Precise power marketing method based on machine learning
Through reference feature selection, distance function and cluster quality evaluation, the problems of improper selection of initial cluster centers, inaccurate correlation between data points and cluster centers and difficult parameter configuration in traditional power marketing clustering methods are solved, and the problem of improper selection of initial cluster centers is improved. The accuracy and stability of clustering enhance the accuracy of power marketing strategies.
Patent Information
- Application Number
- CN202411101933.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-12
- Publication Date
- 2024-11-12
AI Technical Summary
Traditional power marketing clustering methods have low accuracy, distinction, stability and effectiveness caused by improper selection of initial clustering centers, and purity, accuracy, stability and reliability caused by inaccurate correlation between data points and clustering centers. Low, and the problem of difficult clustering parameters.
Determine the initial clustering range and boundaries through benchmark feature selection and fine data space division, design distance function and local density calculations to improve the correlation between data points and cluster centers, and optimize clustering using cluster quality evaluation and optimal guided search function Class parameters.
It improves the accuracy, distinction, stability and reliability of clustering, reduces the blindness of the clustering process, improves the configuration efficiency and flexibility of clustering parameters, and enhances the accuracy of power marketing strategies.
Smart Images

Figure CN118941319A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power marketing, and specifically to a precision power marketing method based on machine learning. Background Art
[0002] The precision power marketing method based on machine learning is to use machine learning technology and big data analysis and processing technology to conduct in-depth analysis and mining of a large amount of power user data, build accurate power marketing templates, match user needs, and formulate precise marketing strategies. However, the traditional power marketing clustering method has problems such as low accuracy, discrimination, stability and effectiveness of clustering due to improper selection of initial clustering centers; low purity, accuracy, stability and reliability of clustering due to inaccurate correlation between data points and clustering centers and unreasonable distribution of data points; and difficult configuration of clustering parameters. Summary of the invention
[0003] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a precise power marketing method based on machine learning. In view of the problems of low accuracy, discrimination, stability and effectiveness of clustering caused by improper selection of initial clustering centers in traditional power marketing clustering methods, this scheme more accurately determines the scope and boundaries of initial clustering through benchmark feature selection and fine data space division, avoiding arbitrariness; clearly calculates the initial width and final width of space division, making space division more precise, better adapting to the distribution characteristics of data, and improving the accuracy and discrimination of clustering; takes into account the data characteristics of each region, reduces the blindness of initial clustering center selection, and makes the clustering process more stable and effective; in view of the problems of inaccurate correlation between data points and clustering centers and unreasonable distribution of data points in traditional power marketing clustering methods, the purity, accuracy, stability and reliability of clustering are improved. In order to solve the problem of low accuracy, this scheme designs a distance function to more accurately reflect the degree of association between data points and cluster centers; the calculation of association makes the distribution of data points more reasonable and precise; the calculation of local density can better identify the distribution characteristics of data and improve the purity and accuracy of clustering; the cluster center update formula comprehensively considers the factors of association and local density, so that the adjustment of cluster centers is more in line with the actual situation of data and the stability and reliability of clustering are improved; in view of the problem that clustering parameters are difficult to configure in traditional power marketing clustering methods, this scheme uses a comprehensive clustering quality assessment, comprehensively considers stability, inter-cluster and intra-cluster differences, and configures the optimal number of clusters in a more systematic and objective manner; the design of the optimal guided search function makes the search for clustering parameters more efficient, and conducts different degrees of detailed exploration at different stages, which improves the ability to configure the optimal parameter combination and increases the flexibility and adaptability of the search.
[0004] The technical solution adopted by the present invention is as follows: a precise power marketing method based on machine learning, the method comprising the following steps:
[0005] Step S1: data collection;
[0006] Step S2: data preprocessing;
[0007] Step S3: Select the initial cluster center;
[0008] Step S4: clustering;
[0009] Step S5: Optimizing clustering parameters;
[0010] Step S6: Power marketing strategy generation.
