Shop profit prediction system based on big data analysis

Abstract

The invention relates to the technical field of big data analysis, and discloses a shop profit prediction system based on big data analysis, which comprises a multi-dimensional acquisition module and an intelligent prediction module. According to the shop profit prediction system based on big data analysis, operation management data of a shop and dynamic change data of a consumer market are acquired through a multi-dimensional acquisition module and are classified to form a data set, an intelligent prediction module sets a time axis, and standardization processing is performed on an operation data set and a market data set; the intelligent prediction module analyzes the store operation state of each node, generates a single-day data set, sets a monitoring period with a fixed time length, generates a period data set, and then constructs gradient boosting tree models in sequence according to a time sequence, each model corrects and records prediction errors of a previous time node model, and the intelligent prediction precision is high. And the intelligent prediction module performs attribution analysis on the time axis, the single-day data set and the periodic data set through a large language model, and generates optimization suggestions, so that the optimization management efficiency is high.

Description

Technical Field

[0001] The present invention relates to the technical field of big data analysis, and in particular to a store profit prediction system based on big data analysis. Background Art

[0002] Store profit refers to the net income obtained by a store through the sale of goods or provision of services in a certain period of time, that is, the balance after deducting the total cost from the total income. It is an important indicator to measure the operating efficiency of a store and reflects the profitability and market competitiveness of the store. Store profit includes three parts: operating profit, net investment income and net non-operating income and expenditure. Among them, operating profit is the core part of store profit and usually accounts for the largest proportion of total profit. Accurate profit forecasting helps enterprises formulate market entry strategies, product development plans and marketing strategies, etc., and provides direction guidance for the long-term development of enterprises. By predicting profits, enterprises can allocate resources more reasonably, such as funds, manpower, and material resources, improve resource utilization efficiency, reduce costs, and enhance profitability. Profit forecasting helps enterprises identify potential risks, such as changes in market demand and rising costs, so as to formulate countermeasures in advance and reduce losses caused by uncertainty. Target profit provides a quantitative standard for the performance evaluation of enterprises, helps to motivate employees to improve their work enthusiasm, and promotes the improvement of the overall performance of enterprises. Store profit forecasting methods include cost-volume-profit analysis, relevant ratio method, and factor measurement method. The cost-volume-profit analysis method analyzes the impact of changes in one factor on other factors based on the changing relationship between cost, business volume and profit. The core is to find the break-even point. The relevant ratio law predicts the profit during the planning period based on the inherent relationship between profit and related indicators. The factor measurement method predicts the profit amount of the enterprise during the planning period based on the various factors that affect the change of profit during the planning period.

[0003] At present, when traditional store profit forecasting systems based on big data analysis process large amounts of data, they may ignore or simplify key influencing factors, and the selected forecast basis is not sufficient. In addition, traditional analysis models are relatively simple and cannot fully consider the correlation between various complex factors, resulting in low profit forecast accuracy and difficulty in adjustment and optimization according to actual conditions. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a store profit prediction system based on big data analysis, which has the advantages of high intelligent prediction accuracy and high optimization management efficiency. It solves the problem that the traditional store profit prediction system based on big data analysis has insufficient prediction basis and low profit prediction accuracy.

[0005] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a store profit prediction system based on big data analysis, comprising a multi-dimensional acquisition module and an intelligent prediction module; The multi-dimensional acquisition module is composed of a store data unit, a market data unit and a pre-processing unit. The store data unit collects business data sets through a network connection to a database, and the business data sets include store business management data. The market data unit collects market data sets through a network connection to a big data platform, and the market data sets include dynamic change data of the consumer market. The pre-processing unit collects business data sets through a network connection to a big data platform, and the market data sets include dynamic change data of the consumer market. , standardize the business data set and market data set, and transmit them to the intelligent prediction module; The intelligent prediction module consists of a business analysis unit, a fusion analysis unit and a decision management unit. The business analysis unit , analyze the store operating status of each node and generate the corresponding single-day data group , and then set a fixed-length monitoring cycle , generate the corresponding periodic data group The fusion analysis unit is based on the time axis , Single-day data group and periodic data groups , build a gradient boosted tree model , and uses a Bayesian optimizer for optimization and updating. The decision management unit uses a large language model to target the time axis , Single-day data group and periodic data groups Conduct attribution analysis and generate optimization recommendations.

