Method and system for improving accurate cost collection based on artificial intelligence

Abstract

The present invention relates to the field of data processing technology, and more specifically, to a method and system for improving the accuracy of cost aggregation based on artificial intelligence, the method comprising: obtaining electricity cost data, generating cost data points according to the cost categories to which the electricity cost data belongs, constructing a set of cost data points, clustering the set of cost data points, calculating the importance of the cost data points of each dimension in the set of cost data points, obtaining the proportion of the importance in the data of each dimension as the weight of the data of each dimension, calculating the eccentricity of the initial clustering center; based on the weight and the eccentricity of the initial clustering center, obtaining the availability of the initial clustering center, completing clustering according to the availability, obtaining clustering results, comparing the existing cost aggregation results with the clustering results, and reviewing and comparing the differences. The present invention evaluates the initial clustering center, and the clustering results are more accurate, thereby improving the accuracy of cost aggregation.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology. More specifically, the present invention relates to a method and system for accurately improving cost collection based on artificial intelligence. Background Art

[0002] Enterprises in the power industry usually have complex cost structures and special resource management. In order to achieve the goals of optimizing resource management, improving operational efficiency, and enhancing financial transparency, enterprises in the power industry usually require highly accurate cost aggregation.

[0003] The existing Chinese patent application document with publication number CN112948441A discloses a multidimensional data collection method and device for financial data, which is used to solve the technical problems of low efficiency and high development cost of existing financial data collection methods. The method includes: determining a pre-built dimension data table for each dimension in the financial data; selecting a query dimension from the dimension, and building a query solution table according to the query dimension; according to several query dimensions in the query solution table, respectively obtaining data corresponding to the corresponding query dimension from the data source table composed of financial documents, and storing the data in a temporary table; processing the data in the temporary table, and synchronizing the processing results to the collection table corresponding to the query solution table.

[0004] The above method only solves how to query the aggregation results of data of different dimensions. However, in the aggregation process, cost data needs to be clustered. However, cost aggregation often involves a large amount of financial data, and some data may be missed or recorded twice, which affects the accuracy of cost aggregation. The K-means++ algorithm is currently used to measure whether the cost of each category being aggregated is misclassified or repeatedly aggregated. However, the K-means++ algorithm randomly selects an initial clustering center, and selects the remaining initial clustering centers based on the initial clustering center, and then performs clustering iterations. The clustering centers located at the edge of the data space are not conducive to the iterative process of the K-means++ clustering algorithm, affecting the clustering effect, and further affecting the comparison between the clustering results and the cost aggregation results. Summary of the invention

[0005] In order to solve the problem of whether the spatial position in the initial cluster is conducive to the iteration of the clustering algorithm when the K-means++ clustering algorithm clusters the set of cost data points, the present invention provides solutions in the following aspects.

[0006] In the first aspect, a method for accurately improving cost aggregation based on artificial intelligence includes: obtaining electricity cost data, generating cost data points according to the cost category to which the electricity cost data belongs, constructing a set of cost data points, clustering the set of cost data points, and selecting any cost data point as an initial clustering center; calculating the importance of the cost data points of each dimension in the set of cost data points, obtaining the proportion of the importance in the data of each dimension as the weight of the data of each dimension, calculating the number of data points in each part when the set of cost data points is divided and the distance between the initial clustering center as the eccentricity of the initial clustering center; calculating the availability of the initial clustering center based on the weight and the eccentricity of the initial clustering center, , where Indicates the availability of the initial cluster center, represents the number of cost data points in the cost data point set, Indicates the number of dimensions of data contained in each cost data point. Represents the cost data point The weight of the dimension, Indicates The cost data point is the distance from the initial cluster center to the The distance in dimensions, Indicates the eccentricity of the initial cluster center, An exponential function representing a natural constant; clustering is performed according to the availability to obtain a clustering result, and the existing cost aggregation result is compared with the clustering result to review and compare the difference.

[0007] The effect is that by considering the importance of each cost data point in multiple dimensions and determining the weight of each dimension data, the contribution and importance of the cost data point in clustering can be more accurately evaluated. The eccentricity of the initial cluster center point is determined according to the number of data points contained in each part and the distance between the cost data point and the pre-selected initial cluster center in the cost data point set. According to the relative position and absolute position of the initial cluster center in the data space, it is judged whether the position of the pre-selected first initial cluster center is conducive to selecting other initial cluster centers that are conducive to cluster iteration, so as to make the clustering result more accurate.

