Intelligent pricing management system based on machine learning and data analysis

Abstract

The invention relates to the technical field of intelligent pricing, in particular to an intelligent pricing management system based on machine learning and data analysis, and aims to improve the purity and consistency of data and lay an accurate data foundation for subsequent analysis through cleaning and format unified processing of market original data and elimination of redundant information. Through a non-dominated sorting genetic algorithm, the influence weight of each data point is analyzed, comprehensive quantitative modeling of price influence factors is realized, a model basis with better interpretation and adaptability is provided for price prediction, and market changes and competition patterns are dynamically captured by combining a linear regression model, so that the price prediction accuracy is improved. And the price is ensured to be dynamically and accurately adjusted along with the market through a fine adjustment strategy, the dynamic matching capability between the price and the market fluctuation is realized, the adaptability of the strategy under different market conditions is verified and optimized through a multi-scene simulation test and market feedback monitoring, and the high efficiency and stability of the price adjustment strategy in practical application are ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent pricing, and particularly to an intelligent pricing management system based on machine learning and data analysis. Background Art

[0002] The core of the intelligent pricing technology field lies in realizing automatic price adjustment and precise matching through data analysis and model prediction, according to multi-dimensional factors such as market demand, user behavior, and competitive environment, so as to improve revenue or market competitiveness.

[0003] The purpose of the intelligent pricing management system based on machine learning and data analysis is to comprehensively evaluate and optimize the complex influencing factors in the pricing process by establishing a data-driven price model, adapt to market volatility and consumer preferences, achieve price optimization, improve revenue, and enhance user satisfaction.

[0004] The existing technology lacks an efficient cleaning and screening mechanism, resulting in redundant data affecting the analysis quality, increasing the computational burden. It is difficult to effectively quantify complex multi-dimensional influencing factors in model construction, and the analysis and adjustment strategies for market dynamics are often based on static parameter configurations, lacking dynamic response capabilities and unable to quickly adapt to the frequent fluctuations of the market environment. There is also a lack of a systematic method for multi-scenario testing, unable to comprehensively evaluate the stability and adaptability of strategies under different conditions, resulting in the implementation effect of strategies being difficult to comprehensively guarantee. Summary of the Invention

[0005] The purpose of the present invention is to solve the deficiencies existing in the prior art and propose an intelligent pricing management system based on machine learning and data analysis.

[0006] To achieve the above purpose, the present invention adopts the following technical solutions: The intelligent pricing management system based on machine learning and data analysis includes:

[0007] Initial data processing module: According to the market raw input data, perform data cleaning and format unification processing, remove useless formats and characters, and generate a cleaned data set;

[0008] Key data screening module: Based on the cleaned data set, use a Bloom filter to exclude known irrelevant data items, concentrate on processing potentially valid transaction data and pricing influencing factor data, and generate a screened key data set;

[0009] Price model construction module: Based on the screened key data set, use the non-dominated sorting genetic algorithm to analyze and determine the influence weights of each data point on the price, construct a price influence model, and generate a preliminary price prediction model;

[0010] Price Adjustment Strategy Module: Through the preliminary price prediction model, using a linear regression model, analyze market dynamics and competitor strategies, and ensure that prices reflect market trends through model fine-tuning to form a dynamic price adjustment strategy;

[0011] Price Strategy Simulation and Testing Module: According to the dynamic price adjustment strategy, simulate market scenarios, evaluate the performance of the strategy under various market conditions through multi-scenario testing, confirm the market adaptability of the dynamic price adjustment strategy, and generate a market adaptability feature set;

[0012] Price Implementation Monitoring Module: Monitor the performance of the dynamic price adjustment strategy, capture market feedback in real time, adjust the dynamic price adjustment strategy to match market changes, and output a price implementation benefit assessment.

[0013] As a further solution of the present invention, the initial data processing module includes a data cleaning sub-module, a format unification sub-module, and a data set generation sub-module, where:

[0014] Data Cleaning Sub-module: Based on the original market input data, remove blank records and duplicate records, delete data fields that do not meet the specifications by matching field types and structures, and at the same time screen key fields and delete noise fields to generate normalized data content;

[0015] Format Unification Sub-module: Based on the normalized data content, parse the field structure and adjust the field order, perform standard type conversion on numerical fields and text fields, and unify the time format and unit format to generate uniformly formatted data;

[0016] Data Set Generation Sub-module: Based on the uniformly formatted data, integrate transaction data and pricing data through field mapping and data grouping, extract the required fields and remove irrelevant fields, and at the same time divide the field structure to output the cleaned data set.

[0017] As a further solution of the present invention, the key data screening module includes a Bloom filter sub-module, a data screening sub-module, and a key data generation sub-module, where:

[0018] Bloom Filter Sub-module: Based on the cleaned data set, eliminate known redundant fields through field matching, mark irrelevant fields and filter duplicate data, and at the same time exclude invalid data to generate a screened data set;

[0019] Data Screening Sub-module: Based on the screened data set, parse the transaction data and pricing fields, screen the pricing-related fields, remove the irrelevant fields in the transaction data, and match the screened fields through association analysis to generate screened data content;

[0020] Key data generation sub-module: Based on the screened data content, re-order the screened fields by field association grouping, integrate the transaction fields and pricing fields, establish a complete data table structure through content extraction, and output the screened key data set.

[0021] As a further solution of the present invention, the price model construction module includes a data weight analysis sub-module, a model parameter construction sub-module, and a price prediction model generation sub-module, wherein:

[0022] Data weight analysis sub-module: Based on the screened key data set, analyze the data fields and statistically calculate the association between data points and price changes item by item, calculate the impact of each group on price changes through grouped data, match the change range of each data point with historical price records and calculate the impact value, and generate the price impact weight;

[0023] Model parameter construction sub-module: Based on the price impact weight, adopt the non-dominated sorting genetic algorithm, screen the fields according to the weight and arrange them in the order of impact, establish an association relationship for the data fields and define linear or non-linear combinations, match the weight data points with the field corresponding rules by constructing a relationship table, summarize the key parameters and set the parameter calculation rules, and establish a price impact parameter set;

[0024] Price prediction model generation sub-module: Based on the price impact parameter set, analyze the parameter structure and gradually construct the model relationship, input the screened data set layer by layer for calculation and adjust the data output order, proofread the matching degree between the data output result and historical price changes, correct the parameters, and generate a preliminary price prediction model.

