Intelligent system for fitting behavior data acquisition and inventory optimization

Abstract

The invention belongs to the technical field of inventory management, and relates to a try-on behavior data acquisition and inventory optimization intelligent system, which comprises a radio frequency data acquisition module for generating a try-on behavior original data set; the strength grade judgment module is used for judging the strength grade of the try-on behavior; the satisfaction detection module triggers a high-satisfaction event mark and generates a try-on satisfaction index; the health index generation module is used for superposing a high-satisfaction event proportion and then performing reverse proportion operation with the actual purchase quantity to generate a try-on conversion health index; the dynamic early warning module monitors the continuous overrun state of the try-on conversion health index, triggers an inventory reverse early warning signal and sends a trigger instruction to the electronic price tag management module; and the execution response module is used for generating an allocation suggestion list containing cross-store allocation logic for the multiple early warning commodities. According to the method, the problem that a dynamic inventory optimization system with the try-on value as the core is difficult to support in a traditional mode is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of inventory management and relates to an intelligent system for collecting try-on behavior data and optimizing inventory. Background Art

[0002] The retail industry currently faces a technological gap between collecting data on try-on behavior and converting it into commercial value. Traditional inventory management systems rely on limited historical sales data and manual experience to determine replenishment cycles, making it difficult to accurately capture consumers' dynamic preferences during the actual try-on process. This leads to potential sales losses for high-try-on-volume products due to lagging display logic or irrational inventory distribution, creating an information gap between the value of try-on and inventory strategies.

[0003] Existing solutions often rely on offline questionnaires and static radio frequency identification (RFID) technology. Some systems estimate try-on popularity by counting fitting room usage or shelf retrieval times. Others leverage basic RFID technology to track clothing flow and use the signal strength of fixed-position sensors to determine product movement. While these approaches can establish a preliminary behavioral data framework in localized scenarios, their granularity is limited to discrete records of spatial movement.

[0004] In response to the above problems, traditional methods are difficult to support a dynamic inventory optimization system centered on the value of try-on, and there is an urgent need to break through the intelligent judgment mechanism of multimodal data fusion. Summary of the Invention

[0005] In order to solve the above problems, the present invention provides an intelligent system for collecting try-on behavior data and optimizing inventory.

[0006] The intelligent system for collecting trial behavior data and optimizing inventory includes:

[0007] The RF data acquisition module collects RF signal data of the try-on action in real time through an RF sensor array with adjustable mounting angles installed in the fitting room. This data is combined with the reflection signal change characteristics of the wireless RF tags pre-installed on the clothes to generate the original dataset of the try-on behavior.

[0008] The intensity level determination module is used to receive the switch status data of the fitting room door magnetic sensor and the clothing flow time data from the shelf to the fitting room, and conduct a linkage analysis on the original dataset of the try-on behavior to determine the intensity level of the try-on behavior;

[0009] The satisfaction detection module uses a pressure-sensing carpet to obtain the consumer's dwell time in front of the fitting mirror, calculates the proportion based on the duration threshold corresponding to the intensity level of the fitting behavior, triggers a high satisfaction event marker, and generates a fitting satisfaction index;

[0010] A health index generation module extracts the number of in-depth try-ons from the try-on behavior intensity level and weights it according to a preset weight coefficient. After adding the proportion of high satisfaction events, it performs an inverse proportional operation with the actual purchase volume to generate a try-on conversion health index.

[0011] A dynamic early warning module adjusts the industry benchmark threshold according to the product category identifier, monitors the continuous exceeding of the try-on conversion health index, triggers an inventory reverse early warning signal, and sends a trigger instruction to the electronic price tag management module;

[0012] The execution response module is used to adjust the display priority of electronic price tags through wireless communication links after receiving inventory reverse warning signals, and generate a transfer suggestion list including cross-store transfer logic for products with multiple warnings.

[0013] A further solution of the present invention generates a raw dataset of try-on behaviors, comprising the following steps:

[0014] The infrared ranging device obtains the consumer's body parameters and calculates the optimal coverage angle of the radio frequency sensor based on the preset human body structure theory calculation formula;

[0015] driving the radio frequency sensor array to rotate to the optimal coverage angle and continuously receiving a sequence of intensity changes of reflected signal fluctuations;

[0016] The peak-to-valley difference of the reflected signal is extracted from the intensity change sequence of the reflected signal fluctuation as the action amplitude, and the time interval between two reflected signal drops is used as the duration feature to generate the original dataset of the try-on behavior.

[0017] A further solution of the present invention is to drive the RF sensor array to rotate to the optimal coverage angle, comprising the following steps:

[0018] The infrared ranging device installed at the entrance of the fitting room obtains the consumer's shoulder width and height parameters, and calculates the optimal coverage angle of the RF sensor based on these parameters;

[0019] Using the geometric relationship between the height of the center point of the human torso and the half value of the shoulder width, the coverage angle formula is defined;

[0020] The pitch angle of the sensor component is adjusted in real time through a mechanical transmission mechanism so that the radio frequency signal covers the consumer's main movement areas.

[0021] A further solution of the present invention is to determine the intensity level of the try-on behavior, comprising the following steps:

[0022] The door magnetic sensor monitors the open and closed status of the test room door. When the door is detected to be closed, a timer is started to record the duration of the door closing.

[0023] The time it takes for clothes to be taken out of the shelf and into the fitting room is calculated by taking the difference between the reading time of the RFID tag when the clothes are removed from the shelf and the recognition time after entering the fitting room.

[0024] If the clothing flow time in the closed state exceeds the preset time threshold and the movement amplitude index does not reach the amplitude threshold, it is marked as a shallow try-on; if the difference in the numerical value of the radio frequency signal strength drop detected twice or more continuously during the closed state exceeds the preset critical value, it is marked as a deep try-on.

