Methods for Identifying Risks in Engineering Project Audit

Abstract

The present invention relates to the field of audit management technology, and in particular to a method for identifying audit risks of engineering projects. The method for identifying audit risks of engineering projects comprises the following steps: step 1: constructing an operation set; step 2: preprocessing data in the operation set to obtain undetermined risk points and their values; step 3: constructing relationship edges between undetermined risk points to determine the direction of the relationship between risk points; step 4: outputting a causal relationship network of risk points, and providing an engineering project audit risk identification method that does not rely on professional auditors and can identify audit risks of engineering projects.

Description

Technical Field

[0001] The present invention relates to the technical field of audit management, and in particular to a method for identifying audit risks of engineering projects. Background Art

[0002] So far, the identification of audit risks in domestic and foreign engineering projects mainly relies on audit management personnel to review materials, obtain interviews and interpret them. This requires experienced professionals to make judgments based on the relationship between materials, does not fully utilize the information in the management information system, and is limited in scientificity and accuracy. How to discover audit management knowledge from data without relying on professional auditors and improve the efficiency of risk point identification has become an audit management problem that needs to be solved urgently. Solving this problem improves the reusability of enterprise data and can provide scientific support for audit departments and corporate decision-making. Summary of the invention

[0003] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a method for identifying project audit risks that does not rely on professional auditors and can identify project audit risks.

[0004] The technical solution adopted by the present invention to solve the technical problem is: a method for identifying risks in engineering project audit, comprising the following steps:

[0005] Step 1: Build an operation set;

[0006] Step 2: Preprocess the data in the operation set to obtain the pending risk points and their values;

[0007] Step 3: Build the relationship edges between the undetermined risk points and determine the direction of the relationship between the risk points;

[0008] Step 4: Output the risk point causal network.

[0009] In step 1, project life cycle data, project settlement data and audit report corpus are collected, and the calculation formula for constructing the operation set U is as follows:

[0010]

[0011] In the formula (1), n ​​is the total number of engineering projects in audit management, U represents the total set of operations, P1, P2…P n They are the 1st, 2nd,…nth engineering projects respectively.

[0012] In the step 2, text mining technology is used to pre-process the data in the operation set to obtain the pending risk points and their values;

[0013] The step 2 includes the following sub-steps:

[0014] 2-1: Segment the audit report corpus, remove Chinese stop words, and obtain the audit report keyword set W;

[0015] 2-2: Refer to the TF-IDF algorithm for any keyword w in W p Calculate weights and obtain keyword frequency TI p ;

[0016] 2-3: TI the keyword frequency in descending order p Sorting, screening the keywords whose total weight accounts for 80% as the names of pending risk points, forming the set of pending risk points R′;

[0017] 2-4: According to the audit business rules, obtain the pending risk point value P for each project from the project life cycle data and project settlement data j , if it meets the audit rules, it takes 0, otherwise it takes 1, which is expressed as follows:

[0018] P j =P j (R j ′1,R j ′2…R j ' k ) (6);

[0019] Among them, Publication (6) represents the k undetermined risk points and their values ​​of the jth project;

[0020] In step 2-1, the calculation formula of the audit report keyword set W is as follows:

[0021] W={W1,W2…W n}={w1,w2…w m}, W1∪W2…W n =W(2);

[0022]

[0023] In the above formula (2), n is the total number of engineering projects in audit management, W represents the total set of audit report keywords for n projects, W1, W2…W n are the key word sets of the audit reports corresponding to the 1st, 2nd, ...nth engineering projects respectively, m represents the total number of key words in W; in Publication (3), l is the number of key words in the audit report of the jth engineering project.

[0024] In step 2-2, the keyword frequency TI p The calculation formula is as follows:

[0025]

[0026] Among them, in Announcement (4), npj The keyword w p In W j The frequency of occurrence in k n pj It is W j Total number of keywords in |j:w p ∈W j | contains the keyword w p The number of engineering projects.

