A digital audit doubt point extraction method and system based on power big data platform

Abstract

The present invention belongs to the technical field of power grid information resources and management methods, and specifically relates to a digital audit doubt point extraction method based on a power big data platform, comprising the following steps: S1: using the power big data platform to collect transaction item set data information from different databases; S2: using two binary numbers to encode the storage status and association status of each transaction item in a transaction item set k; S3: using the Apriori algorithm to iteratively obtain a frequent k-item set; S4: constructing a fitness function to evaluate the closeness of the relationship between transaction items in the frequent k-item set, and determining the maximum fitness function value as the initial optimal position; S5: using the shark intelligent algorithm to optimize the Apriori algorithm association rule, that is, optimizing the fitness function to find the optimal transaction item combination; S6: in the frequent k-item set, the Apriori algorithm is used to extract the most frequent transaction item combination. max Find the optimal transaction item combination in the item set, go to S2, reverse the code, and obtain the optimal transaction item combination association process, that is, the transaction item combination with the strongest doubt.

Description

Technical Field

[0001] The present invention belongs to the technical field of power grid information resources and management methods, and specifically relates to a digital audit doubt point extraction method and system based on a power big data platform. Background Art

[0002] Digital auditing, which uses big data to improve audit efficiency and accuracy, offers significant advantages for audits involving numerous projects and a wide range of systems. The most important goal of an audit is to identify suspicious projects and uncover potential issues behind their implementation. However, traditional on-site audit management practices pose challenges such as difficulty controlling audit quality and delayed feedback. Furthermore, power grid companies face numerous projects with complex interconnectedness, making manual audits inefficient and difficult to effectively establish strong connections. This makes it difficult to uncover suspicious points during project implementation.

[0003] As the management and operation of State Grid Electric Power Company are becoming more and more information-based, the formation of the power big data platform has solved this problem to a certain extent; but the audit method that is still mostly adopted is manual audit as the main method and information technology as the auxiliary method, that is, common project audit doubts are determined by human experience, and information technology is used to find out whether the above doubts are standardized, which makes it difficult to discover certain hidden risks and new risks; at the same time, in the process of data mining with existing information technology, each audit project includes multiple business subsystems. The business data generated by the business subsystems are large in quantity, with many table items, high similarity between data, and mutual management between data. When using the power big data platform for auditing, the Apriori association algorithm is used to scan the original database in large quantities, and the exhaustive method is used to select each set. Multiple scans continuously generate a huge number of sub-sets, which puts a great test on the device memory and is not suitable for the existing power big data platform. Summary of the Invention

[0004] The present invention provides a digital audit doubt point extraction method based on the electric power big data platform. The shark intelligent algorithm is used to optimize the Apriori association rule, and the Apriori iterative search for strongly associated frequent k-item sets is transformed into a multidimensional space optimization process. That is, the influence of Apriori exhaustive enumeration on memory is solved through iterative optimization, making it suitable for the memory of the existing electric power big data platform.

[0005] A method for extracting suspicious points from digital audits based on a power big data platform includes the following steps:

[0006] S1: Use the power big data platform to collect data information of transaction item sets from different databases;

[0007] S2: Use two binary digits to encode the storage status and association status of each transaction item in the transaction item set k;

[0008] S3: Use Apriori algorithm to iteratively obtain frequent k-item sets;

[0009] S4: Construct a fitness function to evaluate the closeness of the relationship between the items in the frequent k-item set, and determine the initial optimal position with the maximum fitness function value;

[0010] S5: Use the Shark Intelligent Algorithm to optimize the association rules of the Apriori algorithm, that is, optimize the fitness function as the objective function to find the optimal combination of transaction items;

[0011] S51: Set the items of the frequent k-item set to M, initialize the position of the transaction item combination and the initial optimization speed;

[0012] S52: Combined with the fitness function, the shark intelligent algorithm is used to find the optimal combination of transaction items in the frequent k-item set, and then go to S3 until the optimal combination of transaction items in the frequent k-item set is found;

[0013] S6: In frequent k max -Find the optimal transaction item combination in the item set, go to S2, use two-bit binary number coding to perform code inversion, and then obtain the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt.

