Power marketing recommendation method and system based on big data
By introducing feedback mechanisms and marketing strategy effect prediction in power marketing recommendations, combining substitute populations, evolution factors and local search algorithms, the problem that power marketing recommendation solutions cannot meet user needs and unstable clustering results is solved, and the accuracy of recommendations is improved and user satisfaction.
Patent Information
- Application Number
- CN202411108196.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-15
AI Technical Summary
The existing power marketing recommendation method has the problem of only big data classification for users and not predicting the results of the power marketing plan, which leads to the power marketing recommendation plan that cannot meet the user's needs, and the built-in parameters of the clustering method are improperly set, resulting in clustering. The results are unstable, affecting the quality of clustering results.
Introduce feedback mechanism and marketing strategy effect prediction, use user feedback data to classify users again, and adjust the power marketing recommendation plan based on the marketing strategy effect prediction results, and use substitute populations, evolution factors and local search to improve the algorithm to obtain clustering parameters to find stability cluster variation range.
It improves the accuracy and effectiveness of power marketing recommendations, enhances users' recognition and satisfaction with power products and services, and improves the accuracy of user classification results.
Smart Images

Figure CN118967211A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of power marketing recommendation, and specifically relates to a power marketing recommendation method and system based on big data. Background Art
[0002] With the increasingly fierce competition in the energy market and the advancement of information technology, big data technology has become more mature and widely used. Against this background, power marketing recommendation methods based on big data have emerged. Generally, power marketing recommendation methods use data analysis and machine learning technology to provide each customer with personalized power packages and service recommendations based on their power consumption history and needs. However, in the existing methods suitable for power marketing recommendation, there are problems such as only big data classification of users and no prediction of the results of power marketing plans, which leads to the inability of power marketing recommendation plans to meet user needs; in the existing clustering methods suitable for power marketing recommendation, there are problems such as improper built-in parameter settings, which leads to unstable clustering results, thus affecting the quality of clustering results. 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 method and system for power marketing recommendation based on big data. In view of the fact that the existing methods suitable for power marketing recommendation only classify big data for users and do not predict the results of power marketing plans, resulting in the power marketing recommendation plans being unable to meet the needs of users, this solution creatively introduces a feedback mechanism and marketing strategy effect prediction, classifies users again through user feedback data, and adjusts the power marketing recommendation plans according to the results of marketing strategy effect prediction, thereby improving the accuracy and effectiveness of marketing recommendations, thereby enhancing user recognition and satisfaction with power products and services; in view of the fact that the existing clustering methods suitable for power marketing recommendation have improper built-in parameter settings, resulting in unstable clustering results, thereby affecting the quality of clustering results, this solution uses substitute populations, evolutionary factors and local searches to improve the algorithm for obtaining clustering parameters, finds a stable cluster change range, and improves the accuracy of user classification results.
[0004] The invention provides a method and system for recommending power marketing based on big data, and the method comprises the following steps:
[0005] Step S1: data collection;
[0006] Step S2: data preprocessing;
[0007] Step S3: user classification;
[0008] Step S4: Marketing strategy formulation;
[0009] Step S5: Prediction of marketing strategy effect;
[0010] Step S6: Power marketing recommendation results.
[0011] Furthermore, in step S1, the data collection is specifically to obtain power marketing raw data from the power system by collection; the power marketing raw data includes user data, power product data and user feedback data, the user data includes product user electricity consumption data, user power equipment usage data, user personal information and user historical payment record data, the power product data includes power price packages and related power equipment product data, and the user feedback data includes power marketing plan selection data and user satisfaction data on power recommended plans.
[0012] Furthermore, in step S2, the data preprocessing is used to preprocess the original data, and specifically includes the following steps:
[0013] Step S21: numerical filtering, for removing missing values, abnormal values and duplicate values from the numerical data to obtain a power marketing filtered data set;
[0014] Step S22: One-hot encoding conversion, used to convert the original data into digital form, specifically encoding the power marketing original data set by one-hot encoding method to obtain the power marketing coded data set;
[0015] Step S23: feature selection, which is used to extract effective user features for power marketing recommendation from the original data, specifically, using an adaptive feature selection method to perform feature selection on the power marketing preliminary data set to obtain a feature selection data set;
[0016] Step S24: performing data preprocessing, specifically performing data preprocessing on the power marketing original data set through the numerical filtering, one-hot encoding conversion and feature selection to obtain a preliminary power marketing data set.