[0027] In step 2-3, the undetermined risk point set R′ is expressed as follows:

[0028] R′={R′1,R′2…R′ k} (5);

[0029] Among them, k in Publication (5) is the number of undetermined risk points.

[0030] The step three includes the following sub-steps:

[0031] 3-1: Construct a complete graph G(V,E) between audit risk points, V = {v1,v2…,v n}, where node v j ∈V is a node in G, whose attribute value is v j =(P1(R′ 1j ),P2(R′ 2j )…P n (R′ nj )}, n is the number of nodes, E={e1,e2…,e m}, where edge e k (v i ,v j )∈E is an edge in G, m is the number of edges;

[0032] 3-2: Screen the edges on the complete graph;

[0033] 3-3: Determine the direction of the edges on the complete graph.

[0034] The step 3-2 includes the following sub-steps:

[0035] 3-2-1: Input G = (V, E), let l = -1; G′ = G. Set O as the node set and initialize it to an empty set. Set O contains two subsets, O(i, j) and O(j, i), which are used to store the node pairs v i ,v j or node pair v j ,v i A set of nodes whose values ​​are independent of each other;

[0036] 3-2-2: l = l + 1;

[0037] 3-2-3: For any node pair v i ,v j , take out the node v i The neighbor node set adj(G′,v i ), and let the set K = adj(G′,v i )\v j ∪adj(G′,v i )\v i ;

[0038] 3-2-4: If |K|>l, check the nodes v one by one k Can we make the node v i ,v j The values ​​of are independent of each other. If possible, delete the edge between the two points. Otherwise, add K to the sets O(i,j) and O(j,i).

[0039] Repeat steps 3-2-2 to 3-2-4 until the size of all adjacent point sets in the graph is less than l.

[0040] The step 3-3 includes the following sub-steps:

[0041] 3-3-1: Input G′ and O set;

[0042] 3-3-2: For any node pair v i ,v j , where i, j are natural numbers, if its neighbors, Then update G′ and delete the undirected edge e(v i ,v k ) and e(v k ,v j ), and add directed edge e <v i ,v k > and e <v j ,v k >.

[0043] 3-3-3: Update G′ according to the following rules to convert the remaining undirected edges into directed edges:

[0044] Rule 1: If there are both e <v i ,v k > and e(v k ,v j ), then delete e(v k ,v j ), increase e <v j ,v k >.

[0045] Rule 2: If there are both e <vi ,v k > and e <v k ,v j >, then delete e(v i ,v j ), increase e <v i ,v j >.

[0046] Rule 3: If there is e <v i ,v k1 > and e <v k1 ,v j >, and e <v i ,v k2 > and e <v k2 ,v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i ,v j >.

[0047] Rule 4: If there is e <v i ,v k1 > and e <v k1 ,v k2 >, and e <v k1 ,v k2 > and e <v k2 ,v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i ,v j >.

[0048] The step 4 includes the following sub-steps:

[0049] 4-1: Output G′(V,E);

[0050] 4-2: Non-isolated nodes in the network G′ are audit risk points;

[0051] 4-3: The edges in the network G′ represent the causal relationship between risk points.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] 1. Using project life cycle data, project settlement data and audit report corpus, audit risks in engineering projects can be automatically identified;

[0054] 2. This method has the ability to process multi-source project audit data and risk identification, and provides risk point association and disposal guidance for risk management in audit management; it uses text processing technology to quantify project life cycle data, project settlement data and audit report corpus data, improves the efficiency of audit risk identification, and provides an effective method;

[0055] 3. Analyzing the network diagram constructed by risk points and their causal relationships can reveal the way risks spread and provide traceability of risk points and their causes. It is a guide for audit management business when professionals lack experience and plays an important role in project audit risk management. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of screening edges on a complete graph according to Embodiment 2 of the present invention.

[0057] Figure 2 It is a schematic diagram of determining the edge direction on a complete graph according to embodiment 2 of the present invention.

