Power marketing data analysis method and system

Abstract

The invention provides a power marketing data analysis method and system, and the method comprises the steps: collecting the consumption data of power consumers, and constructing a power consumer behavior analysis model to analyze the consumption data; according to an output result of the consumer behavior prediction model, making an optimal pricing strategy to realize maximized sales benefits; performing real-time effect evaluation on each marketing activity, and dynamically adjusting a pricing strategy according to evaluation; by combining the marketing effect and the existing pricing strategy, a dynamic optimization strategy is designed, and the pricing and marketing strategies can be timely adjusted according to the market change, so that the overall benefit is maximized; and according to the consumption data of each consumer, in combination with the pricing strategy after the dynamic optimization strategy is executed, personalized recommendation is realized to respond to market changes and user behaviors. According to the method, the competitiveness of the electric power company is improved by optimizing the marketing strategy, and finally the user experience and the electric power marketing effect are improved.

Description

Technical Field

[0001] The present invention belongs to the field of power marketing data analysis, and in particular, relates to a power marketing data analysis method and system. Background Art

[0002] With the gradual opening of the global power market, power companies are facing unprecedented market competition pressure. Under the traditional power pricing model, power companies often set electricity prices based on historical power consumption data and fixed pricing strategies. Although this pricing model is effective to a certain extent, it lacks a deep understanding of consumer needs, behaviors and psychology, and has a weak ability to respond to market changes, resulting in unsatisfactory results in power marketing activities.

[0003] The current power marketing system mostly relies on static analysis methods based on historical data, often ignoring consumers' psychological behavior and decision-making preferences. Under this model, power pricing and marketing strategies are too rigid and lack flexibility. For example, many power companies rely solely on pre-set price ranges and promotional plans to attract users, without making differentiated pricing or promotional strategies based on the characteristics of different consumer groups, resulting in inefficient marketing results and a decline in user participation.

[0004] In addition, most existing power marketing systems lack real-time feedback mechanisms and dynamic adjustment capabilities. Power companies are usually unable to respond in a timely manner to changes in the real-time market environment or consumer behavior, which often lags behind in the effectiveness of marketing activities and fails to fully tap consumer needs. For example, in certain promotional activities, if the price changes do not take into account consumers' loss aversion or price sensitivity, it may provoke negative reactions from consumers, resulting in counterproductive marketing effects.

[0005] In this context, behavioral economics, as a discipline that studies human decision-making behavior, has gradually been introduced into the field of power marketing. The theories of loss aversion, anchoring effect, and choice preference in behavioral economics provide new perspectives for power marketing, especially in understanding the irrational factors in consumer decision-making. However, the existing power marketing system often only adopts simplified behavioral economics principles in pricing (such as regular price adjustments), while ignoring how to combine consumers' decision-making psychology in real time and accurately to make personalized recommendations and dynamic optimization. With the complexity of the power market, power companies are in urgent need of a system that can perform real-time analysis and dynamic optimization, which can not only use behavioral economics principles to deeply understand consumer decisions, but also evaluate and adjust marketing activities in real time through adaptive algorithms, so as to truly achieve precision marketing and pricing.

[0006] At present, although some innovative technologies such as machine learning and data mining have been introduced into the field of power marketing, attempting to improve marketing effectiveness through large-scale data analysis, most systems are still limited to static data analysis and lack a response mechanism to dynamic market environments and changes in consumer behavior. In addition, existing technologies still find it difficult to achieve real-time feedback and adaptive adjustments, which makes it difficult for power companies to respond quickly and flexibly when facing changing market demands, resulting in waste of resources and inefficient marketing results. Summary of the invention

[0007] The purpose of the present invention is to propose a method and system for analyzing power marketing data. Through real-time feedback mechanism, personalized recommendation and dynamic adjustment, the system can better adapt to the complex and changing market environment, provide accurate power pricing solutions, and improve the competitiveness of power companies by optimizing marketing strategies, ultimately improving user experience and power marketing effects.

[0008] In order to achieve the above object, a first aspect of the present invention provides a power marketing data analysis method, the method comprising:

[0009] S1. Collect consumption data of electricity consumers and build a behavior analysis model for electricity consumers to analyze the consumption data, including:

[0010] Collect and pre-process the consumption data of electricity consumers, construct the behavioral characteristics of consumers based on the pre-processed data and divide consumers into multiple groups, and construct a consumer behavior prediction model based on the supervised learning model based on the behavioral characteristics and groups to predict consumers' reactions to different electricity pricing strategies or promotional activities, and output the response probability of each consumer;

[0011] S2. Formulate an optimal pricing strategy based on the output results of the consumer behavior prediction model to achieve maximum sales benefits, including:

[0012] Designing a demand function according to the response probability of each consumer, wherein the demand function represents the demand quantity of the consumer under a specific price, and constructing an optimization model based on the demand function of each consumer to maximize the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process;

[0013] Execute the pricing strategy output by the optimization model and collect response data from consumer groups. Through the Q-learning method, dynamically update the pricing strategy based on the sales effect of each adjustment.

[0014] Design a multi-dimensional evaluation function to cyclically evaluate the overall benefits of all pricing strategies and ultimately select the optimal pricing execution;

[0015] S3. Conduct real-time effect evaluation on each marketing campaign and dynamically adjust pricing strategies based on the evaluation. After each marketing cycle, feedback on marketing effects will be input into the pricing strategy for the next cycle for adjustment.

[0016] S4. Design dynamic optimization strategies by combining marketing results with existing pricing strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits;

[0017] S5. Based on the consumption data of each consumer, combined with the pricing strategy after executing the dynamic optimization strategy, personalized recommendations are implemented to respond to market changes and user behavior.

[0018] Furthermore, the consumers are divided into multiple groups using a K-means clustering algorithm, and the consumers are divided into multiple groups according to their behavioral characteristics, wherein each group represents a specific consumption behavior pattern. The goal of the K-means clustering algorithm is to minimize the difference in the consumer feature vectors within the group, so that the consumer behaviors of the same group are most similar, which is expressed as:

[0019]

[0020] Among them, C k represents the center of the kth group, representing the typical consumption behavior characteristics of the group; X i represents the behavioral feature vector of consumer i, and N represents the total number of consumers;

[0021] The output of the consumer behavior prediction model based on the supervised learning model is expressed as:

[0022] P(response i |X i ,C k )=f(X i ,C k )

[0023] Among them, P (response i ) represents the probability of consumer i responding to a certain pricing or promotion activity; f(X i ,C k ) represents the consumer characteristics X i and group characteristics C k The model predicts the probability of the consumer's response based on these inputs.

