Integrated new retail smart store POS system

Abstract

The invention relates to an integrated new retail smart store POS system, and belongs to the technical field of POS systems, the integrated new retail smart store POS system comprises an integrated POS system, a data modeling analysis system and an intelligent decision-making system, store retail data stored in the integrated POS system is utilized to construct a classification and regression tree model to generate a decision-making tree so as to analyze relevance among retail data variables, determining preferences of customers to commodities, and determining an importance sequence of explanatory variables by calculating weight differences among different explanatory variables in a modeling process; and pruning the decision tree by using the relationship between the surface error rate, the gain rate and the Gini coefficient. The method solves the problem that in the prior art, the modeling efficiency is low due to the fact that a model needs to be repeatedly constructed for multiple times to determine the importance sequence of explanatory variables; the problem of low precision caused by overfitting of the decision tree is effectively solved, meanwhile, the interpretation variables which are irrelevant or small in relevance are removed, and the interpretation ability of the decision tree is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of POS systems, and specifically relates to an integrated new retail smart store POS system. Background Art

[0002] With the rapid development of technology and the changes in consumer demand, the traditional retail industry is facing huge transformation challenges. Traditional POS systems can no longer meet retailers' needs for efficient, intelligent and integrated store management. Therefore, the integrated new retail smart store POS system came into being. It can integrate multiple functional modules such as sales management, inventory management, and member management to improve store operation efficiency, optimize customer experience, and help retailers achieve intelligent and efficient store management and enhance competitiveness.

[0003] POS systems can record and store a large amount of retail data such as sales volume, sales amount, and customer flow. Statistics and analysis of this data can help retailers gain a deeper understanding of sales, customer needs, and market trends, thereby formulating more effective marketing strategies. However, most POS systems focus mainly on descriptive analysis of retail data, but lack in-depth statistical analysis and model building capabilities, which makes it impossible to accurately analyze the correlation between retail data variables, making it difficult to provide support for marketing strategies. Summary of the invention

[0004] The present invention provides an integrated new retail smart store POS system, which solves the problem that traditional POS systems lack in-depth statistical analysis and model building capabilities and are difficult to analyze the correlation between retail data variables, and realizes intelligent and efficient management of smart stores.

[0005] To achieve the above object, the present invention provides the following technical solutions: An integrated new retail smart store POS system, characterized by: comprising an integrated POS system, a data modeling and analysis system, and an intelligent decision-making system, wherein: The integrated POS system is used to manage the retail business of the new retail smart store and store retail data, wherein the retail data includes customer data and product data. The customer data includes the customer's age, occupation and gender; the product data includes the category, cost, sales price and sales volume of the products purchased by the customer; Data modeling and analysis system, used to perform statistical analysis and modeling on retail data, and analyze the correlation between retail data variables, including customer preference analysis module and commodity pricing analysis module; The intelligent decision-making system is used to provide intelligent decision-making suggestions for user marketing work based on the output results of the data modeling and analysis system, including customer promotion module and product pricing module.

[0006] Furthermore, in the data modeling and analysis system, the customer preference analysis module is used to analyze the customer's consumption behavior and provide suggestions for the customer promotion strategy of the product. The method adopted is to analyze the influence of the customer's age, occupation and gender on the customer's purchase decision by building a classification and regression tree model, including the following steps: A1. Determine the model structure: determine the explanatory variables and response variables for building the classification and regression tree models, where the customer's age, occupation and gender are the explanatory variables and the category of goods purchased by the customer is the response variable; A2, generate a decision tree: sort the importance of the explanatory variables, perform binary recursive decomposition of the response variable according to the order of importance of the explanatory variables, until the Gini coefficient after decomposition does not change, and generate a decision tree; A3, decision tree pruning: adjust the model parameters, and prune the decision tree based on the relationship between the Gini coefficient and the model parameters to obtain the optimal subtree; the model parameters are the surface error rate gain rate.