[0017] Furthermore, in step S3, the user classification is specifically to classify the user data using a clustering method, including the following steps:
[0018] Step S31: performing a DBSCAN clustering operation on the user data;
[0019] Step S32: Optimizing clustering parameters, including the following steps:
[0020] Step S321: Initialize parameters, population size N, maximum number of iterations t max , search for the lower limit lb and the upper limit ub;
[0021] Step S322: Initializing the population, including the following steps:
[0022] Step S3221: Generate N chaotic sequence values, and initialize the random value to y0; the formula used is as follows:
[0023] ;
[0024] In the formula, represents the i-th chaotic sequence value, represents the i+1th chaotic sequence value;
[0025] Step S3222: Chaotic sequence value conversion search individual; the formula is as follows:
[0026] ;
[0027] In the formula, represents the position of the i-th individual;
[0028] Step S323: population division, calculation of individual fitness values f in the population i , the silhouette coefficient of the clustering result after DBSCAN clustering of user data based on individual positions is used as the fitness value of the individual; the positions and fitness values of all search individuals in the population are copied to the substitute population; the individuals are sorted from best to worst according to the fitness value, and the individual fa with the best fitness and its position X are defined a , the individual fb with the second best fitness and its position X b , the third best individual fc and its position X c , and store its position and fitness value in the substitute population;
[0029] Step S324: Update the search individual position; the formula used is as follows:
[0030] ;
[0031] ;
[0032] ;
[0033] ;
[0034] ;
[0035] In the formula, X1 represents the position of the i-th individual after moving to fa, X2 represents the position of the i-th individual after moving to fb, and X3 represents the position of the i-th individual after moving to fc. represents the position of the i-th individual in the t-th generation population, represents the position of the i-th individual in the t+1 generation population, r1, r2 and r3 represent randomness parameters in the range of [0,1], is a smoothing term; t represents the number of iterations;
[0036] Step S325: Obtain the evolution factor; the formula used is as follows:
[0037] ;
[0038] In the formula, Indicates the evolution factor of the changed T distribution, which is used to adjust the degree of evolution. represents the minimum evolution factor, Indicates the maximum value of evolution factor;
[0039] Step S326: perform evolution operation; the formula used is as follows:
[0040] ;
[0041] ;
[0042] ;
[0043] In the formula, represents the position of the i-th individual in the t+1-th generation mutation population, represents the position of the i-th individual in the t+1 generation evolution population, Pc represents the mutation probability, which is a fixed parameter. represents the fitness value of the position of the i-th individual in the t-th generation population, and r4 represents the randomness parameter in the range of [0,1];
[0044] Step S327: Replacement strategy: re-evaluate the fitness values of all search individuals in the population, compare the fitness values of the evolved search individuals with the original search individuals, and retain the search individuals with the highest fitness values in the current population; perform a one-to-one fitness value comparison between each search individual in the current population and the corresponding search individual in the substitute population, and if the fitness value of the search individual in the current population is higher than that of the search individual in the corresponding substitute population, replace the search individual in the corresponding substitute population;
[0045] Step S328: Local search: select 50% of the individuals with the best fitness for local search, and generate new individuals to replace the eliminated individuals; the formula used is as follows:
[0046] ;
[0047] In the formula, represents the position vector of the local search individual, represents the position vector of the current search individual, Represents the position vector of the nearest neighboring search individual of the current search individual, Represents the fitness value of the position vector of the adjacent search individual, represents the fitness value of the position vector of the current search individual, and r5 represents the randomness parameter in the range of [0,1];
[0048] Step S329: Replace the individual with the best fitness value: Determine the fitness value of the local search individual. The fitness value is better than The fitness value of Replacement of replacement population , otherwise it is discarded; fa, fb and fc are compared with the fitness values of the search individuals in the substitute population, and the better search individuals in the substitute population update fa, fb and fc;
[0049] Step S3210: search determination, specifically, by constructing a search termination condition, performing search determination on the optimal individual position, and obtaining the optimal individual position data setting;
[0050] The search termination conditions include threshold termination and iteration termination;
[0051] The threshold termination is specifically to set the fitness threshold, when the individual fitness value f i When it is above the fitness threshold, the hyperparameter search is completed;
[0052] The iteration termination specifically refers to terminating the iteration and obtaining the optimal individual position when the maximum number of iterations is reached;
[0053] The optimal individual position includes the eps parameter and the MinPts parameter;
[0054] Step S33: User classification: Use the clustering algorithm with optimized eps and MinPts parameters to classify different users in different dimensions to obtain user category data.
[0055] Further, in step S4, the marketing strategy formulation is specifically formulating a personalized power marketing strategy text for the user based on the user classification result;
[0056] The personalized power marketing strategy text content mainly includes various products and value-added services with power as the core for users, covering power supply services, energy-saving products and smart grid recommendation solutions;
[0057] The personalized power marketing strategy text formulation process specifically selects a suitable marketing strategy text template based on the user classification results, adds user personal information and recommended solutions to the selected template, combines user classification information and fills in specific content to generate the final personalized power marketing strategy text.