[0025] A further solution of the present invention, shallow and deep testing, comprises the following steps:

[0026] The priority of deep try-on is higher than that of shallow try-on. If both shallow try-on and deep try-on conditions are met in the same try-on cycle, the deep try-on will prevail and overwrite the original marking result. Behaviors that do not meet both shallow try-on and deep try-on conditions are classified as normal browsing.

[0027] A further solution of the present invention generates a try-on satisfaction index, comprising the following steps:

[0028] A pressure-sensing carpet is installed in the area between the fitting room exit and the fitting room mirror. When a consumer finishes trying on clothes and leaves the fitting room, the carpet's grid-shaped pressure sensor array detects their position and duration in front of the mirror, providing real-time information on their stay time.

[0029] When the dwell time exceeds the set ratio of the corresponding try-on intensity level, a high satisfaction event mark is triggered;

[0030] The pressure distribution and dwell time of consumers are collected through pressure-sensing carpets. The pressure distribution includes the distance between the pressure points of both feet and the offset of the center of gravity, which can be used to distinguish different consumers.

[0031] A composite parameter is constructed based on the ratio of the number of in-depth try-ons to the number of high-satisfaction events. The number of in-depth try-ons is multiplied by the in-depth try-on weight coefficient to obtain a weighted in-depth try-on quantitative value. The in-depth try-on quantitative value is added to the number of high-satisfaction events and divided by the actual number of purchases to obtain the final try-on satisfaction index.

[0032] A further solution of the present invention is to generate a depth test weight coefficient, comprising the following steps:

[0033] The depth of the try-on weight coefficient is determined by the contribution of the try-on behavior to sales;

[0034] A linear regression model is used to model historical data. The historical records of a single consumer's trial duration, movement amplitude, satisfaction index, and product sales are used as a single dataset. The single datasets of multiple consumers are combined into a training set to pre-train the linear regression model.

[0035] The model's input variables include try-on duration, movement amplitude, and satisfaction index, and the output variable is product sales. The calculation of the in-depth try-on weight coefficient satisfies the following formula:

[0036]

[0037] Among them, α is the depth test weight coefficient; n is the number of samples; T i The value of the number of times product i is tried on in depth; G i It represents the proportion of sales of product i to total sales.

[0038] A further solution of the present invention generates a try-on conversion health index, comprising the following steps:

[0039] Count the number of marked high-satisfaction events and the weighted in-depth try-on quantitative value, and convert it into a proportion of all try-on events for the current product. Add the number of high-satisfaction events and the weighted in-depth try-on quantitative value to obtain the total try-on value score;

[0040] The total score of the try-on value and the number of actual purchases of the same product obtained simultaneously are reversely proportionally calculated to generate an index value that reflects the degree of deviation between the product try-on value and sales performance, which is the try-on conversion health index.

[0041] A further solution of the present invention triggers an inventory reverse warning signal, comprising the following steps:

[0042] Extract the category identifier from the product master data and adjust the warning threshold based on the preset category sensitivity parameters. The stored benchmark threshold for try-on conversion of different categories of products is multiplied by the category correction coefficient.

[0043] The try-on conversion health index of the same product is calculated on a rolling basis every day. If the index value exceeds the adjusted warning threshold for multiple consecutive days, an inventory reverse warning signal is triggered;

[0044] The inventory reverse warning signal is an electronic warning sign automatically generated by the system, which contains the slow-moving product code, warning level, and recommended handling measures, and is then pushed to the store management terminal via an encrypted message.

[0045] A further solution of the present invention generates a transfer suggestion list including cross-store transfer logic, including the following steps:

[0046] After confirming that a product has triggered an alert, an encrypted instruction is sent to the electronic price tag on the shelf where the product is located via a wireless communication link;

[0047] After receiving the instruction, the electronic price tag increases the display exposure frequency of the corresponding product according to the preset priority rules, which is manifested by enlarging the font of the product name, adding a dynamic flashing logo, and expanding the display area of ​​the product description information;

[0048] For products that have triggered warnings more than once in a row, the product's try-on conversion health index in other stores will be further retrieved. If the try-on index of at least two other stores in the same city is lower than the adjusted warning threshold, a cross-store transfer recommendation list will be generated, including the recommended transfer-out store code, transfer-in store code, and recommended transfer quantity.

[0049] In summary, the present invention has the following beneficial technical effects:

[0050] 1. By using an adjustable RF sensor array to match the consumer's body shape in real time, combined with analysis of reflected signal fluctuations from clothing RFID tags, the system can accurately capture detailed characteristics such as the amplitude and duration of movements during the try-on process. The sensor array dynamically adjusts its angle based on shoulder width and height parameters acquired through infrared ranging, ensuring complete coverage of the movement trajectories of consumers of different body types. This solves the signal loss problem caused by fixed installations of traditional monitoring equipment and provides accurate data support for subsequent behavioral analysis.

[0051] 2. Based on a multi-dimensional correlation analysis of fitting room door status, clothing flow duration, and RF signal fluctuations, the system categorizes fitting behavior into three levels: general browsing, shallow fitting, and deep fitting. By setting scientific time thresholds and signal fluctuation thresholds, combined with judgment logic validated by historical data, it accurately identifies deep fitting behaviors with purchase potential. This hierarchical identification mechanism effectively distinguishes between mere interest and actual purchase intent, helping merchants develop targeted service strategies.