[0007] Furthermore, in step A2, the importance of the explanatory variables is sorted, and the response variable is binary recursively decomposed according to the importance order of the explanatory variables until the decomposed Gini coefficient does not change, and a decision tree is generated, including the following steps: A21, calculate the weight difference of different explanatory variables: calculate the weight of customers of different ages, occupations and genders in a commodity category relative to all customers in the commodity category, and calculate the variance of the weights of different explanatory variables as the weight difference of different explanatory variables; A22, ranking of explanatory variable importance: ranking the importance of explanatory variables according to the difference in weights of different explanatory variables, where the greater the difference in weights of the explanatory variables, the greater the importance; A23, perform binary recursive decomposition on the response variable to generate a decision tree: according to the order of importance of the explanatory variables, select the explanatory variable with the greatest importance as the root node of the decision tree, decompose the response variable data set into two subsets, and the root node is the point of the first decomposition of the decision tree; then repeat the methods of steps A21-A22 for the samples of the two subsets to identify the order of importance of the explanatory variables of the corresponding subsets, select the explanatory variable with the greatest importance as the node again, and decompose each subset sample into two subsets; calculate the Gini coefficient in the process of binary recursive decomposition of the decision tree, and when the Gini coefficient no longer changes, no longer decompose; the calculation formula of the Gini coefficient is: , Where Gini is the Gini coefficient, which is used to measure the probability that a randomly selected sample is misclassified as other categories. The smaller the Gini is, the higher the purity of the node is; K is the number of categories for classifying the data set, K=2; P nis the ratio of the sample size of the nth category to the total sample size.

[0008] Furthermore, in step A3, pruning the generated decision tree comprises the following steps: A31, according to the order of binary recursive decomposition of the decision tree, sequentially calculate the subtree Gini coefficient and the surface error rate gain rate after each decomposition, and the surface error rate gain rate is calculated by the formula: , Where α is the surface error rate gain rate. The smaller α is, the more times the decision tree is decomposed and the more nodes the tree has. R(t) is the error cost of a tree with t as a single node. The tree is pruned at this node. R(t) = r(t)×p(t), r(t) is the error rate of node t, and p(t) is the proportion of the number of samples on node t to the total number of samples. T t is the number of leaf nodes contained in the subtree, R(T t ) represents the error cost of the subtree T with t as the root node, ,in m is the number of nodes, i =1,2,……, m ; A32, compare the surface error rate gain rate and Gini coefficient of different decomposition times of decision trees. When the surface error rate gain rate decreases and the reduction of the Gini coefficient becomes smaller, the subtree with the corresponding decomposition number is the best subtree after pruning the decision tree.

[0009] Furthermore, in the data modeling and analysis system, the commodity pricing analysis module is connected to the customer preference analysis module to analyze the relationship between the commodity price and sales volume preferred by the customer and provide suggestions for the formulation of commodity prices. The method adopted is to determine the relationship between commodity price and sales volume through a polynomial fitting model, including the following steps: B1, determine the model structure: determine the independent variables and dependent variables of the polynomial fitting model according to the response variables corresponding to the terminal nodes of the decision tree in step A2; B2, adjust the parameters of the model to obtain a model with good fitting performance; B3, determine the optimal product pricing: Based on the determined polynomial fitting relationship and combined with the product cost, calculate the optimal product price with the highest profit.

[0010] Furthermore, in step B1, the model structure is a polynomial fitting model constructed with commodity price as the independent variable and sales volume as the dependent variable, and its formula is: , In the formula, y For sales volume;x is the commodity price; a0, a1, a2, ..., a n is the fitting parameter; n is the order of the model.

[0011] Furthermore, in step B2, the adjustment of model parameters is to determine the appropriate order of the model using the AIC criterion, and determine the fitting parameters that minimize the sum of squares of the model residuals.

[0012] Furthermore, in step B3, the optimal commodity price is the commodity price with the highest profit, and its calculation formula is: , In the formula, z a is the profit; b is the cost of goods; y For sales volume; x For commodity prices, z When the value is extreme x Solve it and you can get the best price for the product.

