CN-121981802-A - Edge internet of things perceived unmanned retail terminal commodity recommendation method and system

CN121981802ACN 121981802 ACN121981802 ACN 121981802ACN-121981802-A

Abstract

The invention relates to the technical field of unmanned retail recommendation and discloses an edge-based internet-of-things-aware unmanned retail terminal commodity recommendation method and system, wherein the method comprises the steps of extracting consumption events of retail commodities and consumption tracks of retail areas from consumer behavior data; the method comprises the steps of extracting associated consumption weights among different retail commodities and independent consumption weights of the retail commodities by utilizing a commodity interaction sequence mining mode, calculating to obtain visibility indexes and interaction probabilities of different display positions in the unmanned retail terminal, constructing a reward function for improving purchase intention and purchase conversion rate of consumers, and solving to obtain a retail commodity display layout scheme of the unmanned retail terminal. According to the invention, by combining edge internet of things perception and an interactive mining mode of retail goods and display positions, a layout scheme of the unmanned retail terminal can be dynamically generated based on real interactive behaviors of consumers and the value of the display positions, so that the purchase intention expression effect and the actual purchase conversion rate are improved.

Inventors

WANG SHILIANG
Tan Lingye
HU PAN
XU YUHAO
LIU HANFANG
ZHANG YANYUN
LI YUNZHE
WANG JIALE

Assignees

济南大学

Dates

Publication Date: 20260505
Application Date: 20260120

Claims (9)