[0058] Figure 3 It is a schematic diagram of output G′(V, E) of embodiment 2 of the present invention. DETAILED DESCRIPTION

[0059] The embodiments of the present invention are further described below in conjunction with the accompanying drawings:

[0060] Example 1

[0061] like Figures 1 to 3 As shown, the following steps are included:

[0062] Step 1: Construct an operation set; collect project life cycle data, project settlement data, and audit report corpus from the management information system, engineering project construction site, and audit business site by exporting, taking photos, and filling in. The calculation formula for constructing the operation set U is as follows:

[0063]

[0064] In the formula (1), n ​​is the total number of engineering projects in audit management, U represents the total set of operations, P1, P2…P n They are the 1st, 2nd,…nth engineering projects respectively.

[0065] Step 2: pre-processing the data in the operation set to obtain the pending risk points and their values; in step 2, the data in the operation set is pre-processed using text mining technology to obtain the pending risk points and their values;

[0066] The step 2 includes the following sub-steps:

[0067] 2-1: Segment the audit report corpus, remove Chinese stop words, and obtain the audit report keyword set W; in step 2-1, the calculation formula of the audit report keyword set W is as follows:

[0068] W={W1,W2…W n}={w1,w2…w m}, W1∪W2…W n =W(2);

[0069]

[0070] In the above formula (2), n is the total number of engineering projects in audit management, W represents the total set of audit report keywords for n projects, W1, W2…W n are the key word sets of the audit reports corresponding to the 1st, 2nd, ...nth engineering projects respectively, m represents the total number of key words in W; in Publication (3), l is the number of key words in the audit report of the jth engineering project.

[0071] 2-2: Refer to the TF-IDF algorithm for any keyword w in W p Calculate weights and obtain keyword frequency TI p In step 2-2, the keyword frequency TI p The calculation formula is as follows:

[0072]

[0073] Among them, in Announcement (4), n pj The keyword w p In W j The frequency of occurrence in k n pj It is W j Total number of keywords in |j:w p ∈W j | contains the keyword w p The number of engineering projects.

[0074] The TF-IDF algorithm is a text mining algorithm used to evaluate the importance of a word to a document set or one of the documents in a corpus. Its assumption is that the importance of a word increases in direct proportion to the number of times it appears in a document, but at the same time it decreases in inverse proportion to the frequency of its appearance in the corpus. It is suitable for calculating the importance of words when there are a large number of documents in the corpus set. In this method, there are a large number of audit report texts, so this algorithm is highly applicable.

[0075] 2-3: TI the keyword frequency in descending order pSorting, screening the keywords whose total weight accounts for 80% as the names of the pending risk points, forming the pending risk point set R′; in the step 2-3, the pending risk point set R′ is expressed as follows:

[0076] R′={R′1,R′2…R′ k} (5);

[0077] Among them, k in Publication (5) is the number of undetermined risk points.

[0078] 2-4: According to the audit business rules, obtain the pending risk point value P for each project from the project life cycle data and project settlement data j , if it meets the audit rules, it takes 0, otherwise it takes 1, which is expressed as follows:

[0079] P j =P j (R′ j1 ,R′ j2 …R′ jk ) (6);

[0080] Among them, Publication (6) represents the k undetermined risk points and their values ​​for the j-th project.

[0081] Step 3: Construct relationship edges between undetermined risk points and determine the direction of the relationship between risk points; Step 3 includes the following sub-steps:

[0082] 3-1: Construct a complete graph G(V,E) between audit risk points for the project, where node v j ∈V is a node in G, whose attribute value is v j =(P1(R′ 1j ),P2(R′ 2j )…P n (R′ nj )}, n is the number of nodes. E={e1,e2…,e m}, where edge e k (v i ,v j )∈E is an edge in G and m is the number of edges.

[0083] 3-2: Screening the edges on the complete graph; the step 3-2 includes the following sub-steps:

[0084] 3-2-1: Input G = (V, E), let l = -1; G′ = G. Set O as the node set and initialize it to an empty set. Set O contains two subsets, O(i, j) and O(j, i), which are used to store the node pairs v i ,v j or node pair v j ,vi A set of nodes whose values ​​are independent of each other;

[0085] 3-2-2: l = l + 1;

[0086] 3-2-3: For any node pair v i ,v j , where i and j are node numbers, take out node v i The neighbor node set adj(G′,v i ), and let the set K = adj(G′,v i )\v j ∪adj(G′,v i )\v i ;

[0087] 3-2-4: If |K|>l, check the nodes v one by one k Can we make the node v i ,v j The values ​​of are independent of each other. If possible, delete the edge between the two points. Otherwise, add K to the sets O(i,j) and O(j,i).