[0024] Furthermore, the function of demand elasticity correction is expressed as:

[0025]

[0026] Among them, Demandi Indicates that the price P i Next, the quantity demanded by consumer i; represents the basic demand of consumer i, the demand without price intervention; P i represents electricity pricing; α: demand elasticity coefficient, which indicates the consumer's sensitivity to price changes; P (response i |X i ,C k ) represents the consumer i’s price P i The probability of response;

[0027] The objective function of the optimization model is expressed as:

[0028]

[0029] Among them, P i Demand represents electricity pricing; i represents consumer i at price P i The demand under i |X i ,C k ) represents the response probability of consumer i; Budget i represents the budget of consumer i; λ represents the Lagrange multiplier, which represents the impact of the budget constraint on the optimization process;

[0030] The Q-learning method is used to dynamically update the pricing strategy based on the sales effect of each adjustment, which is expressed as:

[0031]

[0032] Among them, Q(P k ,t) indicates that at time point t, the price P k The strategy value represents the expected benefit of the pricing strategy; R t represents the actual reward at time t; γ represents the discount factor, which controls the importance of future benefits; α represents the learning rate, which controls the speed of policy update; P k ′ represents other pricing strategies at time point t;

[0033] The multi-dimensional evaluation function is expressed as:

[0034] E(P k )=w1·Consumer Acceptance(P k )+w2·Profit(P k )+w3

[0035] MarketShare(P k)+w4·Customer Satisfaction(P k )

[0036] Among them, ConsumerAcceptance(P k ) represents the pricing strategy P k Profit(P k ) represents the pricing strategy P k Total profit brought by; Market Share (P k ) represents the pricing strategy P k Impact on market share; Customer Satisfaction (P k ) represents the consumer’s satisfaction with the pricing; w1, w2, w3, w4 represent the weights of each evaluation indicator, which are adjusted according to business objectives.

[0037] Furthermore, in order to avoid excessive fluctuations in pricing strategies during real-time adjustment, a regularization term is introduced into the optimization model to constrain the range of pricing changes, which is expressed as:

[0038]

[0039] Among them, P k Indicates pricing strategy; Demand represents the initial pricing strategy; i represents consumer i at price P k β1, β2 represent the regularization coefficients, which are used to balance pricing smoothness and demand volatility.

[0040] Furthermore, in S3, the evaluation result quantifies the activity effect based on a comprehensive marketing effect function, which is expressed as:

[0041] R k (t) = γ1·Sales k (t)+γ2·Participation k (t)+γ3·Satisfaction k (t)

[0042] Among them, Sales k (t) represents the sales volume of activity k at time t; Participation k (t) represents the number of participants in activity k; Satisfaction k (t) represents consumer satisfaction, which may come from survey data or automatic scoring; γ1, γ2, and γ3 represent the weights of each indicator, reflecting the importance of each factor in effect evaluation.

[0043] Furthermore, the pricing strategy is adjusted based on the feedback through the real-time effect evaluation results, which is expressed as:

[0044] P k (t+1)=P k (t)+η·R k (t)

[0045] Among them, P k (t) represents the pricing strategy at time t; P k (t+1) represents the pricing strategy for the next cycle, based on R k (t) is adjusted; η represents the adjustment coefficient, which controls the magnitude of pricing changes;

[0046] At the same time, in order to avoid excessive fluctuations in pricing strategies and ensure that pricing adjustments are within a reasonable range, a time decay mechanism and a time regularization term are introduced into the pricing strategy. The time decay mechanism is expressed as:

[0047] P k (t+1)=P k (t)+η·R k (t)·(1-β t )

[0048] Where β represents the time decay factor, β∈[0,1). As time goes by, the influence of historical feedback gradually weakens; t represents the time step; P k (t+1) represents the pricing with time decay mechanism added;

[0049] The time regularization term is expressed as:

[0050]

[0051] Among them, λ represents the regularization coefficient, which controls the adjustment amplitude; represents the initial pricing strategy, which is used to constrain pricing changes; P k (t+2) represents the pricing with the time regularization term added.

[0052] Furthermore, the S4 specifically includes:

[0053] The design objective function represents the balance between the benefits and stability brought by pricing adjustments and is expressed as:

[0054]

[0055] in, represents the optimization objective function, which means that by adjusting the pricing strategy P k The comprehensive benefits after (t+1); R k(t) represents the marketing effectiveness evaluation at time t, measuring the performance of the current strategy; P k (t) represents the pricing scheme at the current time t; P k (t+1) represents the pricing plan at the next moment; α represents the weight factor of marketing effect, reflecting the importance of marketing effect in the objective function; γ represents the penalty factor of pricing stability, controlling the amplitude of pricing changes; |P k (t+1)-P k (t)| p The penalty term for pricing changes, p is a positive integer, which penalizes large pricing changes to avoid excessive price fluctuations;

[0056] The adaptive gradient descent method is used to calculate the direction and step size of the pricing adjustment, so that the pricing adjustment can be adjusted automatically according to market feedback, specifically including:

[0057] First, calculate the objective function for the pricing scheme P k The gradient of (t+1) is expressed as:

[0058]

[0059] Then, the pricing scheme P is updated according to the gradient descent method k (t+1), expressed as

[0060]

[0061] in, Represents the objective function About Pricing Plan P k The gradient of (t+1); Represents marketing effect R k (t) Price P k The sensitivity of (t+1) reflects the impact of pricing on marketing effectiveness; η(t) represents the step size at the current moment, which controls the amplitude of pricing adjustment; p represents the power of pricing change, which controls the intensity of the penalty term for pricing stability;

[0062] Design a feedback mechanism to ensure that each pricing adjustment is optimized based on the results of the previous cycle, expressed as:

[0063]

[0064] Among them, the objective function is updated according to the real-time evaluation report after each feedback.

[0065] Furthermore, in order to ensure the stability of the pricing strategy, a stability constraint is introduced in the optimization process. The stability constraint ensures that the amplitude of each pricing adjustment will not be too large, and avoids the instability of the market due to frequent pricing changes, which is expressed as:

[0066] |P k (t+1)-P k (t)|≤ΔP max

[0067] Among them, ΔP max Indicates the maximum pricing change range to ensure that pricing adjustments do not exceed the set range, thereby avoiding excessive price fluctuations.

[0068] Furthermore, the S5 specifically includes:

[0069] Combining user behavior and real-time market feedback, we build a personalized recommendation scoring model to dynamically optimize recommendation results.

[0070] It is expressed as:

[0071] R i,j (t) = λ1·S i,j (t)+λ2·M j (t)+λ3·H i,j (t)+λ4·ΔM j (t)

[0072] Among them, R i,j (t) represents the personalized recommendation score of user i for product j, which changes over time; S i,j (t) represents the similarity between user i and item j, which is calculated based on historical purchase or browsing behavior using collaborative filtering or deep learning methods; M j (t) represents the market performance score of product j at time t, based on the optimized pricing strategy P k (t), reflects the market demand for goods and the effectiveness of marketing activities; H i,j (t) represents the personalized preference score of user i for product j, based on the user's historical data and preferences; ΔM j (t) represents the change in the market performance of commodity j, measuring the immediate impact of market fluctuations on commodity demand;

[0073] The dynamic feedback mechanism is introduced to dynamically adjust the recommendation results according to market fluctuations, changes in user preferences and real-time feedback on pricing strategies, which can be expressed as:

[0074]

[0075] Among them, ΔR i,j (t) represents the adjustment amount of user i’s recommendation score for product j, reflecting the impact of changes in market and user preferences on recommendations; It represents a measure of the change in the market performance of product j, indicating the immediate impact of pricing adjustments or marketing activities on product performance; It represents the rate at which user i's preference for product j changes, reflecting the user's response speed to recommendations;

[0076] A multi-objective optimization framework is designed to optimize user loyalty and customer experience, balancing the priorities of different objectives, expressed as:

[0077]

[0078] in, represents the overall optimization objective function, which combines the optimization of sales volume and customer loyalty; It indicates the optimization goal based on sales volume, mainly improving sales conversion rate by adjusting the recommendation system; It represents the optimization goal based on customer loyalty, reflecting the impact of the recommendation system on the long-term relationship of users;

[0079] Whenever the market or user behavior changes, new feedback is received and the recommendations are adjusted based on this feedback.

