A method for commodity layout based on the shopping behavior of customers in shopping malls and supermarkets

Abstract

The present invention provides a commodity layout method based on the shopping behavior of customers in a supermarket and a hypermarket, which includes obtaining a commodity data set of commodities picked up and not put back on all the shelves in the supermarket and a hypermarket; calculating an optimal commodity combination association rule set according to the commodity data set; supplementing the optimal commodity combination association rule set, constructing and correcting an optimal shopping path in the supermarket and a hypermarket; calculating a heat estimation value of each grid in the supermarket and a hypermarket; and performing commodity layout according to the planned optimal shopping path, the optimal commodity combination, and the shopping heat area. The present invention uses RFID to realize the interaction between customers and commodities in the supermarket and a hypermarket, provides comprehensive data in multiple dimensions from shopping paths to commodity selection, etc., and transcends the limitations of simple point-of-sale data or video surveillance; by analyzing the commodity combinations purchased by users, planning the shopping paths in the supermarket and a hypermarket at the same time, and combining with the shopping hot areas in the supermarket and a hypermarket, the present invention realizes commodity layout, thereby improving the accurate positioning of commodity placement and thus improving the accurate layout of commodities.

Description

Technical Field

[0001] The present invention relates to the technical field of product layout, and in particular to a product layout method based on the shopping behavior of customers in a supermarket or shopping mall. Background Art

[0002] In modern retail, the importance of customer behavior analysis has become increasingly prominent, especially in a complex shopping environment such as a large supermarket or shopping mall.

[0003] For the placement of products in a supermarket or shopping mall, traditional methods mainly rely on sales data or direct observation to understand customer behavior, and then place products according to the purchasing behavior of customers. This method has obvious limitations; moreover, these methods often cannot provide real-time and comprehensive data, thus affecting the merchant's ability to quickly respond to market dynamics and customer needs.

[0004] Patent CN109102039A discloses a mobile intelligent supermarket system based on RFID technology. This system is based on the Android platform, developed in the Java language, designed with a C / S architecture for the system structure, and at the same time, through the target recognition and information collection functions of RFID technology, it realizes functions such as product information search, supermarket navigation, and intelligent shopping in the system, providing consumers with convenient shopping guidance functions.

[0005] Although the above patent can achieve shopping navigation, it cannot provide strategies for merchants to layout products according to customer shopping behavior, and the product layout in a supermarket or shopping mall can still only be achieved by traditional methods. Summary of the Invention

[0006] Aiming at the deficiencies of the prior art, the present invention provides a product layout method based on the shopping behavior of customers in a supermarket or shopping mall. The present invention realizes the precise layout of products through the combined analysis of products, path planning, and shopping hot zone analysis.

[0007] The technical solution of the present invention is: a product layout method based on the shopping behavior of customers in a supermarket or shopping mall, including the following steps:

[0008] S1), obtaining a product data set D of products that are picked up and not put back on all the shelves in the supermarket or shopping mall;

[0009] S2), calculating the association rules of products on each shelf according to the product data set D, and screening out the optimal product combination association rule set R that meets the requirements;

[0010] S3), calculating the supplementary association rules of different products on different shelves, and screening out the optimal product combination supplementary association rule set R' that meets the requirements; then integrating the association rule set R and the supplementary association rule R' to obtain a new optimal product combination association rule set R new ;

[0011] S4), Divide the shopping mall into multiple nodes and construct the optimal shopping path in the shopping mall;

[0012] S5), Correct the optimal shopping path obtained in step S4) according to the actual shopping path of the user to obtain the corrected optimal shopping path;

[0013] S6), Divide the shopping mall into a two-dimensional grid and calculate the heat estimation value of each grid in the shopping mall;

[0014] S7), Add the corrected optimal shopping path obtained in step S5) to the grid in step S6), and arrange the new optimal commodity combination association rule set R * obtained in step S3) in the grid where the heat estimation value is greater than the preset expected threshold ξ new for the commodities in the optimal commodity combination.