[0011] Furthermore, in step S1, the data collection is to collect users' electricity consumption data, electricity consumption patterns, customer type data, historical payment data, equipment information data, weather-related data, and geographic area data; the electricity consumption data includes monthly electricity consumption, daily electricity consumption, peak and valley electricity consumption, and electricity consumption time periods; the customer types include residential users, commercial users, and industrial users; the historical payment data includes the user's payment time, amount, and arrears; the weather-related data includes temperature and humidity; and the geographic area data includes the user's geographical location and regional characteristics.
[0012] Furthermore, in step S2, the data preprocessing specifically includes the following steps:
[0013] Step S21: Data collation, checking the integrity of the data, filling in missing values, checking for outliers in the data, correcting and removing data by comparing with historical data and data of similar users, standardizing the data, and converting the data into a unified metric;
[0014] Step S22: Generate an electricity marketing portrait, analyze the user's electricity usage pattern, determine the characteristics of electricity usage habits, summarize the electricity usage characteristics and needs based on customer types, extract payment credit information from payment data, infer electricity demand based on equipment information, analyze electricity usage differences based on geographical area data, and comprehensively construct a user electricity marketing portrait that includes electricity usage habits, consumption capacity, credibility, and potential demand characteristics, and add this portrait data to the feature data.
[0015] Furthermore, in step S3, the selecting of the initial cluster center specifically includes the following steps:
[0016] Step S31: benchmark feature selection, selecting two features for determining the initial cluster center position, and finding the maximum and minimum values of the two data dimensions in the two features;
[0017] Step S32: Calculate the data space division factor, which is expressed as follows:
[0018] ;
[0019] Among them, v represents the number of clusters, and denote the spatial partitioning factors for the first and second dimensions, respectively, Indicates taking the absolute value, Indicates taking the minimum value;
[0020] Step S33: Determine the initial width of the space division, and determine the initial width of the space division by the space division factor, which is expressed as follows:
[0021] ;
[0022] ;
[0023] Among them, Y1 represents the partition width determined for the first data dimension, and Y2 represents the partition width determined for the second data dimension. Indicates the maximum value of the first data dimension, Indicates the minimum value of the first data dimension, Indicates the maximum value of the second data dimension, represents the minimum value of the second data dimension, a1 represents the adjustment factor of the first data dimension, a2 represents the adjustment factor of the second data dimension, and e represents a natural constant;
[0024] Step S34: space division, based on the determined space division width, divide the two dimensions until there is a last piece of space that is not enough to be divided, and then evenly distribute the space to the divided space to complete the final division. The final divided space width is expressed as follows:
[0025] ;
[0026] ;
[0027] Wherein, Y1f and Y2f represent the final division widths of the first and second dimensional spaces, respectively, and D1 and D2 represent the number of division regions of the first and second dimensional spaces, respectively;
[0028] Step S35: Determine the initial cluster center. First calculate the average position in each divided area, and then calculate the overall average position of all data points in the entire data space. The average position of each divided area is averaged with the overall average position to obtain the comprehensive average position corresponding to each area. Randomly select A comprehensive average positions as the initial cluster center.
[0029] Furthermore, in step S4, the clustering specifically includes the following steps:
[0030] Step S41: Design a distance function, which is expressed as follows:
[0031] ;
[0032] Among them, i represents the index of the data point, j represents the index of the cluster center, represents the location of data point i, represents the location of cluster center j, represents the distance from data point i to cluster center j, It represents the Euclidean distance between the data point closest to data point i and the data point closest to cluster center j. represents the Euclidean distance from data point i to the nearest data point, represents the Euclidean distance from the cluster center j to the nearest data point, Represents the square of the Euclidean distance between data point i and cluster center j;
[0033] Step S42: Calculate the degree of association, which is expressed as follows:
[0034] ;
[0035] Among them, k represents the index of the cluster center, represents the correlation between data point i and cluster center j, represents the number of cluster centers, Represents the distance between the location of data point i and the location of cluster center j, represents the distance between the location of data point i and the location of cluster center k, and w represents the adjustment factor;
[0036] Step S43: assigning data points, calculating the correlation between each data point and each cluster center, and assigning the data point to the cluster with the highest correlation;
[0037] Step S44: Calculate the local density, expressed as follows:
[0038] ;
[0039] in, represents the local density of data point i, represents the local area width factor, Indicates the nearest data point i A set of U data points, sl represents the index of a data point in the set of U data points, represents the position of the data point sl in the set of U data points, Indicates that it belongs to the symbol, Represents the Euclidean distance from data point i to data point sl;
[0040] Step S45: removing noise data points, setting a threshold value of the correlation between the data point and the cluster center and a threshold value of the local density of the data point, and removing the data point when both the correlation between the data point and the cluster center and the local density of the data point are smaller than the threshold value;
[0041] Step S46: cluster center update, expressed as follows:
[0042] ;
[0043] in, represents the updated cluster center, G represents the number of data points clustered around cluster center j, e represents a natural constant, represents the smoothing factor;
[0044] Step S47: Clustering, repeating steps S43-S46 until the clustering result no longer changes.