[0088] Repeat steps 3-2-2 to 3-2-4 until the size of all neighbor point sets in the graph is less than l. The Fisher Z Test method is used for the independence test in 3-2-4;

[0089] Fisher Z Test method, also known as F test method, was proposed by British statistician Fisher, also known as variance homogeneity test. It mainly compares the variance of two groups of data to determine whether their precision is significantly different, and has a wide range of applications. In this method, this test can be used to quickly identify whether the difference in variance between two nodes is significant.

[0090] 3-3: Determine the edge direction on the complete graph. Step 3-3 includes the following sub-steps:

[0091] 3-3-1: Input G′ and O set;

[0092] 3-3-2: For any node pair v i ,v j If its neighbor, Then update G′ and delete the undirected edge e(v i ,v k ) and e(v k ,v j ), and add directed edge e <v i ,v k > and e <v j ,v k >.

[0093] 3-3-3: Update G′ according to the following rules to convert the remaining undirected edges into directed edges:

[0094] Rule 1: If there are both e <v i ,v k > and e(v k ,v j ), then delete e(v k ,v j ), increase e <v j ,v k >.

[0095] Rule 2: If there are both e <v i ,v k > and e <v k ,v j >, then delete e(v i ,v j ), increase e <v i ,v j >.

[0096] Rule 3: If there is e <v i ,v k1 > and e <v k1 ,v j >, and e <v i ,v k2 > and e <v k2 ,v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i ,v j >.

[0097] Rule 4: If there is e <v i ,v k1 > and e <v k1 ,v k2 >, and e <v k1 ,v k2 > and e <v k2 ,v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i ,v j >.

[0098] Step 4: Output the risk point causal relationship network. Step 4 includes the following sub-steps:

[0099] 4-1: Output G′(V,E);

[0100] 4-2: Non-isolated nodes in the network G′ are audit risk points;

[0101] 4-3: The edges in the network G′ represent the causal relationship between risk points.

[0102] Example 2

[0103] Step 1: Collect project life cycle data, project settlement data and third-party intermediary audit reports to build an operation set;

[0104] From the management information system, project construction site, and audit business site, we collected 154 medium and low voltage distribution network project life cycle data, project settlement data, and audit report corpus by exporting, taking photos, and filling in order to construct an operation set:

[0105]

[0106] Step 2: Use text mining technology to pre-process the data in the operation set to obtain the pending risk points and their values:

[0107] 2-1: Segment the audit report corpus, remove Chinese stop words, and obtain the audit report keyword set W:

[0108] W={W1,W2…W 154}={w1,w2…w 3019}, W1∪W2…W 154 =W(2);

[0109] W j ={files, information, projects, commencement and completion, visa form, time} (3);

[0110] 2-2: For any keyword w in W p Calculate weights and obtain keyword frequency TI p :

[0111] The frequency of the keyword "visa form":

[0112] 2-3: TI the keyword frequency in descending order p Sorting, screening keywords with a total weight of 80%, combining the audit business to determine the names of pending risk points, and forming a set of pending risk points R′:

[0113] R′

[0114] ={Project feasibility study depth, implementation before project approval, bidding documents, notice of winning the bid, contract signing, construction progress, supervision qualification test report, hidden works, list of materials supplied by Party A, accurate plan of materials supplied by Party A, list of materials supplied by Party B, legality of materials supplied by Party B, completion drawings, calculation of quantities, acceptance management, scrapping procedures, demolition plan, settlement process, settlement amount, final settlement}(5)

