Digital commercial operation system and digital commercial operation method thereof
Through the digital commercial operation system, using a variety of data analysis and machine learning technologies, the problems of insufficient market potential awareness and competitive advantages in the formulation of traditional sales strategies are solved, precise market grading and strategic management are achieved, and the sales team is enhanced. Efficiency and market adaptability.
Patent Information
- Application Number
- CN202410829127.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-25
- Publication Date
- 2024-11-01
AI Technical Summary
There are problems in the process of setting traditional sales targets and formulating action plans, such as insufficient awareness of market potential, insufficient driving force of sales teams, and insufficient competitive advantages of sales strategies, mainly due to the lack of systematic data analysis tools and artificial intelligence prediction and analysis tools. This leads to inadequate capture of market dynamics and insufficiently accurate strategy formulation.
A digital business operation system was designed, including market hierarchical division modules, market benchmarking analysis modules at the same level, organizational structure benchmarking recommendation modules, channel product portfolio execution recommendation modules, action plan formulation modules and execution review and prediction modules, using Tencent Map Technology such as data, coefficient of variation method, PCA dimensionality reduction, KMeans clustering, CRITIC, entropy value method, collaborative filtering, deep learning and gradient enhancer modeling to realize market grading, strategy formulation and execution effect monitoring.
The digital management closed loop of marketing strategies has been realized, the market hierarchy adaptability and strategy accuracy have been improved, the sales team's motivation and competitiveness have been enhanced, and the strategy can be adjusted in a timely manner to respond to market changes.
Smart Images

Figure CN118886932A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to a digital business operation system and a digital business operation method thereof, and belongs to the technical field of combining artificial intelligence with business operations. Background Art
[0002] In the process of traditional sales goal setting and action plan formulation, there are indeed some challenges caused by technical limitations and imperfect methodology. The following is a further elaboration on the shortcomings you mentioned:
[0003] 1. Lack of awareness of market potential
[0004] The traditional model that relies on manual analysis is often limited by personal experience and the breadth and depth of information obtained. Although sales staff have direct contact with customers and can obtain first-line feedback, this feedback is often partial and subjective, and it is difficult to fully reflect the potential and trends of the entire market. The lack of systematic data analysis tools and models to quantify market potential means that companies may miss untapped market opportunities or have deviations in the assessment of certain market segments. In addition, personal judgment is easily affected by factors such as emotions and biases and is not objective enough.
[0005] 2. The sales team’s driving force is not fully utilized
[0006] The sales team's incentive mechanism, training resources, and workflow design all directly affect its effectiveness. In the past, relying on manual analysis may make the sales strategy less accurate, resulting in unclear goals or insufficient motivation for the sales team to execute. At the same time, the lack of immediate data feedback and personalized performance guidance makes it difficult for sales staff to quickly adjust strategies to respond to market changes, affecting their performance and career growth. In addition, if information sharing within the team is not smooth, it will also affect the overall collaborative efficiency and the stimulation of innovative thinking.
[0007] 3. Lack of competitive advantage in sales strategy
[0008] In a rapidly changing market environment, understanding trends is the key to developing effective sales strategies. Relying on traditional analytical methods, companies may not be able to capture market dynamics, competitor movements, and subtle changes in consumer preferences in a timely manner, thus missing the best time to adjust strategies. The lack of predictive analysis tools based on big data and artificial intelligence has caused companies to rely more on historical experience and intuition when formulating sales strategies, rather than scientific predictions of future trends. This limits the ability of companies to innovate sales models, optimize product portfolios and pricing strategies, and makes it difficult for them to stand out from the competition. Summary of the invention
[0009] In order to overcome the defects of the prior art, the present invention provides a digital business operation system and method thereof. The technical solution of the present invention is:
[0010] A digital business operation system, comprising:
[0011] A market level division module, wherein the market level division module is used to realize the division of market levels;
[0012] A peer market benchmarking analysis module, which is used by business decision makers to understand the market status and explore market opportunities;
[0013] An organizational structure benchmarking recommendation module, which is used to help business decision makers understand the gap between current staffing, staff efficiency and the benchmark market;
[0014] A channel product combination execution recommendation module, which is used to understand the gap between the actual product combination sold by each channel and the optimal product combination recommended by the algorithm, and help formulate an action plan;
[0015] An action plan formulation module, which is used by business decision makers to formulate monthly goals and action strategies to achieve the monthly goals;
[0016] The execution review and prediction module is used to monitor the execution effect of business strategies through the indicator dashboard and manage the implementation of business strategies in real time.