[0052] 3. Integrating in-depth correlation analysis of the number of try-ons, satisfaction indicators, and actual purchase volume, the Try-on Conversion Health Index provides real-time insights into the degree of alignment between product try-on popularity and sales results. This index utilizes weighted calculations and inverse proportionality, dynamically adjusting warning thresholds based on the sensitivity of different product categories. This allows for rapid identification of high-try-on, low-conversion products, providing a quantitative basis for inventory optimization decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. The drawings are used to provide a further understanding of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0054] Figure 1 It is a schematic diagram of the framework in the embodiment of the present application.

[0055] Figure 2 It is a schematic diagram of the process flow in the embodiment of this application.

[0056] Figure 3 It is a curve chart that discloses the inverse relationship between the try-on conversion health index and the actual purchase amount in the embodiment of the present application. DETAILED DESCRIPTION

[0057] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] The following is combined with Figure 1-Figure 3 The preferred embodiments of the present invention are described in detail.

[0059] Refer to the attached Figure 1-2 As shown, the present invention proposes an intelligent system for collecting try-on behavior data and optimizing inventory, including the following modules:

[0060] The RF data acquisition module collects RF signal data of the try-on action in real time through an RF sensor array with adjustable mounting angles installed in the fitting room. This data is combined with the reflection signal change characteristics of the wireless RF tags pre-installed on the clothes to generate the original dataset of the try-on behavior.

[0061] The intensity level determination module is used to receive the switch status data of the fitting room door magnetic sensor and the clothing flow time data from the shelf to the fitting room, and conduct a linkage analysis on the original dataset of the try-on behavior to determine the intensity level of the try-on behavior;

[0062] The satisfaction detection module uses a pressure-sensing carpet to obtain the consumer's dwell time in front of the fitting mirror, calculates the proportion based on the duration threshold corresponding to the intensity level of the fitting behavior, triggers a high satisfaction event marker, and generates a fitting satisfaction index;

[0063] A health index generation module extracts the number of in-depth try-ons from the try-on behavior intensity level and weights it according to a preset weight coefficient. After adding the proportion of high satisfaction events, it performs an inverse proportional operation with the actual purchase volume to generate a try-on conversion health index.

[0064] A dynamic early warning module adjusts the industry benchmark threshold according to the product category identifier, monitors the continuous exceeding of the try-on conversion health index, triggers an inventory reverse early warning signal, and sends a trigger instruction to the electronic price tag management module;

[0065] The execution response module is used to adjust the display priority of electronic price tags through wireless communication links after receiving inventory reverse warning signals, and generate a transfer suggestion list including cross-store transfer logic for products with multiple warnings.

[0066] In one embodiment of the present invention, generating a raw dataset of try-on behaviors includes the following steps:

[0067] An RF sensor array with adjustable mounting angle is deployed inside the fitting room. When a consumer enters the fitting room carrying clothing with a wireless RF tag, the RF sensor array triggers angle adjustment based on the consumer's body data to ensure that the RF signal covers the consumer's main movement area.

[0068] Specifically, the consumer's shoulder width and height parameters are obtained through an infrared ranging device installed at the entrance of the fitting room. Based on these parameters, the optimal coverage angle of the radio frequency sensor is calculated to meet the following formula:

[0069]

[0070] Wherein, θ represents the optimal coverage angle of the RF sensor; h is the installation height of the RF sensor array; H is the height of the consumer; W is the consumer's shoulder width; 0.4H is the height of the center point of the consumer's torso, which is approximately the position of the consumer's navel and is obtained based on human anatomy theory; 0.5W is half the consumer's shoulder width; (180° / π) is the unit conversion factor used to convert radians to degrees; arctan() belongs to the inverse tangent function, and the result of the function is in radians by default. It needs to be multiplied by (180° / π) for unit conversion to convert radians to degrees.

[0071] Based on the calculated results, the RF sensor array rotates to the set optimal coverage angle, ensuring that the effective coverage of the RF signal matches the consumer's body area. When a consumer enters a fitting room wearing clothing with a RFID tag, the RF sensor array continuously transmits detection signals and receives a sequence of reflected signal fluctuations from the RFID tag. By analyzing the amplitude and time interval of the reflected signal intensity, it extracts movement amplitude and duration indicators, generating a raw dataset of movement amplitude and duration for the try-on behavior.

[0072] Among them, the action amplitude and duration are defined by quantifying the peak-to-valley difference of the reflected signal intensity change. The signal peak-to-valley difference is extracted from the signal fluctuation sequence reflected by the wireless radio frequency tag as the action amplitude, and the time interval between two signal drops reflected by the wireless radio frequency tag is used as the duration feature.

[0073] A wireless radio frequency tag is a passive radio frequency identification tag attached to a specific location on clothing (such as a collar tag or waistline). The intensity of its reflected signal changes regularly with the relative motion between the tag and the radio frequency sensor.

[0074] The RF sensor array with adjustable mounting angle is a sensor group composed of multiple RF sensors installed in a non-fixed manner on the side walls of the fitting room. The orientation angle of the sensor's emitting surface is changed through a mechanical transmission mechanism to adapt to the motion capture needs of consumers with different body shapes.

[0075] For example, when a consumer with a height of 175 cm and a shoulder width of 45 cm enters the fitting room, the infrared ranging device measures her body parameters, and the radio frequency sensor array rotates to the set optimal coverage angle based on the solution results. According to the solution results, the third radio frequency sensor is instructed to tilt upward 12° and the fifth sensor downward 8° to cover her torso activity area.

[0076] When a consumer tries on a sweatshirt with a wireless radio frequency tag and raises his arm, the reflected signal strength drops from the baseline value of 120dBμV to 85dBμV and lasts for 1.2 seconds. The amplitude of the movement is judged to be 35dBμV and the duration is 1.2 seconds. It is recorded as a complete upper limb extension behavior and stored in the original dataset of the try-on behavior.