[0045] Furthermore, in step S5, the optimization of clustering parameters specifically includes the following steps:
[0046] Step S51: Determine optimization parameters, including the number of clusters, adjustment factors and smoothing factors;
[0047] Step S52: clustering quality assessment, expressed as follows:
[0048] ;
[0049] Among them, v represents the number of clusters, Indicates the clustering quality when the number of clusters is v, represents the quality assessment weight, Stab(v) represents the clustering stability when the number of clusters is v, which is the maximum number of times two data points are assigned to the same cluster during the clustering process. represents the between-cluster sum of squares, represents the within-cluster sum of squares, represents the total number of data points;
[0050] Step S53: determining the number of clusters, and selecting the number of clusters with the highest clustering quality within the value range of the number of clusters as the final number of clusters;
[0051] Step S54: constructing a parameter search space, and constructing a parameter search space of an adjustment factor and a smoothing factor under the background of determining the number of clusters;
[0052] Step S55: Design a search guide factor, reduce the search step size as the number of searches increases, and seek detailed exploration, as shown below:
[0053] ;
[0054] Where t represents the number of parameter searches, T represents the maximum number of parameter searches, represents the search guide factor during the t+1th parameter search, and e represents a natural constant;
[0055] Step S56: Design a random optimal guided search function, which is expressed as follows:
[0056] ;
[0057] in, It represents the parameter position searched by the random optimal guided search function during the t+1th parameter search. represents the parameter position searched by the random optimal guided search function during the t-th parameter search, Indicates the parameter position with the highest fitness at this time, represents the average value of the position of the initial search cluster, r1 represents a random number between 0 and 1;
[0058] Step S57: Cluster parameter search, set the parameter fitness qualified value, perform parameter search through the random optimal guided search function, calculate the fitness of the searched parameters, if there are parameters higher than the qualified value, set the parameters to the optimized parameter output; if the maximum number of parameter searches is reached, search again; otherwise, increase the search number by one and perform the next search.
[0059] Furthermore, in step S6, the power marketing strategy is generated by clustering the user data, and after the clustering is completed, the power consumption characteristic data and power consumption marketing portrait of the user in the same cluster are obtained, and the power consumption marketing strategy of the corresponding user is generated according to the power consumption characteristic data and the power consumption marketing portrait to carry out precise power marketing. At the same time, feedback data is collected and updated to the clustered cluster to continuously enrich and improve the cluster.
[0060] The beneficial effects achieved by the present invention using the above scheme are as follows:
[0061] (1) In view of the problem that the traditional power marketing clustering method has low clustering accuracy, discrimination, stability and effectiveness due to improper selection of initial clustering centers, this scheme more accurately determines the scope and boundaries of the initial clustering through benchmark feature selection and fine data space division, thus avoiding arbitrariness; explicitly calculates the initial width and final width of the spatial division, making the spatial division more precise and better adapted to the distribution characteristics of the data, thus improving the accuracy and discrimination of clustering; and considers the data characteristics of each region, reducing the blindness of the initial clustering center selection, making the clustering process more stable and effective.