[0058] Furthermore, in step S5, the marketing strategy effect prediction is specifically to predict the effect of the power marketing recommendation plan for the user in the marketing strategy formulation, which specifically includes the following steps:
[0059] Step S51: Establishing a marketing strategy effect prediction model, including the following steps:
[0060] Step S511: Initialize model parameters, set the number of model hidden units, input vector dimension, output vector dimension and number of training iterations;
[0061] Step S512: Forget gate calculation process, the formula used is as follows:
[0062] ;
[0063] In the formula, f t represents the output of the forget gate; X t Indicates input data; H t-1 Indicates the hidden state at the last moment; W xf and W hf represents the weight matrix corresponding to the forget gate; b f represents the bias parameter of the forget gate; σ represents the Sigmoid function;
[0064] Step S513: Input gate calculation process, the formula used is as follows:
[0065] ;
[0066] In the formula, i t represents the output of the input gate; W xi and W hi represents the corresponding input gate weight matrix; b i Represents the bias parameter of the input gate;
[0067] Step S514: Calculate new cell state candidate values , the formula used is as follows:
[0068] ;
[0069] Where W xc and W hc represents the corresponding cell unit state weight matrix; b c Bias parameter representing the cell unit state;
[0070] Step S515: Cell unit status update process, the formula used is as follows:
[0071] ;
[0072] In the formula, ct Indicates the cell state at the current moment; c t-1 Indicates the cell state at the previous moment;
[0073] Step S516: Output gate calculation process, the formula used is as follows:
[0074] ;
[0075] In the formula, o t Represents the output of the output gate; W xo and W ho It represents the corresponding output gate weight matrix; b o Represents the bias parameter of the output gate;
[0076] Step S517: hidden state calculation process, the formula used is as follows:
[0077] ;
[0078] Where tanh represents the hyperbolic tangent function; h t Indicates the hidden state at the current moment;
[0079] Step S518: Get the current prediction result , the formula used is as follows:
[0080] ;
[0081] Where b v The bias parameter representing the predicted value;
[0082] Step S52: text data cleaning, specifically, representing each text segment in the text data of the power marketing plan selection situation data and the marketing strategy formulation plan content in the user feedback data as a vector, and obtaining historical text feature data and real-time text feature data respectively;
[0083] Step S53: Model training, specifically including the following steps:
[0084] Step S531: data labeling, specifically, dividing the user satisfaction data on the power recommendation scheme in the user feedback data into two labels, namely, satisfactory recommendation and unsatisfactory recommendation;
[0085] Step S532: effect prediction model training, specifically using the historical text feature data to train the marketing strategy effect prediction model to obtain a trained effect prediction model;
[0086] Step S54: predicting the effect of a marketing strategy, specifically using the real-time text feature data as input data of a trained effect prediction model to obtain a marketing strategy effect prediction result.
[0087] Further, in step S6, the power marketing recommendation result, specifically the user's feedback on the power recommendation plan, specifically includes the following steps:
[0088] Step S61: power marketing recommendation, specifically, power marketing recommendation is made based on the marketing strategy effect prediction results. If the prediction results are met, the power marketing plan is recommended to the user through phone calls and text messages. If the prediction results are not met, the user is reclassified based on user feedback data;
[0089] Step S62: Optimization of recommendation schemes, specifically, clear feedback from users on power recommendation schemes, accurate classification of user categories, continuous adjustment of user classification results and optimization of power marketing recommendation schemes, to improve the accuracy and effectiveness of personalized recommendations.
[0090] The technical solution adopted by the present invention is as follows: the power marketing recommendation system based on big data provided by the present invention includes a data acquisition module, a data preprocessing module, a user classification module, a marketing strategy formulation module, a marketing strategy effect prediction module and a power marketing recommendation result module;
[0091] The data acquisition module acquires power marketing raw data from the power system by acquisition, wherein the power marketing raw data includes user data, power product data and user feedback data, and sends the power marketing raw data to the data preprocessing module;
[0092] The data preprocessing module receives the original power marketing data sent by the data acquisition module, performs data preprocessing on the original power marketing data to obtain preliminary power marketing data, and sends the preliminary power marketing data to the user classification module;
[0093] The user classification module receives the preliminary power marketing data sent by the data preprocessing module, and performs user classification on the user data to obtain a user classification data set, and sends the user classification data set to the marketing strategy formulation module;
[0094] The marketing strategy formulation module receives the user classification data set sent by the user classification module, formulates a personalized power marketing strategy text according to the user classification result in the user classification data set, obtains the marketing strategy text content, and sends the marketing strategy text content to the marketing strategy effect prediction module;
[0095] The marketing strategy effect prediction module receives the marketing strategy text content sent by the marketing strategy formulation module, and performs effect prediction on the power marketing recommendation plan of the user in the marketing strategy formulation, obtains the marketing strategy effect prediction result, and sends the marketing strategy effect prediction result to the power marketing recommendation result module;
[0096] The power marketing recommendation result module receives the marketing strategy effect prediction result sent by the marketing strategy effect prediction module, and recommends personalized power marketing plans to users, obtains power marketing plans and user feedback data, and sends the user feedback data to the data acquisition module.
[0097] The beneficial effects achieved by the present invention using the above scheme are as follows:
[0098] (1) In view of the problem that the existing power marketing recommendation methods and systems based on big data only classify users’ big data and fail to predict the results of power marketing plans, the power marketing recommendation plans cannot meet the needs of users. This solution creatively introduces a feedback mechanism and marketing strategy effect prediction, classifies users again through user feedback data, and adjusts the power marketing recommendation plans according to the results of marketing strategy effect prediction, thereby improving the accuracy and effectiveness of marketing recommendations and enhancing users’ recognition and satisfaction with power products and services.