[0077] In one embodiment of the present invention, determining the intensity level of the try-on behavior includes the following steps:

[0078] The original dataset of the try-on behavior generated by the radio frequency data acquisition module is linked with the fitting room switch status data and the duration data of the clothes being taken out to analyze and determine the intensity level of the try-on behavior.

[0079] Specifically, the door magnetic sensor monitors the open and closed status of the fitting room door. When the door is detected to be closed, a timer is started to record the duration of the door closing. The difference between the reading time of the wireless radio frequency tag on the clothing when it is removed from the shelf and the recognition time after entering the fitting room is used to calculate the flow time of the clothing from being taken out to the fitting room. The two time data are combined with the movement amplitude in the radio frequency data acquisition module to perform the try-on intensity determination:

[0080] If the clothing flow time in the closed state exceeds the preset time threshold and the movement amplitude index does not reach the amplitude threshold, it is marked as a shallow try-on;

[0081] If the difference between two or more consecutive drops in RF signal strength during door closing exceeds a preset threshold, it will be marked as a deep penetration test; for example, the difference between two consecutive drops in signal strength exceeds 30% of the baseline value and the time interval is less than 2 seconds;

[0082] Deep try-on has a higher priority than shallow try-on. If both conditions are met at the same time in the same try-on cycle, the deep try-on will prevail and overwrite the original marking result. Behaviors that do not meet the above conditions are classified as normal browsing.

[0083] Among them, the critical value is obtained through historical data analysis, for example, the average value of the trial fluctuation amplitude corresponding to the successful purchase of goods in the past 30 days plus 1.5 times its standard deviation is selected.

[0084] The time threshold and amplitude threshold are derived through historical data analysis. For example, through historical data analysis, a sampling of try-on data from 100 stores in the past 6 months found that more than 70% of shallow try-on behaviors (such as simply taking clothes without actually wearing them) corresponded to a shelf-to-fitting room flow time of 28-32 seconds, and the median of 30 seconds was taken as the threshold.

[0085] The fitting room switch status data represents the binary signal collected by the door magnetic sensor installed between the fitting room door frame and door panel. The door opening and closing status is determined by the change in magnetic field strength (such as open is 0 and closed is 1), which is used to define the spatial boundary of the try-on behavior.

[0086] The duration data of clothing being taken out is based on the time difference positioning technology of the RFID reader in the shelf area and the fitting room. The total time taken for the wireless RFID tag to be taken off the shelf and enter the fitting room is calculated, and the clothing flow process is recorded with second-level accuracy.

[0087] For example, after the consumer closes the fitting room door, the clothing flow is recorded as 42 seconds, exceeding the expert-preset time threshold of 30 seconds. At the same time, the movement amplitude indicator output by the RF data acquisition module shows a maximum fluctuation of 28dBμV, which is lower than the expert-preset amplitude threshold of 40dBμV. In this case, it can be judged as a shallow try-on. If the RF signal strength is detected twice within the next 10 seconds, the calculated amplitude difference is 43dBμV and the interval is only 1.5 seconds. It is superimposed and judged as a deep try-on, triggering priority data recording.

[0088] In one embodiment of the present invention, generating a try-on satisfaction index includes the following steps:

[0089] Based on the intensity level of the trying-on behavior determined by the intensity level determination module and combined with the analysis of consumers' retention behavior in front of the fitting mirror, a trying-on satisfaction index is generated.

[0090] Specifically, a pressure-sensing carpet is installed in the area between the fitting room exit and the fitting room mirror. When a consumer finishes trying on clothes and leaves the fitting room, the grid-shaped pressure sensor array of the pressure-sensing carpet detects their position and duration in front of the mirror. The dwell time data is acquired in real time and correlated with the intensity level of the try-on behavior (normal browsing, shallow try-on, deep try-on) output by the intensity level determination module:

[0091] If the dwell time exceeds 50% of the duration of the corresponding try-on intensity level, it is considered a high satisfaction event, otherwise it is marked as a low satisfaction event. Finally, the proportion of high satisfaction events and the try-on intensity level are weighted and calculated to generate the try-on satisfaction index.

[0092] Among them, the dwell time data refers to the time difference from the first time the pressure-sensing carpet detects foot pressure to the last time the pressure disappears, accurate to 0.1 second, and is used to quantify the consumer's attention to the trial results.

[0093] A pressure-sensing carpet is a detection device consisting of a flexible circuit board and distributed pressure sensors installed on the floor. It records dwell time by sensing changes in the body's gravity distribution. Its output signal is a binary on-off sequence of contacts. For multiple consumers, pressure distribution patterns (such as the distance between the pressure points of both feet and the offset of the center of gravity) are used to distinguish between different consumers, ensuring that dwell time data corresponds to a single individual.

[0094] The try-on satisfaction index is a composite parameter constructed based on the ratio of the number of in-depth try-ons to the number of high-satisfaction events. The calculation rule is: the number of in-depth try-ons is multiplied by the in-depth try-on weight coefficient to obtain a weighted in-depth try-on quantitative value. The in-depth try-on quantitative value is added to the number of high-satisfaction events, and the result is divided by the actual number of purchases to obtain the final try-on satisfaction index. If the actual number of purchases is zero, the maximum value is assigned +1 or the product is ignored.

[0095] The deep try-on weight coefficient is determined by the contribution of try-on behavior to sales. A linear regression model is used to model historical data. The historically recorded data of individual consumers' try-on duration, movement amplitude, satisfaction index, and product sales are used as a single dataset. Multiple single datasets of consumers are combined into a training set to pre-train the linear regression model. The model's input variables include try-on duration, movement amplitude, and satisfaction index, and the output variable is product sales.