[0006] Preferably, the expression of the business data set is , to Indicates the store's first day to the The business management data for the day includes total sales, number of customers, operating costs and inventory quantity. Indicates the time point when store operation and management data is obtained.

[0007] Preferably, the expression of the market data set is , to Indicates that the consumer market has changed from the first to the Dynamic change data, including advertising and marketing activity records and consumer questionnaire records, Indicates the time point for obtaining data on dynamic changes in the consumer market.

[0008] Preferably, the standardization process is as follows: Preferably, the standardization process is as follows: The operating data is collected in chronological order, from the first day to the The operating and management data of the day are inserted into the timeline in sequence , generate the corresponding time nodes, and then collect the market data in chronological order, from the first to the second The dynamically changing data is inserted into the timeline in sequence , generate corresponding time nodes, among which the dynamic change data of the consumer market and the store operation and management data at the same time point correspond to the same time node.

[0009] Preferably, the single-day data set The calculation process is as follows: According to the timeline , extract time nodes management data and time nodes The total sales of , the time node The number of customers is marked as , the time node The operating costs are marked as , the time node The inventory quantity of ; In the formula, Indicates time node The average order value, Indicates time node Gross profit, Represents the ratio of gross profit to total sales, which is the time node Gross profit margin, Indicates time node The inventory turnover rate, Indicates time node Single-day data set.

[0010] Preferably, the periodic data set The calculation process is as follows: According to the time axis , statistical monitoring cycle During the period, the management data of all time nodes will be monitored. During this period, the sum of total sales at all time nodes is marked as , the monitoring cycle The number of time nodes during the period is marked as , the same monitoring cycle as last year During this period, the sum of total sales at all time nodes is marked as , the previous monitoring cycle During this period, the sum of total sales at all time nodes is marked as ;

[0011] In the formula, Indicates monitoring cycle Average sales during the period, Indicates the monitoring period last year and the current monitoring period The change in the sum of total sales between Indicates the monitoring period last year and the current monitoring period The year-on-year growth rate of the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The change in the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The month-on-month growth rate of the sum of total sales.

[0012] Preferably, the gradient boosting tree model The construction process is as follows: Gradient Boosting Tree Model It includes several decision nodes and leaf nodes, where each decision node contains a test threshold for filtering data that exceeds the test threshold, and each leaf node contains the final prediction result. The test thresholds include the customer unit price threshold, gross profit margin threshold, inventory turnover threshold, average sales threshold, year-on-year growth rate threshold, and month-on-month growth rate threshold. The prediction results include the customer unit price prediction value, gross profit margin prediction value, inventory turnover prediction value, average sales prediction value, year-on-year growth rate prediction value, and month-on-month growth rate prediction value. The test results all use the average value. According to the timeline , Single-day data group and periodic data groups , build the gradient boosting tree model in chronological order , each time node corresponds to a gradient boosting tree model , each gradient boosted tree model Both will correct and record the gradient boosting tree model of the previous time node prediction error.

[0013] Preferably, the optimization and updating process is as follows: In each gradient boosted tree model In the construction process, the Bayesian optimizer is used to arbitrarily select decision nodes and leaf nodes for evaluation. If the point gradient boosting tree model of two adjacent time nodes When the test results of all leaf nodes are achieved, the Bayesian optimizer stops evaluating and marks the gradient boosting tree model with the optimal solution. .

[0014] Preferably, the attribution analysis process is as follows: According to the timeline , extract time nodes Dynamic changes in the consumer market data, using large language models to analyze time nodes Advertising and marketing activity records are used to form marketing hot words, which include holiday names, product names, and discount levels. Large language models are used to analyze time nodes. Consumer questionnaire records are compiled into demand keywords, including product functions, cost-effectiveness and after-sales service, and then a large language model is used to analyze the time nodes. Marketing hot words, demand keywords and single-day data groups at the same time point , Periodic Data Group of relevance.

[0015] Preferably, if the number of repetitions of a single marketing hot word is consistent with the single-day data set All the values ​​in are proportional, indicating that the marketing hot words are highly correlated with the store's operating status. It is recommended that the store plan its business methods based on the marketing hot words. If the number of repetitions of a single demand keyword is proportional to the periodic data group Any value in the is inversely proportional, indicating that the demand keywords are highly correlated with the store's operating status. It is recommended that the store optimize its service content based on the demand keywords.