[0013] Furthermore, the intelligent decision-making system includes a customer promotion module and a commodity pricing module, wherein: The customer promotion module is connected to the customer preference analysis module in the data analysis and modeling system. It is used to input the collected customer age, occupation and gender data into the already constructed classification and regression tree model, output the customer's preferred products through the decision tree algorithm, and promote the products online or offline accordingly based on the output results; The commodity pricing module is connected to the commodity pricing analysis module in the data analysis and modeling system, and is used to output the results of the polynomial fitting model determined by commodity pricing and sales volume, combined with the calculation formula of the optimal commodity price, to determine the optimal commodity price when the profit is highest.

[0014] The beneficial effects of the present invention are: (1) Using the store retail data stored in the integrated POS system, a classification and regression tree model is constructed to generate a decision tree to analyze the correlation between retail data variables and determine customer preferences for products. During the modeling process, the weight differences between different explanatory variables are calculated to determine the importance order of the explanatory variables, thereby solving the problem of low modeling efficiency caused by the need to repeatedly construct models to determine the importance order of the explanatory variables in the prior art. (2) The relationship between the surface error rate gain rate and the Gini coefficient is used to prune the decision tree, which effectively solves the problem of low precision caused by overfitting of the decision tree. In the process, irrelevant or less relevant explanatory variables are eliminated, thereby improving the explanatory power of the decision tree. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0016] Figure 1 This is a structural block diagram of the integrated new retail smart store POS system in the present invention; Figure 2 A classification and regression tree model image provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0018] See also Figure 1 The present invention provides an integrated new retail smart store POS system, which is characterized by: comprising an integrated POS system, a data modeling and analysis system and an intelligent decision-making system, wherein: The integrated POS system is used to manage the retail business of the new retail smart store and store retail data, wherein the retail data includes customer data and product data. The customer data includes the customer's age, occupation and gender; the product data includes the category, cost, sales price and sales volume of the products purchased by the customer; Data modeling and analysis system, used to perform statistical analysis and modeling on retail data, and analyze the correlation between retail data variables, including customer preference analysis module and commodity pricing analysis module; The intelligent decision-making system is used to provide intelligent decision-making suggestions for user marketing work based on the output results of the data modeling and analysis system, including customer promotion module and product pricing module.

[0019] In this embodiment, the collected data should have a sufficient sample size and cover a variety of retail scenarios to make the data samples more representative and statistically significant. Secondly, the collected data samples need to be preprocessed before modeling, including data outlier processing, missing value processing, duplicate sample elimination and other data cleaning methods.

[0020] See also Figure 2 Furthermore, in the data modeling and analysis system, the customer preference analysis module is used to analyze the customer's consumption behavior and provide suggestions for the customer promotion strategy of the product. The method adopted is to analyze the influence of customer age, occupation and gender on customer purchase decision by building a classification and regression tree model, including the following steps: A1. Determine the model structure: determine the explanatory variables and response variables for building the classification and regression tree models, where the customer's age, occupation and gender are the explanatory variables and the category of goods purchased by the customer is the response variable; A2, generate a decision tree: sort the importance of the explanatory variables, perform binary recursive decomposition of the response variable according to the order of importance of the explanatory variables, until the Gini coefficient after decomposition does not change, and generate a decision tree; A3, decision tree pruning: adjust the model parameters, and prune the decision tree based on the relationship between the Gini coefficient and the model parameters to obtain the optimal subtree; the model parameters are the surface error rate gain rate.