[0096] The calculation of the depth test weight coefficient satisfies the following formula:

[0097]

[0098] Among them, α is the depth test weight coefficient; n is the number of samples; T i The value representing the number of times product i is tried on in depth, obtained through the strength level determination module; G i It represents the proportion of sales of product i to total sales, satisfying the following formula:

[0099]

[0100] Among them, S i represents the sales volume of product i, represents the total sales of all products, and m represents the total number of products.

[0101] For example, a consumer is judged by the intensity level judgment module to be in-depth try-on, and when the try-on duration is 180 seconds, the time he stays in front of the fitting mirror is recorded as 95 seconds. Since 95 seconds exceeds 50% of 180 seconds, it satisfies the requirement of a stay time exceeding 50% of the duration of the corresponding try-on behavior intensity level and is marked as a high satisfaction event. If the consumer completes 3 in-depth try-ons on the same day and triggers 2 high satisfaction judgments, and the actual purchase quantity is 1 piece, the in-depth try-on weight coefficient is 0.6, and the try-on satisfaction index is calculated to be 3.8, which effectively distinguishes between "ineffective try-ons" with a large number of try-ons but low satisfaction and "effective try-ons" with high conversion.

[0102] In one embodiment of the present invention, generating a try-on conversion health index includes the following steps:

[0103] Based on the original data of try-on behavior output by the radio frequency data acquisition module, the try-on behavior intensity level output by the intensity level judgment module, and the try-on satisfaction index output by the satisfaction detection module, a try-on conversion health index is established to measure the consistency between the product try-on value and sales performance.

[0104] Specifically, first obtain the weighted depth try-on quantitative value in the satisfaction detection module, and at the same time count the number of high satisfaction events marked in the satisfaction detection module, and convert it into a proportion value of all try-on events of the current product. Add the above two values ​​to obtain the total try-on value score, and then perform an inverse proportional operation on the total try-on value score and the actual number of purchases of the same product obtained simultaneously by the health index generation module to generate an index value reflecting the degree of deviation between the product try-on value and sales performance.

[0105] If the calculated result exceeds the industry benchmark threshold, it will be determined as a high-try-on-low-conversion risk product and a warning signal will be generated.

[0106] The industry benchmark threshold is set based on industry standards and expert experience. The percentage value, representing the consumer's subjective approval, is calculated by dividing the number of high-satisfaction events marked by the satisfaction detection module by the total number of times the same product has been tried on.

[0107] Refer to the attached Figure 3 The try-on conversion health index is a quantitative indicator that reflects the degree of match between consumers' attention to the product and the actual purchase volume. An increase in its value indicates that the try-on behavior has not been effectively converted into sales (abnormal state).

[0108] The reverse proportional operation is a division operation with the total score of the try-on value as the numerator and the actual number of purchases as the denominator. The core logic is that the lower the purchase volume, the higher the index value will be calculated for the same total score of the try-on value.

[0109] For example, for a particular pair of jeans, the intensity level determination module is set to count 20 in-depth try-ons on that day, the satisfaction detection module records 15 high satisfaction events, and the health index generation module obtains 5 actual purchases. Assuming a weighting coefficient of 0.6, the in-depth try-on quantification value is 20 × 0.6 = 12, the high satisfaction ratio is 15 / 20 = 0.75, and the total try-on value score is 12 + 0.75 = 12.75.

[0110] The inverse proportional calculation yields a try-on conversion health index of 12.75 / 5 = 2.55. Based on industry standards and expert experience, the industry benchmark threshold is set at 2.0, identifying products with a high try-on-low conversion risk and generating a warning signal.

[0111] In one embodiment of the present invention, triggering an inventory reverse warning signal includes the following steps:

[0112] Based on the try-on conversion health index generated by the health index generation module and product category characteristics, inventory risks are dynamically judged and early warnings are triggered.

[0113] Specifically, the category identifier (women's or men's) is extracted from the product master data, and the warning threshold is adjusted based on the preset category sensitivity parameters: the stored try-on conversion baseline threshold for each category (for example, a universal try-on conversion baseline threshold of 2.0) is multiplied by the category correction factor. For example, the correction factor for women's wear is set to 0.8, and the correction factor for men's wear is set to 1.0, meaning that the women's wear threshold is 20% lower than the men's wear threshold to reflect the different sensitivity of try-on conversions.

[0114] The try-on conversion health index of the same product is calculated on a rolling basis every day. If the index value exceeds the adjusted warning threshold for three consecutive days, an inventory reverse warning signal is triggered.

[0115] Among them, the inventory reverse warning signal is an electronic warning sign automatically generated by the system, which contains the slow-moving product code, warning level, and recommended handling measures, and is then pushed to the store management terminal via encrypted messages.

[0116] The category sensitivity parameter reflects the sensitivity of different product categories to the differences in try-on behavior and sales conversion. It is determined by analyzing the standard deviation ratio of the try-on and purchase conversion rates of women's clothing and men's clothing in historical data. Since try-on behavior has a more significant impact on purchasing decisions in the women's clothing category, the correction coefficient for women's clothing is set lower than that for men's clothing.

[0117] The benchmark threshold for try-on conversions is based on historical data statistics, the distribution of product try-on conversion health indices over the past 6-12 months, and the median conversion efficiency of similar products in the industry. Using the percentile method, we identify the top 20% of products with high try-on and low conversion risk as the benchmark threshold.

[0118] For example, the base threshold for a women's dress is 2.0, and the women's clothing category correction factor is set to 0.8. The adjusted warning threshold is calculated to be 2.0 × 0.8 = 1.6. If the try-on conversion health index is 1.8, 1.9, and 2.1 for three consecutive days, exceeding the adjusted warning threshold, the system generates an inventory reverse warning signal. For the same period, the base threshold for men's jackets is 2.0, and the adjusted warning threshold is calculated to be 2.0 × 1 = 2.0.