[0099] (2) In view of the problem that improper built-in parameter settings exist in general clustering methods in existing power marketing recommendation methods and systems based on big data, resulting in unstable clustering results, thus affecting the quality of clustering results, this scheme adopts substitute population, evolutionary factor and local search to improve the algorithm for obtaining clustering parameters, find a stable cluster variation range, and improve the accuracy of user classification results. BRIEF DESCRIPTION OF THE DRAWINGS
[0100] Figure 1 A schematic diagram of the flow of the power marketing recommendation method based on big data provided by the present invention;
[0101] Figure 2 A schematic diagram of a power marketing recommendation system based on big data provided by the present invention;
[0102] Figure 3 A schematic diagram of the process of user classification in step S3;
[0103] Figure 4 This is a flowchart of the marketing strategy effect prediction in step S5;
[0104] Figure 5 This is a flow chart of power marketing recommendation results in step S6;
[0105] 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
[0106] 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.
[0107] 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.
[0108] Example 1, see Figure 1 The invention provides a power marketing recommendation method based on big data, which comprises the following steps:
[0109] Step S1: data collection;
[0110] Step S2: data preprocessing;
[0111] Step S3: user classification;
[0112] Step S4: Marketing strategy formulation;
[0113] Step S5: Prediction of marketing strategy effect;
[0114] Step S6: Power marketing recommendation results.
[0115] Example 2, see Figure 1 and Figure 2 In step S1, the data collection is used to collect the original data required for power marketing recommendation, specifically, to obtain the power marketing original data from the power system through collection;
[0116] The power marketing raw data includes user data, power product data and user feedback data;
[0117] The user data includes product user electricity consumption data, user power equipment data, user personal information and user historical payment record data;
[0118] The power product data includes power price packages and related power equipment product data, and the user feedback data includes power marketing plan selection data and user satisfaction data on power recommended plans.
[0119] Example 3, see Figure 1and Figure 2 This embodiment is based on the above embodiment. In step S2, the data preprocessing is used to preprocess the original data, and specifically includes the following steps:
[0120] Step S21: numerical filtering, for removing missing values, abnormal values and duplicate values from the numerical data to obtain a power marketing filtered data set;
[0121] Step S22: One-hot encoding conversion, used to convert the original data into digital form, specifically encoding the power marketing original data set by one-hot encoding method to obtain the power marketing coded data set;
[0122] Step S23: feature selection, which is used to extract effective user features for power marketing recommendation from the original data, specifically, using an adaptive feature selection method to perform feature selection on the power marketing preliminary data set to obtain a feature selection data set;
[0123] Step S24: performing data preprocessing, specifically performing data preprocessing on the power marketing original data set through the numerical filtering, one-hot encoding conversion and feature selection to obtain a preliminary power marketing data set.
[0124] Example 4, see Figure 1 , Figure 2 and Figure 3 This embodiment is based on the above embodiment. In step S3, the user classification is specifically to classify the user data using a clustering method, including the following steps:
[0125] Step S31: performing a DBSCAN clustering operation on the user data;
[0126] Step S32: Optimizing clustering parameters, including the following steps:
[0127] Step S321: Initialize parameters, population size N, maximum number of iterations t max , search for the lower limit lb and the upper limit ub;
[0128] Step S322: Initializing the population, including the following steps:
[0129] Step S3221: Generate N chaotic sequence values, and initialize the random value to y0; the formula used is as follows:
[0130] ;
[0131] In the formula, represents the i-th chaotic sequence value, represents the i+1th chaotic sequence value;
[0132] Step S3222: Chaotic sequence value conversion search individual; the formula is as follows:
[0133] ;
[0134] In the formula, represents the position of the i-th individual;
[0135] Step S323: population division, calculation of individual fitness values f in the population i , the silhouette coefficient of the clustering result after DBSCAN clustering of user data based on individual positions is used as the fitness value of the individual; the positions and fitness values of all search individuals in the population are copied to the substitute population; the individuals are sorted from best to worst according to the fitness value, and the individual fa with the best fitness and its position X are defined a , the individual fb with the second best fitness and its position X b , the third best individual fc and its position X c , and store its position and fitness value in the substitute population;
[0136] Step S324: Update the search individual position; the formula used is as follows:
[0137] ;
[0138] ;
[0139] ;
[0140] ;
[0141] ;
[0142] In the formula, X1 represents the position of the i-th individual after moving to fa, X2 represents the position of the i-th individual after moving to fb, and X3 represents the position of the i-th individual after moving to fc. represents the position of the i-th individual in the t-th generation population, represents the position of the i-th individual in the t+1 generation population, r1, r2 and r3 represent randomness parameters in the range of [0,1], is a smoothing term; t represents the number of iterations;
[0143] Step S325: Obtain the evolution factor; the formula used is as follows:
[0144] ;
[0145] In the formula, Indicates the evolution factor of the changed T distribution, which is used to adjust the degree of evolution. represents the minimum value of the evolution factor, Indicates the maximum value of evolution factor;
[0146] Step S326: perform evolution operation; the formula used is as follows:
[0147] ;
[0148] ;
[0149] ;
[0150] In the formula, represents the position of the i-th individual in the t+1-th generation mutation population, represents the position of the i-th individual in the t+1 generation evolution population, Pc represents the mutation probability, which is a fixed parameter. represents the fitness value of the position of the i-th individual in the t-th generation population, and r4 represents the randomness parameter in the range of [0,1];
[0151] Step S327: Replacement strategy: re-evaluate the fitness values of all search individuals in the population, compare the fitness values of the evolved search individuals with the original search individuals, and retain the search individuals with the highest fitness values in the current population; perform a one-to-one fitness value comparison between each search individual in the current population and the corresponding search individual in the substitute population, and if the fitness value of the search individual in the current population is higher than that of the search individual in the corresponding substitute population, replace the search individual in the corresponding substitute population;