[0014] Furthermore, in S2, the two binary digits include:

[0015] The first binary digit indicates whether the transaction item exists; 1 indicates existence, and 0 indicates non-existence;

[0016] The second binary number indicates the association status of the transaction item; 1 indicates that the transaction item is the antecedent of the rule, and 0 indicates that the transaction item is the consequent of the rule.

[0017] Furthermore, in S3, the specific process of obtaining frequent k-item sets using the Apriori algorithm includes the following steps:

[0018] S31: Setting support threshold and confidence threshold;

[0019] S32: traverse the data information of the transaction items in the frequent k-item set, count the support of k transaction items, and obtain the candidate k-item set;

[0020] S33: Calculate the support of k transaction items, remove transaction items whose support is not greater than the support threshold, and obtain a frequent k-item set;

[0021] S34: Go to S32 until k=k max, no new frequent itemsets are generated, that is, there are no itemsets with support less than the support threshold, and the iteration ends.

[0022] Furthermore, in S31, the support represents the frequency of occurrence of transaction item X and transaction item Y in transaction item set k, and is used to reflect the universality of the association between transaction items. Its calculation expression is:

[0023] support(X→Y)=P(X∩Y);

[0024] Confidence indicates how frequently transaction item Y appears in the included transaction X. It is used to reflect the reliability of the association between transaction items. Its calculation expression is:

[0025] Confidence(X→Y)=P(X∩Y) / P(X);

[0026] Among them, when support(X→Y) and confidence(X→Y) both meet the support threshold and confidence threshold, X→Y is considered to be a strong association rule.

[0027] Furthermore, in S4, the calculation expression of the fitness function f(X→Y) is:

[0028] f(X→Y)=support(X→Y)·Confidence(X→Y)

[0029] Among them, support(X→Y) is the support of association rule X→Y, and Confidence(X→Y) is the confidence of association rule X→Y.

[0030] Furthermore, in S51, the initialization process includes:

[0031] Position initialization, the initial position of the transaction item combination in the frequent k-item set is:

[0032]

[0033] Where M is the total number of transaction item sets in the frequent k-item set;

[0034] in, , which is the first i The position vector of the transaction item combination in the frequent k-item set; n is the i transaction items j The total number of is the initial position of the transaction item combination in the frequent k-item set;

[0035] Initialization of optimization speed: The speed vector of the system initialization to find the combination of transaction items in the frequent k-item set is:

[0036]

[0037] in, , which is to find the first i The velocity vector of the transaction item combination; m For the i transaction items j The total number of To find the first frequent k-item in the set i Transaction item set j The speed at which transaction items are combined.

[0038] Furthermore, in S52, the process of the shark intelligent optimization algorithm finding the optimal transaction item combination in the frequent k-item set includes:

[0039] S521: Use the Shark Intelligent Optimization Algorithm to perform local optimization on the transaction items at the extreme points of the fitness function, and sort the new positions found by the size of the fitness function value, retaining the positions of the first half of the data as the best positions;

[0040] S522: When the fitness function between the two iterations is less than the set threshold, the chaotic map is used to jump out of the local optimum and go to S3 for iteration until the set number of iterations is completed, that is, k=k max .

[0041] Furthermore, in the above S521, when the Shark Intelligent Algorithm is used for optimization, the system searches for the optimal combination of transaction items in the frequent k-item set, and the calculation expression for finding the speed update is:

[0042]

[0043] Where, represents the first frequent k-item set i Transaction item set j The speed of the transaction item combination; f(x) is the fitness function; is a random number uniformly distributed between [0,1]; ;

[0044] Among them, the shark's movement methods during the optimization process include:

[0045] Forward motion simulates the shark moving towards the extreme point of the fitness function, that is, the system moves towards the position of the transaction item combination at the extreme point of the fitness function, and the new position generated by the forward motion The calculation expression is:

[0046]

[0047] Where, is the position of the transaction item composition during the k-th iteration, is the speed at which the system finds transaction item combinations during the kth iteration, is the time of the kth iteration;

[0048] Rotation search is used to avoid the local optimal position of the extreme point of the fitness function found by the forward motion, and then the rotation search is used to achieve local optimization. The new position generated by the rotation search in the k+1th iteration is The calculation expression is:

[0049]

[0050] Where p=(1,2,...,P), P is the number of transaction item combinations in each stage of the position search, and the position search of point P is exist nearby; is a random number uniformly distributed between [-1,1];

[0051] When the system finds a better point during the shark's rotating search, it removes that point and continues searching.