[0062] (2) In view of the problems of inaccurate correlation between data points and cluster centers and unreasonable distribution of data points in traditional power marketing clustering methods, which lead to low purity, accuracy, stability and reliability of clustering, this scheme designs a distance function to more accurately reflect the correlation between data points and cluster centers; the calculation of correlation makes the distribution of data points more reasonable and precise; the calculation of local density can better identify the distribution characteristics of data and improve the purity and accuracy of clustering; the cluster center update formula comprehensively considers the correlation and local density factors, so that the adjustment of cluster centers is more in line with the actual situation of data, and the stability and reliability of clustering are improved.
[0063] (3) In view of the problem that clustering parameters are difficult to configure in traditional power marketing clustering methods, this scheme uses a comprehensive clustering quality evaluation to comprehensively consider stability, inter-cluster and intra-cluster differences, and configures the optimal number of clusters in a more systematic and objective manner. It also designs an optimal guided search function to make the search for clustering parameters more efficient, conducts different degrees of detailed exploration at different stages, improves the ability to configure the optimal parameter combination, and increases the flexibility and adaptability of the search. BRIEF DESCRIPTION OF THE DRAWINGS
[0064] Figure 1 A flow chart of the precision power marketing method based on machine learning provided by the present invention;
[0065] Figure 2 is a schematic flow chart of step S3;
[0066] Figure 3 is a schematic flow chart of step S4;
[0067] Figure 4 It is a flowchart diagram of step S5.
[0068] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION
[0069] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.
[0070] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.
[0071] Example 1, see Figure 1 The invention provides a precision power marketing method based on machine learning, which comprises the following steps:
[0072] Step S1: Data collection, collecting users' electricity consumption data, electricity consumption patterns, customer type data, historical payment data, equipment information data, weather-related data, and geographic area data;
[0073] Step S2: Data preprocessing, organizing the data, and comprehensively constructing a user electricity marketing portrait that includes electricity usage habits, consumption capacity, credibility, and potential demand characteristics;
[0074] Step S3: Selecting the initial cluster center by selecting the benchmark feature, calculating the data space partition factor, determining the initial width of the space partition, and space partitioning;
[0075] Step S4: clustering, clustering is performed by designing a distance function, calculating the degree of association, removing noise data points, and updating the cluster center;
[0076] Step S5: Optimizing clustering parameters, optimizing clustering parameters by determining optimization parameters, clustering quality assessment, determining the number of clusters, constructing parameter search space, designing search guide factors, designing random optimal guide search functions and clustering parameter search;
[0077] Step S6: Generate power marketing strategy, cluster the user data, obtain the power consumption characteristic data and power consumption marketing portrait of the users in the same cluster, generate the power consumption marketing strategy for the corresponding users, and conduct precise power marketing.
[0078] Example 2, see Figure 1This embodiment is based on the above embodiment. In step S1, the data collection is to collect the user's electricity consumption data, electricity consumption rules, customer type data, historical payment data, equipment information data, weather-related data, and geographical area data; the electricity consumption data includes monthly electricity consumption, daily electricity consumption, peak and valley electricity consumption, and electricity consumption time period; the customer types include residential users, commercial users, and industrial users; the historical payment data includes the user's payment time, amount, and arrears; the weather-related data includes temperature and humidity; and the geographical area data includes the user's geographical location and regional characteristics.
[0079] Example 3, see Figure 1 This embodiment is based on the above embodiment. In step S2, the data preprocessing specifically includes the following steps:
[0080] Step S21: Data collation, checking the integrity of the data, filling in missing values, checking for outliers in the data, correcting and removing data by comparing with historical data and data of similar users, standardizing the data, and converting the data into a unified metric;
[0081] Step S22: Generate an electricity marketing portrait, analyze the user's electricity usage pattern, determine the characteristics of electricity usage habits, summarize the electricity usage characteristics and needs based on customer types, extract payment credit information from payment data, infer electricity demand based on equipment information, analyze electricity usage differences based on geographical area data, and comprehensively construct a user electricity marketing portrait that includes electricity usage habits, consumption capacity, credibility, and potential demand characteristics, and add this portrait data to the feature data.