[0152] Step S328: Local search: select 50% of the individuals with the best fitness for local search, and generate new individuals to replace the eliminated individuals; the formula used is as follows:
[0153] ;
[0154] In the formula, represents the position vector of the local search individual, represents the position vector of the current search individual, Represents the position vector of the nearest neighboring search individual of the current search individual, Represents the fitness value of the position vector of the adjacent search individual, represents the fitness value of the position vector of the current search individual, and r5 represents the randomness parameter in the range of [0,1];
[0155] Step S329: Replace the individual with the best fitness value: Determine the fitness value of the local search individual. The fitness value is better than The fitness value of Replacement of replacement population , otherwise it is discarded; fa, fb and fc are compared with the fitness values of the search individuals in the substitute population, and the better search individuals in the substitute population update fa, fb and fc;
[0156] Step S3210: search determination, specifically, by constructing a search termination condition, performing search determination on the optimal individual position, and obtaining the optimal individual position data setting;
[0157] The search termination conditions include threshold termination and iteration termination;
[0158] The threshold termination is specifically to set the fitness threshold, when the individual fitness value f i When it is above the fitness threshold, the hyperparameter search is completed;
[0159] The iteration termination specifically refers to terminating the iteration and obtaining the optimal individual position when the maximum number of iterations is reached;
[0160] The optimal individual position includes the eps parameter and the MinPts parameter;
[0161] Step S33: User classification: Use the clustering algorithm with optimized eps and MinPts parameters to classify different users in different dimensions to obtain user category data.
[0162] By performing the above operations, the existing clustering method suitable for power marketing recommendation has the problem that the built-in parameters are improperly set, resulting in unstable clustering results, thus affecting the quality of clustering results. This scheme adopts substitute population, evolutionary factor and local search to improve the algorithm for obtaining clustering parameters, find a stable cluster change range, and improve the accuracy of user classification results.
[0163] Example 5, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, in step S4, the marketing strategy formulation is specifically formulating a personalized power marketing strategy text for the user based on the user classification result;
[0164] The personalized power marketing strategy text content mainly includes various products and value-added services with power as the core for users, covering power supply services, energy-saving products and smart grid recommendation solutions;
[0165] The personalized power marketing strategy text formulation process specifically selects a suitable marketing strategy text template based on the user classification results, adds user personal information and recommended solutions to the selected template, combines user classification information and fills in specific content to generate the final personalized power marketing strategy text.
[0166] Example 6, see Figure 1 , Figure 2 and Figure 4 This embodiment is based on the above embodiment. In step S5, the marketing strategy effect prediction is specifically to predict the effect of the power marketing recommendation plan for the user in the marketing strategy formulation, which specifically includes the following steps:
[0167] Step S51: Establishing a marketing strategy effect prediction model, including the following steps:
[0168] Step S511: Initialize model parameters, set the number of model hidden units, input vector dimension, output vector dimension and number of training iterations;
[0169] Step S512: Forget gate calculation process, the formula used is as follows:
[0170] ;
[0171] In the formula, f t represents the output of the forget gate; X t Indicates input data; H t-1 Indicates the hidden state at the last moment; W xf and W hf represents the weight matrix corresponding to the forget gate; b f represents the bias parameter of the forget gate; σ represents the Sigmoid function;
[0172] Step S513: Input gate calculation process, the formula used is as follows:
[0173] ;
[0174] In the formula, i t represents the output of the input gate; W xi and W hi represents the corresponding input gate weight matrix; b i Represents the bias parameter of the input gate;
[0175] Step S514: Calculate new cell state candidate values , the formula used is as follows:
[0176] ;
[0177] Where W xc and W hc represents the corresponding cell unit state weight matrix; b c Bias parameter representing the cell unit state;
[0178] Step S515: Cell unit status update process, the formula used is as follows:
[0179] ;
[0180] In the formula, c t Indicates the cell state at the current moment; c t-1 Indicates the cell state at the previous moment;
[0181] Step S516: Output gate calculation process, the formula used is as follows:
[0182] ;
[0183] In the formula, o t Represents the output of the output gate; W xo and W ho It represents the corresponding output gate weight matrix; b o Represents the bias parameter of the output gate;
[0184] Step S517: hidden state calculation process, the formula used is as follows:
[0185] ;
[0186] In the formula, tanh represents the hyperbolic tangent function; h t Indicates the hidden state at the current moment;
[0187] Step S518: Get the current prediction result , the formula used is as follows:
[0188] ;
[0189] Where b v The bias parameter representing the predicted value;
[0190] Step S52: text data cleaning, specifically, representing each text segment in the text data of the power marketing plan selection situation data and the marketing strategy formulation plan content in the user feedback data as a vector, and obtaining historical text feature data and real-time text feature data respectively;
[0191] Step S53: Model training, specifically including the following steps:
[0192] Step S531: data labeling, specifically, dividing the user satisfaction data on the power recommendation scheme in the user feedback data into two labels, namely, satisfactory recommendation and unsatisfactory recommendation;
[0193] Step S532: effect prediction model training, specifically using the historical text feature data to train the marketing strategy effect prediction model to obtain a trained effect prediction model;
[0194] Step S54: predicting the effect of a marketing strategy, specifically using the real-time text feature data as input data of a trained effect prediction model to obtain a marketing strategy effect prediction result.