[0052] Furthermore, in S522, the chaotic mapping adopts the Logstic chaotic mapping, which generates new random solutions and thus jumps out of the local optimum. Its expression is:

[0053]

[0054] Where: μ∈(0,4].

[0055] A system for extracting doubtful points from digital audits based on a power big data platform, comprising:

[0056] The data module is used to collect data information from the large power platform and encode it into two-bit binary numbers;

[0057] Model module, used to obtain frequent k-item sets and build fitness functions;

[0058] The optimization module is used by the Shark Intelligent Algorithm to optimize the fitness function and obtain the optimal transaction item combination;

[0059] The output module is used to decode the code for obtaining the optimal transaction item combination and output the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt.

[0060] The beneficial effects of the present invention are:

[0061] The present invention searches for and mines the correlations between various power information, finds the correlations and anomalies between seemingly unrelated audit data, analyzes the deep-seated reasons for the anomalies, compares the corresponding rules of normal relationships with other relevant data, and discovers and analyzes audit doubts; adopts the shark intelligent algorithm to optimize the Apriori association rule, transforms the Apriori iterative search for strongly associated frequent k-item sets into a multi-dimensional space optimization process, that is, solves the impact of Apriori exhaustive enumeration on memory through iterative optimization; and optimizes the calculation of the Apriori algorithm fitness function through the shark intelligent algorithm, thereby enhancing the model's fitting ability and improving the accuracy of finding the optimal transaction item combination. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a schematic diagram of the process of the present invention;

[0063] Figure 2 Schematic diagram of the process of obtaining frequent k-itemsets;

[0064] Figure 3 A flowchart for the Shark Intelligent Optimization Algorithm to find the optimal combination of transaction items in a frequent k-item set;

[0065] Figure 4 Schematic diagram of the system of the present invention. DETAILED DESCRIPTION

[0066] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0067] It should be noted that the experimental methods described in the following embodiments are conventional methods unless otherwise specified, and the reagents and materials are commercially available unless otherwise specified; in the description of the present invention, the terms "horizontal", "longitudinal", "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation to the present invention.

[0068] Furthermore, terms such as "horizontal," "vertical," and "overhanging" do not necessarily imply that a component must be absolutely horizontal or overhanging, but rather that it can be slightly tilted. For example, "horizontal" simply means that its direction is more horizontal than "vertical," and does not mean that the structure must be completely horizontal, but rather that it can be slightly tilted.

[0069] It should also be noted that, in the description of this application, unless otherwise expressly specified or limited, the terms "disposed," "installed," "connected," and "connected" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in this application based on the specific circumstances.

[0070] Figure 1 The figure shows a method for extracting suspicious points from digital audits based on a power big data platform, which includes the following steps:

[0071] S1: Use the power big data platform to collect data information of transaction item sets from different databases;

[0072] S2: Use two binary digits to encode the storage status and association status of each transaction item in the transaction item set k;

[0073] The two-digit binary number specifically includes:

[0074] The first binary digit indicates whether the transaction item exists; 1 indicates existence, and 0 indicates non-existence;

[0075] The second binary number indicates the association status of the transaction item; 1 indicates that the transaction item is the antecedent of the rule, and 0 indicates that the transaction item is the consequent of the rule.

[0076] S3: Use the Apriori algorithm to iteratively obtain frequent k-item sets, construct a fitness function to evaluate the closeness of the relationship between transaction items in the frequent k-item sets, and determine the maximum fitness function value as the initial optimal position;

[0077] like Figure 2 As shown, the specific process includes the following steps:

[0078] S31: Setting support threshold and confidence threshold;

[0079] S32: traverse the data information of the transaction items in the frequent k-item set, count the support of k transaction items, and obtain the candidate k-item set;

[0080] S33: Calculate the support of k transaction items, remove transaction items whose support is not greater than the support threshold, and obtain a frequent k-item set;

[0081] S34: Go to S32 until k=k max , no new frequent itemsets are generated, that is, there are no itemsets with support less than the support threshold, and the iteration ends.