[0015] Preferably, in step S1), constructing the commodity data set specifically includes the following steps:

[0016] S11), Install corresponding RFID tags on each commodity and the shopping cart, and obtain the number of times pu i that each commodity is picked up or the number of times pb i of being put back through the RFID tags;

[0017] S12), Calculate its heat rate η i according to the number of times pu i that each commodity is picked up or the number of times pb i of being put back:

[0018]

[0019] In the formula, pu i is the number of times commodity i is picked up within time t; pb i is the number of times commodity i is put back within time t;

[0020] S13), Multiple commodities that are picked up and not put back on the same shelf within time t are used as a sequence T j * , T j * = {item j,1 , item j,2 , item j,3 ,..., item j,n}; where, T j * is the jth shelf; item j,n is the n commodities picked up and not put back on the jth shelf;

[0021] Then count all the shelves to obtain a dataset D = {T1 * , T2 * ,..., T m *}, which contains all the items picked up from the shelves within time t and not put back.

[0022] Preferably, in step S2), the calculation of the optimal item combination association rule set R specifically includes the following steps:

[0023] S21), Scan the dataset D = {T1 * , T2 * ,..., T m *} of items and count the occurrence frequency p i,j of each item to obtain the frequent item set L of all items;

[0024] S22), For each item i in the frequent item set L, create a head pointer head(i) pointing to all the nodes in the FP-tree that contain this item; then starting from an empty tree, for each sequence T * in the dataset D of items, construct the corresponding FP-tree in the order of the frequent item set L and insert the items into the FP-tree accordingly; if the corresponding child node exists, increment the count; if not, create a new child node;

[0025] S23), Traverse each frequent item item in the FP-tree of sequence T * to generate the conditional pattern base D item , and then recursively construct the conditional FP-tree FP tree (D item ), and repeat this process for each conditional FP-tree until no more frequent items can be generated; obtain the frequent item set of each sequence T * ;

[0026] S24), For each frequent item set X, generate all possible association rules X → Y; that is, when item X is purchased, item Y is purchased simultaneously, and then calculate the confidence Confidence of each association rule:

[0027]

[0028] In the formula, Support(X ∪ Y) represents the frequency of the union of item set X and item set Y in the dataset, and Support(X) represents the frequency of item set X in the dataset;

[0029] S25), Compare the confidence Confidence of each association rule with the preset minimum confidence threshold ρ * to screen out the optimal product combination association rule set R that meets the requirements, that is, the optimal product combination set.

[0030] Preferably, in step S3), the calculation of the optimal product combination supplementary association rule set R' specifically includes the following steps:

[0031] S31), Obtain multiple products that pass through the RFID at the cashier at the same time, and use them as a product supplementary sequence addT j * ={additem j,1 , additem j,2 , additem j,3 ,..., additem j,q}; additem j,q represents the qth product in the jth product supplementary sequence; then count all the product supplementary sequences within time t to obtain the product dataset D' supplemented within time t = {addT1 * , addT2 * ,..., addT m *};

[0032] S32), Then process the supplemented product dataset D' according to the method of steps S21)-S24) to obtain the optimal product combination supplementary association rule set R';

[0033] S33), Determine whether the optimal product combination supplementary association rule set R' is consistent with the optimal product combination association rule set R. If not, update the minimum confidence threshold of the supplementary rules and update the supplementary association rule set R' according to the minimum confidence threshold ; where, the calculation formula of the minimum confidence threshold is:

[0034]

[0035] In the formula, a, b, c are preset parameters, and e is the natural exponent.

[0036] Preferably, in step S3), the integration of the new optimal product combination association rule set R new is:

[0037] R new = R’ ∪ R;

[0038] And the obtained optimal product association rule set R newThe commodity combinations are sorted according to the popularity rate η i in descending order.