[0082] Example 4, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In step S3, the selecting of the initial cluster center specifically includes the following steps:
[0083] Step S31: benchmark feature selection, selecting two features for determining the initial cluster center position, and finding the maximum and minimum values of the two data dimensions in the two features;
[0084] Step S32: Calculate the data space division factor, which is expressed as follows:
[0085] ;
[0086] Among them, v represents the number of clusters, and denote the spatial partitioning factors for the first and second dimensions, respectively, Indicates taking the absolute value, Indicates taking the minimum value;
[0087] Step S33: Determine the initial width of the space division, and determine the initial width of the space division by the space division factor, which is expressed as follows:
[0088] ;
[0089] ;
[0090] Among them, Y1 represents the partition width determined for the first data dimension, and Y2 represents the partition width determined for the second data dimension. Indicates the maximum value of the first data dimension, Indicates the minimum value of the first data dimension, Indicates the maximum value of the second data dimension, represents the minimum value of the second data dimension, a1 represents the adjustment factor of the first data dimension, a2 represents the adjustment factor of the second data dimension, and e represents a natural constant;
[0091] Step S34: space division, based on the determined space division width, divide the two dimensions until there is a last piece of space that is not enough to be divided, and then evenly distribute the space to the divided space to complete the final division. The final divided space width is expressed as follows:
[0092] ;
[0093] ;
[0094] Wherein, Y1f and Y2f represent the final division widths of the first and second dimensional spaces, respectively, and D1 and D2 represent the number of division regions of the first and second dimensional spaces, respectively;
[0095] Step S35: Determine the initial cluster center. First calculate the average position in each divided area, and then calculate the overall average position of all data points in the entire data space. The average position of each divided area is averaged with the overall average position to obtain the comprehensive average position corresponding to each area. Randomly select A comprehensive average positions as the initial cluster center.
[0096] By performing the above operations, in order to address the problems of low clustering accuracy, discrimination, stability and effectiveness caused by improper selection of initial clustering centers in traditional power marketing clustering methods, this solution more accurately determines the scope and boundaries of initial clustering through benchmark feature selection and fine data space division, avoiding arbitrariness; explicitly calculates the initial width and final width of spatial division to make spatial division more precise, better adapt to the distribution characteristics of data, and improve the accuracy and discrimination of clustering; takes into account the data characteristics of each region, reduces the blindness of the initial clustering center selection, and makes the clustering process more stable and effective.
[0097] Example 5, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In step S4, the clustering specifically includes the following steps:
[0098] Step S41: Design a distance function, which is expressed as follows:
[0099] ;
[0100] Among them, i represents the index of the data point, j represents the index of the cluster center, represents the location of data point i, represents the location of cluster center j, represents the distance from data point i to cluster center j, It represents the Euclidean distance between the data point closest to data point i and the data point closest to cluster center j. represents the Euclidean distance from data point i to the nearest data point, represents the Euclidean distance from the cluster center j to the nearest data point, Represents the square of the Euclidean distance between data point i and cluster center j;
[0101] Step S42: Calculate the degree of association, which is expressed as follows:
[0102] ;
[0103] Among them, k represents the index of the cluster center, represents the correlation between data point i and cluster center j, represents the number of cluster centers, Represents the distance between the location of data point i and the location of cluster center j, represents the distance between the location of data point i and the location of cluster center k, and w represents the adjustment factor;
[0104] Step S43: assigning data points, calculating the correlation between each data point and each cluster center, and assigning the data point to the cluster with the highest correlation;
[0105] Step S44: Calculate the local density, expressed as follows:
[0106] ;
[0107] in, represents the local density of data point i, represents the local area width factor, Indicates the nearest data point i A set of U data points, sl represents the index of a data point in the set of U data points, represents the position of the data point sl in the set of U data points, Indicates that it belongs to the symbol, Represents the Euclidean distance from data point i to data point sl;
[0108] Step S45: removing noise data points, setting a threshold value of the correlation between the data point and the cluster center and a threshold value of the local density of the data point, and removing the data point when both the correlation between the data point and the cluster center and the local density of the data point are smaller than the threshold value;
[0109] Step S46: cluster center update, expressed as follows:
[0110] ;
[0111] in, represents the updated cluster center, G represents the number of data points clustered around cluster center j, e represents a natural constant, represents the smoothing factor;
[0112] Step S47: Clustering, repeating steps S43-S46 until the clustering result no longer changes.