[0082] The support of the association rule X→Y represents the frequency of occurrence of transaction item X and transaction item Y in the transaction item set k, which is used to reflect the universality of the association between transaction items. Its calculation expression is:

[0083] support(X→Y)=P(X∩Y);

[0084] The confidence of the association rule X→Y indicates how frequently transaction item Y appears in the included transaction X. It is used to reflect the reliability of the association between transaction items. Its calculation expression is:

[0085] Confidence(X→Y)=P(X∩Y) / P(X);

[0086] When support(X→Y) and confidence(X→Y) both meet the support threshold and confidence threshold, X→Y is considered to be a strong association rule.

[0087] Among them, the number of codes for each transaction item in the frequent k-item set depends on the number of remaining transaction items in the frequent 1-item set. By using two-bit binary number encoding, the number of codes for each transaction item in the frequent k-item set is twice the number of remaining transaction items in the frequent 1-item set.

[0088] S4: Construct a fitness function to evaluate the closeness of the relationship between each k transaction items in the frequent k-item set, and determine the maximum fitness function value as the initial optimal position;

[0089] The fitness function is expressed by the confidence and support of each k transaction items, and its calculation expression is:

[0090] f(X→Y)=support(X→Y)·Confidence(X→Y)

[0091] Among them, support(X→Y) is the support of association rule X→Y, and Confidence(X→Y) is the confidence of association rule X→Y.

[0092] S5: Use the Shark Intelligent Algorithm to optimize the association rules of the Apriori algorithm, that is, optimize the fitness function as the objective function to find the optimal combination of transaction items;

[0093] Among them, the shark intelligent algorithm searches for the best by simulating the shark's foraging behavior; it simulates prey by generating a set of initial odor particles in the search domain; it assumes that the initial position of the shark during the search is the position of any odor particle, and the concentration of the odor particle represents the distance between the shark and the prey. The stronger the odor, the closer the distance between the shark and the prey, which is represented by the fitness function; the shark preferentially approaches the odor particles with higher odor concentration, that is, the optimization process is the process of moving towards the odor particles with higher odor particle concentration.

[0094] S51: Set the items of the frequent k-item set to M, initialize the position of the transaction item combination and the initial speed of finding the optimal transaction item combination;

[0095] In this embodiment, the scent particles in the shark intelligent algorithm are combinations of transaction items in the frequent k-item set, and the shark in the shark intelligent algorithm is a system for finding combinations of transaction items in the frequent k-item set.

[0096] Position initialization, the initial position of the transaction item combination in the frequent k-item set is:

[0097]

[0098] Where M is the total number of transaction item sets in the frequent k-item set;

[0099] in, , which is the position vector of the transaction item combination in the i-th frequent k-item set; n Set the transaction item for the i-th transaction item j The total number of is the initial position of the transaction item combination in the frequent k-item set;

[0100] Initialization of optimization speed: The speed vector of the system initialization to find the combination of transaction items in the frequent k-item set is:

[0101]

[0102] in, , which is to find the first i The velocity vector of the transaction item combination; m For the i transaction items j The total number of To find the first frequent k-item in the set i Transaction item set j The speed at which transaction items are combined.

[0103] Taking frequent 1-itemset as an example, the initialization process includes:

[0104] Position initialization, the initial position of the transaction item combination in the frequent 1-item set is:

[0105]

[0106] Where M is the total number of transaction item sets in the frequent 1-item set;

[0107] in, , which is the first i The position vector of the transaction item combination in the frequent 1-item set; n For the i transaction items j The total number of is the initial position of the transaction item combination in the frequent 1-item set;

[0108] Initialization of optimization speed: The speed vector of the system initialization to find the combination of transaction items in the frequent 1-item set is:

[0109]

[0110] in, , which is to find the first i The velocity vector of the transaction item combination; n For the i transaction items j The total number of To find the first frequent 1-item i Transaction item set j The speed at which transaction items are combined.