[0115] 2-4: According to the audit business rules, obtain the value of the pending risk point from the project life cycle data and project settlement data for each engineering project. If it meets the audit rules, it is 0, otherwise it is 1:

[0116] P j =P j (0,1,1,1,1,0,0,0,0,1,1,1,0,0,0,1,0,1,1,1,1,1,0) (6);

[0117] Among them, Publication (6) represents the k undetermined risk points and their values ​​of the jth project;

[0118] Step 3: Construct the relationship edges between the undetermined risk points and determine the direction of the relationship between the risk points:

[0119] 3-1: Construct a complete graph G(V,E) between audit risk points for the project, V = {v1,v2…,v n154}, where v j =(1,1,1,1,1,0,0,0,1,1,0,0,1,0,1,0,1,0,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,1,1 The output is as follows: ,1,1,1,1,1,1,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,0,0,0,0,0,0,1,1,1,1,1,0,0,1,1,1,1,0,0,1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,1,1,1,1,0,0,0,0,1,1,1,1,0,0,0,0,1,1,1,1,1).

[0120] 3-2: Reference Figure 1 , filter the edges on the complete graph;

[0121] 3-3: Reference Figure 2 , determine the edge direction on the complete graph;

[0122] Step 4: Output risk point causal relationship network:

[0123] (1) Output G′(V,E), such as Figure 3 As shown;

[0124] (2) The non-isolated nodes in the network G′ are non-audit risk points. There are 20 audit risk points in this application.

[0125] (3) The edges in the network G′ represent the causal relationship between risk points.

[0126] The diffusion pattern and key risk points of the audit risk subsystem can be obtained by comprehensively ranking the macro-modularity and micro-network structure indicators as shown in Table 1. There are 3 key risk points in subsystem 1, and their proximity centrality values ​​are relatively large. Therefore, the risk relationship distance of subsystem 1 (consisting of green nodes) is relatively close, that is, when a problem occurs in the key risk point, it will spread to its neighboring risk points. There are 3 key risk points in subsystem 2 (consisting of purple nodes), and their betweenness centrality values ​​are relatively large. This shows that there is a high probability of risk transmission in this subsystem, that is, if a problem occurs in the key risk point, other risk points in the subsystem will have problems one by one along the relationship link. Subsystem 3 (consisting of red nodes) has only 3 nodes, only 1 key risk point, and only edges pointing to it in the causal network. Therefore, the risk generation of this risk point is delayed, that is, when the risk of subsequent nodes occurs, the contract management risk is prominent.

[0127] Table 1 Subsystem risk propagation model

[0128]

[0129] The present invention takes into account the problem of risk point identification and relationship judgment in audit management business. The calculation results show that it is possible to identify audit risk points and risk relationships in engineering projects, improve the reuse of engineering project data, and provide decision-making guidance for audit management.