[0017] The specific division steps of the market level division module are as follows:
[0018] 2.1. By collecting Tencent Maps GDP, population, sales points, disposable income, population density, sales point density, passenger flow basic data, as well as market consumption level, population age characteristics, and channel label data, we sorted out the correlation indicators with the beverage industry, and used business rules to divide the sorted indicators into three categories: people, finance, and market. First, we normalized the data of each indicator. Where xi is the ith value of each label, min(x) is the minimum value of each label, and max(x) is the maximum value of each label. That is, each label is normalized to the maximum and minimum values, and the ratio of the difference between each value and the minimum value of the label to the maximum and minimum values of the label is calculated to convert the data to between [0,1].
[0019] 2.2. By coefficient of variation method Calculate the weight of each label. The coefficient of variation reflects the fluctuation range of the data along the average value. The larger the coefficient of variation, the greater the data volatility and the more information it carries, so the corresponding weight will also be greater.
[0020] 2.3. Sort by scores from high to low, and select the top four labels with the highest weights in the three categories of people, finance, and market, to form 12-dimensional data as the basis for classification;
[0021] 2.4. Perform PCA dimensionality reduction on the 12-dimensional data, select features containing 90% of the information, perform KMeans clustering on the reduced-dimensional data, and divide the data into k1 categories; perform a second clustering on the 12-dimensional data of the largest class after the first clustering using the same method as the first clustering, and divide the data into k2 categories;
[0022] The third clustering method is the same as the first clustering method for the 12-dimensional data of the largest class after the second clustering, and the data is divided into k3 classes;
[0023] 2.5. Integrate three clusterings. The first clustering provides k1-1 categories, the second clustering provides k2-1 categories, and the third clustering provides k3 categories to obtain the clustering results.
[0024] 2.6. Denormalize the classification results and restore the dimensions;
[0025] 2.7. Perform weighted summation on the cluster centers of the clustering results, where the weights are calculated using the coefficient of variation method in step 2.2, and sort the summation results to obtain the categories of districts and counties;
[0026] 2.8. Output the classification results.
[0027] The specific analysis steps of the same-level market benchmarking analysis module are as follows: benchmarking indicator selection, combing the enterprise-level indicator system, analyzing the existing result indicators and execution process indicators to screen out a list of indicators, and then scoring the indicators by combining CRITIC, entropy method, and coefficient of variation method. The weight values of each indicator are obtained by using CRITIC, entropy method, and coefficient of variation method respectively, and the weight values obtained by the three methods are averaged to obtain the final weight recorded as u_i; then grouping by people, goods, and venues, retaining the top 4 indicators and corresponding weights u_i in each group, and finally selecting the benchmarking indicators.
[0028] The specific benchmarking recommendation steps of the organizational structure benchmarking recommendation module are: automatically obtaining the current organizational structure, benchmarking it with the target management structure, and intelligently recommending rectification requirements based on business management rules.
[0029] The specific recommendation steps of the market channel execution guidance recommendation module are: through historical transaction data and market share external data, collaborative filtering and deep learning algorithm recommendation models are used to deduce the optimal product combination for each channel, and combined with business management strategies, the product combination for each channel is recommended to provide a calculation basis for business goal setting.
[0030] The specific action plan formulation steps of the action plan formulation module are as follows: based on the construction of a sales target calculation model, benchmarking is achieved by various core indicators, and the gaps with the benchmark market are output respectively. According to the gaps between each indicator and the benchmark market, each indicator is raised to the same level as the benchmark market by the back-end regression algorithm model, and the operating income target value is calculated. Combined with the optimal product combination output by the channel execution recommendation module, the business target value for each category is calculated.
[0031] The specific prediction steps of the execution review and follow-up prediction module are: using a gradient boosting machine model as a prediction model, and the gradient boosting machine model includes:
[0032] Initialization model: The gradient boosting machine model is initialized to the average value of the target variable, calculated as:
[0033] Among them, F0(x) is the initial model, x is the input data, γ is the constant value of the initial prediction, L is the loss function, and y i is the i-th target value, n is the total number of samples; calculate the residual and train the weak learner: in each iteration, calculate the negative gradient of the current gradient boosting machine model, that is, the residual, and then use the residual to train a new weak prediction model. The negative gradient calculation formula is:
[0034]
[0035] Among them, r it is the residual of the i-th sample in the t-th iteration;
[0036] Update expression: According to the gradient boosting machine model F of the previous iteration t-1 (x) Update the current gradient boosting machine model F(x), expressed as: F(x) = F t-1 (x);
[0037] Update the model: Use the weak learner h obtained from residual training t (x) Update the gradient boosting machine model to improve the overall prediction performance of the gradient boosting machine model. The model update formula is:
[0038] F t (x) = F t-1 (x)+v·h t (x);
[0039] Among them, h t (x) is the weak learner trained at the tth iteration, and v is the learning rate.