1. An edge internet of things perceived unmanned retail terminal commodity recommendation method, which is characterized by comprising the following steps: s1, acquiring consumer behavior data by using multi-mode sensing equipment deployed in an unmanned retail terminal, and extracting consumption events of retail goods and consumption tracks of retail areas from the consumer behavior data to obtain a consumption event set of the retail goods and a consumption track set of the retail areas in the unmanned retail terminal; s2, extracting the associated consumption weight among different retail commodities and the independent consumption weight of the retail commodities by utilizing a commodity interaction sequence mining mode according to the consumption event set of the retail commodities; S3, according to the consumption track set of the retail area, calculating to obtain the visibility indexes and the interaction probabilities of different display positions in the unmanned retail terminal; And S4, constructing a reward function for improving the purchase intention and the purchase conversion rate of consumers according to the associated consumption weight among different retail commodities, the independent consumption weight of the retail commodities, the visibility index of different display positions and the interaction probability, and solving the reward function by adopting an adaptive layout optimization algorithm based on reinforcement learning to obtain a retail commodity display layout scheme of the unmanned retail terminal.
2. The method for recommending goods in an unmanned retail terminal perceived by an edge internet of things as recited in claim 1, wherein in the step S1, the collecting consumer behavior data by using a multi-mode sensing device deployed in the unmanned retail terminal includes: S11, disposing multi-mode sensing equipment consisting of infrared sensing sensors and RFID readers on the edge side of the unmanned retail terminal, wherein the infrared sensing sensors are respectively disposed in front of the unmanned retail terminal and in a commodity retail channel inside the unmanned retail terminal; S12, acquiring a stay infrared signal flow representing a stay state of a consumer and a consumption infrared signal flow representing a consumption behavior of the consumer by using the infrared sensing sensor, wherein the infrared sensing sensor of a commodity retail channel arranged in front of and inside the unmanned retail terminal sequentially acquires the stay infrared signal flow and the consumption infrared signal flow respectively; s13, acquiring the tag change state of the retail commodity tag in the unmanned retail terminal in real time by utilizing the RFID reader-writer, and constructing the tag change state of the retail commodity tag into an RFID signal stream; and S14, taking the stay infrared signal flow, the consumption infrared signal flow and the RFID signal flow as consumer behavior data.
3. The method for recommending goods at an unmanned retail terminal perceived by an edge internet of things as recited in claim 2, wherein in step S1, the consumer behavior data is extracted from consumption events of the retail goods and consumption trajectories of the retail area, and further comprising: S15, calculating the signal energy change rate of the signal acquisition time in the stay infrared signal flow according to the stay infrared signal flow, identifying the stay initial time when a consumer arrives at the unmanned retail terminal area and the stay end time when the consumer leaves the unmanned retail terminal, and taking the time period between the stay initial time and the stay end time as a time range corresponding to a primary consumption event and a section of consumption track; S16, extracting a tag state transition to a vanishing retail commodity set in a time period between the stay initial time and the stay end time according to the RFID signal stream, wherein the tag state transition to vanishing represents that the retail commodity is purchased; taking the time period between the stay initial time and the stay end time as an event range of a consumption event, taking the retail commodity set as a consumption commodity of the consumption event to form a primary consumption event, constructing all the currently extracted consumption events as a consumption event set of retail commodities in an unmanned retail terminal, and synchronously recording the moment that the label state of the retail commodity in the consumption event is converted into vanish; S17, calculating the signal energy change rate of signal acquisition time in the consumption infrared signal stream according to the consumption infrared signal stream, counting the number of signal acquisition time when the absolute value of the signal energy change rate is higher than a preset change rate threshold, and taking the number of signal acquisition time as the operation number of the consumer for taking and putting back the retail commodity; S18, extracting a retail commodity set in a consumption event corresponding to a time period between the stay initial time and the stay end time, sorting retail commodities in the retail commodity set according to the sequence of converting the label state into disappearance, and obtaining display positions of the sorted retail commodities to form a display position sequence; And S19, taking the signal acquisition time number and the display bit sequence as a section of consumption track, and constructing all the currently extracted consumption tracks into a consumption track set of a retail area in the unmanned retail terminal.
4. The method for recommending goods at an unmanned retail terminal perceived by an edge internet of things as claimed in claim 1, wherein in the step S2, the associated consumption weights among different retail goods and the individual consumption weights of the retail goods are extracted by using a goods interaction sequence mining mode, and further comprising: S21, according to the consumption event set of the retail commodity, taking the time period between the stay initial time and the stay end time associated with the consumption event as a consumption time period, extracting the time length of the consumption time period as the consumption time length of the consumption event, calculating the average value of the consumption time lengths of all the consumption events, calculating the weight factor of the consumption event by using the average value of the consumption time length, and weighting the retail commodity in each consumption event to obtain the independent consumption weight of the retail commodity; S22, acquiring the moment when the label states of different retail commodities in the same consumption event are converted into disappearing, and calculating to obtain a moment correlation factor between any two retail commodities; S23, acquiring the co-occurrence times of different retail commodities in a consumption event, calculating to obtain the preliminary association degree between any two retail commodities by adopting an improved co-occurrence frequency algorithm, and weighting the preliminary association degree by utilizing the preliminary association degree to obtain the associated consumption weight between the retail commodities.
5. The method for recommending goods at an unmanned retail terminal perceived by an edge internet of things as claimed in claim 4, wherein in the step S21, the weight factor of the consumption event is obtained by calculating the average value of the consumption time, and the retail goods in each consumption event are weighted to obtain the individual consumption weight of the retail goods, and the calculation formula of the individual consumption weight is as follows: ; Wherein, the Representing a nth retail item Is used to determine the individual consumption weights of the (c) for the (c), A weight factor representing the mth consumption event, Representing the duration of consumption of the mth consumption event, Representing the average of the duration of consumption of all consumption events, M representing the total number of consumption events, Indicating the number of categories of retail goods, Representing a retail item collection Whether or not there is a nth retail item Is used as a basis for the discriminant function of (a), Representing a set of retail items in an mth consumption event, if the set of retail items There is a nth retail item Then 1, Otherwise Is 0.
6. The method for recommending goods in an unmanned retail terminal perceived by an edge internet of things as claimed in claim 1, wherein the step S3 is performed to calculate the visibility index and the interaction probability of different display positions in the unmanned retail terminal, and the method comprises the following steps: S31, calculating to obtain the display weights of different display positions in the consumption track in the unmanned retail terminal according to the sequence of the display positions in the consumption track; S32, integrating the indication function value and the display weight of whether the display bit appears in the display bit sequence, and calculating to obtain a visibility index of the display bit in the unmanned retail terminal; And S33, weighting the indication function value of whether the display bit appears in the display bit sequence according to the signal acquisition time number in the consumption track, and generating the interaction probability of the display bit in the unmanned retail terminal.
7. The method for recommending goods at an unmanned retail terminal perceived by an edge internet of things as recited in claim 1, wherein the step S4 is configured to promote a reward function of consumer purchase intention and purchase conversion, the expression of the reward function is: ; Wherein, the Representing a bonus function that is based on the received data, Representing the weighting factor of the bonus function, Representing a retail merchandise display layout scheme to be solved The prize function value corresponding to the retail merchandise display layout plan a is represented, Represent the first Individual display position Is used for the visual index of (a), Represent the first Individual display position Is used to determine the interaction probability of the (c) in the (c) system, Indicating the number of display bits to be displayed, Representing the first item in retail merchandise display layout scheme A Individual display position The retail item to be displayed is a display, Representing retail goods Is used to determine the individual consumption weights of the (c) for the (c), Representing retail goods The associated consumption weight between them, Representing a retail item of the u-th type, K represents the number of categories of retail goods, Representing retail products in retail product display layout scheme A Is provided with a display position of the display screen, Representing retail goods The presentation bit distance constraint between.
8. The method for recommending goods in an unmanned retail terminal based on edge internet of things perception according to claim 7, wherein in the step S4, the adaptive layout optimization algorithm based on reinforcement learning is adopted to solve the reward function, so as to obtain a retail goods display layout scheme of the unmanned retail terminal, and the method further comprises: s41, initializing and generating a retail commodity display layout scheme as a current state; S42, selecting a layout adjustment action which is optimal and possibly improves the rewarding function value from a pre-trained display layout adjustment action space, adjusting the current state according to the selected layout adjustment action, generating a next time state, taking the next time state as an input value of the rewarding function, and calculating to obtain the rewarding function value of the next time state; S43, updating a Q value between the current state and the layout adjustment action by adopting a Q-learning algorithm, if the updating amplitude is smaller than a preset threshold value or the updating frequency of the Q value reaches a preset maximum iteration frequency, indicating that the algorithm converges, taking the next time state as a retail commodity display layout scheme of the unmanned retail terminal obtained by solving, otherwise, turning to a step S44; S44, taking the next moment state as the current state, recording the update times of the Q value, and returning to the step S42.
9. An edge internet of things perceived unmanned retail terminal commodity recommendation system, which is characterized by comprising a data processing module, an event extraction module, a consumption mining module and a layout scheme recommendation module, so as to realize the edge internet of things perceived unmanned retail terminal commodity recommendation method as set forth in any one of claims 1-8.