[0008] In one embodiment, calculating the importance of the cost data points of each dimension in the cost data point set includes:

[0009] Calculate the variance of the data of all cost data points in each dimension, construct a histogram of each dimensional data, obtain the maximum column height and the minimum column height in the histogram of each dimensional data, and use the ratio of the difference between the maximum column height and the minimum column height to the number of data contained in each dimensional data as the importance of each data point.

[0010] The effect is that the variance measures the dispersion of data points in each dimension. A high variance indicates that the dimension has a greater impact on the cost. The importance of each data point can be obtained by visualizing it through a histogram.

[0011] In one embodiment, the importance level further includes:

[0012] Select any one of the cost data points in the cost data point set as the target data point, and calculate the distance between the target data point and other data points;

[0013] According to the number of data of each target data point in a preset local range, the local density of the target data point is calculated, the local density is normalized, the average local density of the cost data point set is obtained, and the ratio of the local density of the target data point to the average density is calculated as the importance of the target data point.

[0014] In one embodiment, the weights of the dimensional data satisfy the following relationship:

[0015] ,

[0016] In the formula, Indicates The weight of the dimension, Indicates The variance of the dimension data, Indicates The maximum column height of the histogram corresponding to the dimension data, Indicates The minimum column height of the histogram corresponding to the dimension data, Indicates the number of dimensions of data contained in each cost data point. Represents the number of cost data points in the cost data point set.

[0017] The effect is that by calculating the variance of the dimensional data, the larger the variance, the wider the distribution on the dimension, and the greater the impact of the dimension on the difference between the cost data points, it is conducive to better capturing the diversity and differences of the data, so that the initial cluster center selected subsequently is more representative.

[0018] In one embodiment, clustering the set of cost data points includes:

[0019] Randomly selecting an initial cluster center according to the cost data point set, dividing the cost data point set into two parts according to the initial cluster center, and counting the number of data points contained in each part;

[0020] The distance between other cost data points and the initial cluster center is calculated, and the next cluster center is selected based on the squared probability of the distance.

[0021] In one embodiment, the degree of eccentricity satisfies the following relationship:

[0022] ,

[0023] In the formula, Indicates the eccentricity of the initial cluster center, represents the initial cluster center The number of data points in the part of the data set that is smaller than the initial cluster center when the dimension divides the data point set into two parts. represents the initial cluster center The number of data points in the part of the data set that is larger than the initial cluster center when the dimension divides the data point set into two parts. Represents the cost data point The weight of the dimension, represents the number of cost data points in the cost data point set, Indicates the number of dimensions of data contained in each cost data point. represents the initial cluster center The value of the dimension data.

[0024] The effect is that by determining the eccentricity of the initial cluster center point, the subsequent initial cluster centers determined based on the initial cluster center are more likely to evenly cover the entire data space, thereby improving the quality of the clustering results.

[0025] In one embodiment, clustering is performed according to the availability to obtain a clustering result, including:

[0026] In response to the availability of the initial cluster center being less than the preset degree threshold, a cost data point is randomly selected as the initial cluster center. Otherwise, if the availability is greater than or equal to the preset degree threshold, it is used as the subsequent initial cluster center for clustering iteration to obtain the clustering result.

[0027] In the second aspect, a system for accurately improving cost aggregation based on artificial intelligence includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned method for accurately improving cost aggregation based on artificial intelligence is implemented.

[0028] The present invention has the following effects:

[0029] 1. The present invention obtains the importance of each dimension of the cost data point through the data of all cost data points in the cost data point set in each dimension, and then obtains the weight of each dimension. The importance of each dimension when measuring the pre-selected first initial clustering center can be determined according to the data characteristics of each dimension of the cost data.

[0030] 2. The present invention can use the first initial clustering center which is more conducive to the subsequent clustering iteration process by evaluating the initial clustering center, thereby making the clustering result of the cost data more accurate. When compared with the cost aggregation result, it is more likely to identify the misclassified data in the cost aggregation result, thereby improving the accuracy of cost aggregation. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the detailed description below with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0032] Figure 1 It is a method flow chart of steps S1 to S4 in the method for accurately improving cost aggregation based on artificial intelligence in an embodiment of the present invention.