[0080] In another aspect of the present invention, a power marketing data analysis system is provided, the system comprising:

[0081] The power marketing data collection module is used to collect the consumption data of power consumers and build a power consumer behavior analysis model to analyze the consumption data, including:

[0082] Collect and pre-process the consumption data of electricity consumers, construct the behavioral characteristics of consumers based on the pre-processed data and divide consumers into multiple groups, and construct a consumer behavior prediction model based on the supervised learning model based on the behavioral characteristics and groups to predict consumers' reactions to different electricity pricing strategies or promotional activities, and output the response probability of each consumer;

[0083] The user behavior prediction module is used to formulate the optimal pricing strategy according to the output results of the consumer behavior prediction model to achieve the maximum sales benefit, including:

[0084] Designing a demand function according to the response probability of each consumer, wherein the demand function represents the demand quantity of the consumer under a specific price, and constructing an optimization model based on the demand function of each consumer to maximize the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process;

[0085] Execute the pricing strategy output by the optimization model and collect response data from consumer groups. Through the Q-learning method, dynamically update the pricing strategy based on the sales effect of each adjustment.

[0086] Design a multi-dimensional evaluation function to cyclically evaluate the overall benefits of all pricing strategies and ultimately select the optimal pricing execution;

[0087] The power pricing module is used to evaluate the effect of each marketing activity in real time and dynamically adjust the pricing strategy based on the evaluation. After each marketing cycle, the feedback on the marketing effect will be input into the pricing strategy of the next cycle for adjustment;

[0088] The power market adjustment module is used to design dynamic optimization strategies by combining marketing effects and existing pricing strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits;

[0089] The personalized data recommendation module is used to implement personalized recommendations in response to market changes and user behaviors based on the consumption data of each consumer, combined with the pricing strategy after executing the dynamic optimization strategy.

[0090] The beneficial technical effects of the present invention are at least as follows:

[0091] First, the present invention constructs a decision-making behavior model for power consumers by introducing theories such as loss aversion and anchoring effect in behavioral economics. The model can capture the psychological reactions of consumers when faced with different electricity prices, promotions, and pricing schemes, and provide scientific support for power marketing strategies. By understanding consumers' irrational decisions, the present invention can more accurately predict consumers' reactions to power products, and conduct personalized pricing and recommendations accordingly, thereby effectively improving the accuracy of pricing and the participation in marketing activities.

[0092] Secondly, in response to the problem of lagging feedback in traditional power marketing activities, the present invention introduces an adaptive algorithm to analyze the effects of marketing activities in real time and adjust marketing strategies based on real-time data. This mechanism can adjust power pricing or recommended promotional activities based on consumers' immediate reactions (such as acceptance of discounts, emotional feedback on electricity price changes, etc.), thereby ensuring that marketing activities can quickly adapt to market changes and improve marketing effectiveness.

[0093] In addition, the present invention combines the principles of behavioral economics and adaptive algorithms to achieve dynamic pricing and marketing recommendations based on the personalized needs of consumers. By analyzing consumers' historical behavior data and real-time feedback, the system can accurately recommend power products that meet their psychological characteristics and price sensitivity to each user, ensuring that each marketing campaign can effectively stimulate the purchase intention of the target user group, thereby improving user participation and satisfaction.

[0094] In summary, the present invention effectively overcomes several shortcomings of existing power marketing technologies. Through real-time feedback mechanism, personalized recommendation and dynamic adjustment, the system can better adapt to the complex and changing market environment, provide accurate power pricing solutions, and improve the competitiveness of power companies by optimizing marketing strategies, ultimately improving user experience and power marketing effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0095] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0096] Figure 1 The present invention is a flowchart of a power marketing data analysis method according to an embodiment of the present invention.

[0097] Figure 2 This is a framework diagram of a power marketing data analysis system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0098] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0099] like Figure 1 As shown, an embodiment of the present invention provides a method for analyzing power marketing data, the method comprising the following steps:

[0100] S1. Collect consumption data of electricity consumers and build a behavior analysis model for electricity consumers to analyze the consumption data.

[0101] Specifically, the original electricity consumer data set is collected, including historical electricity consumption records, consumer behavior data (such as history of participation in promotional activities), user portrait information, etc.

[0102] First, the data is cleaned to remove missing values, outliers, and irrelevant information. Then, consumer behavior characteristics are extracted, such as historical electricity consumption trends, sensitivity to price changes, frequency of participation in different types of promotional activities, etc. Feature engineering techniques are used to construct new features (for example, price sensitivity indicators, time preferences, etc.), which provide necessary inputs for subsequent modeling. Formula:

[0103] X i =f(consumption history, price sensitivity, promotion participation, user profile)

[0104] Among them, X i Represents the behavioral feature vector of consumer i. This feature vector will be used as the input for subsequent clustering analysis and consumer behavior prediction model.

[0105] Furthermore, consumer segmentation based on cluster analysis:

[0106] The consumer behavior feature set X output from the first step i , including historical consumption records, response patterns for participating in promotional activities, etc. Use the K-means clustering algorithm to divide consumers into multiple groups (K groups) based on their behavioral characteristics. Each group represents a specific consumption behavior pattern, such as a price-sensitive consumer group, a highly loyal consumer group, etc. The goal of clustering is to minimize the differences in consumer feature vectors within a group, so that the behaviors of consumers in the same group are more similar. Formula:

[0107]

[0108] Among them, C k represents the center of the kth group (cluster center), representing the typical consumption behavior characteristics of the group. i Represents the behavioral feature vector of consumer i. The clustering algorithm optimizes the group division of consumers by minimizing the distance between the consumer feature vector within the group and the group center.

[0109] Furthermore, the consumer behavior prediction model is constructed:

[0110] The consumer group C obtained after clustering k and the consumer characteristics X in each group i Based on the clustering results, supervised learning models (such as random forest or XGBoost) are used to build a behavior prediction model for each consumer group. The goal is to predict consumers' responses to different electricity pricing strategies or promotions (such as whether to participate, whether to accept discounts, etc.). During the model training process, the input includes the consumer's historical behavior characteristics and group characteristics, and the output is the probability of each consumer's response to a specific pricing or marketing campaign. Formula:

[0111] P(response i |X i ,C k )=f(X i ,C k )

[0112] Among them, P (response i ) represents the probability of consumer i responding to a certain pricing or promotion activity. i Represents the characteristic vector of consumer i (historical consumption, price sensitivity, promotion participation, etc.). k represents the group k to which consumer i belongs. Each group corresponds to a set of behavioral features, which describes the common characteristics of consumers in the group. i ,C k ) represents the consumer characteristics X i and group characteristics C kThe model predicts the probability of the consumer's response based on these inputs.

