Clock inventory management system based on Internet of Things

Abstract

The invention discloses a clock inventory management system based on the Internet of Things. The clock inventory management system comprises a sensing module, a transmission module, a processing module, a control module and a decision module. The sensing module is used for collecting user terminal data and clock inventory data based on the Internet of Things. The transmission module is used for performing remote data transmission by using an MQTT protocol, encrypting and deploying a local server through HTTPS, and automatically identifying and preprocessing abnormal data; the processing module is used for performing preprocessing and feature extraction on the data sensed by the sensing module based on the Internet of Things; the control module is used for classifying the data acquired by the sensing module, performing hierarchical prediction based on a clock demand prediction model and adjusting the inventory quantity; and the decision module is used for setting a prediction board, adjusting safety inventory parameters, predicting cost change in real time and outputting a cost-service balance curve.

Description

Technical Field

[0001] The present invention relates to the technical field of inventory management, and in particular to a watch inventory management system based on the Internet of Things. Background Art

[0002] In recent years, the watch industry has been undergoing a systemic transformation in its inventory management model in response to consumer trends and global competition. Traditional, manually-reliant inventory management methods, plagued by information lags and inefficient resource allocation, are ill-suited to the demands of high-variety, small-batch production and the complex collaborative requirements of cross-border supply chains. The development of IoT technology has provided a breakthrough solution for the industry. By digitally reshaping inventory management logic and building a global awareness network, it enables real-time monitoring and dynamic optimization of inventory status. Leveraging the collaboration of smart devices and cloud platforms, companies can gain real-time visibility into inventory quantities, movement paths, and environmental parameters, improving inventory transparency. Data modeling enables demand forecasting and intelligent replenishment, reducing the risk of inventory overhangs and stock-outs. It also promotes the digitization of the entire process, from raw material procurement to finished product warehousing, effectively shortening product delivery cycles by dynamically adjusting production plans and inventory strategies. Despite the promising application prospects of this technology, the industry still faces challenges such as limited applicability, high investment costs, and inadequate system compatibility.

[0003] At present, the Chinese invention patent with application number CN202310922208.8 discloses an inventory management and analysis system based on the industrial Internet of Things, including a visual unit, a user terminal, a central control module, a commodity reserve unit, a data feedback module, a quality inspection module and an online delivery unit. The visual unit and the visual unit are bidirectionally electrically connected; the user terminal and the central control module are connected via Ethernet signals, and the central control module and the commodity reserve unit are electrically connected; the central control module and the data feedback module are electrically connected, the data feedback module and the user terminal are signal-connected, the data feedback module and the quality inspection module are electrically connected, the quality inspection module and the online delivery unit are electrically connected, and the online delivery unit and the user terminal are signal-connected. This system can greatly improve the authenticity of the factory areas included, increase customers' favorability towards commodity manufacturers, avoid delays, wrong shipments and missed shipments of commodities within the delivery time, and reduce the risk of order destruction. However, traditional inventory management systems in existing technologies mostly focus on basic data such as inventory quantity and location, and have the problem of "only managing goods but not people" being out of touch with demand. At the same time, they do not control inventory quality and use single-dimensional prediction. The prediction is complex and the output data attributes are single. In addition, they are unable to flexibly and dynamically predict watch inventory in different directions based on hierarchical indexes. Summary of the Invention

[0004] The technical problem solved by the present invention is that traditional inventory management systems in the existing technology focus more on basic data such as inventory quantity and location, and have the problem of "only managing goods but not people" being out of touch with demand. At the same time, they do not control inventory quality and use single-dimensional prediction, which is complex and has a single output data attribute. In addition, they are unable to flexibly and dynamically predict watch inventory in different directions based on hierarchical indexes.