Claims

1. A method for identifying audit risks of engineering projects, characterized in that: The following steps are involved: Step 1: Construct an operation set; In step 1, collect project life cycle data, project settlement data and audit report corpus, and construct the operation set U using the following calculation formula: Where n in formula (1) is the total number of engineering projects in audit management, U represents the total operation set, P1, P2…P n They are the 1st, 2nd, …nth engineering projects respectively; Step 2: pre-processing the data in the operation set to obtain the pending risk points and their values; in step 2, the data in the operation set is pre-processed using text mining technology to obtain the pending risk points and their values; The step 2 includes the following sub-steps: 2-1: Segment the audit report corpus, remove Chinese stop words, and obtain the audit report keyword set W; 2-2: Refer to the TF-IDF algorithm for any keyword w in W p Calculate weights and obtain keyword frequency TI p ; 2-3: TI the keyword frequency in descending order p Sorting, screening the keywords whose total weight accounts for 80% as the names of pending risk points, forming the set of pending risk points R′; 2-4: Obtain the value P of the undetermined risk point from the project life cycle data and project settlement data for each engineering project j , if it meets the audit rules, it takes 0, otherwise it takes 1, which is expressed as follows: P j =P j (R′ j1 ,R′ j2 …R′ jk ) (6); Among them, formula (6) represents the k undetermined risk points and their values ​​of the jth project; Step 3: Construct relationship edges between undetermined risk points and determine the direction of the relationship between risk points; Step 3 includes the following sub-steps: 3-1: Construct a complete graph G(V,E) between audit risk points, where node v j ∈V is a node in G, whose attribute value is v j =(P1(R′ 1j ),P2(R′ 2j )…P n (R′ nj )}, n is the number of nodes, E={e1,e2…,e m }, where edge e k (v i ,v j )∈E is an edge in G, m is the number of edges; 3-2: Screening the edges on the complete graph; the step 3-2 includes the following sub-steps: 3-2-1: Input G = (V, E), let l = -1; G′ = G, set O is a set of nodes, and initialize it to an empty set. Set O includes two subsets, O(i, j) and O(j, i), which are used to store the node pairs v i ,v j or node pair v j ,v i Nodes whose values ​​are independent of each other; 3-2-2：l＝l+1； 3-2-3: For any node pair v i ,v j , where i and j are node numbers, take out node v i The neighbor node set adj(G′,v i ), and let the set K = adj(G′,v i )\v j ∪adj(G′,v i )\v i ; 3-2-4: If |K|>l, check the nodes v one by one k Can we make the node v i ,v j The values ​​of are independent of each other. If possible, delete the edge between the two points. Otherwise, add K to the sets O(i,j) and O(j,i). Repeat steps 3-2-2 to 3-2-4 until the size of all adjacent point sets in the graph is less than l; 3-3: Determine the edge direction on the complete graph; the step 3-3 includes the following sub-steps: 3-3-1: Input G′ and O set; 3-3-2: For any node pair v i ,v j , where i, j are natural numbers, if its neighbors, Then update G′ and delete the undirected edge e(v i, v k ) and e(v k, v j ), and add directed edge e <v i, v k > and e <v j, v k > 3-3-3: Update G′ according to the following rules to convert the remaining undirected edges into directed edges: Rule 1: If there are both e <v i, v k > and e(v k, v j ), then delete e(v k, v j ), increase e <v j, v k > Rule 2: If there are both e <v i, v k > and e <v k, v j >, then delete e(v i ,v j ), increase e <v i, v j > Rule 3: If there is e <v i, v k1 > and e <v k1, v j >, and e <v i, v k2 > and e <v k2, v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i, v j > Rule 4: If there is e <v i, v k1 > and e <v k1, v k2 >, and e <v k1, v k2 > and e <v k2, v j >, and v k1 and v k2 If they are not adjacent, delete e(v i ,v j ), increase e <v i, v j > Step 4: Output the risk point causal network.

2. The method for identifying project audit risks according to claim 1, characterized in that: In step 2-1, the calculation formula of the audit report keyword set W is as follows: W={W1,W2…W n }={w1,w2…w m },W1∪W2…W n =W(2); Where n in formula (2) is the total number of engineering projects in audit management, W represents the total set of audit report keywords for n projects, W1, W2…W n are the key words sets of the audit reports corresponding to the 1st, 2nd, ...nth engineering projects respectively, m represents the total number of keywords in W; in formula (3), l is the number of keywords in the audit report of the jth engineering project.

3. The method for identifying project audit risks according to claim 2 is characterized in that: In step 2-2, the keyword frequency TI p The calculation formula is as follows: Among them, n in formula (4) pj The keyword w p In W j The frequency of occurrence in k n pj It is W j Total number of keywords in |j:w p ∈W j | contains the keyword w p The number of engineering projects.

4. The method for identifying project audit risks according to claim 3 is characterized in that: In step 2-3, the undetermined risk point set R′ is expressed as follows: R′={R′1,R′2…R′ k } (5); Wherein, k in formula (5) is the number of undetermined risk points.

5. The method for identifying project audit risks according to claim 1 is characterized in that: The step 4 includes the following sub-steps: 4-1: Output G′(V,E); 4-2: Non-isolated nodes in the network G′ are audit risk points; 4-3: The edges in the network G′ represent the causal relationship between risk points.

Images