[0113] Furthermore, the prediction model output obtained from step 3 is the actual response data of the consumer. The prediction model is evaluated using the cross-validation method, and the prediction effect of the model is quantified by calculating indicators such as AUC (Area Under Curve) and F1 Score. Based on these evaluation indicators, the hyperparameters of the model are adjusted to optimize the model performance and ensure that the model can show the best prediction accuracy in different consumer groups. Formula:

[0114]

[0115] Among them, AUC represents the area under the curve, which indicates the ability of the model to predict the correct response. TPR (True Positive Rate) represents the proportion of consumers who truly respond are correctly classified. FPR (False Positive Rate) represents the proportion of consumers who are incorrectly classified and do not respond.

[0116] This step combines the analysis of consumer behavior characteristics based on behavioral economics with the clustering-based clustering method to more accurately predict the responses of different consumer groups to electricity pricing or promotional activities. Based on cluster analysis, a supervised learning model is used to build a personalized behavior prediction model for each group, thereby realizing a customized electricity marketing strategy. In addition, through the combined tuning of cross-validation and AUC indicators, the wide applicability and accuracy of the model in different consumer groups are ensured, thereby improving the effectiveness and efficiency of electricity marketing activities.

[0117] S2. Formulate an optimal pricing strategy based on the output results of the consumer behavior prediction model to achieve maximum sales benefits.

[0118] Specifically, in this step, the goal is to develop the best pricing strategy based on consumer behavior prediction to achieve maximum sales benefits. The present invention adopts a dynamic model that combines consumer response prediction and optimized pricing to design electricity prices through a dynamic pricing optimization algorithm and Lagrange multiplier method. Specifically, the model dynamically adjusts electricity pricing to maximize revenue while ensuring that the needs of different consumer groups are met, taking into account consumer demand elasticity and pricing sensitivity, while meeting budget constraints.

[0119] Step 1: Construct the objective function:

[0120] According to the response probability P (response i |X i ,C k), design a demand function for each group. The demand function represents the demand quantity of consumers at a specific price P. i , demand and pricing show a certain relationship, usually following the law of demand elasticity. The model of the present invention uses a function modified by demand elasticity to represent the change of demand with pricing:

[0121]

[0122] Among them, Demand i Indicates that the price P i Under this condition, the demand of consumer i is. represents the basic demand of consumer i, the demand without price intervention. Pi represents the electricity pricing. α represents the demand elasticity coefficient, which indicates the consumer's sensitivity to price changes. P(response i |X i ,C k ) represents the consumer i’s price P i In this demand function, the present invention adds demand elasticity α and the probability of consumer response to reflect the impact of electricity pricing on different consumers.

[0123] Step 2: Optimize the objective function:

[0124] In order to optimize electricity pricing, the present invention needs to maximize the total revenue. Considering the budget constraint and the demand elasticity of each consumer, the present invention constructs an optimization model with the goal of maximizing the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process. The objective function is as follows:

[0125]

[0126] Among them, P i Indicates the electricity price. i represents consumer i at price P i The demand under i |X i ,C k ) represents the response probability of consumer i. i represents the budget of consumer i. λ represents the Lagrange multiplier, which represents the impact of the budget constraint on the optimization process. The optimization model takes into account the demand, response probability and budget constraints of different consumer groups, so that an electricity pricing strategy that can maximize revenue and meet the budget can be formulated. The pricing strategy P from step 1 k and response data for each consumer group. These data include actual demand and response at different pricing levels.

[0127] Furthermore, after implementing the pricing strategy, market feedback data (such as actual demand, consumer response) will be used to adjust the pricing strategy in real time. The present invention uses the Q-learning method in the reinforcement learning algorithm to update the pricing strategy based on the sales effect of each adjustment (such as revenue, participation, etc.), ensuring that the system can be dynamically adjusted according to real-time feedback. This adjustment process improves the adaptability of pricing by introducing additional regularization terms.

[0128] Step 1: Monitor real-time sales feedback:

[0129] After implementing the pricing strategy, collect data on consumers’ response to electricity pricing, including actual demand, sales, number of participants, etc. Use these data to update the effectiveness of the pricing strategy and ensure that each adjustment reflects the latest market response.

[0130] Step 2: Adjust pricing based on Q-learning algorithm:

[0131] Based on real-time feedback, the present invention uses the Q-learning algorithm to update the value of the pricing strategy. The Q-learning algorithm represents the effect of each pricing strategy through a value function and updates the strategy based on market feedback. The update formula is as follows:

[0132]

[0133] Among them, Q(P k ,t) indicates that at time point t, the price P k The strategy value of R represents the expected benefit of the pricing strategy. t P represents the actual reward (such as income, participation, etc.) at time t. γ represents the discount factor, which controls the importance of future benefits. α represents the learning rate, which controls the speed of policy update. k ' represents other pricing strategies at time point t. In this Q-learning algorithm, a dynamic adjustment mechanism based on market feedback is added to ensure that the pricing strategy can be flexibly adjusted according to actual conditions and optimize the power sales effect.

[0134] Step 3: Add regularization terms to ensure pricing stability:

[0135] In order to avoid excessive fluctuations in pricing strategies during real-time adjustments, a regularization term is introduced into the model to constrain the range of pricing changes. This regularization term takes into account the smoothness of pricing changes and ensures pricing stability. The regularization objective can be expressed as:

[0136]

[0137] Among them, P k Indicates pricing strategy. Indicates the initial pricing strategy. i represents consumer i at price P k β1,β2 represent regularization coefficients, which are used to balance pricing smoothness and demand volatility. The introduction of this regularization term helps to limit excessive changes in pricing strategies, improve the stability of strategies, and avoid price fluctuations caused by excessive market feedback adjustments.

[0138] Further, input the final pricing strategy P from step 2 k , and real-time feedback data from consumers, such as demand, sales, etc. At the same time, after optimizing the pricing strategy, it is necessary to evaluate the comprehensive effect of the strategy to ensure that it can achieve the best results in multiple dimensions. The present invention adopts a multi-dimensional evaluation function, comprehensively considering factors such as consumer acceptance, profit, market share, etc., evaluates the overall benefits of the pricing strategy, and finally selects the optimal pricing execution.

[0139] Step 1: Collect evaluation metrics:

[0140] After implementing the pricing strategy, collect evaluation indicators of different dimensions, including consumer acceptance, market share, profit, customer feedback, etc. Evaluate the effectiveness of the pricing strategy by quantifying these indicators.

[0141] Step 2: Construct a comprehensive evaluation function:

[0142] Based on the collected multi-dimensional evaluation indicators, the present invention constructs a comprehensive evaluation function E(P k ), and weighted processing is performed on different evaluation dimensions. The comprehensive evaluation function is as follows:

[0143] E(P k )=w1·Consumer Acceptance(P k )+w2·Profit(P k )+w3

[0144] MarketShare(P k )+w4·Customer Satisfaction(P k )

[0145] Among them, ConsumerAcceptance(P k ) represents the pricing strategy P k Profit(P k ) represents the pricing strategy P k Total profit brought by Market Share(P k ) represents the pricing strategy P kImpact on market share. k ) represents the consumer's satisfaction with the pricing. w1, w2, w3, w4 represent the weights of various evaluation indicators, which are adjusted according to business objectives. This comprehensive evaluation function takes various indicators into consideration to ensure that the pricing strategy can both increase the company's profits and improve consumer satisfaction, and ultimately provide the best decision for electricity pricing.

[0146] Step 3: Execute the optimal pricing strategy:

[0147] According to the comprehensive evaluation results, select the optimal pricing strategy P k During the implementation process, we continuously monitor the market and consumer feedback to ensure that the pricing strategy can be further adjusted according to the actual situation.