[0016] Compared with the prior art, the present invention provides a store profit prediction system based on big data analysis, which has the following beneficial effects: 1. The present invention connects the database and the big data platform through a multi-dimensional acquisition module to obtain the store's business management data and the dynamic change data of the consumer market, and classifies them into business data sets and market data sets, and then sets a unified time axis , standardize the business data set and market data set, and transmit them to the intelligent prediction module, through a unified time axis Align heterogeneous data to ensure data comparability in the time dimension, solve the problem of inconsistent time of multi-source data, and lay the foundation for subsequent modeling. The intelligent prediction module uses the time axis to , analyze the store operating status of each node and generate the corresponding single-day data group , accurately capture daily operating status, help quickly identify abnormal daily fluctuations, and then set a fixed-duration monitoring cycle , generate the corresponding periodic data group , providing data support for periodic business strategy adjustments, the intelligent prediction module , Single-day data group and periodic data groups , build the gradient boosting tree model in chronological order , each time node corresponds to a gradient boosting tree model , each gradient boosted tree model Both will correct and record the gradient boosting tree model of the previous time node The prediction error of each time node is corrected based on the previous error, forming a closed-loop optimization process. In the construction process, the Bayesian optimizer is used to arbitrarily select decision nodes and leaf nodes for evaluation. If the point gradient boosting tree model of two adjacent time nodes When the test results of all leaf nodes are achieved, the Bayesian optimizer stops evaluating and marks the gradient boosting tree model with the optimal solution. , dynamically adjust model parameters, quickly converge to the optimal solution, avoid overfitting problems, enhance the model's generalization ability for new data, and achieve high intelligent prediction accuracy.

[0017] 2. The present invention uses an intelligent prediction module to use a large language model to target the time axis , Single-day data group and periodic data groups Conduct attribution analysis based on the timeline , extract time nodes Dynamic changes in the consumer market data, using large language models to analyze time nodes Advertising and marketing activity records are used to form marketing hot words, which include holiday names, product names, and discount levels. Large language models are used to analyze time nodes. Consumer questionnaire records are compiled into demand keywords, including product functions, cost-effectiveness and after-sales service, and then a large language model is used to analyze the time nodes. Marketing hot words, demand keywords and single-day data groups at the same time point , Periodic Data Group If the number of repetitions of a single marketing buzzword is related to the single-day data set All the values ​​in are proportional, indicating that the marketing hot words are highly correlated with the store's operating status. It is recommended that the store plan its business methods based on the marketing hot words. If the number of repetitions of a single demand keyword is proportional to the periodic data group Any value in is inversely proportional, indicating that the demand keywords are highly correlated with the store operating status. It is recommended that stores optimize service content based on demand keywords, reduce subjective decision-making bias, improve long-term business stability, and help stores optimize inventory management and cost control, thereby optimizing management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the system of the present invention. DETAILED DESCRIPTION