[0113] By performing the above operations, in view of the problems of low purity, accuracy, stability and reliability of clustering caused by inaccurate correlation between data points and cluster centers and unreasonable distribution of data points in traditional power marketing clustering methods, this scheme designs a distance function to more accurately reflect the correlation between data points and cluster centers; the calculation of correlation makes the distribution of data points more reasonable and precise; the calculation of local density can better identify the distribution characteristics of data and improve the purity and accuracy of clustering; the cluster center update formula comprehensively considers the correlation and local density factors, so that the adjustment of cluster centers is more in line with the actual situation of data, and the stability and reliability of clustering are improved.
[0114] Example 6, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In step S5, the optimization of clustering parameters specifically includes the following steps:
[0115] Step S51: Determine optimization parameters, including the number of clusters, adjustment factors and smoothing factors;
[0116] Step S52: clustering quality assessment, expressed as follows:
[0117] ;
[0118] Among them, v represents the number of clusters, Indicates the clustering quality when the number of clusters is v, represents the quality assessment weight, Stab(v) represents the clustering stability when the number of clusters is v, which is the maximum number of times two data points are assigned to the same cluster during the clustering process. represents the between-cluster sum of squares, represents the within-cluster sum of squares, represents the total number of data points;
[0119] Step S53: determining the number of clusters, and selecting the number of clusters with the highest clustering quality within the value range of the number of clusters as the final number of clusters;
[0120] Step S54: constructing a parameter search space, and constructing a parameter search space of an adjustment factor and a smoothing factor under the background of determining the number of clusters;
[0121] Step S55: Design a search guide factor, reduce the search step size as the number of searches increases, and seek detailed exploration, as shown below:
[0122] ;
[0123] Where t represents the number of parameter searches, T represents the maximum number of parameter searches, represents the search guide factor during the t+1th parameter search, and e represents a natural constant;
[0124] Step S56: Design a random optimal guided search function, which is expressed as follows:
[0125] ;
[0126] in, It represents the parameter position searched by the random optimal guided search function during the t+1th parameter search. represents the parameter position searched by the random optimal guided search function during the t-th parameter search, Indicates the parameter position with the highest fitness at this time, represents the average value of the position of the initial search cluster, r1 represents a random number between 0 and 1;
[0127] Step S57: Cluster parameter search, set the parameter fitness qualified value, perform parameter search through the random optimal guided search function, calculate the fitness of the searched parameters, if there are parameters higher than the qualified value, set the parameters to the optimized parameter output; if the maximum number of parameter searches is reached, search again; otherwise, increase the search number by one and perform the next search.
[0128] By performing the above operations, in view of the problem that clustering parameters are difficult to configure in traditional power marketing clustering methods, this scheme uses comprehensive clustering quality evaluation, comprehensively considers stability, inter-cluster and intra-cluster differences, and configures the optimal number of clusters in a more systematic and objective manner; designs the optimal guided search function to make the search for clustering parameters more efficient, conducts different degrees of detailed exploration at different stages, improves the ability to configure the optimal parameter combination, and increases the flexibility and adaptability of the search.
[0129] Embodiment 7, see Figure 1 This embodiment is based on the above embodiment. In step S6, the power marketing strategy is generated by clustering the user's data. After the clustering is completed, the power consumption characteristic data and power consumption marketing portrait of the user in the same cluster are obtained, and the power consumption marketing strategy of the corresponding user is generated according to the power consumption characteristic data and the power consumption marketing portrait to perform precise power marketing. At the same time, feedback data is collected and updated to the clustered cluster to continuously enrich and improve the cluster.
[0130] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0131] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.
[0132] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.