[0005] To solve the above technical problems, the present invention provides the following technical solution: a watch inventory management system based on the Internet of Things, comprising a sensing module, a transmission module, a processing module, a control module, and a decision module: The sensing module is used to collect user terminal data and watch inventory data based on the Internet of Things; The transmission module is used to use the MQTT protocol for remote data transmission, encrypt user privacy and supply chain data through HTTPS, deploy the local server of the watch inventory management network system, and automatically identify and pre-process abnormal data; The processing module is used to preprocess and extract features of the data perceived by the perception module based on the Internet of Things. The preprocessing includes data cleaning and data completion. The feature extraction includes extracting time features and device features through a pre-trained neural network. The control module is used to perform k-means classification on the data acquired by the perception module, perform hierarchical forecasting based on the watch demand forecasting model, add long-term and short-term influencing factors and demand influencing factors, and adjust the inventory quantity; The decision module is used to set up a forecast dashboard, adjust safety stock parameters, predict cost changes in real time and output a cost-service balance curve.

[0006] Preferably, the perception module includes: User terminal data and watch inventory data are collected based on the Internet of Things. The user terminal data includes user usage scenarios, user watch replacement cycle, dial usage preferences and user behavior data. The user behavior data includes dial switching frequency, smart function activation time, maintenance work order data and watch spare parts replacement cycle. The watch inventory data includes watch spare parts electronic tags and watch spare parts environment. The watch spare parts electronic tags include spare parts location, spare parts precision unit and spare parts inventory quantity. Watch spare parts include movement, dial and strap. The watch spare parts environment includes warehouse temperature, warehouse humidity and light intensity.

[0007] Preferably, the transmission module includes: The MQTT protocol is used for remote data transmission, user privacy and supply chain data are encrypted through HTTPS, and a local server of the watch inventory management network system is deployed to automatically identify and pre-process abnormal data in the transmission chain of the local server of the watch inventory management network system. Processing abnormal data includes filtering sensor noise and discarding garbled data packets.

[0008] Preferably, the processing module includes: Preprocess and extract features of the data perceived by the perception module based on the Internet of Things; Preprocessing includes data cleaning and data completion. Data cleaning involves removing demand records for discontinued models and outliers that exceed the preset factory settings for watch usage data, such as battery consumption >5 times per day. Data completion involves filling missing values ​​using the sliding window mean method. Feature extraction includes extracting time features and device features through a pre-trained neural network. The time features include holiday identifiers, promotional activity date labels, and peak factors of the watch sales seasonality curve. The device features include watch factory timestamps, watch usage time, watch repair times, watch repair cycles, and user watch usage frequency.

[0009] Preferably, the control module includes a forecasting unit and an inventory control unit: The prediction unit includes: Based on the data obtained by the perception module, k-means classification is performed to obtain Class A, Class B and Class C watches, where Class A includes first-value watches and first-turnover watches, Class B includes second-value watches and second-turnover watches, and Class C includes third-value watches and third-turnover watches. The data obtained by the perception module is subjected to k-means classification for a second time to obtain Class A, Class B and Class C watches, where Class A represents first-flowing watches, Class B represents second-flowing watches, and Class C represents third-flowing watches. The data obtained by the perception module is subjected to k-means classification for a third time to obtain detailed labels for watches, where the detailed labels for watches include price range, brand, type and inventory range of watches, where the types of watches include mechanical watches, quartz watches and smart watches. The classified Class A, Class B, Class C, Class A, Class B, Class C and detailed labels for watches are stored as index nodes in the watch initial information storage queue in the background of the watch inventory management system.

[0010] Preferably, the watch demand forecasting model is optimized based on the watch initial information storage queue of each watch and the data processed by the processing module, and the optimization process includes: Initialize a watch demand forecasting model, which adopts a hierarchical forecasting structure. The first layer forecasts the Class A, Class B, and Class C watches. The second layer forecasts the demand for Class A, Class B, and Class C watches based on the demand for Class A, Class B, and Class C watches. The third layer forecasts the in-depth demand for detailed watch labels based on the forecast output of the second layer. In addition, long-term and short-term impact factors and continuous and interrupted demand impact factors are added to the watch demand forecasting model.