[0033] Figure 2 It is a schematic diagram of dividing a set of cost data points according to one dimension data in the cost data points in the method for accurately improving cost aggregation based on artificial intelligence in an embodiment of the present invention.

[0034] Figure 3 This is a schematic diagram of data distribution when the variances of data in two dimensions are similar in the method for accurately improving cost aggregation based on artificial intelligence in an embodiment of the present invention.

[0035] Figure 4 It is a schematic diagram of different distances in the data space between initial cluster centers with the same degree of eccentricity and other data points in the cost data point set in the artificial intelligence-based cost aggregation precision improvement method in an embodiment of the present invention.

[0036] Figure 5 It is a structural block diagram of a cost aggregation and precision improvement system based on artificial intelligence in an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0038] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0039] Specific implementation scenario: Since the cost structure of enterprises in the power industry is highly complex, in order to facilitate the processing of various categories of cost data, the corresponding cost data points are constructed according to the cost categories of each level to which the cost data belongs.

[0040] Reference Figure 1 The method for improving the accuracy of cost collection based on artificial intelligence includes steps S1 to S4, which are as follows:

[0041] S1: Obtain electricity cost data, generate cost data points according to the cost categories to which the electricity cost data belongs, construct a set of cost data points, cluster the set of cost data points, and select any cost data point as the initial cluster center.

[0042] To further illustrate, in the power industry, the cost of an enterprise is divided into multiple dimensions, such as fuel cost, operating cost, etc. Fuel cost is taken as the first dimension and operating cost is taken as the second dimension. The amount of a certain cost data is 5,000 yuan, which is marked as: fuel cost-coal purchase cost, operating cost-generator operating cost. If coal purchase cost is the first category of fuel cost and generator operating cost is the fourth category of operating cost, then the initial data point corresponding to this cost data is (1, 4, 5000), and all initial data points are normalized by the maximum value of each dimension among all initial data points. For example, among all initial data points, there are initial data points (10, 2, 2000), (2, 50, 2000) and (5, 2, 12000). Among all initial data points, 10 is the largest in the first dimension, 50 is the largest in the second dimension, and 12000 is the largest in the third dimension. Then, after normalizing the initial data point (1, 4, 5000), we get ( , , )but( , , ) is the cost data point corresponding to the initial data point (1, 4, 5000)), and the cost data point is obtained. All cost data are converted into cost data points to form a set of cost data points.

[0043] An initial cluster center is randomly selected according to the cost data point set, and the cost data point set is divided into two parts according to the initial cluster center. Specifically, refer to Figure 2, the left figure shows two parts after the cost data point set is divided by one dimension data, and the right figure shows two parts after the cost data point set is divided by another dimension data, and the number of data points contained in each part is counted respectively;

[0044] Calculate the distance between other cost data points and the initial cluster center, and select the next cluster center based on the square probability of the distance.

[0045] It should be noted that the K-means++ clustering algorithm is used in this embodiment. Since the K-means++ clustering algorithm treats data of all dimensions equally when processing multidimensional data, in actual applications, data of different dimensions have different importance. Therefore, when measuring the availability of the first initial clustering center by treating data of all dimensions equally, it may lead to misjudgment of the availability of the first initial clustering center. Therefore, the present invention obtains the importance of each dimension of the cost data point through the data of all cost data points in each dimension in the cost data point set, and then obtains the weight of each dimension through the importance of each dimension.

[0046] S2: Calculate the importance of the cost data points of each dimension in the cost data point set, obtain the proportion of the importance in the data of each dimension as the weight of the data of each dimension, and calculate the distance between the number of data points in each part when the cost data point set is divided and the initial cluster center as the eccentricity of the initial cluster center.

[0047] Calculate the variance of the data of all cost data points in each dimension, construct a histogram of each dimensional data, obtain the maximum and minimum column heights in the histogram of each dimensional data, and use the ratio of the difference between the maximum and minimum column heights to the number of data contained in each dimensional data as the importance of each data point.