[0148] By combining consumer behavior models, dynamic pricing optimization, real-time feedback mechanisms, and multi-dimensional evaluation functions, this solution can not only maximize the revenue from electricity sales, but also ensure that the pricing strategy is balanced in multiple dimensions. The innovative introduction of demand elasticity correction terms, real-time adjustment regularization terms, and comprehensive evaluation functions makes the electricity pricing strategy more refined and dynamic, and can respond quickly to market changes and consumer demand. This solution provides a comprehensive and efficient pricing optimization framework for power marketing, which can effectively enhance the competitiveness of power companies in complex market environments.

[0149] S3. Conduct real-time effectiveness evaluation on each marketing activity and dynamically adjust the pricing strategy based on the evaluation. After each marketing cycle, feedback on the marketing effectiveness will be input into the pricing strategy of the next cycle for adjustment.

[0150] Specifically, in this step, the present invention needs to conduct real-time effect evaluation on each marketing activity. The marketing effect is not only measured by sales volume, but also by consumer participation and satisfaction. Therefore, a comprehensive marketing effect function R is set. k (t) is used to quantify the effect of the activity. This function combines multiple factors, such as sales volume, consumer participation and satisfaction, and gives different weights to each factor. The formula is as follows:

[0151] R k (t) = γ1·Sales k (t)+γ2·Participation k (t)+γ3·Satisfaction k (t)

[0152] Among them, Sales k (t) represents the sales volume of activity k at time t.k (t) represents the number of participants in activity k. k (t) represents consumer satisfaction, which may come from survey data or automatic scoring. γ1,γ2,γ3 represent the weights of each indicator, reflecting the importance of each factor in the effect evaluation. k (t), the present invention can comprehensively measure the effectiveness of marketing activities, rather than relying solely on a single indicator (such as sales volume). This comprehensive evaluation provides a basis for subsequent pricing adjustments.

[0153] Furthermore, through the real-time effect evaluation results R k (t), the present invention can adjust the pricing strategy P based on feedback k (t). In order to ensure that the pricing strategy can flexibly respond to market changes, the present invention performs adaptive adjustments in the following ways:

[0154] P k (t+1)=P k (t)+η·R k (t)

[0155] Among them, P k (t) represents the pricing strategy at time t. k (t+1) represents the pricing strategy for the next cycle, based on R k (t) is adjusted. η is the adjustment coefficient, which controls the magnitude of pricing changes.

[0156] In this model, the pricing strategy P k (t) will be based on real-time feedback R k (t) Make fine adjustments. If the effect of a certain activity is good, the price of the activity will be appropriately increased; otherwise, it will be adjusted down.

[0157] Furthermore, in order to avoid excessive fluctuations in pricing strategies and ensure that pricing adjustments are within a reasonable range, the present invention adds a regularization term and a time decay mechanism.

[0158] Time decay mechanism: As time goes by, the impact of historical feedback gradually weakens. In order to ensure that the system is more sensitive to the current market response, the present invention introduces a time decay factor β to adjust the impact of historical effects on pricing strategies.

[0159] P k (t+1)=P k (t)+η·R k (t)·(1-β t )

[0160] Where β represents the time decay factor, β∈[0,1). As time goes by, the influence of historical feedback gradually weakens. t represents the time step.

[0161] Regularization term: In order to avoid drastic changes in pricing, the present invention adds a regularization term to constrain the adjustment range of the pricing strategy.

[0162]

[0163] Among them, λ represents the regularization coefficient, which controls the adjustment range. represents the initial pricing strategy, which is used to constrain pricing changes. Under the action of the regularization term, pricing adjustments will not deviate too far from the initial strategy, thus maintaining pricing stability.

[0164] Furthermore, after each marketing cycle, the marketing effect feedback R k (t) will be input into the pricing adjustment model of the next cycle. Through this mechanism, the pricing strategy can be continuously updated according to market reactions, ensuring that pricing adjustments are real-time and flexible.

[0165] The core of this mechanism is to promptly feed the feedback from each marketing activity into the adjustment process of pricing strategy, so as to quickly respond to market changes and optimize overall revenue.

[0166] By introducing real-time effect evaluation and adaptive adjustment mechanisms, the goal of this step is to ensure close linkage between marketing activities and pricing strategies. The time decay mechanism and regularization term introduced effectively control the volatility of pricing while ensuring a sensitive response to market dynamics. In this process, all adjustments are based on the output of the previous step (pricing strategy P k (t)), and continuously optimizes through feedback to ensure the continuity and stability of the solution. This solution has important applications in the field of power marketing, and can provide real-time optimization for sales strategies and improve the efficiency and effectiveness of marketing activities.

[0167] S4. By combining marketing effectiveness with existing pricing strategies, design dynamic optimization strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits.

[0168] Specifically, in order to dynamically optimize the strategy, the present invention first needs to design an objective function, which takes into account both marketing effectiveness and pricing stability. The present invention uses the following objective function to represent the balance between the benefits and stability brought about by pricing adjustments:

[0169]

[0170] in, represents the optimization objective function, which means that by adjusting the pricing strategy Pk The comprehensive benefits after (t+1). k (t) represents the marketing effectiveness evaluation at time t, measuring the performance of the current strategy. k (t) represents the pricing scheme at the current time t. k (t+1) represents the pricing plan at the next moment. α represents the weight factor of marketing effect, reflecting the importance of marketing effect in the objective function. γ represents the penalty factor of pricing stability, controlling the magnitude of pricing changes. |P k (t+1)-P k (t)| p The penalty term for pricing changes, p is a positive integer (e.g., p=2), which avoids excessive price fluctuations by penalizing larger pricing changes.

[0171] The core of this objective function is to balance the marketing effect R k (t) and pricing stability |P k (t+1)-P k (t)|, where the penalty term for pricing stability has a power term p, which is to ensure that pricing changes will not be too drastic due to small changes in the optimization process. The design of this term effectively prevents excessive risk of price fluctuations.

[0172] Furthermore, the objective function The maximization of requires the use of a suitable optimization method. The present invention uses the gradient descent method to calculate the direction and step size of the pricing adjustment. However, due to the complex market environment and the nonlinear characteristics of the feedback mechanism, the simple gradient descent method may not provide sufficient dynamic response. Therefore, the present invention introduces an adaptive step size η(t) so that the pricing adjustment can be adjusted automatically according to market feedback:

[0173] First, calculate the objective function for the pricing scheme P k Gradient at (t+1):

[0174]

[0175] Then, the pricing scheme P is updated according to the gradient descent method k (t+1):

[0176]

[0177] in, Represents the objective function About Pricing Plan P k The gradient of (t+1). Represents marketing effect R k (t) Price P kThe sensitivity of (t+1) reflects the impact of pricing on marketing effectiveness. η(t) represents the step size at the current moment, which controls the amplitude of pricing adjustment. p represents the power of pricing change, which controls the intensity of the pricing stability penalty term. k The adjustment of (t+1) is guided by the gradient of the objective function, and the step size η(t) is adaptively adjusted based on historical feedback. If the market fluctuates violently, η(t) will increase to accommodate the rapid adjustment needs; otherwise, the step size will be reduced to avoid overreaction.