Claims
1. A precision power marketing method based on machine learning, characterized in that: The method comprises the following steps: Step S1: Data collection, collecting users' electricity consumption data, electricity consumption patterns, customer type data, historical payment data, equipment information data, weather-related data, and geographic area data; Step S2: Data preprocessing, organizing the data, and comprehensively constructing a user electricity marketing portrait that includes electricity usage habits, consumption capacity, credibility, and potential demand characteristics; Step S3: Selecting the initial cluster center by selecting the benchmark feature, calculating the data space partition factor, determining the initial width of the space partition, and space partitioning; Step S4: clustering, clustering is performed by designing a distance function, calculating the degree of association, removing noise data points, and updating the cluster center; Step S5: Optimizing clustering parameters, optimizing clustering parameters by determining optimization parameters, clustering quality assessment, determining the number of clusters, constructing parameter search space, designing search guide factors, designing random optimal guide search functions and clustering parameter search; Step S6: Generate power marketing strategy, cluster the user data, obtain the power consumption characteristic data and power consumption marketing portrait of the users in the same cluster, generate the power consumption marketing strategy for the corresponding users, and conduct precise power marketing.
2. The precision power marketing method based on machine learning according to claim 1 is characterized by: In step S3, the selection of the initial cluster center specifically includes the following steps: Step S31: benchmark feature selection, selecting two features for determining the initial cluster center position, and finding the maximum and minimum values of the two data dimensions in the two features; Step S32: Calculate the data space division factor, which is expressed as follows: ; Among them, v represents the number of clusters, and denote the spatial partitioning factors for the first and second dimensions, respectively, Indicates taking the absolute value, Indicates taking the minimum value; Step S33: Determine the initial width of the space division, and determine the initial width of the space division by the space division factor, which is expressed as follows: ; ; Among them, Y1 represents the partition width determined for the first data dimension, and Y2 represents the partition width determined for the second data dimension. Indicates the maximum value of the first data dimension, Indicates the minimum value of the first data dimension, Indicates the maximum value of the second data dimension, represents the minimum value of the second data dimension, a1 represents the adjustment factor of the first data dimension, a2 represents the adjustment factor of the second data dimension, and e represents a natural constant; Step S34: space division, based on the determined space division width, divide the two dimensions until there is a last piece of space that is not enough to be divided, and then evenly distribute the space to the divided space to complete the final division. The final divided space width is expressed as follows: ; ; Wherein, Y1f and Y2f represent the final division widths of the first and second dimensional spaces, respectively, and D1 and D2 represent the number of division regions of the first and second dimensional spaces, respectively; Step S35: Determine the initial cluster center. First calculate the average position in each divided area, and then calculate the overall average position of all data points in the entire data space. The average position of each divided area is averaged with the overall average position to obtain the comprehensive average position corresponding to each area. Randomly select A comprehensive average positions as the initial cluster center.
3. The precision power marketing method based on machine learning according to claim 1 is characterized by: In step S4, the clustering specifically includes the following steps: Step S41: Design a distance function, which is expressed as follows: ; Among them, i represents the index of the data point, j represents the index of the cluster center, represents the location of data point i, represents the location of cluster center j, represents the distance from data point i to cluster center j, It represents the Euclidean distance between the data point closest to data point i and the data point closest to cluster center j. represents the Euclidean distance from data point i to the nearest data point, represents the Euclidean distance from the cluster center j to the nearest data point, Represents the square of the Euclidean distance between data point i and cluster center j; Step S42: Calculate the degree of association, which is expressed as follows: ; Among them, k represents the index of the cluster center, represents the correlation between data point i and cluster center j, represents the number of cluster centers, Represents the distance between the location of data point i and the location of cluster center j, represents the distance between the location of data point i and the location of cluster center k, and w represents the adjustment factor; Step S43: assigning data points, calculating the correlation between each data point and each cluster center, and assigning the data point to the cluster with the highest correlation; Step S44: Calculate the local density, expressed as follows: ; in, represents the local density of data point i, represents the local area width factor, Indicates the nearest data point i A set of U data points, sl represents the index of the data point in the set of U data points, represents the position of the data point sl in the set of U data points, Indicates that it belongs to the symbol, Represents the Euclidean distance from data point i to data point sl; Step S45: removing noise data points, setting a threshold value of the correlation between the data point and the cluster center and a threshold value of the local density of the data point, and removing the data point when both the correlation between the data point and the cluster center and the local density of the data point are smaller than the threshold value; Step S46: cluster center update, expressed as follows: ; in, represents the updated cluster center, G represents the number of data points clustered around cluster center j, e represents a natural constant, represents the smoothing factor; Step S47: Clustering, repeating steps S43-S46 until the clustering result no longer changes.