[0040] A digital business operation method based on a digital business operation system comprises the following steps:
[0041] a. Divide the market into different levels;
[0042] b. Business decision makers understand the market status and explore market opportunities;
[0043] c. Understand the gap between current staffing, staff efficiency and benchmark market;
[0044] d. Understand the gap between the actual product mix sold in each channel and the optimal product mix recommended by the algorithm, and help formulate an action plan;
[0045] e. Business decision makers set monthly goals and action strategies to achieve them;
[0046] f. Monitor the effectiveness of business strategy execution through the indicator dashboard and manage the implementation of business strategies in real time.
[0047] The advantages of the present invention are: to realize the digitization of marketing strategies, to form a closed management loop for the formulation, implementation and evaluation of marketing strategies, and to solve the traditional inefficient and uncompetitive strategies. The fusion algorithm of multiple clustering is based on the creation of actual business models, which can significantly improve the adaptability and rationality of market classification at the application level. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 It is a schematic diagram of the structure of the present invention.
[0049] Figure 2 It is a flow chart of the specific division steps of the market level division module of the present invention.
[0050] Among them, 1-market level division module; 2-peer market benchmarking analysis module; 3-organizational structure benchmarking recommendation module; 4-channel product portfolio execution recommendation module; 5-action plan formulation module; 6-execution review and forecast module. DETAILED DESCRIPTION
[0051] The present invention will be further described below in conjunction with specific embodiments, and the advantages and features of the present invention will become clearer as the description proceeds. However, these embodiments are exemplary only and do not constitute any limitation to the scope of the present invention. It should be understood by those skilled in the art that the details and forms of the technical solution of the present invention may be modified or replaced without departing from the spirit and scope of the present invention, but these modifications and replacements all fall within the scope of protection of the present invention.
[0052] See also Figure 1 and Figure 2The present invention relates to a digital business operation system, comprising: a market level division module 1, the market level division module 1 is used to realize the division of market levels; a peer market benchmarking analysis module 2, the peer market benchmarking analysis module 2 is used for business decision makers to understand the market status and explore market opportunities; an organizational structure benchmarking recommendation module 3, the organizational structure benchmarking recommendation module 3 is used to help business decision makers understand the gap between current staffing, staff efficiency and benchmark market; a channel product combination execution recommendation module 4, the channel product combination execution recommendation module 4 is used to understand the gap between the actual product combination sold by each channel and the optimal product combination recommended by the algorithm, and help formulate an action plan; an action plan formulation module 5, the action plan formulation module 5 is used for business decision makers to formulate monthly goals and action strategies to achieve monthly goals; an execution review and prediction module 6, the execution review and prediction module 6 is used to monitor the execution effect of business strategies through indicator dashboards, and manage the implementation of business strategies in real time.
[0053] The specific division steps of the market level division module 1 are as follows:
[0054] 2.1. By collecting Tencent Maps GDP, population, sales points, disposable income, population density, sales point density, passenger flow basic data, as well as market consumption level, population age characteristics, and channel label data, we sorted out the correlation indicators with the beverage industry, and used business rules to divide the sorted indicators into three categories: people, finance, and market. First, we normalized the data of each indicator. Where xi is the ith value of each label, min(x) is the minimum value of each label, and max(x) is the maximum value of each label. That is, each label is normalized to the maximum and minimum values, and the ratio of the difference between each value and the minimum value of the label to the maximum and minimum values of the label is calculated. The data is converted to [0,1] to eliminate the factors that cause different results due to large differences in values between different labels. For example, ratio data is usually between [-1,1], and consumer expenditure data reaches billions. When calculating the distance between different markets, it is easy to eliminate the influence of small value labels on the results, which leads to different results.
[0055] 2.2. By coefficient of variation method Calculate the weight of each label. The coefficient of variation reflects the fluctuation range of the data along the average value. The larger the coefficient of variation, the greater the data volatility and the more information it carries, so the corresponding weight will also be greater.