[0048] To further explain, the variance of the data of all cost data points in the cost data point set in each dimension is calculated. For example, there are two cost data points in the cost data point set, each data point has three dimensions, one data point is {0.1, 0.2, 0.3}, and the other data point is {0.12, 0.22, 0.21}, then the variance of 0.1 and 0.12, the variance of 0.2 and 0.22, and the variance of 0.3 and 0.21 are calculated, and there are three variances in total for three dimensions. The histogram of the data of each dimension is counted, and there are as many histograms as there are dimensions. The height of the inner column in each histogram is calculated, and the weight of each dimension is calculated according to the variance of the data of all cost data points in the cost data point set in each dimension and the column height of the histogram corresponding to each dimension.

[0049] Specifically, the weight of each dimension data satisfies the following relationship:

[0050] ,

[0051] In the formula, Indicates The weight of the dimension, Indicates The variance of the dimension data, Indicates The maximum column height of the histogram corresponding to the dimension data, Indicates The minimum column height of the histogram corresponding to the dimension data, Indicates the number of dimensions of data contained in each cost data point. Represents the number of cost data points in the cost data point set.

[0052] Further explanation: the larger the variance of a dimension of the cost data point, the wider the distribution of the cost data point on this dimension, and this dimension has a greater impact on the differences between data points. Therefore, when evaluating the availability of the first initial cluster center, the dimension with a larger variance should have a higher degree of importance. The variance of the dimension data The larger the dimension is, the higher the importance of the dimension. The variance of the dimension data The smaller the variance, the lower the importance of the dimension. However, the variance cannot reflect the degree of separation between data points. For example, when the data of two dimensions have similar variances, the data of one dimension is more dispersed, and the data of the other dimension is more concentrated and divided into two clusters. Figure 3 , dimensions with a higher degree of separation have a higher degree of importance;

[0053] The difference between the column heights in the histogram can be used to measure the degree of separation of the data in this dimension to a certain extent. The difference in column height of the histogram corresponding to the dimension data The larger the The higher the separation degree of the dimensional data, the The difference in column height of the histogram corresponding to the dimension data The smaller the The lower the degree of separation of the dimensional data, the easier it is to calculate the number of cost data points in the cost data point set. right Normalize, the larger the value, the corresponding The higher the importance of the dimension, the The first dimension has a higher weight, and vice versa. The smaller the dimension, the lower the weight.

[0054] It is further explained that in the K-means++ clustering algorithm, the selection of the first initial cluster center has a certain degree of randomness. When the first initial cluster center is located in different positions, it has different effects on the selection of subsequent initial cluster centers. For example, if the first initial cluster center is selected in a remote area in the data set, the subsequent selection of initial cluster centers may deviate from the area where the data distribution is concentrated, affecting the subsequent clustering process. Therefore, it is necessary to calculate the degree of eccentricity of the first initial cluster center.

[0055] The process of obtaining the eccentricity of the initial cluster center includes: randomly selecting a cost data point in the cost data point set as the pre-selected first initial cluster center, dividing the cost data point set into two parts according to the data of each dimension of the pre-selected first initial cluster center, and referring to Figure 2 , and count the number of data points contained in each of the two parts of the cost data set divided according to each dimensional data. If the cost data point contains only two dimensions, then the data set is divided twice according to the values ​​of the two dimensions of the pre-selected first initial clustering center. Each division divides the data set into two parts, and counts the number of data points contained in each of the two parts at each division.

[0056] Specifically, the degree of eccentricity satisfies the following relationship:

[0057] ,

[0058] In the formula, Indicates the eccentricity of the initial cluster center, represents the initial cluster center The number of data points in the part of the data set that is smaller than the initial cluster center when the dimension divides the data point set into two parts. represents the initial cluster center The number of data points in the part of the data set that is larger than the initial cluster center when the dimension divides the data point set into two parts. Represents the cost data point The weight of the dimension, represents the number of cost data points in the cost data point set, Indicates the number of dimensions of data contained in each cost data point. represents the initial cluster center The value of the dimension data.

[0059] Further explanation: The difference between the number of data points in the two parts of the cost data point set is obtained, and then the average value of the difference between the number of data points in the two parts of the cost data point set is obtained. The larger the average value of the difference is, the higher the first initial cluster center is. The more the data point deviates from the center position in each dimension, the greater the eccentricity of the corresponding first initial cluster center. On the contrary, the smaller the average value of the difference is, the first initial cluster center is The closer to the center of the data point in each dimension, the smaller the eccentricity;

[0060] All data are normalized, and the value range of each dimension data is [0, 1], so the center position of each dimension is 0.5, so The first initial cluster center The distance between a dimension and the center of the dimension.