[0178] Furthermore, in order to ensure the stability of the pricing strategy, the present invention introduces a stability constraint in the optimization process. This constraint ensures that the amplitude of each pricing adjustment will not be too large, thus avoiding the market from generating unstable factors due to frequent pricing changes. The specific constraint form is:

[0179] |P k (t+1)-P k (t)|≤ΔP max

[0180] Among them, ΔP max Indicates the maximum price change range, ensuring that the price adjustment does not exceed the set range, thus avoiding excessive price fluctuations. This constraint ensures that the price will not fluctuate drastically each time the strategy is adjusted, so that the continuity and stability of marketing activities are guaranteed. In this way, the stability of the pricing strategy can be controlled, while ensuring that the marketing strategy can be adaptively adjusted according to actual market feedback.

[0181] Furthermore, the core of this step is to make real-time adjustments based on market feedback. k (t) and pricing scheme P k (t) real-time data, constantly adjust the parameters α and γ of the objective function, as well as the step size η(t). For example, when the market feedback is relatively stable, the present invention may choose to increase γ to reduce excessive fluctuations; and when the market fluctuates violently, the step size η(t) may be adjusted to enhance the responsiveness to market changes.

[0182] Furthermore, the complete optimization framework includes a feedback mechanism, through which the present invention can ensure that each pricing adjustment is optimized according to the effect of the previous cycle. In actual operation, the update of pricing strategy is not only based on immediate feedback, but also gradually approaches the optimal pricing strategy through iterative update of the objective function. The model realizes multi-cycle continuous optimization representation through the following update formula

[0183]

[0184] In this formula, the objective function is updated according to the real-time evaluation report after each feedback, so as to fine-tune the pricing strategy and ensure that the strategy can always maximize the marketing effect as the market environment changes.

[0185] Through this dynamic optimization process of strategy, the present invention successfully realizes a closed-loop pricing and marketing optimization system. Marketing Effect R k (t) and pricing scheme P k The feedback mechanism of (t) interacts with each other, and finally makes the pricing strategy P k (t) Ability to dynamically adjust based on market feedback and optimize the objective function Maximize overall benefits.

[0186] In this process, the adaptive optimization framework, gradient descent method and stability constraints proposed in this invention effectively ensure the robustness of the pricing strategy and make real-time adjustments at various stages of marketing activities. Through continuous optimization and feedback, the system can make the best pricing adjustment in each time period, thereby effectively improving market response speed and customer satisfaction.

[0187] S5. Based on the consumption data of each consumer, combined with the pricing strategy after executing the dynamic optimization strategy, personalized recommendations are implemented to respond to market changes and user behavior.

[0188] Specifically, in this step, the present invention will describe the design of the personalized recommendation system in detail, and combine it with the dynamic optimization results of the pricing strategy in the previous step to ensure that the recommendation system can respond to market changes and user behavior. The present invention will reduce the number of unnecessary formulas, highlight the innovative parts, and combine the market feedback mechanism and user personalized needs to optimize the effect of the recommendation system.

[0189] Furthermore, a personalized recommendation scoring model is constructed, in which the core of personalized recommendation is to combine user behavior and real-time market feedback to dynamically optimize the recommendation results. On this basis, the present invention designs a more concise and innovative scoring function model, which will reflect user behavior changes and market fluctuations. Personalized recommendation scoring:

[0190] R i,j (t) = λ1·S i,j (t)+λ2·M j (t)+λ3·H i,j (t)+λ4·ΔM j (t)

[0191] Among them, R i,j (t) represents the personalized recommendation score of user i for product j, which changes over time. i,j(t) represents the similarity between user i and item j, which is calculated based on historical purchase or browsing behavior using collaborative filtering or deep learning methods. j (t) represents the market performance score of product j at time t, based on the optimized pricing strategy P k (t), reflects the market demand for goods and the effectiveness of marketing activities. i,j (t) represents the personalized preference score of user i for product j, based on the user's historical data and preferences. ΔM j (t) represents the change in the market performance of commodity j, measuring the immediate impact of market fluctuations on commodity demand. This item is an innovative design that can capture the short-term fluctuations in commodity market performance caused by pricing and marketing activity adjustments.

[0192] Understandably, the introduction of ΔM j (t) is an innovative term used to capture the immediate changes in market performance. This term reflects the market response of a product in a short period of time after a pricing adjustment or marketing campaign. This ensures that the recommendation system does not only rely on static historical data, but can also dynamically adjust the recommendation results according to real-time market changes.

[0193] Furthermore, in order to ensure that the recommendation system can respond to market and user changes in real time, the present invention introduces a dynamic feedback mechanism for adjusting the recommendation score. This mechanism will dynamically adjust the recommendation results based on market fluctuations, changes in user preferences, and real-time feedback on pricing strategies.

[0194]

[0195] Among them, ΔR i,j (t) represents the adjustment amount of user i’s recommendation score for product j, reflecting the impact of changes in market and user preferences on recommendations. It represents a measure of the change in the market performance of product j, indicating the immediate impact of pricing adjustments or marketing activities on product performance. It measures the rate at which user i’s preference for item j changes, and reflects the user’s response speed to recommendations.

[0196] It is understood that the present invention uses relatively varied forms (e.g. ), avoiding the interference of absolute changes and enabling the feedback mechanism to better adapt to market fluctuations and changes in user behavior in different periods.

[0197] Furthermore, in practical applications, the present invention not only focuses on sales volume, but also optimizes user loyalty and customer experience. Therefore, the present invention proposes a multi-objective optimization framework to balance the priorities of different objectives. Multi-objective optimization:

[0198]

[0199] in, Represents the overall optimization objective function, which combines the optimization of sales volume and customer loyalty. It represents an optimization goal based on sales volume, which mainly improves sales conversion rate by adjusting the recommendation system. It represents the optimization goal based on customer loyalty, reflecting the impact of the recommendation system on the long-term relationship of users.

[0200] It can be understood that in the optimization framework of this patent, the present invention balances the two key goals of sales and customer loyalty. This method can ensure that the recommendation system not only adjusts recommendations for short-term benefits, but also enhances the long-term loyalty of users.

[0201] Furthermore, whenever the market or user behavior changes, the recommendation system will receive new feedback and adjust the recommendation results based on this feedback. In order to enhance the flexibility of the system, the present invention combines the real-time optimization strategy in the previous steps, and each recommendation result will be dynamically adjusted according to the actual response of the user and market changes.

[0202] Furthermore, each interaction between the user and the recommendation result (such as clicking, purchasing, etc.) will be input into the system as feedback, and the user's rating or recommendation priority of the product will be adjusted. Through this dynamic feedback mechanism, the present invention can ensure the accuracy of the recommendation results and can quickly adapt to changes in the market environment and user behavior.

[0203] In this step, the pricing strategy optimization results are combined with the personalized recommendation system to propose an innovative personalized recommendation model. j (t)), dynamic user preference adjustment mechanism, and multi-objective optimization framework, we designed a recommendation system that can quickly respond to market changes and balance short-term sales and long-term customer loyalty. The personalized recommendation system can continuously improve the effect in different market environments and ensure the accuracy and adaptability of recommendations.