[0056] 2.3. Sort by scores from large to small, select the top 4 labels in the three categories of people, finance, and market, and form 12-dimensional data as the basis for classification; use the indicators with higher scores as the input indicators of the market classification model of this category, and use the confirmed indicators to apply multiple KMeans clustering to build a market hierarchical division model, and then use the silhouette coefficient evaluation method combined with business rules to evaluate the model to ensure the credibility of the model output results. Each KMeans clustering performs PCA principal component analysis on the county data that needs to be clustered, reduces the dimension of the data while minimizing information loss, selects labels containing more than 90% of the information, and retains the market with the largest volume in the clustering results. The next clustering is performed on the market with the largest volume in the clustering results as needed to optimize the volume gap between markets at different levels.
[0057] 2.4. Perform PCA dimensionality reduction on the 12-dimensional data, select features containing 90% of the information, perform KMeans clustering on the reduced-dimensional data, and divide the data into k1 categories; perform a second clustering on the 12-dimensional data of the largest class after the first clustering using the same method as the first clustering, and divide the data into k2 categories;
[0058] The third clustering method is the same as the first clustering method for the 12-dimensional data of the largest class after the second clustering, and the data is divided into k3 classes;
[0059] 2.5. Integrate three clusterings. The first clustering provides k1-1 categories, the second clustering provides k2-1 categories, and the third clustering provides k3 categories to obtain the clustering results.
[0060] 2.6. Denormalize the classification results and restore the dimensions;
[0061] 2.7. Perform weighted summation on the cluster centers of the clustering results, where the weights are calculated using the coefficient of variation method in step 2.2, and sort the summation results to obtain the categories of districts and counties;
[0062] 2.8. Output the classification results.
[0063] The specific analysis steps of the same-level market benchmarking analysis module 2 are as follows:
[0064] Select benchmark indicators, sort out the enterprise-level indicator system, analyze the existing result indicators and execution process indicators to screen out a list of indicators, and then use CRITIC, entropy method, and coefficient of variation method to score the indicators. Finally, select benchmark indicators. The backend relies on the data middle platform base Hadoop, Spark and StarRocks distributed storage and distributed computing capabilities.
[0065] The specific benchmarking recommendation steps of the organizational structure benchmarking recommendation module 3 are: automatically obtaining the current organizational structure, benchmarking it with the target management structure, and intelligently recommending rectification requirements based on business management rules.
[0066] The specific recommendation steps of the market channel execution guidance recommendation module 4 are: through historical transaction data and market share external data, collaborative filtering and deep learning algorithm recommendation models are used to deduce the best product combination for each channel, and combined with business management strategies, the product combination for each channel is recommended to provide a calculation basis for business goal setting.
[0067] The goal of recommending the best SKU combination is to find the most suitable products for recommendation in each store based on a large number of candidate product SKUs, combine these products into orders and recommend them to the stores, so as to improve the conversion rate of the stores to the recommended orders.
[0068] The best product recommendation method based on collaborative filtering of store groups is as follows:
[0069] Get sales data for all stores;
[0070] According to the business indicators (Cube level, contract type, channel type, store sales); cluster all stores into N categories (N≤3000), called store groups and recorded as u;
[0071] For each group of stores u, the top 3 stores are selected as candidates based on their similarity between each other.
[0072] The Pearson correlation coefficient calculation formula (as shown below) is used to calculate the store group and SKU purchase similarity matrix;
[0073]
[0074] Based on the Top 3 store groups, the score prediction of the SKUs that have not been purchased / recommended by the target store group is calculated as follows:
[0075]
[0076] Sort the recommended / purchased SKUs from high to low based on the predicted scores, and take the top K items for recommendation, where the business requirement is 15≤K≤30;
[0077] For each store group, essential SKUs and recommended SKUs are eliminated, and then TopN SKUs are selected from the TopK of the group to form new TopK SKUs for recommendation to each store in the group. The business requirement 15≤K≤30 still needs to be met.
[0078] The best product recommendation method based on neural network is as follows:
[0079] Obtain sales data of all stores and pre-process them; obtain category and attribute data of all SKUs and pre-process them; obtain DRTM statistical data and store LBS data and pre-process them;
[0080] Perform embedding processing on store and product category features; perform normalization processing on numerical features; build a deep learning model and use the recall rate of the top 3 product combinations to evaluate the model. In the product combination recommended to the store, the business requirement of SKU number 15≤K≤30 still needs to be met.