[0111] S52: Use the shark intelligent algorithm to find the optimal transaction item combination in the frequent k-item set and go to S3; the specific steps are as follows: Figure 3 As shown;

[0112] S521: Use the Shark Intelligent Optimization Algorithm to perform local optimization on the transaction items at the extreme points of the fitness function, and sort the new positions found by the size of the fitness function value, retaining the positions of the first half of the data as the best positions;

[0113] When using the shark intelligent algorithm for optimization, when the shark tracks the odor, its speed increases with the increase of the odor concentration. The speed of the shark in each dimension is obtained, that is, the calculation expression for finding the speed update in the process of finding the optimal transaction item combination in the frequent k-item set is:

[0114]

[0115] Where, represents the first frequent k-item set i Transaction item setj The speed of the transaction item combination; f(x) is the fitness function; is a random number uniformly distributed between [0,1]; .

[0116] Among them, the shark's movement methods during the optimization process include:

[0117] The forward movement makes the shark move towards the extreme point of the fitness function, that is, the system moves towards the position of the transaction item combination at the extreme point of the fitness function. The new position generated by the forward movement The calculation expression is:

[0118]

[0119] Where, is the position of the transaction item composition during the k-th iteration, is the speed at which the system finds transaction item combinations during the kth iteration, is the time of the kth iteration;

[0120] Rotation search is used to avoid the local optimal position of the extreme point of the fitness function found by the forward motion, and then the rotation search is used to achieve local optimization. The new position generated by the rotation search in the k+1th iteration is The calculation expression is:

[0121]

[0122] Where p=(1,2,...,P), P is the number of transaction item combinations in each stage of the position search, and the position search of point P is exist nearby; is a random number uniformly distributed between [-1,1].

[0123] When the system finds a point with a stronger scent during the simulated shark's rotating search, it removes that point and continues searching.

[0124] Considering that the fitness function should be maximized, the first 50% of all transaction item combinations found in the frequent k-item set are retained. The position corresponding to each transaction item combination is used as the starting point of the next search process. The process of forward movement and rotation search is repeated until the number of iterations reaches the maximum number of iterations, that is, k=k max .

[0125] S522: In order to avoid the shark intelligent algorithm, that is, the system falls into the local optimal solution in the process of searching for the optimal transaction item combination and is difficult to converge, when the fitness function value between the previous and next two iterations is less than the set threshold, the chaotic mapping is used to generate a new random solution to jump out of the local optimal solution, and then go to S3 for iteration to search for the next frequent item set until the set number of iterations is completed, that is, k=k max .

[0126] Among them, the chaotic map adopts the Logstic chaotic map, and its expression is:

[0127]

[0128] Where μ∈(0,4].

[0129] S6: In frequent k max -Find the optimal transaction item combination in the item set, go to S2, use two-bit binary number coding to perform code inversion, and then obtain the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt.

[0130] like Figure 4 As shown, a system for extracting doubtful points from digital audits based on a power big data platform as described above includes:

[0131] The data module is used to collect data information from the large power platform and encode it into two-bit binary numbers;

[0132] Model module, used to obtain frequent k-item sets and build fitness functions;

[0133] The optimization module is used by the Shark Intelligent Algorithm to optimize the fitness function and obtain the optimal transaction item combination;

[0134] The output module is used to decode the code for obtaining the optimal transaction item combination and output the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt.

[0135] Example 2

[0136] In this embodiment, the process of using the digital audit doubt point extraction method includes:

[0137] T1: Collect information about event names, responsible persons, and manufacturers through the power big data platform, as shown in Table 1:

[0138] Table 1 Partial information of the power big data platform

[0139]

[0140] Among them, Maintenance 1 to Maintenance 6 in the event names are represented by t1 to t6 respectively, the responsible persons A, B, C, D, and Zhang San are represented by f1 to f5 respectively, and the manufacturers 1 to 4 are represented by d1 to d4 respectively.

[0141] T2: The first binary number of each transaction item in Table 1 is shown in Table 2:

[0142] Table 2 Binary representation of the first digit

[0143]

[0144] Encode each transaction item in Table 1 with two binary digits, such as:

[0145] The two-digit binary code of maintenance 1 (t1) in the event name is 11, indicating that the transaction item exists and is the rule antecedent.

[0146] The two-digit binary code of f1, which is the person in charge, is 10, indicating that the transaction item exists and is the consequent of the rule;

[0147] The two-digit binary code of manufacturer 1, ie, d1, among the manufacturers is 01, indicating that the transaction item does not exist.

