnel on the business system, the click stream with timestamp is collected from the process of data entry to the result obtained through the module, and the operation sequence including the click position or content and the name with input data and the evaluation of whether the result obtained by the module lacks input data is obtained, so as to build a training data set; The operation sequence model is obtained by training the Seq2Seq algorithm model. The relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the module results lack input data is obtained through the operation sequence model. The evaluation required by the module is set, and the standard operation sequence analyzed by each module is output through the operation sequence model. Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained through the evaluation of different requirements in different modules. The relationship between each data name and the module and the operation order in the standard operation sequence are determined according to the standard operation sequence under the evaluation of different requirements of different modules. Based on the relationship between each data type and the module, the relationship between each data node in the data type of the data knowledge graph in the business process multivariate knowledge graph and each module node in the module knowledge graph is reassigned to obtain an optimized business process multivariate knowledge graph; Build a business system module recommendation model based on the operation sequence model and optimized business process multi-knowledge graph; Based on the data names and operation steps input by business personnel, the business system module recommendation model is used to calculate the possible recommended operation sequences and matching module functions of business personnel, and the module ranking is recommended to business personnel. Different prompts are marked for the data names in the business personnel's operation page according to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes.

2. The module recommendation method of the business system according to claim 1, characterized in that: The data knowledge graph of the business system constructed according to the data of the business system includes: Extract data from business systems using APIs, direct database connections or log files, and perform data cleansing; Define the node H of the data knowledge graph with the data name; Define the relationship R in the data knowledge graph based on the relationship between nodes; Define the node attribute S by the node type; According to the defined data knowledge graph node H, relationship R and node attribute S, the data of the business system after data cleaning is mapped to construct a data knowledge graph (H, S, R).

3. The module recommendation method of the business system according to claim 2, characterized in that: The construction of the module knowledge graph of the business system according to the module data of the business system includes: Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data according to each module; Define the node V of the module knowledge graph with the module name; Define the relationship K in the module knowledge graph based on the relationship between the node V and the data name in the module data; Define the attribute P of the node V with the function of the module; The module data is mapped according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct the module knowledge graph (V, P, K).

4. The module recommendation method of the business system according to claim 3, characterized in that: The constructing of a business process multivariate knowledge graph according to the data knowledge graph and the module knowledge graph includes: Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

5. The module recommendation method of the business system according to claim 4, characterized in that: The operation sequence model obtained by training the Seq2Seq algorithm model includes: The input of the Seq2Seq algorithm model is a click stream with a timestamp, and the output is a score level of the evaluation of whether the module results in the lack of input data; The score X of the evaluation of whether the result obtained by the module is missing input data includes: X≤1: there is a lot of missing data, and the overall impact is large; 1<X≤2: there is some missing data, and the overall impact is large; 2<X≤3: there is some missing data, and the overall impact is small; 3<X≤4: there is some missing data, and there is no impact; 4<X≤5: there is no missing data, and there is no impact; The loss function of the Seq2Seq algorithm model ; in, For the evaluation of manual annotation results, is the model output, are model parameters, is the regularization coefficient, and m is the operation sequence with timestamp.

6. The module recommendation method of the business system according to claim 5, characterized in that: The optimized business process multivariate knowledge graph includes: Based on the operation sequence model, the standard operation sequence under the evaluation of different requirements of different modules is obtained through the evaluation of different requirements of different modules. The top three standard operation sequences with the highest scores and the least data names are selected from the standard operation sequences of the evaluation of each requirement in each different module in the standard operation sequence as the selected standard operation sequences for the evaluation of each requirement in different modules; Based on the standard operation sequence after screening based on the evaluation of each requirement in different modules, the relationship K between the node H and the node V in the standard operation sequence after screening is re-assigned according to its evaluation level, and the relationship K between the node H and the node V that appears for the first time is re-assigned based on the following order: (1) Assign the necessary relationship based on the relationship K between node H and node V in the standard operation sequence after screening with an evaluation level of 2<X≤3; (2) assigning an important relationship to the relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level of 3<X≤4; (3) assigning a non-essential relationship to the relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level of 4<X≤5; (4) The relationship K between node H and node V in the standard operation sequence after screening based on the evaluation level X≤2 is assigned an unimportant relationship; Based on the re-assignment of the relationship K between the node H and the node V, the relationship K in the business process multi-knowledge graph is adjusted to obtain an optimized business process multi-knowledge graph.