[0204] like Figure 2 As shown, in another embodiment of the present invention, a power marketing data analysis system is provided, the system comprising:

[0205] The power marketing data collection module 101 is used to collect consumption data of power consumers and build a power consumer behavior analysis model to analyze the consumption data, including:

[0206] Collect and pre-process the consumption data of electricity consumers, construct the behavioral characteristics of consumers based on the pre-processed data and divide consumers into multiple groups, and construct a consumer behavior prediction model based on the supervised learning model based on the behavioral characteristics and groups to predict consumers' reactions to different electricity pricing strategies or promotional activities, and output the response probability of each consumer;

[0207] The user behavior prediction module 102 is used to formulate an optimal pricing strategy according to the output result of the consumer behavior prediction model to achieve maximum sales benefits, including:

[0208] Designing a demand function according to the response probability of each consumer, wherein the demand function represents the demand quantity of the consumer under a specific price, and constructing an optimization model based on the demand function of each consumer to maximize the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process;

[0209] Execute the pricing strategy output by the optimization model and collect response data from consumer groups. Through the Q-learning method, dynamically update the pricing strategy based on the sales effect of each adjustment.

[0210] Design a multi-dimensional evaluation function to cyclically evaluate the overall benefits of all pricing strategies and ultimately select the optimal pricing execution;

[0211] The power pricing module 103 is used to evaluate the effect of each marketing activity in real time and dynamically adjust the pricing strategy based on the evaluation. After each marketing cycle, the feedback on the marketing effect will be input into the pricing strategy of the next cycle for adjustment;

[0212] The power market adjustment module 104 is used to design a dynamic optimization strategy by combining marketing effects and existing pricing strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits;

[0213] The personalized data recommendation module 105 is used to implement personalized recommendations in response to market changes and user behaviors based on the consumption data of each consumer in combination with the pricing strategy after executing the dynamic optimization strategy.

[0214] In summary, the present invention aims to overcome the deficiencies in the prior art and solve the problems of rigid pricing, delayed feedback on marketing activities, and lack of personalized recommendations. Through this innovative solution, power companies can accurately perform pricing and personalized recommendations based on the psychological and behavioral characteristics of consumers, and use adaptive algorithms to dynamically adjust marketing strategies based on real-time data streams, thereby effectively improving marketing effectiveness and user engagement.

[0215] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0216] In addition, for technical details that are not described in detail in this embodiment, reference can be made to the parameter operation method provided in any embodiment of the present invention, and will not be repeated here.

[0217] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system 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 system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0218] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0219] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0220] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for analyzing power marketing data, characterized in that: The method comprises: S1. Collect consumption data of electricity consumers and build a behavior analysis model for electricity consumers to analyze the consumption data, including: Collect and pre-process the consumption data of electricity consumers, construct the behavioral characteristics of consumers based on the pre-processed data and divide consumers into multiple groups, and construct a consumer behavior prediction model based on the supervised learning model based on the behavioral characteristics and groups to predict consumers' reactions to different electricity pricing strategies or promotional activities, and output the response probability of each consumer; S2. Formulate an optimal pricing strategy based on the output results of the consumer behavior prediction model to achieve maximum sales benefits, including: Designing a demand function according to the response probability of each consumer, wherein the demand function represents the demand quantity of the consumer under a specific price, and constructing an optimization model based on the demand function of each consumer to maximize the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process; Execute the pricing strategy output by the optimization model and collect response data from consumer groups. Through the Q-learning method, dynamically update the pricing strategy based on the sales effect of each adjustment. Design a multi-dimensional evaluation function to cyclically evaluate the overall benefits of all pricing strategies and ultimately select the optimal pricing execution; S3. Conduct real-time effect evaluation on each marketing campaign and dynamically adjust pricing strategies based on the evaluation. After each marketing cycle, feedback on marketing effects will be input into the pricing strategy for the next cycle for adjustment. S4. Design dynamic optimization strategies by combining marketing results with existing pricing strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits; S5. Based on the consumption data of each consumer, combined with the pricing strategy after executing the dynamic optimization strategy, personalized recommendations are implemented to respond to market changes and user behavior.

2. A power marketing data analysis method according to claim 1, characterized in that: The K-means clustering algorithm is used to divide consumers into multiple groups according to their behavioral characteristics, wherein each group represents a specific consumption behavior pattern. The goal of the K-means clustering algorithm is to minimize the difference in the consumer feature vectors within the group so that the consumer behaviors of the same group are most similar, which can be expressed as: Among them, C k represents the center of the kth group, representing the typical consumption behavior characteristics of the group; X i represents the behavioral feature vector of consumer i, and N represents the total number of consumers; The output of the consumer behavior prediction model based on the supervised learning model is expressed as: P(response i |X i ,C k )=f(X i ,C k ) Among them, P (response i ) represents the probability of consumer i responding to a certain pricing or promotion activity; f(X i ,C k ) represents the consumer characteristics X i and group characteristics C k The model predicts the probability of the consumer's response based on these inputs.

3. The power marketing data analysis method according to claim 1, characterized in that: The function of demand elasticity correction is expressed as: Among them, Demand i Indicates that the price P i Next, the quantity demanded by consumer i; represents the basic demand of consumer i, the demand without price intervention; P i represents electricity pricing; α: demand elasticity coefficient, which indicates the consumer's sensitivity to price changes; P (response i |X i ,C k ) represents the consumer i’s price P i The probability of response; The objective function of the optimization model is expressed as: Among them, P i Demand represents electricity pricing; i represents consumer i at price P i The demand under i |X i ,C k ) represents the response probability of consumer i; Budget i represents the budget of consumer i; λ represents the Lagrange multiplier, which represents the impact of the budget constraint on the optimization process; The Q-learning method is used to dynamically update the pricing strategy based on the sales effect of each adjustment, which is expressed as: Among them, Q(P k ,t) indicates that at time point t, the price P k The strategy value represents the expected benefit of the pricing strategy; R t represents the actual reward at time t; γ represents the discount factor, which controls the importance of future benefits; α represents the learning rate, which controls the speed of policy update; P k ′ represents other pricing strategies at time point t; The multi-dimensional evaluation function is expressed as: E(P k )=w1·ConsumerAcceptance(P k )+w2·Profit(P k )+w3·Market Share(P k )+w4·Customer Satisfaction(P k ) Among them, ConsumerAcceptance(P k ) represents the pricing strategy P k Profit(P k ) represents the pricing strategy P k Total profit brought by; Market Share (P k ) represents the pricing strategy P k Impact on market share; Customer Satisfaction (P k ) represents the consumer’s satisfaction with the pricing; w1, w2, w3, w4 represent the weights of each evaluation indicator, which are adjusted according to business objectives.

4. A power marketing data analysis method according to claim 3, characterized in that: In order to avoid excessive fluctuations in pricing strategies during real-time adjustments, a regularization term is introduced into the optimization model to constrain the range of pricing changes, which is expressed as: Among them, P k Indicates pricing strategy; Demand represents the initial pricing strategy; i represents consumer i at price P k β1, β2 represent the regularization coefficients, which are used to balance pricing smoothness and demand volatility.

5. The power marketing data analysis method according to claim 1, characterized in that: In S3, the evaluation result quantifies the activity effect based on a comprehensive marketing effect function, which is expressed as: R k (t)=γ1·Sales k (t)+γ2·Participation k (t)+γ3·Satisfaction k (t) Among them, Sales k (t) represents the sales volume of activity k at time t; Participation k (t) represents the number of participants in activity k; Satisfaction k (t) represents consumer satisfaction, which may come from survey data or automatic scoring; γ1, γ2, and γ3 represent the weights of each indicator, reflecting the importance of each factor in effect evaluation.