4. The precision power marketing method based on machine learning according to claim 1 is characterized by: In step S5, the optimization of clustering parameters specifically includes the following steps: Step S51: Determine optimization parameters, including the number of clusters, adjustment factors and smoothing factors; Step S52: clustering quality assessment, expressed as follows: ; Among them, v represents the number of clusters, Indicates the clustering quality when the number of clusters is v, represents the quality assessment weight, Stab(v) represents the clustering stability when the number of clusters is v, which is the maximum number of times two data points are assigned to the same cluster during the clustering process. represents the between-cluster sum of squares, represents the within-cluster sum of squares, represents the total number of data points; Step S53: determining the number of clusters, and selecting the number of clusters with the highest clustering quality within the value range of the number of clusters as the final number of clusters; Step S54: constructing a parameter search space, and constructing a parameter search space of an adjustment factor and a smoothing factor under the background of determining the number of clusters; Step S55: Design a search guide factor, reduce the search step size as the number of searches increases, and seek detailed exploration, as shown below: ; Where t represents the number of parameter searches, T represents the maximum number of parameter searches, represents the search guide factor during the t+1th parameter search, and e represents a natural constant; Step S56: Design a random optimal guided search function, which is expressed as follows: ; in, It represents the parameter position searched by the random optimal guided search function during the t+1th parameter search. represents the parameter position searched by the random optimal guided search function during the t-th parameter search, Indicates the parameter position with the highest fitness at this time, represents the average value of the position of the initial search cluster, r1 represents a random number between 0 and 1; Step S57: Cluster parameter search, set the parameter fitness qualified value, perform parameter search through the random optimal guided search function, calculate the fitness of the searched parameters, if there are parameters higher than the qualified value, set the parameters to the optimized parameter output; if the maximum number of parameter searches is reached, search again; otherwise, increase the search number by one and perform the next search.
5. The precise power marketing method based on machine learning according to claim 1 is characterized by: In step S2, the data preprocessing specifically includes the following steps: Step S21: Data collation, checking the integrity of the data, filling in missing values, checking for outliers in the data, correcting and removing data by comparing with historical data and data of similar users, standardizing the data, and converting the data into a unified metric; Step S22: Generate an electricity marketing portrait, analyze the user's electricity usage pattern, determine the characteristics of electricity usage habits, summarize the electricity usage characteristics and needs based on customer types, extract payment credit information from payment data, infer electricity demand based on equipment information, analyze electricity usage differences based on geographical area data, and comprehensively construct a user electricity marketing portrait that includes electricity usage habits, consumption capacity, credibility, and potential demand characteristics, and add this portrait data to the feature data.
6. The precision power marketing method based on machine learning according to claim 1 is characterized by: In step S1, the data collection is to collect the user's electricity consumption data, electricity consumption patterns, customer type data, historical payment data, equipment information data, weather-related data, and geographic area data; the electricity consumption data includes monthly electricity consumption, daily electricity consumption, peak and valley electricity consumption, and electricity consumption time period; the customer types include residential users, commercial users, and industrial users; the historical payment data includes the user's payment time, amount, and arrears; the weather-related data includes temperature and humidity; and the geographic area data includes the user's geographical location and regional characteristics.
7. The precise power marketing method based on machine learning according to claim 1 is characterized by: In step S6, the power marketing strategy is generated by clustering the user's data. After the clustering is completed, the power consumption characteristic data and power consumption marketing portrait of the user in the same cluster are obtained, and the power consumption marketing strategy of the corresponding user is generated according to the power consumption characteristic data and the power consumption marketing portrait to carry out precise power marketing. At the same time, feedback data is collected and updated to the clustered cluster to continuously enrich and improve the cluster.
Citation Information
Cited By
Precise power marketing prediction method based on machine learning
CN121599707A