[0039] Preferably, in step S4), the following steps are specifically included:

[0040] S41), taking the entrance S and the exit E of the shopping mall as the starting point and the ending point, and presetting a plurality of nodes between the entrance S and the exit E at the same time. Initialize the actual distance g(v) from the entrance S to each node to ∞, the distance g(s) between the entrance S and the exit E to 0, and at the same time two-dimensionally map the entire shopping mall in the rectangular coordinate system; preset an open list to store the explored nodes. Initially, add the starting point S to the open list, and then repeat steps S42)-S45) until the end point E is visited or the open list is empty;

[0041] S42), select the node u with the minimum cost function from the open list for exploration, and the cost function is:

[0042] f(u) = g(u) + h(u);

[0043] In the formula, f(u) is the comprehensive cost of node u, equal to the sum of its actual distance g(u) and the estimated distance h(u), where h(u) = |x u -x E | + |y u -y E |; (x u , y u ) is the coordinate of node u, and (x E , y E ) is the coordinate of the end point E;

[0044] S43), if the node u is the end point E, the search ends and the shortest path is found; if the node u is not the end point E, mark the node u as visited and remove it from the open list;

[0045] S44), for each neighbor node v of node u, update the actual distance g(v) from the starting point S to node v:

[0046] g(v) = min(g(v), g(u) + δ(u, v));

[0047] In the formula, δ(u, v) is the actual distance from node u to node v;

[0048] S45), if the node v is not in the open list, add it to the open list and calculate the estimated distance h(v) of node v; if the end point e is visited, determine the optimal path P from the starting point S to the end point E by backtracking the parent node.

[0049] Preferably, in step S5), the following steps are specifically included:

[0050] S51), Define a series of observation points O = {o1, o2,..., o Z} on the path actually walked by the user. For each observation point o i , find the point p i on the optimal path P that is closest to o i , and calculate the difference distance d i :

[0051]

[0052] S52), Update the open list at each observation point o i , that is, add or update the nodes related to o i in the open list, and consider adding new neighbor nodes;

[0053] Then for each observation point o i , update the cost function f(v') of its corresponding corrected node v' to reflect the impact of the actual path on the cost:

[0054]

[0055] where g(v') is the actual distance from the starting point to node v', h(v') is the estimated distance from node v' to the end point, λ is a weight parameter that adjusts the impact of the actual path observation difference in the cost function, and ω i is the weight for each difference distance d i ;

[0056] S53), Re-select the node with the minimum corrected cost function f(v') as the next exploration node. When the end point is visited, obtain the corrected optimal path P' from the starting point to the end point by backtracking the corrected nodes.

[0057] Preferably, in step S6), the following steps are specifically included:

[0058] S61), Two-dimensionally grid the commercial supermarket to obtain N grids. The center point position of the i-th grid is represented as H i , the pedestrian flow in the i-th grid is represented as G i , and the total number of times all goods in the i-th grid are picked up or put back is z i ;

[0059] S62), Calculate the weight θ i of the i-th grid, that is:

[0060]

[0061] S63), calculate the bandwidth parameter W, i.e.:

[0062]

[0063] where, is the standard deviation of the data, H i is the center point position of the i-th grid, is the average value of all grid center points; IQR = Q3 - Q1 is the interquartile range of the data, Q1 represents the 25% quantile of the data, that is, the value at the 25% position after arranging the data from small to large; Q3 represents the 75% quantile of the data, that is, the value at the 75% position after arranging the data from small to large;

[0064] S64), calculate the heat value estimates of the four corners of each grid i.e.:

[0065]

[0066] In the formula, H i,k is the k-th corner position of the i-th grid, k = {1, 2, 3, 4}; W is the bandwidth parameter, H i is the center point position of the i-th grid, θ i is the weight of the i-th grid;

[0067] Then, according to the heat value estimates of the four corners of the grid calculate the heat value estimate of each grid i.e.:

[0068]

[0069] Preferably, in step S31), by setting RFID at the cashier position, a plurality of commodities passing through the RFID at the cashier are regarded as a sequence, which means that all the commodities in the same shopping cart are regarded as a sequence.