6. A power marketing data analysis method according to claim 5, characterized in that: The pricing strategy is adjusted based on feedback through real-time effect evaluation results, expressed as: P k (t+1)=P k (t)+η·R k (t) Among them, P k (t) represents the pricing strategy at time t; P k (t+1) represents the pricing strategy for the next cycle, based on R k (t) is adjusted; η represents the adjustment coefficient, which controls the magnitude of pricing changes; At the same time, in order to avoid excessive fluctuations in pricing strategies and ensure that pricing adjustments are within a reasonable range, a time decay mechanism and a time regularization term are introduced into the pricing strategy. The time decay mechanism is expressed as: P k (t+1)=P k (t)+η·R k (t)·(1-β t ) Where β represents the time decay factor, β∈[0,1). As time goes by, the influence of historical feedback gradually weakens; t represents the time step; P k (t+1) represents the pricing with time decay mechanism added; The time regularization term is expressed as: Among them, λ represents the regularization coefficient, which controls the adjustment amplitude; represents the initial pricing strategy, which is used to constrain pricing changes; P k (t+2) represents the pricing with the time regularization term added.

7. The power marketing data analysis method according to claim 1, characterized in that: The S4 specifically includes: The design objective function represents the balance between the benefits and stability brought by pricing adjustments and is expressed as: in, represents the optimization objective function, which means adjusting the pricing strategy P k The comprehensive benefits after (t+1); R k (t) represents the marketing effectiveness evaluation at time t, measuring the performance of the current strategy; P k (t) represents the pricing scheme at the current time t; P k (t+1) represents the pricing plan at the next moment; α represents the weight factor of marketing effect, reflecting the importance of marketing effect in the objective function; γ represents the penalty factor of pricing stability, controlling the amplitude of pricing changes; |P k (t+1)-P k (t)| p The penalty term for pricing changes, p is a positive integer, which penalizes large pricing changes to avoid excessive price fluctuations; The adaptive gradient descent method is used to calculate the direction and step size of the pricing adjustment, so that the pricing adjustment can be adjusted automatically according to market feedback, specifically including: First, calculate the objective function for the pricing scheme P k The gradient of (t+1) is expressed as: Then, the pricing scheme P is updated according to the gradient descent method k (t+1), expressed as in, Represents the objective function About Pricing Plan P k The gradient of (t+1); Represents marketing effect R k (t) Price P k The sensitivity of (t+1) reflects the impact of pricing on marketing effectiveness; η(t) represents the step size at the current moment, which controls the amplitude of pricing adjustment; p represents the power of pricing change, which controls the intensity of the penalty term for pricing stability; Design a feedback mechanism to ensure that each pricing adjustment is optimized based on the results of the previous cycle, expressed as: Among them, the objective function is updated according to the real-time evaluation report after each feedback.

8. The power marketing data analysis method according to claim 7, characterized in that: In order to ensure the stability of the pricing strategy, a stability constraint is introduced in the optimization process. The stability constraint ensures that the amplitude of each pricing adjustment will not be too large, avoiding the instability of the market due to frequent pricing changes, which is expressed as: |P k (t+1)-P k (t)|≤ΔP max Among them, ΔP max Indicates the maximum pricing change range to ensure that pricing adjustments do not exceed the set range, thereby avoiding excessive price fluctuations.

9. The power marketing data analysis method according to claim 1, characterized in that: The S5 specifically includes: Combining user behavior and real-time market feedback, a personalized recommendation scoring model is constructed to dynamically optimize the recommendation results, which can be expressed as: R i,j (t)=λ1·S i,j (t)+λ2·M j (t)+λ3·H i,j (t)+λ4·ΔM j (t) Among them, R i,j (t) represents the personalized recommendation score of user i for product j, which changes over time; S i,j (t) represents the similarity between user i and item j, which is calculated based on historical purchase or browsing behavior using collaborative filtering or deep learning methods; M j (t) represents the market performance score of product j at time t, based on the optimized pricing strategy P k (t), reflects the market demand for goods and the effectiveness of marketing activities; H i,j (t) represents the personalized preference score of user i for product j, based on the user's historical data and preferences; ΔM j (t) represents the change in the market performance of commodity j, measuring the immediate impact of market fluctuations on commodity demand; The dynamic feedback mechanism is introduced to dynamically adjust the recommendation results according to market fluctuations, changes in user preferences and real-time feedback on pricing strategies, which can be expressed as: Among them, ΔR i,j (t) represents the adjustment amount of user i’s recommendation score for product j, reflecting the impact of changes in market and user preferences on recommendations; It represents a measure of the change in the market performance of product j, indicating the immediate impact of pricing adjustments or marketing activities on product performance; It represents the rate at which user i's preference for product j changes, reflecting the user's response speed to recommendations; A multi-objective optimization framework is designed to optimize user loyalty and customer experience, balancing the priorities of different objectives, expressed as: in, represents the overall optimization objective function, which combines the optimization of sales volume and customer loyalty; It indicates the optimization goal based on sales volume, mainly improving sales conversion rate by adjusting the recommendation system; It represents the optimization goal based on customer loyalty, reflecting the impact of the recommendation system on the long-term relationship of users; Whenever the market or user behavior changes, new feedback is received and the recommendations are adjusted based on this feedback.

10. A power marketing data analysis system, characterized in that: The system comprises: The power marketing data collection module is used to collect the consumption data of power consumers and build a power consumer behavior analysis model to analyze the consumption data, including: Collect and pre-process the consumption data of electricity consumers, construct the behavioral characteristics of consumers based on the pre-processed data and divide consumers into multiple groups, and construct a consumer behavior prediction model based on the supervised learning model based on the behavioral characteristics and groups to predict consumers' reactions to different electricity pricing strategies or promotional activities, and output the response probability of each consumer; The user behavior prediction module is used to formulate the optimal pricing strategy according to the output results of the consumer behavior prediction model to achieve the maximum sales benefit, including: Designing a demand function according to the response probability of each consumer, wherein the demand function represents the demand quantity of the consumer under a specific price, and constructing an optimization model based on the demand function of each consumer to maximize the total revenue of all consumer groups while ensuring that the budget limit is not exceeded during the electricity sales process; Execute the pricing strategy output by the optimization model and collect response data from consumer groups. Through the Q-learning method, dynamically update the pricing strategy based on the sales effect of each adjustment. Design a multi-dimensional evaluation function to cyclically evaluate the overall benefits of all pricing strategies and ultimately select the optimal pricing execution; The power pricing module is used to evaluate the effect of each marketing activity in real time and dynamically adjust the pricing strategy based on the evaluation. After each marketing cycle, the feedback on the marketing effect will be input into the pricing strategy of the next cycle for adjustment; The power market adjustment module is used to design dynamic optimization strategies by combining marketing effects and existing pricing strategies to ensure that pricing and marketing strategies can be adjusted in a timely manner according to market changes, thereby maximizing overall benefits; The personalized data recommendation module is used to implement personalized recommendations in response to market changes and user behaviors based on the consumption data of each consumer, combined with the pricing strategy after executing the dynamic optimization strategy.

Images