[0195] By performing the above operations, the existing power marketing recommendation method and system based on big data has the problem of only classifying big data for users and not predicting the results of power marketing plans, resulting in the power marketing recommendation plan being unable to meet the needs of users. This solution creatively introduces a feedback mechanism and marketing strategy effect prediction, classifies users again through user feedback data, and adjusts the power marketing recommendation plan according to the results of marketing strategy effect prediction, thereby improving the accuracy and effectiveness of marketing recommendations, thereby enhancing user recognition and satisfaction with power products and services.
[0196] Embodiment 7, see Figure 1 , Figure 2 ,and Figure 5 This embodiment is based on the above embodiment. In step S6, the power marketing recommendation result is specifically the user's feedback on the power recommendation plan, which specifically includes the following steps:
[0197] Step S61: power marketing recommendation, specifically, power marketing recommendation is made based on the marketing strategy effect prediction results. If the prediction results are met, the power marketing plan is recommended to the user through phone calls and text messages. If the prediction results are not met, the user is reclassified based on user feedback data;
[0198] Step S62: Optimization of recommendation schemes, specifically, clear feedback from users on power recommendation schemes, accurate classification of user categories, continuous adjustment of user classification results and optimization of power marketing recommendation schemes, to improve the accuracy and effectiveness of personalized recommendations.
[0199] Embodiment 8, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the power marketing recommendation system based on big data provided by the present invention includes a data acquisition module, a data preprocessing module, a user classification module, a marketing strategy formulation module, a marketing strategy effect prediction module and a power marketing recommendation result module;
[0200] The data acquisition module acquires power marketing raw data from the power system by acquisition, wherein the power marketing raw data includes user data, power product data and user feedback data, and sends the power marketing raw data to the data preprocessing module;
[0201] The data preprocessing module receives the original power marketing data sent by the data acquisition module, performs data preprocessing on the original power marketing data to obtain preliminary power marketing data, and sends the preliminary power marketing data to the user classification module;
[0202] The user classification module receives the preliminary power marketing data sent by the data preprocessing module, and performs user classification on the user data to obtain a user classification data set, and sends the user classification data set to the marketing strategy formulation module;
[0203] The marketing strategy formulation module receives the user classification data set sent by the user classification module, formulates a personalized power marketing strategy text according to the user classification result in the user classification data set, obtains the marketing strategy text content, and sends the marketing strategy text content to the marketing strategy effect prediction module;
[0204] The marketing strategy effect prediction module receives the marketing strategy text content sent by the marketing strategy formulation module, and performs effect prediction on the power marketing recommendation plan of the user in the marketing strategy formulation, obtains the marketing strategy effect prediction result, and sends the marketing strategy effect prediction result to the power marketing recommendation result module;
[0205] The power marketing recommendation result module receives the marketing strategy effect prediction result sent by the marketing strategy effect prediction module, and recommends personalized power marketing plans to users, obtains power marketing plans and user feedback data, and sends the user feedback data to the data acquisition module.
[0206] 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.
[0207] 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.
[0208] 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. Power marketing recommendation method based on big data, Features: The method comprises the following steps: Step S1: data collection; Step S2: data preprocessing; Step S3: user classification; Step S4: Marketing strategy formulation; Step S5: Prediction of marketing strategy effect; Step S6: power marketing recommendation results; In step S1, the data collection is used to collect the original data required for power marketing recommendation, specifically, to obtain the power marketing original data from the power system through collection; the power marketing original data includes user data, power product data and user feedback data, the user data includes product user power consumption data, user power equipment data, user personal information and user historical payment record data, the power product data includes power price packages and related power equipment product data, and the user feedback data includes power marketing plan selection data and user satisfaction data on power recommendation plans; In step S2, the data preprocessing is used to preprocess the original data, specifically, to preprocess the original data set of power marketing to obtain a preliminary data set of power marketing; In step S3, the user classification is used to classify power marketing users, specifically by using a clustering method to classify the user data to obtain a user classification data set; In step S4, the marketing strategy formulation is used for formulating marketing strategy texts, specifically formulating personalized power marketing strategy texts for users based on user classification results to obtain a personalized power marketing strategy text data set; In step S5, the marketing strategy effect prediction is used to predict the effect of the power marketing recommendation plan for the user in the marketing strategy formulation, specifically by establishing a marketing strategy effect prediction model and performing model training, using the trained model to perform effect prediction, and obtaining effect prediction result data; In step S6, the power marketing recommendation result is specifically to optimize the recommendation scheme based on the feedback data of the power recommendation scheme.