[0119] In one embodiment of the present invention, generating a transfer suggestion list including cross-store transfer logic includes the following steps:

[0120] Based on the inventory reverse warning signal triggered by the dynamic warning module, display priority instructions are sent to the electronic price tag through the wireless communication link, automatically increasing the exposure frequency of the corresponding products on the shelf, and at the same time generating a cross-store transfer recommendation list for products that have received more than three warnings.

[0121] Specifically, after the system confirms that a certain product has triggered an early warning, it sends an encrypted instruction to the electronic price tag on the shelf where the product is located through a wireless communication link; after receiving the instruction, the electronic price tag increases the display exposure frequency of the corresponding product according to the preset priority rules, which is manifested by enlarging the font of the product name, adding a dynamic flashing logo, and expanding the display area of ​​the product description information.

[0122] For products that have triggered warnings for more than three consecutive times, further data on the product's try-on conversion health index in other stores will be retrieved. If the try-on index of at least two other stores in the same city is lower than the adjusted warning threshold, a cross-store transfer recommendation list will be generated, including the recommended transfer-out store code, transfer-in store code, and recommended transfer quantity.

[0123] For products that trigger early warnings more than three times in a row, the inventory, try-on index, and geographic location data of all stores in the same city and surrounding cities (such as within a radius of 200 kilometers) are pulled; stores with excessive allocation costs are eliminated. For example, the allocation cost is greater than the gross profit per item × the number of allocations × 30%. The 30% is combined with historical data mining to avoid loss-making allocations.

[0124] The system makes cross-store transfer decisions using the following judgment chain:

[0125] 1. Check whether the number of consecutive warnings for the current product in this store has reached three;

[0126] 2. Obtain the try-on conversion health index of the same product in other stores in the same city;

[0127] 3. Screen target stores that meet the requirements. The transfer-in store index is based on historical data mining of successful transfers. For example, the transfer-in store index ≤ the transfer-out store index × 0.8. The transfer-in store inventory is based on supply chain safety parameters to prevent secondary backlogs caused by excessive transfers. For example, the transfer-in store inventory ≤ the safety stock × 0.5;

[0128] 4. The allocation quantity is determined by the minimum value of the target store's demand gap and the store's available quantity, while ensuring that the store reserves safety stock. The safety stock is calculated as the average daily sales volume of the past 30 days multiplied by the 7-day stocking cycle coefficient.

[0129] Among them, the wireless communication link is a star-type networking communication channel based on LoRa technology, which uses the 433MHz frequency band to transmit electronic price tag control instructions, and supports the simultaneous update of display content of up to 50 electronic price tags in the same shelf area.

[0130] The display exposure frequency of the product is shortened to 60% of that of ordinary products by rewriting the display register value of the electronic price tag. At the same time, a red border is added to the top of the price tag to enhance its visual appeal.

[0131] A cross-store transfer suggestion list includes a formatted data table showing the product SKU code, inventory balance in the transfer store, the target store's try-on conversion health index for similar products, and estimated shipping costs.

[0132] For example, a women's sweater triggered an inventory reverse warning for three consecutive days in store A (the try-on conversion health index was 1.7, 1.8, and 1.9 respectively). The system sent an exposure improvement instruction to the electronic price tag on the shelf where the product was located. The data showed that its weekly exposure times increased from 120 times to 210 times, and the try-on conversion rate increased from 8% to 15%. For products that triggered warnings for more than three consecutive times, the inventory, try-on index, and geographic location data of all stores in the same city and surrounding cities (such as within a radius of 200 kilometers) were pulled. Store D, which exceeded the allocation cost limit, was eliminated. The try-on conversion health index of the product in store B was 1.2, which was lower than the warning threshold of 1.6 adjusted by the dynamic warning module, and there were only 5 pieces left in stock. The try-on conversion health index of the product in store C was 1.3, and there were only 12 pieces left in stock.

[0133] A cross-store transfer suggestion list is generated: Store A transfers 15 sweaters to Store B (10) and Store C (5). Within three days of the transfer, sales of the sweaters at Stores B and C increase by 8 and 6, respectively, eliminating the risk of slow sales.

[0134] After generating a list of cross-store transfer recommendations, manual verification instructions are triggered through a visual interactive interface to conduct multi-dimensional cross-validation of the original data of try-on behavior, the try-on conversion health index, and transfer recommendations to ensure the reliability of inventory optimization decisions.

[0135] Specifically, the raw data on try-on behavior, including movement amplitude and duration indicators, the determined intensity level of the try-on behavior, and the try-on conversion health index calculated from the generated try-on satisfaction index, are integrated into an interactive verification data set in time series and transmitted via an encrypted network to the audit module of the management terminal. Auditors can use a visual interactive interface to drill down to the details of the try-on event, including radio frequency signal fluctuation curves, heat maps of the pressure-sensitive carpet's dwell time, and comparative data on the historical allocation of products in the same category.

[0136] If the classification of try-on behavior is found to be abnormal or the allocation suggestion is inconsistent with the sales trend, the reviewer can manually correct the try-on intensity level mark or adjust the allocation quantity. The corrected data will be transmitted back to the central database after digital signature authentication, triggering the update of the allocation instruction.

[0137] The visual interactive interface is a multi-layered dynamic display panel rendered using WebGL technology, which supports the simultaneous display of the timeline of the try-on behavior and the changes in radio frequency signal strength. Digital signature authentication uses an asymmetric encryption algorithm to identify and tamper-proof the results of manual corrections, ensuring reliable data traceability.