[0019] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] Because the traditional store profit prediction system based on big data analysis may ignore or simplify the key influencing factors when processing a large amount of data, the prediction basis screened out is not sufficient, and the traditional analysis model is relatively simple and cannot fully consider the correlation between various complex factors, resulting in low profit prediction accuracy and difficulty in adjustment and optimization according to actual conditions. Therefore, a store profit prediction system based on big data analysis is provided. Please refer to Figure 1 , a store profit prediction system based on big data analysis, including a multi-dimensional acquisition module and an intelligent prediction module; The multi-dimensional acquisition module consists of a store data unit, a market data unit, and a preprocessing unit. The store data unit collects the business data set through a network connection database. The business data set includes the business management data of the store. The expression of the business data set is , to Indicates the store's first day to the The operating and management data for the day include total sales, number of customers, operating costs and inventory quantity. Indicates the time point for obtaining store operation and management data; The market data unit collects market data sets through the network connection to the big data platform. The market data sets include dynamic change data of the consumer market. The expression of the market data set is , to Indicates that the consumer market has changed from the first to the Dynamic change data, including advertising and marketing activity records and consumer questionnaire records, Indicates the time point for obtaining data on dynamic changes in the consumer market; The preprocessing unit sets a unified time axis , standardize the business data set and market data set, and transmit them to the intelligent prediction module, through a unified time axis Align heterogeneous data to ensure data comparability in the time dimension, solve the problem of time inconsistency of multi-source data, and lay the foundation for subsequent modeling; The standardized processing process is as follows: the business data is collected in chronological order, from the first day to the The operating and management data of the day are inserted into the timeline in sequence , generate the corresponding time nodes, and then collect the market data in chronological order, from the first to the second The dynamically changing data is inserted into the timeline in sequence , generating corresponding time nodes, where the consumer market dynamic change data and store operation and management data at the same time point correspond to the same time node; The intelligent prediction module consists of a business analysis unit, a fusion analysis unit, and a decision management unit. The business analysis unit is based on the timeline. , analyze the store operating status of each node and generate the corresponding single-day data group , and then set a fixed-length monitoring cycle , generate the corresponding periodic data group ; Single-day data set The calculation process is as follows: According to the time axis , extract time nodes management data and time nodes The total sales of , the time node The number of customers is marked as , the time node The operating costs are marked as , the time node The inventory quantity of ; In the formula, Indicates time node The average order value, Indicates time node Gross profit, Represents the ratio of gross profit to total sales, which is the time node Gross profit margin, Indicates time node The inventory turnover rate, Indicates time node The single-day data set accurately captures the daily operating status and helps quickly identify abnormal single-day fluctuations; Periodic Data Group The calculation process is as follows: According to the timeline , statistical monitoring cycle During the period, the management data of all time nodes will be monitored. During this period, the sum of total sales at all time nodes is marked as , the monitoring cycle The number of time nodes during the period is marked as , the same monitoring cycle as last year During this period, the sum of total sales at all time nodes is marked as , the previous monitoring cycle During this period, the sum of total sales at all time nodes is marked as ; In the formula, Indicates monitoring cycle Average sales during the period, Indicates the monitoring period last year and the current monitoring period The change in the sum of total sales between Indicates the monitoring period last year and the current monitoring period The year-on-year growth rate of the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The change in the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The month-on-month growth rate of the sum of total sales between 2017 and 2018 provides data support for periodic business strategy adjustments; the fusion analysis unit is based on the time axis , Single-day data group and periodic data groups , build a gradient boosted tree model , and use the Bayesian optimizer for optimization updates; Gradient Boosted Tree Model The build process is as follows: Gradient Boosted Tree Model It includes several decision nodes and leaf nodes, where each decision node contains a test threshold for filtering data that exceeds the test threshold. Each leaf node contains the final prediction result. The test thresholds include the customer unit price threshold, gross profit margin threshold, inventory turnover threshold, average sales threshold, year-on-year growth rate threshold, and month-on-month growth rate threshold. The prediction results include the customer unit price prediction value, gross profit margin prediction value, inventory turnover prediction value, average sales prediction value, year-on-year growth rate prediction value, and month-on-month growth rate prediction value. The test results all use the average value to effectively capture complex data features and improve prediction accuracy. According to the timeline , Single-day data group and periodic data groups , build the gradient boosting tree model in chronological order , each time node corresponds to a gradient boosting tree model , each gradient boosted tree model Both will correct and record the gradient boosting tree model of the previous time node The model at each time node is corrected based on the previous error, forming a closed-loop optimization process to ensure that the system continues to adapt to the market change; The optimization update process is as follows: In each gradient boosted tree model In the construction process, the Bayesian optimizer is used to arbitrarily select decision nodes and leaf nodes for evaluation. If the point gradient boosting tree model of two adjacent time nodes When the test results of all leaf nodes are achieved, the Bayesian optimizer stops evaluating and marks the gradient boosting tree model with the optimal solution. , dynamically adjust model parameters, quickly converge to the optimal solution, avoid overfitting problems, enhance the model's generalization ability for new data, and achieve high intelligent prediction accuracy; The decision management unit uses a large language model to target the time axis. , Single-day data group and periodic data groups Conduct attribution analysis based on the timeline , extract time nodes Dynamic changes in the consumer market data, using large language models to analyze time nodes Advertising and marketing activity records are used to form marketing hot words, which include holiday names, product names, and discount levels. Large language models are used to analyze time nodes. Consumer questionnaire records are compiled into demand keywords, including product functions, cost-effectiveness and after-sales service, and then a large language model is used to analyze the time nodes. Marketing hot words, demand keywords and single-day data groups at the same time point , Periodic Data Group If the number of repetitions of a single marketing buzzword is related to the single-day data set All the values ​​in are proportional, indicating that the marketing hot words are highly correlated with the store's operating status. It is recommended that the store plan its business methods based on the marketing hot words. If the number of repetitions of a single demand keyword is proportional to the periodic data group Any value in is inversely proportional, indicating that the demand keywords are highly correlated with the store operating status. It is recommended that stores optimize service content based on demand keywords, reduce subjective decision-making bias, improve long-term business stability, and help stores optimize inventory management and cost control, thereby optimizing management efficiency.