[0021] Furthermore, in step A2, the importance of the explanatory variables is sorted, and the response variable is binary recursively decomposed according to the importance order of the explanatory variables until the decomposed Gini coefficient does not change, and a decision tree is generated, including the following steps: A21, calculate the weight difference of different explanatory variables: calculate the weight of customers of different ages, occupations and genders in a commodity category relative to all customers in the commodity category, and calculate the variance of the weights of different explanatory variables as the weight difference of different explanatory variables; A22, ranking of explanatory variable importance: ranking the importance of explanatory variables according to the difference in weights of different explanatory variables, where the greater the difference in weights of the explanatory variables, the greater the importance; A23, perform binary recursive decomposition on the response variable to generate a decision tree: according to the order of importance of the explanatory variables, select the explanatory variable with the greatest importance as the root node of the decision tree, decompose the response variable data set into two subsets, and the root node is the point of the first decomposition of the decision tree; then repeat the methods of steps A21-A22 for the samples of the two subsets to identify the order of importance of the explanatory variables of the corresponding subsets, select the explanatory variable with the greatest importance as the node again, and decompose each subset sample into two subsets; calculate the Gini coefficient in the process of binary recursive decomposition of the decision tree, and when the Gini coefficient no longer changes, no longer decompose; the calculation formula of the Gini coefficient is: , Where Gini is the Gini coefficient, which is used to measure the probability that a randomly selected sample is misclassified as other categories. The smaller the Gini is, the higher the purity of the node is; K is the number of categories for classifying the data set, K=2; P n is the ratio of the sample size of the nth category to the total sample size.

[0022] Furthermore, in step A3, pruning the generated decision tree comprises the following steps: A31, according to the order of binary recursive decomposition of the decision tree, sequentially calculate the subtree Gini coefficient and the surface error rate gain rate after each decomposition, and the surface error rate gain rate is calculated by the formula: , Where α is the surface error rate gain rate. The smaller α is, the more times the decision tree is decomposed and the more nodes the tree has. R(t) is the error cost of a tree with t as a single node. The tree is pruned at this node. R(t) = r(t)×p(t), r(t) is the error rate of node t, and p(t) is the proportion of the number of samples on node t to the total number of samples. T t is the number of leaf nodes contained in the subtree, R(T t ) represents the error cost of the subtree T with t as the root node, ,in m is the number of nodes, i =1,2,……, m ; A32, compare the surface error rate gain rate and Gini coefficient of different decomposition times of decision trees. When the surface error rate gain rate decreases and the reduction of the Gini coefficient becomes smaller, the subtree with the corresponding decomposition number is the best subtree after pruning the decision tree.

[0023] In the prior art, most of them are to obtain a model with high explanatory power by repeatedly sorting the explanatory variables. This process is too cumbersome, and when there are many explanatory variables, huge calculations are required, which greatly reduces the efficiency of modeling. In the modeling process of the present invention, the weight variances of all explanatory variables are first calculated respectively to determine the degree of influence on the response variable, and then the difference between the weight variances of different explanatory variables is identified to sort the importance of the explanatory variables. Here, the larger the weight variance, the greater the difference in the tendency of the explanatory variable to select the commodity category, which means that the influence of the explanatory variable is greater. For example, for some cosmetics, the weight purchased by female customers is often much larger than that of male customers, and the weight variance of gender is also relatively large, which means that the importance of gender as an explanatory variable is also relatively high for the cosmetics. In addition, in binary recursive decomposition, it is often necessary to perform multiple decompositions to maximize the purity of the classification, that is, the Gini coefficient is minimized. This process makes the decision tree complicated and easily leads to overfitting. In addition, the correlation between some explanatory variables and the response variable is unknown, and irrelevant or less relevant variables may be introduced when building the model, so it is necessary to prune the decision tree after generating it.

[0024] In this embodiment, as the parameter value (surface error rate gain rate) decreases, the number of decision tree decompositions increases, and the Gini coefficient decreases accordingly. The pruning process refers to the relationship between the surface error rate gain rate and the Gini coefficient. When the surface error rate gain rate decreases and the magnitude of the reduction in the Gini coefficient becomes smaller, the surface error rate gain rate is the most appropriate value, and the subtree with the corresponding number of decompositions is the best subtree after the decision tree is pruned. Here, with the help of a scatter plot between the surface error rate gain rate and the Gini coefficient, the appropriate parameter value can be determined through the changing trend of the Gini coefficient with the scatter plot in the figure.