[0081] The specific action plan formulation steps of the action plan formulation module 5 are: based on the construction of a sales target calculation model, benchmarking is achieved by various core indicators, and the gaps with the benchmark market are output respectively. According to the gaps between each indicator and the benchmark market, each indicator is handed over to the back-end regression algorithm model to improve each indicator to the same level as the benchmark market, and the operating income target value is calculated. Combined with the optimal product combination output by the channel execution recommendation module, the business target value for each category is calculated.
[0082] The specific prediction steps of the execution review and follow-up prediction module 6 are:
[0083] The gradient boosting machine model is used as the prediction model, and the gradient boosting machine model includes: initialization model: the gradient boosting machine model is initialized to the average value of the target variable, and the calculation formula is: Among them, F0(x) is the initial model, x is the input data, γ is the constant value of the initial prediction, L is the loss function, and y i is the i-th target value, n is the total number of samples; calculate the residual and train the weak learner: in each iteration, calculate the negative gradient of the current gradient boosting machine model, that is, the residual, and then use the residual to train a new weak prediction model. The negative gradient calculation formula is:
[0084]
[0085] Among them, r it is the residual of the i-th sample in the t-th iteration;
[0086] Update expression: According to the gradient boosting machine model F of the previous iteration t-1 (x) Update the current gradient boosting machine model F(x), expressed as: F(x) = F t-1 (x);
[0087] Update the model: Use the weak learner h obtained from residual training t (x) Update the gradient boosting machine model to improve the overall prediction performance of the gradient boosting machine model. The model update formula is:
[0088] F t (x) = F t-1 (x)+v·h t (x);
[0089] Among them, h t(x) is the weak learner trained at the tth iteration, and v is the learning rate.
[0090] The present invention also relates to a digital business operation method based on a digital business operation system, comprising the following steps:
[0091] a. Divide the market into different levels;
[0092] b. Business decision makers understand the market status and explore market opportunities;
[0093] c. Understand the gap between current staffing, staff efficiency and benchmark market;
[0094] d. Understand the gap between the actual product mix sold in each channel and the optimal product mix recommended by the algorithm, and help formulate an action plan;
[0095] e. Business decision makers set monthly goals and action strategies to achieve monthly goals; f. Monitor the effectiveness of business strategy execution through indicator dashboards and manage the implementation of business strategies in real time.
[0096] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.
Claims
1. A digital business operation system, It is characterized in that include: A market level division module, wherein the market level division module is used to realize the division of market levels; A peer market benchmarking analysis module, which is used by business decision makers to understand the market status and explore market opportunities; An organizational structure benchmarking recommendation module, which is used to help business decision makers understand the gap between current staffing, staff efficiency and the benchmark market; A channel product combination execution recommendation module, which is used to understand the gap between the actual product combination sold by each channel and the optimal product combination recommended by the algorithm, and help formulate an action plan; An action plan formulation module, which is used by business decision makers to formulate monthly goals and action strategies to achieve the monthly goals; The execution review and prediction module is used to monitor the execution effect of business strategies through the indicator dashboard and manage the implementation of business strategies in real time.
2. The digital business operation system according to claim 1, characterized in that the specific division steps of the market level division module are: 2.
1. By collecting Tencent Maps GDP, population, sales points, disposable income, population density, sales point density, passenger flow basic data, as well as market consumption level, population age characteristics, and channel label data, we sorted out the correlation indicators with the beverage industry, and used business rules to divide the sorted indicators into three categories: people, finance, and market. First, we normalized the data of each indicator. where x i is the ith value of each label, min(x) is the minimum value of each label, and max(x) is the maximum value of each label. That is, each label is normalized to the maximum and minimum values, and the ratio of the difference between each value and the minimum value of the label to the maximum and minimum values of the label is calculated to convert the data to [0,1]. 2.
2. By coefficient of variation method Calculate the weight of each label. The coefficient of variation reflects the fluctuation range of the data along the average value. The larger the coefficient of variation, the greater the data volatility and the more information it carries, so the corresponding weight will also be greater. 2.
3. Sort by scores from high to low, and select the top four labels with the highest weights in the three categories of people, finance, and market, to form 12-dimensional data as the basis for classification; 2.