[0070] The beneficial effects of the present invention are as follows:

[0071] 1. The present invention uses RFID technology to seamlessly track the interaction between customers and commodities in the entire shopping mall, providing comprehensive data in multiple dimensions such as shopping paths and commodity selections, exceeding the limitations of simple point-of-sale data or video surveillance;

[0072] 2. The present invention analyzes the commodity combinations purchased by users, plans the shopping paths in the shopping mall at the same time, and combines the shopping hot areas in the shopping mall to realize the layout of commodities, thereby improving the accurate positioning of commodity placement and thus improving the accurate layout of commodities. Description of the Drawings

[0073] Figure 1It is the flow framework diagram of the method of the present invention;

[0074] Figure 2 It is the supermarket grid diagram of the method of the present invention. Specific embodiments

[0075] The following further describes the specific embodiments of the present invention with reference to the accompanying drawings:

[0076] As Figure 1 shown, this embodiment provides a commodity layout method based on the shopping behavior of supermarket customers, including the following steps:

[0077] S1), Obtain the commodity dataset D of the commodities picked up and not put back on all the shelves in the supermarket; specifically constructing the commodity dataset specifically includes the following steps:

[0078] S11), Install corresponding RFID tags on each commodity and the shopping cart, and obtain the number of times pu i that each commodity is picked up i or the number of times pb

[0079] that it is put back through the RFID tag; i S12), According to the number of times pu i that each commodity is picked up i or the number of times pb

[0080]

[0081] that it is put back, calculate its heat rate η i : i In the formula, pu

[0082] is the number of times commodity i is picked up within time t; pb j * is the number of times commodity i is put back within time t; j * S13), Take the multiple commodities picked up and not put back on the same shelf within time t as a sequence T j,1 , item j,2 , item j,3 ,..., item j,n}; where, T j * is the jth shelf; item j,n is the n commodities picked up and not put back on the jth shelf;

[0083] Then count all the shelves to obtain the dataset D = {T1 * , T2 *,..., T m *}。

[0084] S2). Calculate the association rules of the commodities on each shelf according to the commodity dataset D, and filter out the optimal commodity combination association rule set R that meets the requirements; specifically:

[0085] S21). Scan the commodity dataset D = {T1 * , T2 * ,..., T m *}, and count the occurrence frequency p i,j of each commodity, so as to obtain all frequent item sets L of the commodities;

[0086] S22). For each commodity i in the frequent item set L, create a head pointer head(i) pointing to all nodes in the FP tree that contain this commodity; then start from an empty tree, for each sequence T in the commodity dataset D * , construct the corresponding FP tree in the order of the frequent item set L, and insert the commodity into the FP tree accordingly; if the corresponding child node exists, increase the count; if it does not exist, create a new child node;

[0087] S23). Traverse each frequent item item in the FP tree of the sequence T * , generate the conditional pattern base D item , and then recursively construct the conditional FP tree FP tree (D item ), and repeat this process for each conditional FP tree until no more frequent items can be generated; obtain the frequent item sets of each sequence T * ;

[0088] S24). For each frequent item set X, generate all possible association rules X → Y; that is, when purchasing commodity X, commodity Y is purchased at the same time, and then calculate the confidence Confidence of each association rule:

[0089]

[0090] where Support(X ∪ Y) represents the frequency of the union of item set X and item set Y in the dataset, and Support(X) represents the frequency of item set X in the dataset;

[0091] S25). Compare the confidence Confidence of each association rule with the preset minimum confidence threshold ρ * to filter out the optimal commodity combination association rule set R that meets the requirements, that is, the optimal commodity combination set.