[0148] That is, the code 111001 represents the association t1→f1.

[0149] Similarly, each transaction item is encoded with two binary digits.

[0150] T3: Use Apriori algorithm to obtain frequent 1-item sets,

[0151] Set the support threshold to 1 / 13, remove the transaction items whose support is not greater than the support threshold, and obtain the frequent 1-item set;

[0152] Among them, transaction item t3, transaction item t4, transaction item t5, transaction item f3, transaction item f4, transaction item d1, and transaction item d2 have support values ​​less than the support threshold and are therefore removed.

[0153] The frequent 1-item set includes: transaction item t1, transaction item t2, transaction item t6, transaction item f1, transaction item f2, transaction item f5, transaction item d3, and transaction item d4; for example, the code of transaction item t1 is [1101010101010101];

[0154] At this point, the dimension of the transaction item is reduced from 15×2=30 to 8×2=16.

[0155] T4: Construct the fitness function, namely:

[0156]

[0157] Calculate each transaction item in the frequent 1-item set to obtain the fitness value of each transaction item.

[0158] T5: Use the shark algorithm to find the optimal transaction item in the frequent 1-item set; the threshold is set to 10 -3 Sort the fitness values ​​by size, and keep the transaction items with the fitness values ​​in the first half for iteration until the optimal transaction item combination is found in the frequent 3-item set, that is, the optimal solution is [0111011001010110];

[0159] Among them, the meanings of the codes in the optimal solution [0111011001010110] are as follows: 01 indicates that transaction item t1 does not exist; 11 indicates that transaction item t2 exists and is the antecedent of the rule; 01 indicates that transaction item t6 does not exist; 10 indicates that transaction item f1 exists and is the consequent of the rule; 01 indicates that transaction item f2 does not exist; 01 indicates that transaction item d3 does not exist; 10 indicates that transaction item d4 exists and is the consequent of the rule.

[0160] T6: Inversely solve the optimal solution and obtain the four strong associations among transaction items t2, f2, and d, that is, the four strong associations among maintenance 2, personnel B, and manufacturer.

[0161] Therefore, in the subsequent audit process, the relevant audit process of Maintenance 2-Personnel B-Manufacturer 4 can be strengthened in management and audit.

[0162] It should be noted that the values ​​of the support threshold, the confidence threshold, and the set threshold in the present invention are selected according to actual conditions.