[0025] As a further solution of the present invention, the non-dominated sorting genetic algorithm is calculated according to the formula:

[0026]

[0027] where: f(x) represents the comprehensive price impact parameter value, x represents the current data field value input, w i represents the price impact weight of the i-th field, g i (x) represents the price impact factor corresponding to the i-th field, h i represents the correlation weight between fields, t i represents the time sensitivity of the field, k i represents the change range of the field, β represents the correlation correction coefficient, α represents the time sensitivity correction coefficient, γ represents the change range correction coefficient, and n represents the total number of screened key fields.

[0028] As a further solution of the present invention, the price adjustment strategy module includes a market dynamics analysis sub-module, a competitive strategy response sub-module, and a dynamic strategy generation sub-module, wherein:

[0029] Market dynamics analysis sub-module: Based on the preliminary price prediction model, adopt a linear regression model, gradually analyze the price change records in the market historical data and extract core data points, group the data points by time dimension, analyze the differences between each group of data points and count the change trends, integrate the grouping results and establish a market dynamics feature list, and generate a market dynamics feature set;

[0030] Competitive strategy response sub-module: Based on the market dynamics feature set, extract the data points that match the competitor's price data, compare the competitor's price records and analyze the change range, classify the data points by similarity and extract key data points, integrate the classification results and construct a competitive response relationship table, and generate competitive strategy response features;

[0031] Dynamic strategy generation sub-module: Based on the competitive strategy response features, match the core data points of the price prediction model and the market dynamics feature set, analyze the influence parameters of the core data points and gradually adjust the model, integrate the adjusted model results and verify them, and generate a dynamic price adjustment strategy.

[0032] As a further solution of the present invention, the linear regression model is in accordance with the formula:

[0033] y t = θ0 + θ1·u t + θ2·v t + θ3·w t + θ4·q t + ∈ t

[0034] Where: y t represents the market price change value corresponding to time t, θ0 represents the intercept term, u t represents the historical trading volume data at time t, v t represents the external market influence factor at time t, w t represents the competitive market price volatility at time t, q t represents the social public opinion intensity factor at time t, θ1 represents the regression coefficient of the trading volume u t θ2 represents the regression coefficient of the external market influence factor v t θ3 represents the regression coefficient of the competitive market price volatility w t θ4 represents the regression coefficient of the social public opinion intensity factor q t ∈ t represents the residual.

[0035] As a further solution of the present invention, the price strategy simulation test module includes a market scenario construction sub-module, a multi-scenario test sub-module, and a strategy adaptability confirmation sub-module, where:

[0036] Market scenario construction sub-module: Based on the dynamic price adjustment strategy, analyze historical market data, split transaction data and pricing data, organize market characteristic data by time period and group them, generate a scenario data set by adjusting the price fluctuation range and sales volume change trend within the time period, and perform parameter matching on the scenario data set to generate a market scenario data set.

[0037] Multi-scenario testing sub-module: Based on the market scenario data set, load different scenario data sets one by one and match the dynamic price adjustment strategy, gradually simulate the strategy execution process and record the data differences before and after price adjustment, statistically analyze the response degree of the scenario data set to the strategy by comparing the performance of key data points, organize the response results and establish a comprehensive performance data table, and generate scenario test performance data.

[0038] Strategy adaptability confirmation sub-module: Based on the scenario test performance data, extract the performance data of the strategy in each scenario data set, confirm the response range of the strategy to each market scenario by calculating the deviation value of the performance data, analyze the correlation between the deviation range and market characteristics and extract stability parameters, and generate a market adaptability feature set.

[0039] As a further solution of the present invention, the price implementation monitoring module includes a market feedback capture sub-module, a strategy dynamic adjustment sub-module, and a benefit evaluation output sub-module, where:

[0040] Market feedback capture sub-module: Based on the dynamic price adjustment strategy, collect real-time market transaction data and competitive price data and classify them by time period, analyze the correlation between sales volume and price change data and extract key feedback data, gradually compare the deviation between the current data and historical data, and generate a market feedback data set by combining the feedback results.

[0041] Strategy dynamic adjustment sub-module: Based on the market feedback data set, analyze the price change trend in the feedback data and match the execution parameters of the dynamic price adjustment strategy, gradually adjust the parameter range and correct the price adjustment model, and test the price fluctuation change of the feedback data by scenario through the corrected model to generate an adjusted strategy parameter set.

[0042] Benefit evaluation output sub-module: Based on the adjusted strategy parameter set, calculate the impact of the parameters on market sales data, analyze the correlation between the sales volume and price change data after parameter adjustment, statistically analyze the deviation value of key data and organize the comprehensive performance results, and output the price implementation benefit evaluation.

[0043] As a further solution of the present invention, gradually adjust the parameter range and correct the price adjustment model, extract the key fields of price changes in the feedback data and screen item by item, classify the parameter fields affecting the implementation effect of the price adjustment strategy, calculate the parameter deviation range by segment and set the adjustment threshold, gradually expand and shrink the adjustment range for the parameters exceeding the threshold, and at the same time group and record the adjustment range according to the time dimension, correct the input rules of the price adjustment model through the grouped records of the parameters, layer by layer analyze the model structure after parameter correction and generate the updated price adjustment model;

[0044] Test the price fluctuation changes of the feedback data scene by scene through the corrected model, load the scene data in the market feedback data set and classify it according to the scene characteristics, input each scene data item by item into the corrected model and record the price adjustment results output by the model, calculate the matching degree and deviation of the price fluctuation of each scene by comparing the output data of the model with the actual feedback price changes in the scene data, gradually summarize the price adjustment effects under different scene conditions, analyze the trend of price fluctuation changes and generate the fluctuation performance data set of the scene-by-scene test.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0046] 1. In the present invention, by cleaning and unifying the format of the original market data, redundant information is eliminated, improving the purity and consistency of the data, laying an accurate data foundation for subsequent analysis;

[0047] 2. In the present invention, through the non-dominated sorting genetic algorithm, analyze the influence weights of each data point, realize the comprehensive quantitative modeling of price influencing factors, and provide a more explanatory and adaptable model foundation for price prediction;