[0092] S3), calculate the replenishment association rules for different products on different shelves, and screen out the optimal product combination replenishment association rule set R' that meets the requirements; then integrate the association rule set R and the replenishment association rule R' to obtain the new optimal product combination association rule set R new ; The specific steps are as follows:

[0093] S31), Obtain multiple products passing through the RFID at the cashier at the same time, and use them as a product replenishment sequence addT j * ={additem j,1 , additem j,2 , additem j,3 ,..., additem j,q}; additem j,q represents the q-th product in the j-th product replenishment sequence; then count all the product replenishment sequences within time t, so as to obtain the replenished product data set D' within time t = {addT1 * , addT2 * ,..., addT m *};

[0094] S32), Then process the replenished product data set D' according to the method in steps S21)-S24) to obtain the optimal product combination replenishment association rule set R';

[0095] S33), Determine whether the optimal product combination replenishment association rule set R' is consistent with the optimal product combination association rule set R. If not, update the minimum confidence threshold of the replenishment rule and update the replenishment association rule set R' according to the minimum confidence threshold ; Among them, the calculation formula of the minimum confidence threshold is:

[0096]

[0097] In the formula, a, b, c are preset parameters, and e is the natural exponent.

[0098] S34), The new optimal product combination association rule set R new is integrated according to the following formula:

[0099] R new = R’ ∪ R;

[0100] And sort the product combinations in the obtained optimal product association rule set R new according to the size of the popularity rate η i .

[0101] S4), Divide the shopping mall into multiple nodes and construct the optimal shopping path in the shopping mall. Specifically, it includes the following steps:

[0102] S41), Take the entrance S and the exit E of the shopping mall as the starting point and the ending point. At the same time, preset multiple nodes between the entrance S and the exit E. Initialize the actual distance g(v) from the entrance S to each node as ∞, and the distance g(s) between the entrance S and the exit E as 0. At the same time, two-dimensionally represent the entire shopping mall in a rectangular coordinate system. At the same time, preset an open list to store the explored nodes. Initially, add the starting point S to the open list. Next, repeat steps S42)-S45) until the ending point E is visited or the open list is empty;

[0103] S42), Select the node u with the minimum cost function from the open list for exploration. The cost function is:

[0104] f(u) = g(u) + h(u);

[0105] In the formula, f(u) is the comprehensive cost of node u, which is equal to the sum of its actual distance g(u) and the estimated distance h(u), where h(u) = |x u -x E | + |y u -y E |; (x u , y u ) is the coordinate of node u, and (x E , y E ) is the coordinate of the ending point E;

[0106] S43), If node u is the ending point E, the search ends and the shortest path is found; if node u is not the ending point E, mark node u as visited and remove it from the open list;

[0107] S44), For each neighbor node v of node u, update the actual distance g(v) from the starting point S to node v:

[0108] g(v) = min(g(v), g(u) + δ(u, v));

[0109] In the formula, δ(u, v) is the actual distance from node u to node v;

[0110] S45), If node v is not in the open list, add it to the open list and calculate the estimated distance h(v) of node v; if the ending point e is visited, determine the optimal path P from the starting point S to the ending point E by backtracking the parent node.

[0111] S5). Modify the optimal shopping path obtained in step S4) according to the actual shopping path of the user to obtain the modified optimal shopping path. The specific steps are as follows:

[0112] S51). Define a series of observation points O = {o1, o2,..., o Z} on the path actually traveled by the user. For each observation point o i , find the point p i on the optimal path P that is closest to o i , and calculate the difference distance d i :

[0113]

[0114] S52). Update the open list at each observation point o i , that is, add or update the nodes related to o i in the open list, and consider adding new neighbor nodes;

[0115] Then for each observation point o i , update the cost function f(v') of its corresponding modified node v' to reflect the impact of the actual path on the cost:

[0116]

[0117] where g(v') is the actual distance from the starting point to node v', h(v') is the estimated distance from node v' to the end point, λ is a weight parameter that adjusts the impact of the actual path observation difference on the cost function, and ω i is the weight for each difference distance d i ;

[0118] S53). Re - select the node with the minimum modified cost function f(v') as the next exploration node. When the end point is visited, obtain the modified optimal path P' from the starting point to the end point by backtracking the modified nodes.