[0025] Furthermore, in the data modeling and analysis system, the commodity pricing analysis module is connected to the customer preference analysis module to analyze the relationship between the commodity price and sales volume preferred by the customer and provide suggestions for the formulation of commodity prices. The method adopted is to determine the relationship between commodity price and sales volume through a polynomial fitting model, including the following steps: B1, determine the model structure: determine the independent variables and dependent variables of the polynomial fitting model according to the response variables corresponding to the terminal nodes of the decision tree in step A2; B2, adjust the parameters of the model to obtain a model with good fitting performance; B3, determine the optimal product pricing: Based on the determined polynomial fitting relationship and combined with the product cost, calculate the optimal product price with the highest profit.

[0026] Furthermore, in step B1, the model structure is a polynomial fitting model constructed with commodity price as the independent variable and sales volume as the dependent variable, and its formula is: , In the formula, y For sales volume; x is the commodity price; a0, a1, a2, ..., a n is the fitting parameter; n is the order of the model.

[0027] Furthermore, in step B2, the adjustment of model parameters is to determine the appropriate order of the model using the AIC criterion, and determine the fitting parameters that minimize the sum of squares of the model residuals.

[0028] Furthermore, in step B3, the optimal commodity price is the commodity price with the highest profit, and its calculation formula is: , In the formula, z a is the profit; b is the cost of goods; y For sales volume; x For commodity prices, z When the value is extreme x Solve it and you can get the best price for the product.

[0029] Furthermore, the intelligent decision-making system includes a customer promotion module and a commodity pricing module, wherein: The customer promotion module is connected to the customer preference analysis module in the data analysis and modeling system. It is used to input the collected customer age, occupation and gender data into the already constructed classification and regression tree model, output the customer's preferred products through the decision tree algorithm, and promote the products online or offline accordingly based on the output results; The commodity pricing module is connected to the commodity pricing analysis module in the data analysis and modeling system, and is used to output the results of the polynomial fitting model determined by commodity pricing and sales volume, combined with the calculation formula of the optimal commodity price, to determine the optimal commodity price when the profit is highest.

[0030] The beneficial effects of the present invention are: (1) Using the store retail data stored in the integrated POS system, a classification and regression tree model is constructed to generate a decision tree to analyze the correlation between retail data variables and determine customer preferences for products. During the modeling process, the weight differences between different explanatory variables are calculated to determine the importance order of the explanatory variables, thereby solving the problem of low modeling efficiency caused by the need to repeatedly construct models to determine the importance order of the explanatory variables in the prior art. (2) The relationship between the surface error rate gain rate and the Gini coefficient is used to prune the decision tree, which effectively solves the problem of low precision caused by overfitting of the decision tree. In the process, irrelevant or less relevant explanatory variables are eliminated, thereby improving the explanatory power of the decision tree.

[0031] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. An integrated new retail smart store POS system, characterized by: It includes an integrated POS system, a data modeling and analysis system, and an intelligent decision-making system, including: The integrated POS system is used to manage the retail business of the new retail smart store and store retail data, wherein the retail data includes customer data and product data. The customer data includes the customer's age, occupation and gender; the product data includes the category, cost, sales price and sales volume of the products purchased by the customer; Data modeling and analysis system, used to perform statistical analysis and modeling on retail data, and analyze the correlation between retail data variables, including customer preference analysis module and commodity pricing analysis module; Intelligent decision-making system, which is used to provide intelligent decision-making suggestions for user marketing work based on the output results of the data modeling and analysis system, including customer promotion module and product pricing module; In the data modeling and analysis system, the customer preference analysis module uses a classification and regression tree model to analyze the impact of customer age, occupation and gender on the category of goods purchased by the customer, including the following steps: A1. Determine the model structure: determine the explanatory variables and response variables for building the classification and regression tree models, where the customer's age, occupation and gender are the explanatory variables and the category of goods purchased by the customer is the response variable; A2, generate a decision tree: sort the importance of the explanatory variables, perform binary recursive decomposition of the response variable according to the order of importance of the explanatory variables, until the Gini coefficient after decomposition does not change, and generate a decision tree; A3, decision tree pruning: adjust the model parameters, and prune the decision tree based on the relationship between the Gini coefficient and the model parameters to obtain the optimal subtree; the model parameters are the surface error rate gain rate.