[0048] 3. In the present invention, by combining the linear regression model, dynamically capture market changes and competition patterns, and ensure the accurate adjustment of prices with market dynamics through the fine-tuning strategy, realizing the dynamic matching ability between prices and market fluctuations;

[0049] 4. In the present invention, through multi-scene simulation tests and market feedback monitoring, verify and optimize the adaptability of the strategy under different market conditions, ensure the efficiency and stability of the price adjustment strategy in practical applications, enable the price adjustment strategy to quickly respond to market changes, and enhance the comprehensive ability of revenue improvement and market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 is the system flow chart of the present invention;

[0051] Figure 2 is the system flow diagram of the present invention;

[0052] Figure 3 This is a schematic diagram of the system framework of the present invention. Detailed implementation manners

[0053] In order to make the objectives, technical solutions and advantages of the present invention more clear and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0054] Please refer to Figure 1 and Figure 2 , the present invention provides a technical solution: an intelligent pricing management system based on machine learning and data analysis includes:

[0055] Initial data processing module: According to the original market input data, perform data cleaning and format unification processing, remove useless formats and characters, and generate a cleaned data set;

[0056] Key data screening module: Based on the cleaned data set, use a Bloom filter to exclude known irrelevant data items, and centrally process potentially valid transaction data and pricing influencing factor data to generate a screened key data set;

[0057] Price model construction module: Based on the screened key data set, use a non-dominated sorting genetic algorithm to analyze and determine the influence weights of each data point on the price, construct a price influence model, and generate a preliminary price prediction model;

[0058] Price adjustment strategy module: Through the preliminary price prediction model, use a linear regression model to analyze market dynamics and competitor strategies, and ensure that the price reflects market trends through model fine-tuning to form a dynamic price adjustment strategy;

[0059] Price strategy simulation test module: According to the dynamic price adjustment strategy, simulate market scenarios, evaluate the performance of the strategy under various market conditions through multi-scenario tests, confirm the market adaptability of the dynamic price adjustment strategy, and generate a market adaptability feature set;

[0060] Price implementation monitoring module: Monitor the performance of the dynamic price adjustment strategy, capture market feedback in real time, adjust the dynamic price adjustment strategy to match market changes, and output a price implementation benefit evaluation.

[0061] Please refer to Figure 3 , the initial data processing module includes a data cleaning sub-module, a format unification sub-module and a data set generation sub-module, where:

[0062] Data cleaning sub-module: Based on the original market input data, blank records and duplicate records are removed. By matching the field types and structures, data fields that do not meet the specifications are deleted. At the same time, key fields are screened and noise fields are deleted to generate normalized data content;

[0063] Format unification sub-module: Based on the normalized data content, the field structure is parsed and the field order is adjusted. Numeric fields and text fields are converted to standard types, and the time format and unit format are unified to generate data in a unified format;

[0064] Dataset generation sub-module: Based on the data in a unified format, transaction data and pricing data are integrated through field mapping and data grouping. The required fields are extracted and irrelevant fields are removed. At the same time, the field structure is divided to output the cleaned dataset;

[0065] Data cleaning sub-module: Based on the original market input data, conditional screening algorithms are used to remove blank records and duplicate records. The specific steps are as follows: Use the Pandas library to operate on the data table. Locate all null value records through the isnull function, and use the dropna method to delete the rows containing null values. The parameter axis is set to 0 to indicate row operation, and the parameter how is set to any to indicate that as long as there is a null value, it will be deleted. Use the duplicated function to mark duplicate rows and delete all duplicate records. The parameter keep is set to first to keep the records that appear for the first time. Perform matching operations on the field types and structures, obtain the field types one by one, and perform field-by-field checks by comparing the field types with the expected standard types. Match the data fields that meet the specifications and delete the fields that do not match. At the same time, perform screening operations on the key fields, use the field list to perform column index matching on the data table, and only keep the fields that appear in the column index. Detect and delete noise fields through statistical analysis methods. Specifically, count the frequency of field values, and fields with a frequency lower than 10% are marked as noise fields and deleted to generate normalized data content;

[0066] Format unification sub-module: Based on the normalized data content, data type conversion algorithms are used to parse the field structure and adjust the field order. Sort the data table according to the specified fields. The sorting parameter ascending is set to True to indicate ascending sorting. Use the astype method to convert numeric fields to floating-point types and text fields to string types. Perform unified operations on the time field format to unify the time field format and clarify the target time format. Perform unified operations on the unit format. Use the string replacement method to standardize the unit identifiers in the field content. Specifically, use the replace method to unify the unit identifiers to standard symbols. Check the type consistency of all fields through field type checks to generate data in a unified format;

[0067] Dataset Generation Sub-module: Based on data in a unified format, it integrates transaction data and pricing data through field mapping and data grouping algorithms. It performs field mapping operations on transaction data and pricing data by field name using the merge method. The parameter how is set to inner for inner join, and the on parameter specifies the matching field name. It groups the mapped data by field, specifies the grouping field using the groupby method, calculates the statistical values within the group for the grouped data, uses the sum method to calculate the total field value, and the mean method to calculate the field mean. It extracts the required fields by field index matching, only retains the fields in the field index list, deletes the irrelevant fields using the drop method, divides the field structure, and completes data division by creating a table structure containing field names, field types, and grouping indexes. The field names in the table structure are recorded as strings, the field types are recorded as floating-point or string types, and the grouping indexes are recorded as integers. Finally, it outputs the cleaned dataset;

[0068] Please refer to Figure 3 , the Key Data Screening Module includes a Bloom Filter Sub-module, a Data Screening Sub-module, and a Key Data Generation Sub-module, where:

[0069] Bloom Filter Sub-module: Based on the cleaned dataset, it eliminates known redundant fields through field matching, marks irrelevant fields, filters duplicate data, and excludes invalid data to generate a screened dataset;

[0070] Data Screening Sub-module: Based on the screened dataset, it parses transaction data and pricing fields, screens pricing-related fields, removes irrelevant fields in the transaction data, and matches the screened fields through association analysis to generate screened data content;