[0138] For example, the system generated a transfer recommendation list for a specific women's jacket, suggesting that 15 jackets be transferred from Store A to Store B. Auditors, using a visual interface, discovered that the try-on conversion health index for similar jackets at Store B over the previous week had approached the threshold. Further review of the raw try-on data revealed that the RF signal fluctuation amplitudes corresponding to three deep try-ons were 15% below the historical average, possibly due to misjudgment caused by sensor angle offset. Auditors manually downgraded these three try-ons to shallow ones. After the system recalculated the try-on conversion health index, Store B's index rose above the threshold, automatically canceling the transfer recommendation and triggering a sensor calibration work order. This correction result is digitally signed and synchronized to all connected systems, ensuring an auditable closed-loop decision chain.

[0139] In the event of a RF sensor array failure or special scenario, the trial wear behavior record is manually entered and converted into standard RF signal simulation data.

[0140] Specifically, if the RF sensor array detects three consecutive self-test failures, manual data collection is automatically triggered, and a try-on behavior record form template is sent to the management terminal. Based on the individual's actual movements, the staff fills in the record form with the start and end time, movement type, and amplitude level. A handheld body scanner is used to obtain the individual's shoulder width and height data. The record form data is parsed by the data conversion engine, which generates a simulated RF signal fluctuation sequence based on the mapping between RF signal strength and movement amplitude. This sequence is then injected into the original try-on behavior data stream.

[0141] The manually entered try-on behavior record form is a structured spreadsheet containing the wearer's body parameters, fields describing the try-on actions, and a timestamp. A handheld body scanner is a mobile terminal device equipped with an infrared ranging module that measures shoulder width and height by emitting invisible light. The data conversion engine is a conversion program based on a table comparing movement type and radio frequency signal strength. It converts the manually recorded movement amplitude levels into corresponding radio frequency signal strength differences and generates a continuous signal fluctuation curve at preset time intervals.

[0142] For example, when a store's fitting room sensor array lost power due to renovations, staff observed a consumer raising their arms twice while trying on a men's jacket. Staff then selected the "upper limb extension" action type in the record sheet and marked the amplitude level as level two. A handheld body scanner measured the wearer's shoulder width as 48cm and height as 178cm. The data conversion engine, based on preset rules, mapped the level two amplitude to a 28dBμV drop in RF signal strength and generated two signal fluctuations lasting 1.5 seconds. This portion of analog data, after being processed by the RF data acquisition module, triggers the in-depth try-on determination of the intensity level determination module, and ultimately participates in the health index generation module's try-on conversion health index calculation and generates inventory warning signals, achieving a seamless connection between manual data and automated processes.

[0143] It should be noted that the formulas described above can translate physical quantities of different attributes into unitless standard values ​​or superimposable parameters of the same dimension through the principle of dimensional consistency and mathematical standardization (e.g., normalization, dimensionless parameter conversion, or unit system unification). This eliminates the interference of different dimensions on the operational logic, allowing the formulas to retain the distribution characteristics of the original data while maintaining mathematical rationality and adaptability to objective laws. The above are merely exemplary embodiments of the present invention and are not intended to limit the scope of the present invention.

[0144] The modules can be implemented in whole or in part through software, hardware, or a combination thereof, supporting hardware embedded in or independent of a processor in a computer device, and also supporting software stored in a memory in a computer device, so that the processor can call and execute operations corresponding to the modules.

[0145] It should be noted that the human body information (including but not limited to human device information and personal information, etc.) and data (including but not limited to data used for analysis, stored data and displayed data, etc.) involved in the present invention are all information and data authorized by the human body or fully authorized by all parties. The collection, use and processing of relevant data require relevant legal standards.

[0146] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention, and should all be included in the scope of protection of the present invention.

Claims

1. An intelligent system for collecting data on try-on behavior and optimizing inventory, characterized by: include: The RF data acquisition module collects RF signal data of the try-on action in real time through an RF sensor array with adjustable mounting angles installed in the fitting room. This data is combined with the reflection signal change characteristics of the wireless RF tags pre-installed on the clothes to generate the original dataset of the try-on behavior. The intensity level determination module is used to receive the switch status data of the fitting room door magnetic sensor and the clothing flow time data from the shelf to the fitting room, and conduct a linkage analysis on the original dataset of the try-on behavior to determine the intensity level of the try-on behavior; The satisfaction detection module uses a pressure-sensing carpet to obtain the consumer's dwell time in front of the fitting mirror, calculates the proportion based on the duration threshold corresponding to the intensity level of the fitting behavior, triggers a high satisfaction event marker, and generates a fitting satisfaction index; A health index generation module extracts the number of in-depth try-ons from the try-on behavior intensity level and weights it according to a preset weight coefficient. After adding the proportion of high satisfaction events, it performs an inverse proportional operation with the actual purchase volume to generate a try-on conversion health index. A dynamic early warning module adjusts the industry benchmark threshold according to the product category identifier, monitors the continuous exceeding of the try-on conversion health index, triggers an inventory reverse early warning signal, and sends a trigger instruction to the electronic price tag management module; The execution response module is used to adjust the display priority of electronic price tags through wireless communication links after receiving inventory reverse warning signals, and generate a transfer suggestion list including cross-store transfer logic for products with multiple warnings.

2. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 1, characterized in that: Generating the original dataset of try-on behavior includes the following steps: The infrared ranging device obtains the consumer's body parameters and calculates the optimal coverage angle of the radio frequency sensor based on the preset human body structure theory calculation formula; driving the radio frequency sensor array to rotate to the optimal coverage angle and continuously receiving a sequence of intensity changes of reflected signal fluctuations; The peak-to-valley difference of the reflected signal is extracted from the intensity change sequence of the reflected signal fluctuation as the action amplitude, and the time interval between two reflected signal drops is used as the duration feature to generate the original dataset of the try-on behavior.

3. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 2, characterized in that: Driving the radio frequency sensor array to rotate to the optimal coverage angle includes the following steps: The infrared ranging device installed at the entrance of the fitting room obtains the consumer's shoulder width and height parameters, and calculates the optimal coverage angle of the RF sensor based on these parameters; Using the geometric relationship between the height of the center point of the human torso and the half value of the shoulder width, the coverage angle formula is defined; The pitch angle of the sensor component is adjusted in real time through a mechanical transmission mechanism so that the radio frequency signal covers the consumer's main movement areas.

4. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 1, characterized in that: Determining the intensity level of the try-on behavior includes the following steps: The door magnetic sensor monitors the open and closed status of the test room door. When the door is detected to be closed, a timer is started to record the duration of the door closing. The time it takes for clothes to be taken out of the shelf and into the fitting room is calculated by taking the difference between the reading time of the RFID tag when the clothes are removed from the shelf and the recognition time after entering the fitting room. If the clothing flow time in the closed state exceeds the preset time threshold and the movement amplitude index does not reach the amplitude threshold, it is marked as a shallow try-on; if the difference in the numerical value of the radio frequency signal strength drop detected twice or more continuously during the closed state exceeds the preset critical value, it is marked as a deep try-on.

5. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 4 is characterized in that: Shallow and deep fitting include the following steps: The priority of deep try-on is higher than that of shallow try-on. If both shallow try-on and deep try-on conditions are met in the same try-on cycle, the deep try-on will prevail and overwrite the original marking result. Behaviors that do not meet both shallow try-on and deep try-on conditions are classified as normal browsing.

6. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 4, characterized in that: Generating a try-on satisfaction index includes the following steps: A pressure-sensing carpet is installed in the area between the fitting room exit and the fitting room mirror. When a consumer finishes trying on clothes and leaves the fitting room, the carpet's grid-shaped pressure sensor array detects their position and duration in front of the mirror, providing real-time information on their stay time. When the dwell time exceeds the set ratio of the corresponding try-on intensity level, a high satisfaction event mark is triggered; The pressure distribution and dwell time of consumers are collected through pressure-sensing carpets. The pressure distribution includes the distance between the pressure points of both feet and the offset of the center of gravity, which can be used to distinguish different consumers. A composite parameter is constructed based on the ratio of the number of in-depth try-ons to the number of high-satisfaction events. The number of in-depth try-ons is multiplied by the in-depth try-on weight coefficient to obtain a weighted in-depth try-on quantitative value. The in-depth try-on quantitative value is added to the number of high-satisfaction events and divided by the actual number of purchases to obtain the final try-on satisfaction index.

7. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 6, characterized in that: The generation of the depth test weight coefficient includes the following steps: The depth of the try-on weight coefficient is determined by the contribution of the try-on behavior to sales; A linear regression model is used to model historical data. The historical records of a single consumer's trial duration, movement amplitude, satisfaction index, and product sales are used as a single dataset. The single datasets of multiple consumers are combined into a training set to pre-train the linear regression model. The model's input variables include try-on duration, movement amplitude, and satisfaction index, and the output variable is product sales. The calculation of the in-depth try-on weight coefficient satisfies the following formula: Among them, α is the depth test weight coefficient; n is the number of samples; T i The value of the number of times product i is tried on in depth; G i It represents the proportion of sales of product i to total sales.

8. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 6, characterized in that: Generating a try-on conversion health index includes the following steps: Count the number of marked high-satisfaction events and the weighted in-depth try-on quantitative value, and convert it into a proportion of all try-on events for the current product. Add the number of high-satisfaction events and the weighted in-depth try-on quantitative value to obtain the total try-on value score; The total score of the try-on value and the number of actual purchases of the same product obtained simultaneously are reversely proportionally calculated to generate an index value that reflects the degree of deviation between the product try-on value and sales performance, which is the try-on conversion health index.

9. The intelligent system for collecting try-on behavior data and optimizing inventory according to claim 8, characterized in that: Triggering an inventory reverse warning signal includes the following steps: Extract the category identifier from the product master data and adjust the warning threshold based on the preset category sensitivity parameters. The stored benchmark threshold for try-on conversion of different categories of products is multiplied by the category correction coefficient. The try-on conversion health index of the same product is calculated on a rolling basis every day. If the index value exceeds the adjusted warning threshold for multiple consecutive days, an inventory reverse warning signal is triggered; The inventory reverse warning signal is an electronic warning sign automatically generated by the system, which contains the slow-moving product code, warning level, and recommended handling measures, and is then pushed to the store management terminal via an encrypted message.

10. The intelligent system for collecting data on try-on behavior and optimizing inventory according to claim 9, characterized in that: Generating a transfer suggestion list that includes cross-store transfer logic involves the following steps: After confirming that a product has triggered an alert, an encrypted instruction is sent to the electronic price tag on the shelf where the product is located via a wireless communication link; After receiving the instruction, the electronic price tag increases the display exposure frequency of the corresponding product according to the preset priority rules, which is manifested by enlarging the font of the product name, adding a dynamic flashing logo, and expanding the display area of ​​the product description information; For products that have triggered warnings more than once in a row, the product's try-on conversion health index in other stores will be further retrieved. If the try-on index of at least two other stores in the same city is lower than the adjusted warning threshold, a cross-store transfer recommendation list will be generated, including the recommended transfer-out store code, transfer-in store code, and recommended transfer quantity.

Images