4. Perform PCA dimensionality reduction on the 12-dimensional data, select features containing 90% of the information, perform KMeans clustering on the reduced-dimensional data, and divide the data into k1 categories; perform a second clustering on the 12-dimensional data of the largest class after the first clustering using the same method as the first clustering, and divide the data into k2 categories; The third clustering method is the same as the first clustering method for the 12-dimensional data of the largest class after the second clustering, and the data is divided into k3 classes; 2.
5. Integrate three clusterings. The first clustering provides k1-1 categories, the second clustering provides k2-1 categories, and the third clustering provides k3 categories to obtain the clustering results. 2.
6. Denormalize the classification results and restore the dimensions; 2.
7. Perform weighted summation on the cluster centers of the clustering results, where the weights are calculated using the coefficient of variation method in step 2.2, and sort the summation results to obtain the categories of districts and counties; 2.
8. Output the classification results.
3. The digital business operation system according to claim 1 or 2, It is characterized in that The specific analysis steps of the peer market benchmarking analysis module are as follows: Select benchmark indicators, sort out the enterprise-level indicator system, analyze the existing result indicators and execution process indicators to screen out a list of indicators, and then use CRITIC, entropy method, and coefficient of variation method to score the indicators. Use CRITIC, entropy method, and coefficient of variation method to get the weight value of each indicator, and average the weight values get by the three methods to get the final weight recorded as u_i; then group by people, goods, and sites, retain the top 4 indicators and corresponding weights u_i in each group, and finally select the benchmark indicators.
4. The digital business operation system according to claim 1 or 2, It is characterized in that The specific benchmarking recommendation steps of the organizational structure benchmarking recommendation module are: automatically obtaining the current organizational structure, benchmarking it with the target management structure, and intelligently recommending rectification requirements based on business management rules.
5. The digital business operation system according to claim 1 or 2, It is characterized in that The specific recommendation steps of the market channel execution guidance recommendation module are: through historical transaction data and market share external data, collaborative filtering and deep learning algorithm recommendation models are used to deduce the optimal product combination for each channel, and combined with business management strategies, the product combination for each channel is recommended to provide a calculation basis for business goal setting.
6. The digital business operation system according to claim 1 or 2, It is characterized in that The specific action plan formulation steps of the action plan formulation module are as follows: based on the construction of a sales target calculation model, benchmarking is achieved by various core indicators, and the gaps with the benchmark market are output respectively. According to the gaps between each indicator and the benchmark market, each indicator is raised to the same level as the benchmark market by the back-end regression algorithm model, and the operating income target value is calculated. Combined with the optimal product combination output by the channel execution recommendation module, the business target value for each category is calculated.
7. The digital business operation system according to claim 1 or 2, It is characterized in that The specific prediction steps of the execution review and follow-up prediction module are as follows: A gradient boosting machine model is used as a prediction model. Includes: Initialization model: The gradient boosting machine model is initialized to the average value of the target variable, calculated as: Among them, F 0 (x) is the initial model, x is the input data, γ is the constant value of the initial prediction, L is the loss function, y i is the i-th target value, n is the total number of samples; calculate the residual and train the weak learner: in each iteration, calculate the negative gradient of the current gradient boosting machine model, that is, the residual, and then use the residual to train a new weak prediction model. The negative gradient calculation formula is: Among them, r it is the residual of the i-th sample in the t-th iteration; Update expression: According to the gradient boosting machine model F of the previous iteration t-1 (x) Update the current gradient boosting machine model F(x), expressed as: F(x) = F t-1 (x); Update the model: Use the weak learner h obtained from residual training t (x) Update the gradient boosting machine model to improve the overall prediction performance of the gradient boosting machine model. The model update formula is: F t (x)=F t-1 (x)+v·h t (x); Among them, h t (x) is the weak learner trained at the tth iteration, and v is the learning rate.
8. A digital business operation method based on the digital business operation system of any one of claims 1 to 7, It is characterized in that The following steps are involved: a. Divide the market into different levels; b. Business decision makers understand the market status and explore market opportunities; c. Understand the gap between current staffing, staff efficiency and benchmark market; d. Understand the gap between the actual product mix sold in each channel and the optimal product mix recommended by the algorithm, and help formulate an action plan; e. Business decision makers set monthly goals and action strategies to achieve them; f. Monitor the effectiveness of business strategy execution through the indicator dashboard and manage the implementation of business strategies in real time.
Citation Information
Patent Citations
Novel retail analysis method and system based on big data
CN116664158A
Smart agricultural trade management system
CN118071382A
Digital marketing operation and management system and method for smart commercial complex
CN118071395A
Marketing optimization system
US20040093296A1