[0071] Key Data Generation Sub-module: Based on the screened data content, it reorders the screened fields by field association grouping, integrates transaction fields and pricing fields, establishes a complete data table structure through content extraction, and outputs the screened key dataset;

[0072] Bloom Filter Sub-module: Based on the cleaned dataset, it uses the Bloom filter algorithm for field matching operations. By constructing a bit array of the Bloom filter, the size of the bit array is set to 2 to the 20th power. It generates the corresponding index positions for the field values using two hashes respectively and writes the index positions into the Bloom filter. It uses the Bloom filter to retrieve and match each input field to check if there is a matching field. During the comparison process, it uses boolean operations to mark irrelevant fields, filters the duplicate data indexes recorded in the Bloom filter, eliminates the duplicate fields at the recorded positions, and simultaneously uses regular expression matching to batch clear redundant field values and excludes invalid data to generate a screened dataset;

[0073] Data Screening Sub-module: Based on the screening data set, the association rule algorithm is used to analyze transaction data and pricing fields. The specific steps include encoding the transaction data and pricing fields. The encoding rule is to map the field names to integer values. The Apriori algorithm is used for frequent item set analysis. The minimum support threshold is set to 0.3, and the confidence threshold is set to 0.7. The associated fields that meet the support and confidence conditions are extracted. The mapping table is used to reverse decode the pricing associated fields to generate the original field values. The fields that do not meet the minimum support threshold are removed. The Pandas library in Python is used to screen the extracted fields, and the unrelated fields in the transaction data are removed row by row. The screening field set is generated by matching through the field value index, and the screened data content is output;

[0074] Key Data Generation Sub-module: Based on the screened data content, the hash grouping algorithm is used to reorder the screening fields by grouping according to field association. The field values are converted into fixed-length hash values and used as the grouping basis. The data rows are reordered through the grouping index generated by the hash value. At the same time, the Pandas library is used to integrate the transaction fields and pricing fields. The concat function is used to concatenate the transaction fields and pricing fields by column. The field matching rule is that the transaction data index is the same as the pricing field index. The content is extracted from the integrated fields using string processing methods, and the numerical fields are converted into floating-point formats and a complete data table structure is established. The table structure is defined as including field names, data types, and grouping indexes. The field names are recorded as character types, the data types are floating-point types or string types, and the grouping indexes are integer types. Finally, the screened key data set is output;

[0075] Please refer to Figure 3 , the price model construction module includes a data weight analysis sub-module, a model parameter construction sub-module, and a price prediction model generation sub-module, where:

[0076] Data Weight Analysis Sub-module: Based on the screened key data set, the data fields are analyzed and the association between data points and price changes is counted item by item. The impact of each group on price changes is calculated by grouping the data. The change range of each data point is matched with the historical price records and the impact value is calculated to generate the price impact weight;

[0077] Model Parameter Construction Sub-module: Based on the price impact weight, the non-dominated sorting genetic algorithm is used. The fields are screened according to the weight and arranged in the order of influence. The association relationships between data fields are established and linear or non-linear combinations are defined. The corresponding rules between weight data points and fields are matched by constructing a relationship table. The key parameters are summarized and the parameter calculation rules are set to establish a set of price impact parameters;

[0078] Price prediction model generation sub-module: Based on the price impact parameter set, analyze the parameter structure, gradually construct the model relationship, input and filter the data set layer by layer for calculation, adjust the data output order, proofread the matching degree between the data output result and the historical price change, correct the parameters, and generate a preliminary price prediction model;

[0079] Data weight analysis sub-module: Based on the filtered key data set, use the Pearson correlation coefficient calculation method to analyze each data field item by item and count the correlation between the data points and the price change. The specific steps are to standardize all fields in the data set, calculate the correlation with the price field for each field in the data table in turn, complete the standardization by subtracting the average value of the field value and normalizing with the standard deviation, execute the pairwise correlation coefficient calculation between fields through the programming library method, filter out the fields with a correlation coefficient greater than 0.5 and record the correlation strength, group and count the data, calculate the specific impact of each group of data on the price change using the group average value, match the change range of each data point item by item in combination with the historical price record, and use the dynamic segmentation method to analyze the correlation strength of each data point within the change range, and finally generate the price impact weight;

[0080] Model parameter construction sub-module: Based on the price impact weight, use the non-dominated sorting genetic algorithm to screen the weights of the fields and sort them according to the influence order. The specific steps are to initialize the sorting of the price impact weights according to the importance of the fields, divide the fields into multiple weight levels to construct different populations, perform the selection operation on the populations, determine the field retention rules by preferentially selecting and eliminating the weights corresponding to the fields, when performing the crossover operation, split and recombine the fields according to their positions to generate a new field arrangement combination, and complete the mutation operation by randomly adjusting the field positions in the field arrangement. Construct a linear relationship for the fields with higher weights, combine the low-weight fields according to non-linear rules, gradually establish a weight relationship table for the fields, match the association rules between the fields to further define the specific relationship between the data points and the fields, finally summarize all the weight data points to generate a key parameter set, and set the data calculation rules to complete the construction of the parameter set, and finally generate the price impact parameter set;

[0081] Price prediction model generation sub-module: Based on the price impact parameter set, the multi-layer perceptron neural network algorithm is used to analyze the parameter set and gradually establish model relationships. The specific steps are to construct the input layer, hidden layer, and output layer of the model. The number of fields in the input layer corresponds to the number of fields in the parameter set. The number of fields in the hidden layer is set to twice that of the input layer. The activation function is selected as a non-linear processing function for processing the input data layer by layer. The forward propagation method is used to calculate the weighted results of data weights and field values. The weight values are adjusted iteratively after random initialization. The matching relationship between the input and the hidden layer is calibrated through the field distribution of the parameter set. The model output is calibrated using historical price records, and the output order is adjusted to conform to the logical relationship of the parameter set. After dynamically correcting the model parameters, the preliminary construction of the prediction model is completed, and finally, a preliminary price prediction model is generated.