7. The module recommendation method of the business system according to claim 6, characterized in that: The method of calculating the possible recommended operation sequence of the business personnel and the functions of the matching modules through the business system module recommendation model based on the data name and operation steps input by the business personnel, and making the module ranking recommendation to the business personnel includes: The standard operation sequence after screening with the relationship K between the node H and the node V in different modules as the necessary relationship is used as the recommended operation sequence of different modules of the business system module recommendation model; The data names and operation steps entered by the business personnel are matched with the nodes H and the order in the recommended operation sequences of different modules. The modules corresponding to the three recommended operation sequences with the highest matching degree are recommended to the business personnel, and the functional attributes of the modules are attached to the recommended modules in the display interface.

8. The module recommendation method of the business system according to claim 7, characterized in that: The prompts for marking different data names in the operation page of the business personnel according to the module selected by the business personnel and based on the connection relationship between the data node and the module node include: According to the module selected by the business personnel, the recommended operation sequence corresponding to the module is obtained, and the node H in the recommended operation that has a necessary relationship with the module is marked as "recommended required"; According to the module selected by the business personnel, a filtered standard operation sequence for the evaluation of each requirement in the module is obtained, and nodes H in the filtered standard operation sequence that have an important relationship with the module are marked as "recommended to be filled in"; nodes H in the filtered standard operation sequence that have a non-essential relationship with the module are marked as "optional to be filled in"; nodes H in the filtered standard operation sequence that have an unimportant relationship with the module are marked as "not necessary to be filled in unless there are other requirements".

9. A module recommendation system for a business system, characterized in that: It includes data knowledge graph generation module, module knowledge graph generation module, business process multi-knowledge graph generation module, standard operation sequence generation module, business process multi-knowledge graph optimization module, business system module recommendation model construction module and recommendation module; The data knowledge graph generation module is used to construct a data knowledge graph of the business system based on the data of the business system; The module knowledge graph generation module is used to construct a module knowledge graph of the business system according to the module data of the business system; The business process multi-knowledge graph generation module is used to construct a business process multi-knowledge graph based on the data knowledge graph and the module knowledge graph; The standard operation sequence generation module is used to collect click streams with timestamps from data entry to the process of obtaining results through the module according to the operation of the business personnel on the business system, obtain an operation sequence including the click position or content and the name of the input data, and an evaluation of whether the result obtained by the module lacks input data, and construct a training data set; The operation sequence model is obtained by training the Seq2Seq algorithm model. The relationship between the data name and time sequence in the operation sequence of each module and the evaluation of whether the module results lack input data is obtained through the operation sequence model. The evaluation required by the module is set, and the standard operation sequence analyzed by each module is output through the operation sequence model. The business process multivariate knowledge graph optimization module is used to obtain a standard operation sequence under the evaluation of different requirements of different modules based on the operation sequence model, determine the relationship between each data name and the module and the operation order in the standard operation sequence according to the standard operation sequence under the evaluation of different requirements of different modules, and reassign the relationship between each data node in the data type of the data knowledge graph in the business process multivariate knowledge graph and each module node in the module knowledge graph based on the relationship between each data type and the module, so as to obtain an optimized business process multivariate knowledge graph; The business system module recommendation model construction module is used to construct a business system module recommendation model based on the operation sequence model and the optimized business process multivariate knowledge graph; The recommendation module is used to calculate the possible recommended operation sequence of the business personnel and the functions of the matching modules through the business system module recommendation model based on the data names and operation steps input by the business personnel, and recommend the module ranking to the business personnel. According to the modules selected by the business personnel and based on the connection relationship between the data nodes and the module nodes, different prompts are marked for the data names in the business personnel's operation page.

10. The module recommendation system of the business system according to claim 9, characterized in that: The data knowledge graph of the business system constructed according to the data of the business system includes: Extract data from business systems using APIs, direct database connections or log files, and perform data cleansing; Define the node H of the data knowledge graph with the data name; Define the relationship R in the data knowledge graph based on the relationship between nodes; Define the node attribute S by the node type; Map the data of the business system after data cleaning according to the defined data knowledge graph node H, relationship R and node attribute S to build a data knowledge graph (H, S, R); The construction of the module knowledge graph of the business system according to the module data of the business system includes: Obtain each module in the business system, and determine the relevant data involved in the operation of each module as module data according to each module; Define the node V of the module knowledge graph with the module name; Define the relationship K in the module knowledge graph based on the relationship between the node V and the data name in the module data; Define the attribute P of the node V with the function of the module; Map the module data according to the node V, relationship K and node attribute P of the defined module knowledge graph to construct the module knowledge graph (V, P, K); The constructing of a business process multivariate knowledge graph according to the data knowledge graph and the module knowledge graph includes: Match the data name in the module data with the node H of the data knowledge graph, use the successfully matched node H as the node of the module knowledge graph, and establish a relationship K with the node V to merge and associate the data knowledge graph and the module knowledge graph to form a multi-knowledge graph of the business process (H, S, R, V, P, K).

Images