[0021] Example 1: In this experiment, a store with a total daily sales of 10,000 yuan was selected as the experimental object. According to statistics, the number of customers in a single day was 100, the operating cost was 6,000 yuan, and the inventory was 500 pieces. The single-day data set of this store is The calculation process is as follows: In the formula, Indicates the average daily customer price. represents the daily gross profit, It represents the ratio of gross profit to total sales, that is, the daily gross profit margin. It represents the daily inventory turnover rate; Gradient Boosted Tree Model In the decision node, the customer unit price threshold is set to 100-200 yuan / person, the gross profit margin threshold is set to 0.3-0.5, and the inventory turnover rate threshold is set to 10-30 times / item. After judgment, the store's single-day data set The daily average order value of 100 is included in the average order value threshold, and the daily gross profit margin is Included in the gross profit margin threshold, inventory turnover threshold included in the inventory turnover threshold, current gradient boosting tree model The prediction results in the leaf nodes (customer unit price prediction value, gross profit margin prediction value, and inventory turnover rate prediction value) are all correct, and the gradient boosting tree model is updated in one day When the customer unit price threshold, gross profit margin threshold and inventory turnover rate threshold are used, the decision nodes corresponding to them do not need to be corrected. Example 2: In this experiment, a store with a total sales of 1 million yuan in February is selected as the experimental object. The store has 10 operating days in February. Therefore, the time axis The corresponding February time nodes are 10. According to statistics, the total sales in February last year was 1.8 million yuan, and the total sales in the previous month (January) was 900,000 yuan. The calculation process is as follows:

[0022] In the formula, represents the average sales in February, It represents the change in the total sales between February last year and February this year. It represents the year-on-year growth rate of the total sales between February last year and the current February. Indicates the change in the total sales between the previous month (January) and the current February. It indicates the month-on-month growth rate of the total sales between the previous month (January) and the current February; Using a large language model to analyze the demand keyword in February is "low cost performance", and the number of repetitions is consistent with the store cycle data group The year-on-year growth rate is inversely proportional, indicating that the demand keyword "low cost-effectiveness" is highly correlated with the current store operating status. It is recommended that stores optimize their service content based on the demand keyword "low cost-effectiveness".

[0023] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A store profit prediction system based on big data analysis, characterized by: Including multi-dimensional acquisition module and intelligent prediction module; The multi-dimensional acquisition module is composed of a store data unit, a market data unit and a pre-processing unit. The store data unit collects business data sets through a network connection to a database, and the business data sets include store business management data. The market data unit collects market data sets through a network connection to a big data platform, and the market data sets include dynamic change data of the consumer market. The pre-processing unit collects business data sets through a network connection to a big data platform, and the market data sets include dynamic change data of the consumer market. , standardize the business data set and market data set, and transmit them to the intelligent prediction module; The intelligent prediction module consists of a business analysis unit, a fusion analysis unit and a decision management unit. The business analysis unit , analyze the store operating status of each node and generate the corresponding single-day data group , and then set a fixed-length monitoring cycle , generate the corresponding periodic data group The fusion analysis unit is based on the time axis , Single-day data group and periodic data groups , build a gradient boosted tree model , and uses a Bayesian optimizer for optimization and updating. The decision management unit uses a large language model to target the time axis , Single-day data group and periodic data groups Conduct attribution analysis and generate optimization recommendations.

2. The store profit prediction system based on big data analysis according to claim 1 is characterized by: The expression of the business data set is , to Indicates the store's first day to the The operating and management data for the day include total sales, number of customers, operating costs and inventory quantity. Indicates the time point when store operation and management data is obtained.

3. The store profit prediction system based on big data analysis according to claim 2 is characterized in that: The expression of the market data set is , to Indicates that the consumer market has changed from the first to the Dynamic change data, including advertising and marketing activity records and consumer questionnaire records, Indicates the time point for obtaining data on dynamic changes in the consumer market.