[0082] Please refer to Figure 3 , the non-dominated sorting genetic algorithm, according to the formula:

[0083]

[0084] where: f(x) represents the comprehensive price impact parameter value, x represents the value of the current data field input, w i represents the price impact weight of the i-th field, g i (x) represents the price impact factor corresponding to the i-th field, h i represents the correlation weight between fields, t i represents the time sensitivity of the field, k i represents the change range of the field, β represents the correlation correction coefficient, α represents the time sensitivity correction coefficient, γ represents the change range correction coefficient, and n represents the total number of selected key fields;

[0085] Execution process: First, analyze the input original data field x through a machine learning model, extract the key fields with a high degree of correlation with price decisions, and use the non-dominated sorting genetic algorithm to determine the price impact weight w i , and reflect the importance of each field to price decisions through the weight w i . Then, for each field, calculate the price impact factor g i (x), which is generated by combining historical data and field characteristics and reflects the direct contribution of the field value to the price. At the same time, further calculate the correlation weight h i of the field. By establishing a correlation matrix between fields, the correlation degree between the field and other fields is quantified, and the normalization method is used to adjust it to the range of [0, 1]. Subsequently, calculate the time sensitivity t i of the field. By analyzing the change trend of the field in the time series and combining methods such as time difference and moving average, its time dependence is quantified. Then calculate the change range k of the fieldi , by analyzing the recent fluctuation range of field values, combining the standard deviation and normalization methods to determine the impact degree of fluctuations on prices, and finally combining the correction coefficients β, α, and γ, which are used to adjust the weight contributions of the correlations between fields, time sensitivity, and fluctuation amplitude respectively, to calculate the comprehensive price impact parameter value f(x), providing support for generating the final price for the intelligent pricing management system.

[0086] Please refer to Figure 3 , the price adjustment strategy module includes a market dynamics analysis sub-module, a competitive strategy response sub-module, and a dynamic strategy generation sub-module, where:

[0087] Market dynamics analysis sub-module: Based on the preliminary price prediction model, using the linear regression model, gradually analyze the price change records in the market historical data and extract the core data points, group the data points by time dimension, analyze the differences between each group of data points and statistically analyze the change trends, integrate the grouping results and establish a market dynamics feature list to generate a market dynamics feature set;

[0088] Competitive strategy response sub-module: Based on the market dynamics feature set, extract the data points that match the competitor price data, compare the competitor price records and analyze the change intervals, classify the data points by similarity and extract the key data points, integrate the classification results and construct a competitive response relationship table to generate competitive strategy response features;

[0089] Dynamic strategy generation sub-module: Based on the competitive strategy response features, match the price prediction model with the core data points of the market dynamics feature set, analyze the impact parameters of the core data points and gradually adjust the model, integrate the adjusted model results and verify them to generate a dynamic price adjustment strategy;

[0090] Market dynamics analysis sub-module: Based on the preliminary price prediction model, use the linear regression algorithm to gradually analyze the price change records in the market historical data and extract the core data points. Sort the price change records in the data table according to the time field, use the sliding window method to segment the data for each time window. The sliding window size is set to 7 days and the step size is set to 1 day. Calculate the slope and intercept of the linear regression model for the data points in each window, use the ordinary least squares method to calculate the regression parameters, screen the core data points after extracting the regression parameters of the data points in the window, mark the data points with the absolute value of the slope greater than 0.05 as core data points, group the core data points by time dimension, use the grouping statistical method to analyze the differences between each group of data points, calculate the change trends of each group through the mean and standard deviation, integrate the trend data of the grouping into a dynamic feature list, and finally generate a market dynamics feature set;

[0091] Competitive Strategy Response Sub-module: Based on the market dynamic feature set, a similarity measurement algorithm is used to extract data points that match the competitor's price data. The cosine similarity method is employed to calculate the similarity between the market dynamic feature data and the competitor's price data. Data points with a similarity greater than 0.8 are marked as matching data points. The competitor's price records are compared and the change intervals are analyzed. The overlapping parts of the two sets of data are matched through the time field, and the start and end values of the price change interval are calculated. The data points are classified according to similarity, and the k-means clustering algorithm is used to classify the matching data points. The number of clusters k is set to 3, and the centroids are initialized as 3 randomly selected data points. The distance between each data point and the centroids is calculated based on similarity and the category is reallocated. The category assignment is iteratively adjusted until the centroids are stable. The key data points in each category are extracted, and the data points with the highest frequency of occurrence in each category are selected through frequency statistics. The classification results are integrated and a competitive response relationship table is constructed, and finally, the competitive strategy response features are generated;

[0092] Dynamic Strategy Generation Sub-module: Based on the competitive strategy response features, a parameter optimization algorithm is used to match the core data points of the price prediction model with the market dynamic feature set. The particle swarm optimization algorithm is used to adjust the model parameters. The size of the particle swarm is initialized to 50, and each particle corresponds to a combination of model parameters. The velocity range of the particles is set to [-1, 1]. The fitness function is defined based on the core data points in the market dynamic feature set, and the mean square error between the model output result and the actual value of the core data points is used as the fitness value. The position and velocity of the particles are iteratively updated, and the movement of the particles is guided by the global optimal solution and the individual optimal solution. The influencing parameters of the core data points are gradually analyzed and the model is adjusted. The price prediction model is updated with the optimized parameters. The adjusted model results are integrated to verify the price prediction data. The cross-validation method is used to split the training set and the test set for training and verification respectively, and finally, a dynamic price adjustment strategy is generated.

[0093] Please refer to Figure 3 , the linear regression model, according to the formula:

[0094] y t = θ0 + θ1·u t + θ2·v t + θ3·w t + θ4·q t + ò t

[0095] where: y t represents the market price change value corresponding to time t, θ0 represents the intercept term, u t represents the historical trading volume data at time t, v t represents the external market impact factor at time t, w t represents the competitive market price volatility at time t, qt The social public opinion intensity factor representing time t, where θ1 represents the trading volume u t is the regression coefficient, and θ2 represents the external market impact factor v t is the regression coefficient, and θ3 represents the competitive market price volatility w t is the regression coefficient, and θ4 represents the social public opinion intensity factor q t is the regression coefficient, ∈ t represents the residual;

[0096] Execution process: First, extract the time series y t from the market historical data, which represents the change values of the market price at different time points and serves as the target variable for prediction. Then, obtain the historical trading volume u t at the corresponding time point t, using the transaction quantity recorded in the database as the input variable. Meanwhile, extract the external market impact factor v t , such as data on macroeconomic indicators and policy changes, and normalize them to the same dimension for model processing. Subsequently, analyze other market price data related to the current market competition and calculate the price volatility w t Next, calculate the social public opinion intensity factor q t , which is used as the input variable after statistically counting the occurrence frequency of relevant keywords within a certain time period and normalizing the processing. Finally, use the least squares method to optimize the regression coefficients θ0, θ1, θ2, θ3, and θ4 in the formula, and determine the optimal coefficient values by minimizing the sum of squared residuals. After all the coefficients of the model are optimized, substitute the data of each time point into the formula for calculation in turn, gradually analyze the price change trend of the data points, generate a list of market dynamic characteristics, and provide accurate market dynamic analysis support for the intelligent pricing management system.