2. According to the big data-based power marketing recommendation method of claim 1, Features: In step S2, the data preprocessing specifically includes the following steps: Step S21: numerical filtering, for removing missing values, abnormal values and duplicate values from the numerical data to obtain a power marketing filtered data set; Step S22: One-hot encoding conversion, used to convert the original data into digital form, specifically encoding the power marketing original data set by one-hot encoding method to obtain the power marketing coded data set; Step S23: feature selection, which is used to extract effective user features for power marketing recommendation from the original data, specifically, using an adaptive feature selection method to perform feature selection on the power marketing preliminary data set to obtain a feature selection data set; Step S24: performing data preprocessing, specifically performing data preprocessing on the power marketing original data set through the numerical filtering, one-hot encoding conversion and feature selection to obtain a preliminary power marketing data set.
3. According to the big data-based power marketing recommendation method of claim 2, Features: In step S3, the user classification specifically includes the following steps: Step S31: performing a DBSCAN clustering operation on the user data; Step S32: Optimizing clustering parameters, including the following steps: Step S321: Initialize parameters, population size N, maximum number of iterations t max , search for the lower limit lb and upper limit ub; Step S322: Initializing the population, including the following steps: Step S3221: Generate N chaotic sequence values, and initialize the random values to y 0 ; The formula used is as follows: ; In the formula, represents the i-th chaotic sequence value, represents the i+1th chaotic sequence value; Step S3222: Chaotic sequence value conversion search individual; the formula is as follows: ; In the formula, represents the position of the i-th individual; Step S323: population division, calculation of individual fitness values f in the population i , the silhouette coefficient of the clustering result after DBSCAN clustering of user data based on individual positions is used as the fitness value of the individual; the positions and fitness values of all search individuals in the population are copied to the substitute population; the individuals are sorted from best to worst according to the fitness value, and the individual fa with the best fitness and its position X are defined a , the individual fb with the second best fitness and its position X b , the third best individual fc and its position X c , and store its position and fitness value in the substitute population; Step S324: Update the search individual position; the formula used is as follows: ; ; ; ; ; In the formula, X1 represents the position of the i-th individual after moving to fa, X2 represents the position of the i-th individual after moving to fb, and X3 represents the position of the i-th individual after moving to fc. represents the position of the i-th individual in the t-th generation population, represents the position of the i-th individual in the t+1 generation population, r1, r2 and r3 represent randomness parameters in the range of [0,1], is a smoothing term; t represents the number of iterations; Step S325: Obtain the evolution factor; the formula used is as follows: ; In the formula, Indicates the evolution factor of the changed T distribution, which is used to adjust the degree of evolution. represents the minimum evolution factor, Indicates the maximum value of evolution factor; Step S326: perform evolution operation; the formula used is as follows: ; ; ; In the formula, represents the position of the i-th individual in the t+1-th generation mutation population, represents the position of the i-th individual in the t+1 generation evolution population, Pc represents the mutation probability, which is a fixed parameter. represents the fitness value of the position of the i-th individual in the t-th generation population, and r4 represents the randomness parameter in the range of [0,1]; Step S327: Replacement strategy: re-evaluate the fitness values of all search individuals in the population, compare the fitness values of the evolved search individuals with the original search individuals, and retain the search individuals with the highest fitness values in the current population; perform a one-to-one fitness value comparison between each search individual in the current population and the corresponding search individual in the substitute population, and if the fitness value of the search individual in the current population is higher than that of the search individual in the corresponding substitute population, replace the search individual in the corresponding substitute population; Step S328: Local search: select 50% of the individuals with the best fitness for local search, and generate new individuals to replace the eliminated individuals; the formula used is as follows: ; In the formula, represents the position vector of the local search individual, represents the position vector of the current search individual, Represents the position vector of the nearest neighboring search individual of the current search individual, Represents the fitness value of the position vector of the adjacent search individual, represents the fitness value of the position vector of the current search individual, and r5 represents the randomness parameter in the range of [0,1]; Step S329: Replace the individual with the best fitness value: Determine the fitness value of the local search individual. The fitness value is better than The fitness value of Replacement of replacement population , otherwise it is discarded; fa, fb and fc are compared with the fitness values of the search individuals in the substitute population, and the better search individuals in the substitute population update fa, fb and fc; Step S3210: search determination, specifically, by constructing a search termination condition, performing search determination on the optimal individual position, and obtaining the optimal individual position data setting; The search termination conditions include threshold termination and iteration termination; The threshold termination is specifically to set the fitness threshold, when the individual fitness value f i When it is above the fitness threshold, the hyperparameter search is completed; The iteration termination specifically refers to terminating the iteration and obtaining the optimal individual position when the maximum number of iterations is reached; The optimal individual position includes the eps parameter and the MinPts parameter; Step S33: User classification: Use the clustering algorithm with optimized eps and MinPts parameters to classify different users in different dimensions to obtain user category data.
4. According to the big data-based power marketing recommendation method of claim 3, Features: In step S4, the marketing strategy formulation is specifically formulating a personalized power marketing strategy text for the user based on the user classification result; The personalized power marketing strategy text content mainly includes various products and value-added services with power as the core for users, covering power supply services, energy-saving products and smart grid recommendation solutions; The personalized power marketing strategy text formulation process specifically selects a suitable marketing strategy text template based on the user classification results, adds user personal information and recommended solutions to the selected template, combines user classification information and fills in specific content to generate the final personalized power marketing strategy text.