[0061] In addition, in another embodiment, any one of the cost data points in the cost data point set is selected as the target data point, and the distance between the target data point and other data points is calculated;

[0062] According to the number of data of each target data point in the preset local range, the local density of the target data point is calculated, the local density is normalized, the average local density of the cost data point set is obtained, and the ratio of the local density of the target data point to the average density is used as the importance of the target data point.

[0063] S3: Based on the weight and the eccentricity of the initial cluster center, the availability of the initial cluster center is calculated.

[0064] It should be noted that although the degree of eccentricity can well measure the degree to which the pre-selected first initial cluster center deviates from the center of the cost data point set and the degree to which it deviates from the center of the data space, the pre-selected first initial cluster center with the same degree of eccentricity may have different distances from other data points in the cost data point set in the data space. Figure 4 , and then the availability of the first initial cluster center is calculated to achieve a better clustering effect.

[0065] Specifically, the availability satisfies the following relationship:

[0066] ,

[0067] In the formula, Indicates the availability of the initial cluster center, represents the number of cost data points in the cost data point set, Indicates the number of dimensions of data contained in each cost data point. Represents the cost data point The weight of the dimension, Indicates The cost data point is the distance from the initial cluster center to the The distance in dimensions, Indicates the eccentricity of the initial cluster center, An exponential function representing a constant of nature.

[0068] Further explanation: Represents the cost data point set in The cost data point is the distance from the initial cluster center to the The components on the dimensions are calculated to obtain the average value of the weighted distance as the availability. The larger the value, the more likely it is that the subsequent clustering centers selected based on the initial clustering center are evenly distributed in the set of cost data points, so as to make the clustering result better. On the contrary, the smaller the availability, the more likely it is that the subsequent clustering centers selected are difficult to be evenly distributed in the set of cost data points.

[0069] At the same time, the degree of eccentricity can also measure whether the cluster center selected as the basis of the first initial clustering is conducive to the subsequent clustering iteration process. The higher the eccentricity of the first initial cluster center, the more the first initial cluster center deviates from the center position of the data space, which is not conducive to the subsequent clustering iteration process. The smaller the availability of the first initial cluster center, the more conducive to the subsequent clustering iteration.

[0070] S4: Complete clustering according to the degree of availability, obtain clustering results, compare the existing cost aggregation results with the clustering results, and review and compare the differences.

[0071] In response to the availability of the initial cluster center being less than the preset degree threshold, a cost data point is randomly selected as the initial cluster center. Otherwise, if the availability is greater than or equal to the preset degree threshold, it is used as the subsequent initial cluster center for clustering iteration to obtain the clustering result.

[0072] Further explanation, the preset degree threshold is set to 0.6, the K-means++ algorithm is a prior art, and the subsequent iterative process is well known to those skilled in the art and will not be described in detail. After clustering is completed, since the clustering algorithm only divides the cost data into several categories, it cannot directly determine which category corresponds to which category of cost aggregation; in the existing cost aggregation results, there are many cost data in one category, and each category in the clustering results also has many data. The cost data in one category in the cost aggregation results are compared with the data in each category in the clustering results. The category of cost aggregation corresponds to the one with the largest intersection between the data in each category of the cost aggregation results. For example: there are data {1, 2, 3, 4, 5} in one category of the cost aggregation, and one category in the clustering results is {1, 2, 7, 8, 9}, and the other category is {6, 7, 8, 9, 10}. Because the intersection of {1, 2, 7, 8, 9} and {1, 2, 3, 4, 5} is larger than the intersection of {1, 2, 3, 4, 5} and {6, 7, 8, 9, 10}, the clustering result {1, 2, 7, 8, 9} is the category corresponding to the cost aggregation {1, 2, 3, 4, 5}. Therefore, by finding that the cost data corresponding to a category in the clustering result has the most data corresponding to the existing cost aggregation result category, this cost aggregation category is matched with this category in the clustering result, and the existing cost aggregation result is compared with the clustering result. The relevant staff will conduct a second confirmation of the cost data of different categories in the cost aggregation result and the clustering result, thereby reducing the review time of the relevant staff on the cost aggregation and improving the accuracy of the cost aggregation.