[0097] Please refer to Figure 3 , the price strategy simulation test module includes a market scenario construction sub-module, a multi-scenario test sub-module, and a strategy adaptability confirmation sub-module, where:

[0098] Market scenario construction sub-module: Based on the dynamic price adjustment strategy, analyze the historical market data, split the transaction data and pricing data, organize the market characteristic data by time period and group them, generate a scenario data group by adjusting the price fluctuation range and sales volume change trend within the time period, and perform parameter matching on the scenario data group to generate a market scenario data set;

[0099] Multi-scenario test sub-module: Based on the market scenario data set, load different scenario data groups one by one and match the dynamic price adjustment strategy, gradually simulate the strategy execution process and record the data differences before and after the price adjustment, statistically analyze the response degree of the scenario data group to the strategy by comparing the performance of key data points, organize the response results and establish a comprehensive performance data table, and generate scenario test performance data;

[0100] Strategy Adaptability Confirmation Sub-module: Based on the scenario test performance data, extract the performance data of the strategy in each scenario data group, confirm the response range of the strategy to each market scenario by calculating the deviation value of the performance data, analyze the correlation between the deviation range and market characteristics and extract stability parameters, and generate a market adaptability feature set;

[0101] Market Scenario Construction Sub-module: Based on the dynamic price adjustment strategy, use the time series decomposition algorithm to analyze historical market data and split transaction data and pricing data. Use the seasonal decomposition method to separate the trend, cycle, and residuals of historical market data. Set the time window to 12 months. Use the separation results to classify transaction data and pricing data into long-term trend items and short-term fluctuation items respectively. Group the market characteristic data by quarter through the time field. Use the grouping function to count the price fluctuation range within the time period, calculate the difference between the maximum and minimum prices in each time period and record it. Analyze the sales volume change trend and calculate the average sales volume change rate in each time period by the moving average method. Mark the time periods with a change rate greater than 5% and generate scenario data groups. Integrate the scenario data groups through parameter matching operations. Use the method of field index and value matching to establish a one-to-one correspondence between fields and data points in the scenario group. Finally, generate a market scenario data set;

[0102] Multi-Scenario Test Sub-module: Based on the market scenario data set, use the discrete event simulation method to load different scenario data groups one by one and match the dynamic price adjustment strategy. Gradually load the parameter settings of each scenario data group through the simulator. Set the simulation time step to 1 day. Gradually record the specific operation data of price adjustment in the scenario. Perform real-time calculations on data fields according to the strategy execution process. Gradually associate the dynamic price adjustment strategy with the data in the scenario data group through the field mapping function. Extract the price change records and sales data after the strategy execution. Use the data comparison function to calculate the differences between key data points before and after the adjustment. Complete statistical analysis by calculating the absolute change rate of the difference value. Classify and organize the degree of response of data points to the strategy according to groups. Summarize the response results of all scenarios to establish a comprehensive performance data table. Finally, generate scenario test performance data;

[0103] Strategy Adaptability Confirmation Sub-module: Based on the scenario test performance data, the deviation analysis method is used to extract the performance data of the strategy in each scenario data group. The adaptability analysis is completed by calculating the deviation values of the strategy performance in each scenario data group. The absolute deviation formula is used to calculate the deviation values of key fields in the performance data item by item. The fields with deviation values greater than 10% are marked as unstable fields. The correlation analysis between the deviation range and market characteristics uses the correlation analysis method to calculate the correlation coefficients between the market characteristic fields and the deviation values field by field. The stability parameters are extracted by screening the fields with correlation coefficients greater than 0.5. Finally, all fields and parameters are summarized to form the market adaptability feature set.

[0104] Please refer to Figure 3 , the price implementation monitoring module includes a market feedback capture sub-module, a strategy dynamic adjustment sub-module, and a benefit evaluation output sub-module, where:

[0105] Market Feedback Capture Sub-module: Based on the dynamic price adjustment strategy, real-time market transaction data and competitive price data are collected and classified by time period. The association between sales volume and price change data is analyzed and key feedback data is extracted. The deviation between the current data and historical data is gradually compared, and the market feedback data set is generated by merging the feedback results;

[0106] Strategy Dynamic Adjustment Sub-module: Based on the market feedback data set, the price change trend in the feedback data is analyzed and the execution parameters of the dynamic price adjustment strategy are matched. The parameter range is gradually adjusted and the price adjustment model is corrected. The price fluctuation changes of the feedback data are tested by scenario through the corrected model, and the adjusted strategy parameter set is generated;

[0107] Benefit Evaluation Output Sub-module: Based on the adjusted strategy parameter set, the impact of the parameters on the market sales data is calculated. By analyzing the association between the sales volume and price change data after parameter adjustment, the deviation values of key data are statistically analyzed and the comprehensive performance results are sorted out, and the price implementation benefit evaluation is output;

[0108] Market feedback capture submodule: Based on the dynamic price adjustment strategy, the streaming data processing method is used to collect real-time market transaction data and competitive price data and classify them by time period. The real-time data is received and classified through the Kafka stream processing framework. The data collection window is set to 1 hour, and the timestamp field is used to group the data by time period. The sales volume and price change related fields are extracted through field filtering operations, and irrelevant fields are removed to simplify the data processing process. The association between sales volume and price change data is analyzed, and the rolling window analysis method is used to calculate the sales volume change rate in each time period. The rate calculation is based on the one-to-one matching of the sales volume increment and the price change ratio within the time interval. By gradually comparing the deviation between the current data and the historical data, the data differentiation calculation method is used to match the current and historical field values one by one and calculate the deviation value. Records with a deviation range greater than 5% are marked as abnormal feedback data. The normal feedback data and abnormal feedback data are integrated through field merging operations to finally generate a market feedback data set;