4. The store profit prediction system based on big data analysis according to claim 3 is characterized by: The standardized processing flow in the preprocessing unit is as follows: The operating data is collected in chronological order, from the first day to the The operating and management data of the day are inserted into the timeline in sequence , generate the corresponding time nodes, and then collect the market data in chronological order, from the first to the second The dynamically changing data is inserted into the timeline in sequence , generate corresponding time nodes, among which the dynamic change data of the consumer market and the store operation and management data at the same time point correspond to the same time node.

5. The store profit prediction system based on big data analysis according to claim 4 is characterized in that: The single-day data set The calculation process is as follows: According to the timeline , extract time nodes management data and time nodes The total sales of , the time node The number of customers is marked as , the time node The operating costs are marked as , the time node The inventory quantity is marked as ; In the formula, Indicates time node The average order value, Indicates time node of gross profit, Represents the ratio of gross profit to total sales, which is the time node Gross profit margin, Indicates time node The inventory turnover rate, Indicates time node Single-day data set.

6. The store profit prediction system based on big data analysis according to claim 5 is characterized by: The periodic data set The calculation process is as follows: According to the time axis , statistical monitoring cycle During the period, the operation and management data of all time nodes will be monitored During this period, the sum of total sales at all time nodes is marked as , the monitoring cycle The number of time nodes during the period is marked as , the same monitoring cycle as last year During this period, the sum of total sales at all time nodes is marked as , the previous monitoring cycle During this period, the sum of total sales at all time nodes is marked as ; In the formula, Indicates monitoring cycle Average sales during the period, Indicates the monitoring period last year and the current monitoring period The change in the sum of total sales between Indicates the monitoring period last year and the current monitoring period The year-on-year growth rate of the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The change in the sum of total sales between Indicates the last monitoring cycle and the current monitoring period The month-on-month growth rate of the sum of total sales.

7. The store profit prediction system based on big data analysis according to claim 6 is characterized by: The gradient boosted tree model The construction process is as follows: Gradient Boosting Tree Model It includes several decision nodes and leaf nodes, where each decision node contains a test threshold for filtering data that exceeds the test threshold, and each leaf node contains the final prediction result. The test thresholds include the customer unit price threshold, gross profit margin threshold, inventory turnover threshold, average sales threshold, year-on-year growth rate threshold, and month-on-month growth rate threshold. The prediction results include the customer unit price prediction value, gross profit margin prediction value, inventory turnover prediction value, average sales prediction value, year-on-year growth rate prediction value, and month-on-month growth rate prediction value. The test results all use the average value. According to the timeline , Single-day data group and periodic data groups , build the gradient boosting tree model in chronological order , each time node corresponds to a gradient boosting tree model , each gradient boosted tree model Both will correct and record the gradient boosting tree model of the previous time node prediction error.

8. The store profit prediction system based on big data analysis according to claim 7 is characterized by: The optimization update process is as follows: In the construction process, the Bayesian optimizer is used to arbitrarily select decision nodes and leaf nodes for evaluation. If the point gradient boosting tree model of two adjacent time nodes When the test results of all leaf nodes are achieved, the Bayesian optimizer stops evaluating and marks the gradient boosting tree model with the optimal solution. .

9. The store profit prediction system based on big data analysis according to claim 8 is characterized by: The attribution analysis process is as follows: According to the timeline , extract time nodes Dynamic changes in the consumer market data, using large language models to analyze time nodes Advertising and marketing activity records are used to form marketing hot words, which include holiday names, product names, and discount levels. Large language models are used to analyze time nodes. Consumer questionnaire records are compiled into demand keywords, including product functions, cost-effectiveness and after-sales service, and then a large language model is used to analyze the time nodes. Marketing hot words, demand keywords and single-day data groups at the same time point , Periodic Data Group of relevance.

10. The store profit prediction system based on big data analysis according to claim 9 is characterized in that: If the number of repetitions of a single marketing buzzword is consistent with the single-day data set All the values ​​in are proportional, indicating that the marketing hot words are highly correlated with the store's operating status. It is recommended that the store plan its business methods based on the marketing hot words. If the number of repetitions of a single demand keyword is proportional to the periodic data group Any value in the is inversely proportional, indicating that the demand keywords are highly correlated with the store's operating status. It is recommended that the store optimize its service content based on the demand keywords.

Images