[0119] S6). Grid the supermarket in two - dimensions and calculate the heat estimate value of each grid in the supermarket. The specific steps are as follows:

[0120] S61). Grid the supermarket in two - dimensions to obtain N grids. The central point position of the i - th grid is represented as H i , the number of people flowing in the i - th grid is represented as G i , and the total number of times all the goods in the i - th grid are picked up is z i ;

[0121] S62). Calculate the weight θ i of the i - th grid, that is:

[0122]

[0123] S63), Calculate the bandwidth parameter W, i.e.:

[0124]

[0125] Wherein, is the standard deviation of the data, H i is the center point position of the i-th grid, is the average value of all grid center points; IQR = Q3 - Q1 is the interquartile range of the data, Q1 represents the 25% quantile of the data, i.e., the value at the 25% position after arranging the data from small to large; Q3 represents the 75% quantile of the data, i.e., the value at the 75% position after arranging the data from small to large;

[0126] S64), Calculate the heat estimations at the four corners of each grid respectively That is:

[0127]

[0128] In the formula, H i,k is the k-th corner position of the i-th grid, k = {1, 2, 3, 4}; W is the bandwidth parameter, H i is the center point position of the i-th grid, θ i is the weight of the i-th grid;

[0129] Then, according to the heat estimations at the four corners of the grid calculate the heat estimation value of each grid That is:

[0130]

[0131] S7), Add the corrected optimal shopping path obtained in step S5) to the grid in step S6), and arrange the optimal commodity combination association rule set R * obtained in step S3) in the grid where the heat estimation value is greater than the preset expected threshold ξ new for the commodities in the optimal commodity combination.

[0132] As Figure 2 shown, the data collected in this embodiment is as follows: T1 * = {A, B, C}, After analysis, the association rule A → C (i.e., indicating that A and C can be used as a commodity combination) is obtained. There is supplementary data addT * = {A, C, D}. After analysis, the finally updated association rules are A → C and C → D. And the heat rate of A is higher than that of D, so it can be considered to give priority to placing the commodity combination A and C.

[0133] Furthermore, the commercial supermarket is two-dimensionally mapped onto a rectangular coordinate system, and the entire commercial supermarket is divided into 9 grids (assuming (1,1) is the entrance of the commercial supermarket and (3,3) is the exit of the commercial supermarket). Through analysis, the optimal path P is obtained: (1,1) → (2,1) → (2,2) → (3,2) → (3,3). Then, it is corrected in combination with the actual shopping path of customers. Through analysis, the corrected optimal path P' is obtained: (1,1) → (1,2) → (2,2) → (2,3) → (3,3). Finally, the heat estimation value of each grid is calculated and marked in the grid, as shown in the appendix Figure 2 As shown, the corrected path P' is added, and grids with heat estimation values higher than the expected threshold ξ * are selected on the corrected path P', and the above-obtained preferentially placeable product combinations are placed therein, or the products that the commercial supermarket wants to promote are placed in these grids.

[0134] The above embodiments and the descriptions in the specification only illustrate the principles and the best embodiments of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed.

Claims

1. A method for commodity layout based on the shopping behavior of supermarket customers, characterized in that, It includes the following steps: S1), Obtain the commodity dataset D of the commodities that have been picked up and not put back on all the shelves in the supermarket; S2), Calculate the association rules of the commodities on each shelf according to the commodity dataset D, and filter out the optimal commodity combination association rule set R that meets the requirements; S3), calculate the replenishment association rules for different products on different shelves, and screen out the optimal product combination replenishment association rule set R' that meets the requirements; then integrate the association rule set R and the replenishment association rule R' to obtain a new optimal product combination association rule set R new ; S4), Divide the supermarket into multiple nodes and construct the optimal shopping path in the supermarket; S5), Correct the optimal shopping path obtained in step S4) according to the actual shopping path of the user to obtain the corrected optimal shopping path; S6), Divide the supermarket into a two-dimensional grid and calculate the heat estimation value of each grid in the supermarket, which specifically includes the following steps: S61), the commercial supermarket is two-dimensionally grid-divided to obtain N grids, and the center point position of the i-th grid is represented as H i , and the pedestrian flow in the i-th grid is represented as G i , and the total number of times all the goods in the i-th grid are picked up or put back is z i ; S62), calculate the weight of the i-th grid That is: S63), Calculate the bandwidth parameter W, that is: Among them, is the standard deviation of the data, H i is the center point position of the i-th grid, is the average value of all grid center points; IQR = Q3 - Q1 is the interquartile range of the data, Q1 represents the 25% quantile of the data, that is, the value at the 25% position after arranging the data from small to large; Q3 represents the 75% quantile of the data, that is, the value at the 75% position after arranging the data from small to large; S64), calculate the heat value estimations of the four corners of each grid respectively That is: Where, H i,k is the k-th corner position of the i-th grid, k = {1, 2, 3, 4}; W is the bandwidth parameter, and H i is the center point position of the i-th grid, is the weight of the i-th grid; S65), and then estimate the heat of the four corners of the grid Calculate the heat estimate value of each grid That is: S7), add the corrected optimal shopping path obtained in step S5) to the grid in step S6), and arrange the new optimal commodity combination association rule set R obtained in step S3) in the grid where the heat estimation value is greater than the preset expected threshold ξ * with the commodities of the optimal commodity combination in new it.