[0109] Strategy dynamic adjustment submodule: Based on the market feedback data set, the gradient descent algorithm is used to analyze the price change trend in the feedback data and match the execution parameters of the dynamic price adjustment strategy. The trend direction is calculated by analyzing the price field change sequence in the feedback data, and the single-step change value is extracted by using the difference operation of the sequence. The change value is input into the gradient descent model to adjust the parameter range. The learning rate is set to 0.01, and the parameter value is updated step by step to correct the price adjustment model. During the model correction process, the price fluctuation changes of the feedback data are tested scenario by scenario. The hierarchical regression test method is used to load each group of data in the feedback data set, and the price fluctuation value after the model adjustment is calculated step by step. The model output value is calibrated layer by layer through scenario testing, and the parameter points that deviate from the original feedback data are eliminated and the parameter range is adjusted to finally generate the adjusted strategy parameter set;

[0110] Benefit evaluation output submodule: Based on the adjusted strategy parameter set, the multivariate linear regression method is used to calculate the impact of the parameters on the market sales data. The adjusted parameter set is used as the independent variable and the market sales volume is used as the dependent variable. The relationship between the parameters and the sales data is analyzed by fitting the regression model. The correlation between the sales volume and the price change data after the parameter adjustment is gradually analyzed. The deviation value of the key data is statistically calculated by calculating the difference in sales volume before and after the adjustment. The contribution of each parameter to the sales volume change is sorted out using the grouping statistical method. A comprehensive performance data table is constructed, and the mean and standard deviation of the deviation value are statistically calculated by parameter field classification. Finally, the price is output to implement the benefit evaluation.

[0111] See also Figure 3, gradually adjust the parameter range and correct the price adjustment model, extract the key fields of price changes in the feedback data and screen them item by item, classify the parameter fields that affect the execution effect of the price adjustment strategy, calculate the parameter deviation range by segments and set the adjustment threshold, gradually expand and shrink the adjustment range for the parameters exceeding the threshold, and at the same time record the adjustment range by grouping according to the time dimension, correct the input rules of the price adjustment model by the grouped records of the parameters, layer by layer analyze the model structure after parameter correction and generate an updated price adjustment model;

[0112] Test the price fluctuation changes of the feedback data by scene through the corrected model, load the scene data in the market feedback data set and classify it according to scene characteristics, input each item of the scene data into the corrected model and record the price adjustment results output by the model, calculate the matching degree and deviation of the price fluctuation of each scene by comparing the model output data with the actual feedback price changes in the scene data, gradually summarize the price adjustment effects under different scene conditions, analyze the trend of price fluctuation changes and generate a fluctuation performance data set for per-scene testing.

[0113] The above is only a preferred embodiment of the present invention, and it is not intended to limit the present invention in other forms. Any person skilled in the art may use the technical content disclosed above to make changes or modifications into equivalent embodiments with equivalent changes and apply them to other fields. However, as long as it does not depart from the technical solution content of the present invention, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent pricing management system based on machine learning and data analysis, characterized in that, The system comprises: Initial data processing module: Based on the original market input data, data cleaning and format unification are performed to remove useless formats and characters and generate a cleaned data set; Key data screening module: Based on the cleaned data set, a Bloom filter is used to exclude known irrelevant data items, and potentially effective transaction data and pricing influencing factor data are centrally processed to generate a screened key data set; Price model building module: Based on the selected key data set, a non-dominated sorting genetic algorithm is used to analyze and determine the influence weight of each data point on the price, build a price influence model, and generate a preliminary price prediction model; Price adjustment strategy module: through the preliminary price forecasting model, the linear regression model is used to analyze the market dynamics and competitor strategies, and the model is fine-tuned to ensure that the price reflects the market trend, forming a dynamic price adjustment strategy; Price strategy simulation test module: Based on the dynamic price adjustment strategy, simulate the market scenario, evaluate the performance of the strategy under various market conditions through multi-scenario testing, confirm the market adaptability of the dynamic price adjustment strategy, and generate a market adaptability feature set; Price implementation monitoring module: monitors the performance of the dynamic price adjustment strategy, captures market feedback in real time, adjusts the dynamic price adjustment strategy to match market changes, and outputs price implementation benefit evaluation.

2. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that, The initial data processing module includes a data cleaning submodule, a format unification submodule and a data set generation submodule, wherein: Data cleaning submodule: Based on the original market input data, remove blank records and duplicate records, delete data fields that do not meet the specifications by matching field types and structures, and filter key fields and delete noise fields to generate standardized data content; Format unification submodule: based on the standardized data content, parse the field structure and adjust the field order, perform standard type conversion on the numerical field and the text field, unify the time format and unit format, and generate unified format data; Dataset generation submodule: Based on the unified format data, integrate transaction data and pricing data through field mapping and data grouping, extract required fields and remove irrelevant fields, divide the field structure, and output the cleaned data set.

3. The intelligent pricing management system based on machine learning and data analysis according to claim 1, wherein, The key data screening module includes a Bloom filtering submodule, a data screening submodule and a key data generating submodule, wherein: Bloom filter submodule: Based on the cleaned data set, known redundant fields are eliminated through field matching, irrelevant fields are marked and duplicate data is filtered, while invalid data is excluded to generate a screening data set; Data screening submodule: based on the screening data set, parsing the transaction data and pricing fields, screening pricing related fields, removing irrelevant fields in the transaction data, matching the screened fields through association analysis, and generating screening data content; Key data generation submodule: Based on the filtered data content, the filtered fields are reordered by field association grouping, transaction fields and pricing fields are integrated, a complete data table structure is established through content extraction, and the filtered key data set is output.

4. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that The price model construction module includes a data weight analysis sub-module, a model parameter construction sub-module, and a price prediction model generation sub-module, where: Data weight analysis sub-module: Based on the filtered key data set, parse the data fields and count the association between data points and price changes item by item. Calculate the impact of each group on price changes through grouped data, match the change range of each data point with the historical price records and calculate the impact value, and generate the price impact weight. Model parameter construction sub-module: Based on the price impact weight, use the non-dominated sorting genetic algorithm to screen fields according to the weight and arrange them in the order of influence, establish an association relationship for the data fields and define linear or non-linear combinations. Match the weight data points with the field correspondence rules by constructing a relationship table, summarize the key parameters and set the parameter calculation rules, and establish a price impact parameter set. Price prediction model generation sub-module: Based on the price impact parameter set, parse the parameter structure and gradually construct the model relationship. Input the filtered data set layer by layer for calculation and adjust the data output order, proofread the matching degree between the data output result and the historical price changes, and correct the parameters to generate a preliminary price prediction model.

5. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that, The non-dominated sorting genetic algorithm follows the formula: Among them: f(x) represents the comprehensive price impact parameter value, x represents the value of the current data field input, w i represents the price impact weight of the i-th field, g i (x) represents the price impact factor corresponding to the i-th field, h i represents the correlation weight between fields, t i represents the time sensitivity of the field, k i represents the change range of the field, β represents the correlation correction coefficient, α represents the time sensitivity correction coefficient, γ represents the change range correction coefficient, and n represents the total number of selected key fields.

6. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that The price adjustment strategy module includes a market dynamics analysis sub-module, a competitive strategy response sub-module, and a dynamic strategy generation sub-module, where: Market dynamics analysis sub-module: Based on the preliminary price prediction model, use the linear regression model to gradually analyze the price change records in the historical market data and extract the core data points. Group the data points according to the time dimension, analyze the differences between the data points in each group and count the change trends, integrate the grouping results and establish a market dynamics feature list, and generate a market dynamics feature set. Competitive strategy response sub-module: Based on the market dynamics feature set, extract the data points that match the competitor's price data, compare the competitor's price records and analyze the change range, classify the data points according to similarity and extract the key data points, integrate the classification results and construct a competitive response relationship table, and generate competitive strategy response features. Dynamic strategy generation sub-module: Based on the competitive strategy response features, match the price prediction model with the core data points of the market dynamics feature set, analyze the impact parameters of the core data points and gradually adjust the model, integrate the adjusted model results and verify them, and generate a dynamic price adjustment strategy.

7. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that The linear regression model follows the formula: y t = θ0 + θ1·u t + θ2·v t + θ3·w t + θ4·q t + ò t where: y t represents the market price change value corresponding to time t, θ0 represents the intercept term, u t represents the historical trading volume data at time t, v t represents the external market impact factor at time t, w t represents the competitive market price volatility at time t, q t represents the social sentiment intensity factor at time t, θ1 represents the regression coefficient of the trading volume u t θ2 represents the regression coefficient of the external market impact factor v t θ3 represents the regression coefficient of the competitive market price volatility w t θ4 represents the regression coefficient of the social sentiment intensity factor q t ∈ t represents the residual.

8. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that The price strategy simulation test module includes a market scenario construction sub-module, a multi-scenario test sub-module, and a strategy adaptability confirmation sub-module, where: Market scenario construction sub-module: Based on the dynamic price adjustment strategy, parse the historical market data and split the transaction data and pricing data. Organize the market feature data by time period and group them. Generate a scenario data group by adjusting the price fluctuation range and sales volume change trend within the time period, and match the parameters of the scenario data group to generate a market scenario data set. Multi-scenario testing sub-module: Based on the market scenario dataset, load different scenario data groups one by one and match the dynamic price adjustment strategy, gradually simulate the strategy execution process and record the data differences before and after price adjustment. By comparing the performance of key data points, statistically analyze the response degree of the scenario data group to the strategy, organize the response results and establish a comprehensive performance data table, and generate scenario test performance data; Strategy adaptability confirmation sub-module: Based on the scenario test performance data, extract the performance data of the strategy in each scenario data group, confirm the response range of the strategy to each market scenario by calculating the deviation value of the performance data, analyze the correlation between the deviation range and market characteristics and extract stability parameters, and generate a market adaptability feature set.

9. The intelligent pricing management system based on machine learning and data analysis according to claim 1, characterized in that The price implementation monitoring module includes a market feedback capture sub-module, a strategy dynamic adjustment sub-module, and a benefit evaluation output sub-module, where: Market feedback capture sub-module: Based on the dynamic price adjustment strategy, collect real-time market transaction data and competitive price data and classify them by time period, analyze the correlation between sales volume and price change data and extract key feedback data, gradually compare the deviation between the current data and historical data, and generate a market feedback dataset by merging the feedback results; Strategy dynamic adjustment sub-module: Based on the market feedback dataset, analyze the price change trend in the feedback data and match the execution parameters of the dynamic price adjustment strategy, gradually adjust the parameter range and correct the price adjustment model, and test the price fluctuation changes of the feedback data for each scenario through the corrected model, and generate an adjusted strategy parameter set; Benefit evaluation output sub-module: Based on the adjusted strategy parameter set, calculate the impact of the parameters on market sales data, analyze the correlation between the sales volume and price change data after parameter adjustment, statistically analyze the deviation value of key data and organize the comprehensive performance results, and output the price implementation benefit evaluation.

10. The intelligent pricing management system based on machine learning and data analysis according to claim 9, characterized in that, For the step of gradually adjusting the parameter range and correcting the price adjustment model, extract the key fields of the price change in the feedback data and screen them item by item, classify the parameter fields that affect the execution effect of the price adjustment strategy, calculate the parameter deviation range by segments and set the adjustment threshold, gradually expand and shrink the adjustment range for the parameters exceeding the threshold, and at the same time group and record the adjustment range by time dimension, correct the input rules of the parameters for the price adjustment model through the grouped records, layer by layer analyze the model structure after parameter correction and generate an updated price adjustment model; For the step of testing the price fluctuation changes of the feedback data for each scenario through the corrected model, load the scenario data in the market feedback dataset and classify them by scenario characteristics, input each scenario data item by item into the corrected model and record the price adjustment results output by the model. By comparing the model output data with the actual feedback price changes in the scenario data, calculate the matching degree and deviation of the price fluctuation for each scenario, gradually summarize the price adjustment effects under different scenario conditions, analyze the trend of price fluctuation changes and generate a fluctuation performance dataset for testing each scenario.

Images