[0073] The present invention also provides a cost collection and precision improvement system based on artificial intelligence. Figure 4 As shown, the system includes a processor and a memory, and the memory stores computer program instructions. When the computer program instructions are executed by the processor, the artificial intelligence-based cost aggregation precision improvement method according to the first aspect of the present invention is implemented.

[0074] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0075] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0076] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0077] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. The method for accurately improving cost collection based on artificial intelligence is characterized by: include: Acquire electricity cost data, generate cost data points according to the cost categories to which the electricity cost data belongs, construct a set of cost data points, cluster the set of cost data points, and select any cost data point as an initial cluster center; The importance of the cost data points of each dimension in the cost data point set is calculated, the proportion of the importance in the data of each dimension is obtained as the weight of the data of each dimension, and the eccentricity of the initial cluster center is calculated. The eccentricity of the initial cluster center satisfies the following relationship: , where Indicates the eccentricity of the initial cluster center, represents the initial cluster center The number of data points in the part of the data set that is smaller than the initial cluster center when the dimension divides the cost data point set into two parts. represents the initial cluster center The number of data points in the part of the data set that is larger than the initial cluster center when the dimension divides the cost data point set into two parts. Represents the cost data point The weight of the dimension, represents the number of cost data points in the cost data point set, Indicates the number of dimensions of data contained in each cost data point. represents the initial cluster center The value of the dimension data, Represents the value of the center position data of each dimension; Based on the weight and the eccentricity of the initial cluster center, the availability of the initial cluster center is calculated, , where Indicates the availability of the initial cluster center, Indicates The cost data point is the distance from the initial cluster center to the The distance in dimensions, Exponential functions representing constants of nature; Clustering is completed according to the availability to obtain a clustering result, and the existing cost aggregation result is compared with the clustering result to review and compare the differences.

2. The method for accurately improving cost collection based on artificial intelligence according to claim 1 is characterized in that: Calculating the importance of the cost data points of each dimension in the cost data point set includes: Calculate the variance of the data of all cost data points in each dimension, construct a histogram of each dimensional data, obtain the maximum column height and the minimum column height in the histogram of each dimensional data, and use the ratio of the difference between the maximum column height and the minimum column height to the number of data contained in each dimensional data as the importance of each data point.

3. The method for accurately improving cost collection based on artificial intelligence according to claim 1 is characterized in that: in, The importance also includes: Select any one of the cost data points in the cost data point set as the target data point, and calculate the distance between the target data point and other data points; According to the number of data of each target data point in a preset local range, the local density of the target data point is calculated, the local density is normalized, the average local density of the cost data point set is obtained, and the ratio of the local density of the target data point to the average density is calculated as the importance of the target data point.

4. The method for accurately improving cost collection based on artificial intelligence according to claim 1 is characterized in that: The weights of the data in each dimension satisfy the following relationship: ， In the formula, Indicates The weight of the dimension, Indicates The variance of the dimension data, Indicates The maximum column height of the histogram corresponding to the dimension data, Indicates The minimum column height of the histogram corresponding to the dimension data, Indicates the number of dimensions of data contained in each cost data point. Represents the number of cost data points in the cost data point set.

5. The method for accurately improving cost collection based on artificial intelligence according to claim 1 is characterized in that: Clustering the set of cost data points includes: Randomly selecting an initial cluster center according to the cost data point set, dividing the cost data point set into two parts according to the initial cluster center, and counting the number of data points contained in each part; The distance between other cost data points and the initial cluster center is calculated, and the next cluster center is selected based on the squared probability of the distance.

6. The method for accurately improving cost collection based on artificial intelligence according to claim 1 is characterized in that: Clustering is completed according to the availability to obtain clustering results, including: In response to the availability of the initial cluster center being less than the preset degree threshold, a cost data point is randomly selected as the initial cluster center. Otherwise, if the availability is greater than or equal to the preset degree threshold, it is used as the subsequent initial cluster center for clustering iteration to obtain the clustering result.

7. The cost collection and precision improvement system based on artificial intelligence is characterized by: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the method for accurately improving cost aggregation based on artificial intelligence according to any one of claims 1 to 6 is implemented.

Images