2. The commodity layout method based on the shopping behavior of supermarket customers according to claim 1 is characterized in that: In step S1), constructing the commodity dataset specifically includes the following steps: S11), install corresponding RFID tags on each commodity and shopping cart, and obtain the number of times pu that each commodity is picked up through the RFID tags i or the number of times pb of being put back i ; S13), a plurality of items picked up and not put back in the same shelf within time t are regarded as a sequence T j * , T j * = {item j,1 , item j,2 , item j,3 ,..., item j,n}; where T j * is the j-th shelf; item j,n is the n items picked up and not put back on the j-th shelf; Then count all the shelves to obtain a data set D = {T1 * , T2 * ,..., T m *} that contains all the items picked up from the shelves within time t and not put back.

3. The commodity layout method based on the shopping behavior of supermarket customers according to claim 2, wherein: In step S2), the calculation of the optimal commodity combination association rule set R specifically includes the following steps: S21), scan the dataset D of goods = {T1 * , T2 * ,..., T m *}, and count the occurrence frequency p i,j of each good, so as to obtain the frequent item set L of all goods; S22) For each item i in the frequent itemset L, create a head pointer head(i) that points to all the nodes in the FP-tree that contain this item; then starting from an empty tree, for each sequence T in the dataset D of items * , construct the corresponding FP-tree in the order of the frequent itemset L and insert the items into the FP-tree accordingly; if the corresponding child node exists, increment the count; if it does not exist, create a new child node; S23), traverse each sequence T * in the FP-tree of each frequent item item, generate the conditional pattern base D item , and then recursively construct the conditional FP-tree FP tree (D item ), and repeat this process for each conditional FP-tree until no more frequent items can be generated; obtain the frequent item sets of each sequence T * ; S24), For each frequent item set X, generate all its possible association rules X→Y; that is, when purchasing commodity X, commodity Y is purchased at the same time, and then calculate the confidence Confidence of each association rule: In the formula, Support(X∪Y) represents the frequency of the union of item set X and item set Y appearing in the dataset, Support(X) represents the frequency of item set X appearing in the dataset; S25), compare the confidence Confidence of each association rule with a preset minimum confidence threshold ρ * to filter the optimal set R of association rules for product combinations that meet the requirements, that is, the optimal set of product combinations.

4. A method for commodity layout based on the shopping behavior of customers in a shopping mall and supermarket according to claim 1, characterized in that: In step S3), the calculation of the optimal commodity combination supplementary association rule set R' specifically includes the following steps: S31), Obtain multiple commodities with RFID passing through the cashier desk simultaneously, and use them as a commodity replenishment sequence addT j * ={additem j,1 ,additem j,2 ,additem j,3 ,...,additem j,q}; additem j,q represents the q-th commodity in the j-th commodity replenishment sequence; then count all the commodity replenishment sequences within time t, so as to obtain the dataset D' of the replenished commodities within time t = {addT1 * ,addT2 * ,...,addT m *}; S32), Then process the supplementary commodity dataset D' according to the method of steps S21)-S24) to obtain the optimal commodity combination supplementary association rule set R'; S33), Determine whether the optimal product combination supplementary association rule set R' is consistent with the optimal product combination association rule set R. If they are inconsistent, update the minimum confidence threshold of the supplementary rules and according to the minimum confidence threshold update the supplementary association rule set R'; where, the calculation formula of the minimum confidence threshold is as follows: where a, b, c are preset parameters, e is the natural exponent, and ρ * is the preset minimum confidence threshold.