Description

Edge internet of things perceived unmanned retail terminal commodity recommendation method and system Technical Field The invention relates to the field of unmanned retail recommendation, in particular to an edge internet of things perceived unmanned retail terminal commodity recommendation method and system. Background With the sustainable development of the internet of things, edge computing and intelligent retail technology, the unmanned retail terminal gradually evolves from an unmanned selling mode with automatic selling as a core in early stage into an intelligent sensing and accurate recommending mode with behavior sensing, decision analysis and dynamic optimizing capabilities. The residence time, the moving path, the goods taking and returning, the final purchasing and other actions of the consumer in the unmanned retail terminal form a complete and continuous consumption process, and the action data not only reflects the instant demand state of the consumer, but also implies key information such as the attention degree, the preference strength, the associated selection habit and the like of the goods. Meanwhile, in practical application, the commodity layout of the traditional retail terminal depends on manual experience and static design, and is difficult to adapt to dynamic changes of consumer behaviors, so that the problems of unbalanced commodity exposure rate, low utilization rate of display space, limited sales conversion rate and the like are caused. Among existing studies, much research is focused on sales forecasting and inventory management. For example, patent CN118428998a proposes a sales prediction method and system for an unmanned retail intelligent cabinet, which uses LSTM model to predict sales volume by collecting, cleaning and extracting historical data, thereby reducing the risk of backout and stagnation. The method has certain advantages in sales prediction precision and system integration, but the core of the method still stays in sales prediction of a time sequence layer, is mainly used for replenishment and inventory decision, and is usually based on cloud centralized modeling, insufficient in real-time perception and instant recommendation support on an edge side, and difficult to map prediction results directly into specific commodity recommendation and display layout adjustment strategies. Aiming at the problem, the invention provides the edge-based internet-of-things-aware commodity recommendation method and system for the unmanned retail terminal, which dynamically optimize the display layout of retail commodities in the unmanned retail terminal, so that the placement position and arrangement mode of the retail commodities can improve the purchase intention of consumers, and further improve the purchase conversion rate of the retail commodities. Disclosure of Invention The invention provides an unmanned retail terminal commodity recommendation method and system based on edge internet of things perception, wherein the infrared and RFID signal stream data are collected at the edge side through multi-mode sensing equipment in the step S1, the dependence on image data is avoided, the refined modeling on the stay, interaction and purchase behaviors of consumers is realized on the premise of protecting privacy, the problems that the conventional unmanned retail system is high in visual recognition dependence and high in deployment cost are solved, the discrete retail commodity taking and purchase behaviors are converted into the computable independent consumption weights and the associated consumption weights through the mining of a consumption event interaction sequence in the step S2, the technical problem that the hidden collocation relation between commodities is difficult to be directly obtained from original perception data is solved, the exposure and interaction strength of different display positions in a consumption path are quantized through the introduction of a display position visual index and interaction probability, the limitation of only relying on a static position or a manual experience evaluation display position value is broken through the unified rewarding function of the characteristics of the retail commodity in the step S4, the optimal target capable of directly reflecting the purchase and conversion effect is constructed, and further the problem that the conventional dynamic consumption intention-adaptive dynamic display method based on the self-adaptive layout-driving and the optimal layout-based on the self-adaptive layout-adaptive display intention-based on the optimal layout algorithm is solved. In order to achieve the above purpose, the invention provides an edge internet of things perceived unmanned retail terminal commodity recommendation method, which comprises the following steps: s1, acquiring consumer behavior data by using multi-mode sensing equipment deployed in an unmanned retail terminal, and extracting consumption events of retail goods