2. The integrated new retail smart store POS system according to claim 1, characterized in that: In step A2, the importance of the explanatory variables is sorted, and the response variable is binary recursively decomposed according to the importance order of the explanatory variables until the Gini coefficient after decomposition does not change, and a decision tree is generated, including the following steps: A21, calculate the weight difference of different explanatory variables: calculate the weight of customers of different ages, occupations and genders in a commodity category relative to all customers in the commodity category, and calculate the variance of the weights of different explanatory variables as the weight difference of different explanatory variables; A22, ranking of explanatory variable importance: ranking the importance of explanatory variables according to the difference in weights of different explanatory variables, where the greater the difference in weights of the explanatory variables, the greater the importance; A23, perform binary recursive decomposition on the response variable to generate a decision tree: according to the order of importance of the explanatory variables, select the explanatory variable with the greatest importance as the root node of the decision tree, and decompose the response variable data set into two subsets, wherein the root node is the point of the first decomposition of the decision tree; then repeat the methods of steps A21-A22 for the samples of the two subsets respectively to identify the order of importance of the explanatory variables of the corresponding subsets, select the explanatory variable with the greatest importance as the node again, and decompose each subset sample into two subsets; calculate the Gini coefficient in the process of binary recursive decomposition of the decision tree, and when the Gini coefficient no longer changes, no longer perform decomposition.

3. The integrated new retail smart store POS system according to claim 1, characterized in that: In step A3, the adjustment of the model parameters and the pruning of the decision tree based on the relationship between the Gini coefficient and the model parameters to obtain the optimal subtree include the following steps: A31, according to the order of binary recursive decomposition of the decision tree, sequentially calculate the subtree Gini coefficient and the surface error rate gain rate after each decomposition, and the surface error rate gain rate is calculated by the formula: ， Where α is the surface error rate gain rate. The smaller α is, the more times the decision tree is decomposed and the more nodes the tree has. R(t) is the error cost of a tree with t as a single node. The tree is pruned at this node. R(t) = r(t)×p(t), r(t) is the error rate of node t, and p(t) is the proportion of the number of samples on node t to the total number of samples. T t is the number of leaf nodes contained in the subtree, R(T t ) represents the error cost of the subtree T with t as the root node, ,in m is the number of nodes, i =1,2,……, m ; A32, compare the surface error rate gain rate and Gini coefficient of different decomposition times of decision trees. When the surface error rate gain rate decreases and the reduction of the Gini coefficient becomes smaller, the subtree with the corresponding decomposition number is the best subtree after pruning the decision tree.

4. The integrated new retail smart store POS system according to claim 1, characterized in that: In the data modeling and analysis system, the commodity pricing analysis module is connected to the customer preference analysis module, and a polynomial fitting model is used to determine the relationship between commodity price and sales volume, and the optimal commodity price is determined in combination with commodity cost, including the following steps: B1, determine the model structure: determine the independent variables and dependent variables of the polynomial fitting model, where the commodity price is the independent variable and the sales volume is the dependent variable; B2, adjust the parameters of the model to obtain a model with good fitting performance; B3, determine the optimal product price: Based on the determined polynomial fitting relationship and the product cost, calculate the optimal product price when the profit is the highest.

5. The integrated new retail smart store POS system according to claim 4, characterized in that: In step B2, the adjustment of model parameters is to determine the appropriate order of the model using the AIC criterion and determine the fitting parameters that minimize the sum of squares of the model residuals.

6. The integrated new retail smart store POS system according to claim 1, characterized in that: The intelligent decision-making system, The customer promotion module is connected to the customer preference analysis module of the data modeling and analysis system, and is used to input the collected customer age, occupation and gender data into the constructed classification and regression tree model, output the customer's preferred products through the decision tree algorithm, and promote the products online or offline according to the output results; The commodity pricing module is connected to the commodity pricing analysis module in the data modeling and analysis system, and is used to output the polynomial fitting model results determined by commodity pricing and sales volume to determine the optimal commodity price.

Images