5. The commodity layout method based on the shopping behavior of supermarket customers according to claim 4, wherein: In step S3), the integration of the new optimal commodity combination association rule set R new is as follows: R new = R' ∪ R; And the obtained optimal commodity association rule set R new The commodity combinations in i are sorted according to the size of the popularity rate η.

6. The commodity layout method based on the shopping behavior of supermarket customers according to claim 5, characterized in that: In step S3), the heat rate η i is calculated according to the number of times pu i and pb i that each product is picked up or put back: wherein, pu i is the number of times product i is picked up within time t; p b i is the number of times product i is put back within time t.

7. A method for commodity layout based on the shopping behavior of customers in a supermarket or shopping mall according to claim 1, characterized in that: In step S4), it specifically includes the following steps: S41), Take the entrance S and the exit E of the supermarket as the starting point and the ending point. At the same time, preset multiple nodes between the entrance S and the exit E. Initialize the actual distance g(v) from the entrance S to each node as ∞, and the distance g(s) between the entrance S and the exit E as 0. At the same time, two-dimensionally map the entire supermarket in a rectangular coordinate system; at the same time, preset an open list to store the explored nodes. Initially, add the starting point S to the open list, and then repeat steps S42)-S45) until the ending point E is visited or the open list is empty; S42), Select the node u with the minimum cost function from the open list for exploration. The cost function is: f(u) = g(u) + h(u); where f(u) is the comprehensive cost of node u, which is equal to the sum of its actual distance g(u) and estimated distance h(u), where h(u) = |x u - x E | + |y u - y E |; (x u , y u ) are the coordinates of node u, and (x E , y E ) are the coordinates of the end point E; S43), If the node u is the ending point E, the search ends and the shortest path is found; if the node u is not the ending point E, mark the node u as visited and remove it from the open list; S44), For each neighbor node v of the node u, update the actual distance g(v) from the starting point S to the node v: g(v) = min(g(v), g(u) + δ(u, v)); In the formula, δ(u, v) is the actual distance from the node u to the node v; S45), If the node v is not in the open list, add it to the open list and calculate the estimated distance h(v) of the node v; if the ending point e is visited, determine the optimal path P from the starting point S to the ending point E by backtracking the parent node.

8. A method for commodity layout based on the shopping behavior of supermarket customers according to claim 1, characterized in that: In step S5), it specifically includes the following steps: S51), define a series of observation points O = {o1, o2,..., o Z} on the path actually walked by the user. For each observation point o i , find the point p i on the optimal path P that is closest to o i , and calculate the difference distance d i : S52), update the open list at each observation point o i i.e., add or update the nodes related to o i in the open list, and consider adding new neighbor nodes; Then for each observation point o i , update the cost function f(v') of its corresponding corrected node v' to reflect the impact of the actual path on the cost: Among them, g(v') is the actual distance from the starting point to node v', h(v') is the estimated distance from node v' to the end point, λ is a weight parameter that adjusts the influence of the actual path observation difference in the cost function, and ω i is the weight for each difference distance d i ; S53), reselect the node with the minimum modified cost function f(v') as the next exploration node. When the end point is visited, the modified optimal path P' from the start point to the end point is obtained by backtracking the modified nodes.

9. A method for commodity layout based on the shopping behavior of supermarket customers according to claim 1, characterized in that: In step S31), by setting up RFID at the cashier position, a sequence of multiple goods passing through the RFID at the cashier refers to all the goods in the same shopping cart as a sequence.

Images