[0163] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A digital audit doubt point extraction method based on the power big data platform, characterized by: The steps include: S1: Utilize the power big data platform to collect data information of transaction item sets from different databases; the data information includes the event name, person in charge, and manufacturer information; S2: Use two binary digits to encode the storage status and association status of each transaction item in the transaction item set k; S3: Use Apriori algorithm to iteratively obtain frequent k-item sets; S4: Construct a fitness function to evaluate the closeness of the relationship between the items in the frequent k-item set, and determine the initial optimal position where the fitness function value is the largest; S5: Use the Shark Intelligent Algorithm to optimize the association rules of the Apriori algorithm, that is, optimize the fitness function as the objective function to find the optimal combination of transaction items; S51: Set the items of the frequent k-item set to M, initialize the position of the transaction item combination and the initial optimization speed; S52: Combined with the fitness function, the shark intelligent algorithm is used to find the optimal combination of transaction items in the frequent k-item set, and then the process goes to S3 until the optimal combination of transaction items in the frequent k-item set is found. S6: In frequent k max - Find the optimal transaction item combination in the item set, go to S2, use two-bit binary code to perform code inversion, and then obtain the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt; In S2, the two binary numbers include: The first binary digit indicates whether the transaction item exists; 1 indicates existence, and 0 indicates non-existence; The second binary number indicates the association status of the transaction item; 1 indicates that the transaction item is the antecedent of the rule, and 0 indicates that the transaction item is the consequent of the rule; In S4, the calculation expression of the fitness function f(X→Y) is: f(X→Y)=support(X→Y)·Confidence(X→Y) Among them, support(X→Y) is the support of association rule X→Y, and Confidence(X→Y) is the confidence of association rule X→Y; In S51, the initialization process includes: Position initialization, the initial position of the transaction item combination in the frequent k-item set is: Where M is the total number of transaction item sets in the frequent k-item set; in, , which is the first i The position vector of the transaction item combination in the frequent k-item set; n For the i transaction items j The total number of is the initial position of the transaction item combination in the frequent k-item set; Initialization of optimization speed: The speed vector of the system initialization to find the combination of transaction items in the frequent k-item set is: in, , which is to find the first i The velocity vector of the transaction item combination; m For the i transaction items j The total number of To find the first frequent k-item in the set i Transaction item set j The speed of combining transaction items; In S52, the process of the shark intelligent optimization algorithm finding the optimal transaction item combination in the frequent k-item set includes: S521: Use the Shark Intelligent Optimization Algorithm to perform local optimization on the transaction items at the extreme points of the fitness function, and sort the new positions found by the size of the fitness function value, retaining the positions of the first half of the data as the best positions; S522: When the fitness function between the two iterations is less than the set threshold, the chaotic map is used to jump out of the local optimum and go to S3 for iteration until the set number of iterations is completed, that is, k=k max ; In the above S521, when the Shark Intelligent Algorithm is used for optimization, the system searches for the optimal combination of transaction items in the frequent k-item set, and the calculation expression for finding the speed update is: Where, represents the first frequent k-item set i Transaction item set j The speed of the transaction item combination; f(x) is the fitness function; is a random number uniformly distributed between [0,1]; ; Among them, the shark's movement methods during the optimization process include: Forward motion simulates the shark moving towards the extreme point of the fitness function, that is, the system moves towards the position of the transaction item combination at the extreme point of the fitness function, and the new position generated by the forward motion The calculation expression is: Where, is the position of the transaction item composition during the k-th iteration, is the speed at which the system finds transaction item combinations during the kth iteration, is the time of the kth iteration; Rotation search is used to avoid the local optimal position of the extreme point of the fitness function found by the forward motion, and then the rotation search is used to achieve local optimization. The new position generated by the rotation search in the k+1th iteration is The calculation expression is: Where p=(1,2,...,P), P is the number of transaction item combinations in each stage of the position search, and the position search of point P is exist nearby; is a random number uniformly distributed between [-1,1]; When the system finds a better point during the shark's rotating search, it removes that point and continues searching.

2. The method for extracting doubtful points from digital audits based on a power big data platform according to claim 1 is characterized by: In S3, the specific process of using the Apriori algorithm to obtain frequent k-item sets includes the following steps: S31: Setting support threshold and confidence threshold; S32: traverse the data information of the transaction items in the frequent k-item set, count the support of k transaction items, and obtain the candidate k-item set; S33: Calculate the support of k transaction items, remove transaction items whose support is not greater than the support threshold, and obtain a frequent k-item set; S34: Go to S32 until k=k max , no new frequent itemsets are generated, that is, there are no itemsets with support less than the support threshold, and the iteration ends.

3. The method for extracting doubtful points from digital audits based on a power big data platform according to claim 2 is characterized by: In S31, the support degree represents the frequency of occurrence of transaction item X and transaction item Y in transaction item set k, and is used to reflect the universality of the association between transaction items. Its calculation expression is: support(X→Y)=P(X∩Y); Confidence indicates how frequently transaction item Y appears in the included transaction X. It is used to reflect the reliability of the association between transaction items. Its calculation expression is: Confidence(X→Y)=P(X∩Y) / P(X); Among them, when support(X→Y) and confidence(X→Y) both meet the support threshold and confidence threshold, X→Y is considered to be a strong association rule.

4. The method for extracting doubtful points from digital audits based on a power big data platform according to claim 1 is characterized by: In S522, the chaotic mapping adopts the Logstic chaotic mapping, which generates new random solutions and thus jumps out of the local optimum. Its expression is: Where: μ∈(0,4].

5. A system for extracting doubtful points from digital audits based on a power big data platform as described in any one of claims 1 to 4, characterized in that: include: The data module is used to collect data information from the large power platform and encode it into two-bit binary numbers; Model module, used to obtain frequent k-item sets and build fitness functions; The optimization module is used by the Shark Intelligent Algorithm to optimize the fitness function and obtain the optimal transaction item combination; The output module is used to decode the code for obtaining the optimal transaction item combination and output the relevant association process of the optimal transaction item combination, that is, the transaction item combination with the strongest doubt.

Images