5. According to the big data-based power marketing recommendation method of claim 4, Features: In step S5, the marketing strategy effect prediction includes the following steps: Step S51: Establishing a marketing strategy effect prediction model, including the following steps: Step S511: Initialize model parameters, set the number of model hidden units, input vector dimension, output vector dimension and number of training iterations; Step S512: Forget gate calculation process, the formula used is as follows: ; In the formula, f t represents the output of the forget gate; X t Indicates input data; H t-1 Indicates the hidden state at the last moment; W xf and W hf represents the weight matrix corresponding to the forget gate; b f represents the bias parameter of the forget gate; σ represents the Sigmoid function; Step S513: Input gate calculation process, the formula used is as follows: ; In the formula, i t represents the output of the input gate; W xi and W hi represents the corresponding input gate weight matrix; b i Represents the bias parameter of the input gate; Step S514: Calculate new cell state candidate values , the formula used is as follows: ; Where W xc and W hc represents the corresponding cell unit state weight matrix; b c Bias parameter representing the cell unit state; Step S515: Cell unit status update process, the formula used is as follows: ; In the formula, c t Indicates the cell state at the current moment; c t-1 Indicates the cell state at the previous moment; Step S516: Output gate calculation process, the formula used is as follows: ; In the formula, o t Represents the output of the output gate; W xo and W ho It represents the corresponding output gate weight matrix; b o Represents the bias parameter of the output gate; Step S517: hidden state calculation process, the formula used is as follows: ; In the formula, tanh represents the hyperbolic tangent function; h t Indicates the hidden state at the current moment; Step S518: Get the current prediction result , the formula used is as follows: ; Where b v The bias parameter representing the predicted value; Step S52: text data cleaning, specifically, representing each text segment in the text data of the power marketing plan selection situation data and the marketing strategy formulation plan content in the user feedback data as a vector, and obtaining historical text feature data and real-time text feature data respectively; Step S53: Model training, specifically including the following steps: Step S531: data labeling, specifically, dividing the user satisfaction data on the power recommendation scheme in the user feedback data into two labels, namely, satisfactory recommendation and unsatisfactory recommendation; Step S532: effect prediction model training, specifically using the historical text feature data to train the marketing strategy effect prediction model to obtain a trained effect prediction model; Step S54: predicting the effect of a marketing strategy, specifically using the real-time text feature data as input data of a trained effect prediction model to obtain a marketing strategy effect prediction result.
6. According to the big data-based power marketing recommendation method of claim 5, Features: In step S6, the power marketing recommendation result specifically includes the following steps: Step S61: power marketing recommendation, specifically, power marketing recommendation is made based on the marketing strategy effect prediction results. If the prediction results are met, the power marketing plan is recommended to the user through phone calls and text messages. If the prediction results are not met, the user is reclassified based on user feedback data; Step S62: Optimization of recommendation schemes, specifically, clear feedback from users on power recommendation schemes, accurate classification of user categories, continuous adjustment of user classification results and optimization of power marketing recommendation schemes, to improve the accuracy and effectiveness of personalized recommendations.
7. A power marketing recommendation system based on big data, used to implement the power marketing recommendation method based on big data as described in any one of claims 1 to 6, Features: It includes data collection module, data preprocessing module, user classification module, marketing strategy formulation module, marketing strategy effect prediction module and power marketing recommendation result module.
8. According to the big data-based power marketing recommendation system of claim 7, Features: The data acquisition module acquires power marketing raw data from the power system by acquisition, wherein the power marketing raw data includes user data, power product data and user feedback data, and sends the power marketing raw data to the data preprocessing module; The data preprocessing module receives the original power marketing data sent by the data acquisition module, performs data preprocessing on the original power marketing data to obtain preliminary power marketing data, and sends the preliminary power marketing data to the user classification module; The user classification module receives the preliminary power marketing data sent by the data preprocessing module, and performs user classification on the user data to obtain a user classification data set, and sends the user classification data set to the marketing strategy formulation module; The marketing strategy formulation module receives the user classification data set sent by the user classification module, formulates a personalized power marketing strategy text according to the user classification result in the user classification data set, obtains the marketing strategy text content, and sends the marketing strategy text content to the marketing strategy effect prediction module; The marketing strategy effect prediction module receives the marketing strategy text content sent by the marketing strategy formulation module, and performs effect prediction on the power marketing recommendation plan of the user in the marketing strategy formulation, obtains the marketing strategy effect prediction result, and sends the marketing strategy effect prediction result to the power marketing recommendation result module; The power marketing recommendation result module receives the marketing strategy effect prediction result sent by the marketing strategy effect prediction module, and recommends personalized power marketing plans to users, obtains power marketing plans and user feedback data, and sends the user feedback data to the data acquisition module.
Citation Information
Cited By
Personalized government affair service recommendation method and system based on artificial intelligence
CN119557518A
Artificial Intelligence-Based Personalized E-Government Service Recommendation Method and System
CN119557518B
Intelligent film and television content recommendation method and system based on artificial intelligence
CN120763360A
Personalized video recommendation method and system based on big data
CN120769084A