[0011] Preferably, the class A, class B, and class C labels are defined as optimization targets for first-level prediction, and the prediction process includes: The mathematical expression of the first-layer prediction objective function is: ; in, For the cycle, is the forecast demand for the t+1th period, is the forecast demand in the tth period, is the actual detection requirement of the t-th cycle, is the smoothing coefficient; Through two smoothing processes, the forecast data trend is obtained. The mathematical expression of the first-level forecast quadratic objective function is: ; ; in, is the trend value of the t-th period, The trend value of the t-1th period, is the trend smoothing coefficient; The mathematical expression of the final prediction output is: ; in, Predict the output value for the first layer.

[0012] Preferably, based on the first-level forecast output value, demand forecasting is performed for Class A, Class B, and Class C watches, and according to the first-level forecasting process, based on the ratio of the classification ratios of Class A, Class B, Class C to Class A, Class B, Class C, the classification ratios and demand characteristics in the first-level forecasting process are changed, and the smoothing coefficient and trend smoothing coefficient are manually adjusted to perform the second-level forecasting for Class A, Class B, and Class C watches; Based on the second-level prediction output value, demand forecast is performed on the watch type corresponding to the watch detailed label. According to the second-level prediction process, based on the ratio of the watch type corresponding to the watch detailed label to the classification ratio of Class A, Class B, and Class C, the classification ratio and demand characteristics in the second-level prediction process are changed, and the smoothing coefficient and trend smoothing coefficient are manually adjusted to perform the third-level prediction of the watch type corresponding to the watch detailed label.

[0013] Preferably, in the forecasting process of the watch demand forecasting model, long-term and short-term impact factors are added, and the smoothing coefficient and trend smoothing coefficient are adjusted to distinguish the weights of short-term and long-term impacts. The differentiation logic includes: Set the smoothing coefficient threshold and trend smoothing coefficient threshold, set the short-term demand fluctuation threshold and long-term demand fluctuation threshold; When the data is adjusted to short-term demand forecast, the smoothing coefficient is set to be greater than the smoothing coefficient threshold and the trend smoothing coefficient is set to be less than the trend smoothing coefficient threshold; When the data is adjusted to long-term demand forecast, the smoothing coefficient is set to be less than the smoothing coefficient threshold and the trend smoothing coefficient is set to be greater than the trend smoothing coefficient threshold; Factors influencing the need to incorporate continuous and interrupted demands include: A Lagrangian residual term is added to the prediction model, and the weight of the Lagrangian residual term is the weighted large number probability distribution of the time feature and the device feature.

[0014] Obtaining end-user growth targets, adjusting inventory quantities based on forecast data from the watch demand forecasting model, monitoring watch shortages in real time, and forecasting future demand by increasing feature weights of watch inventory quantities in the watch demand forecasting model.

[0015] Preferably, the decision module includes: Setting up a forecast dashboard that displays sales trends, forecast error heatmaps, and inventory health by category; The inventory health includes a categorized display of turnover rate and inventory ratio.

[0016] Adjust safety stock parameters, predict cost changes in the next three months in real time, and output the cost-service balance curve.

[0017] The beneficial effects of the present invention are as follows: full-link collaboration connects design-production-inventory-after-sales data, improves out-of-stock response speed, reduces stagnant inventory, saves procurement costs, innovatively integrates user behavior and inventory environment data, and constructs an association mapping of "user needs-product characteristics-inventory status". It provides differentiated storage management for precision spare parts such as movements and straps, avoids inventory losses caused by environmental abnormalities, fills the gap in inventory quality control of traditional solutions, builds a three-level dynamic prediction system, realizes refined classification management, introduces long-term and short-term time influencing factors and continuous and interrupted demand influencing factors into the prediction model, and realizes a paradigm upgrade from "inventory management" to "demand insight". This solution realizes the full-process intelligence of "demand perception-intelligent prediction-dynamic regulation-decision optimization" through the user-product-inventory full-link data closed loop, transforming inventory management from a "cost center" to a "value center", not only solving inventory backlog and out-of-stock problems, but also feeding back product design and supply chain planning through user behavior data, providing the watch industry with an integrated empowerment path from the supply chain execution layer to the strategic layer. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A schematic diagram of the basic flow of a watch inventory management system based on the Internet of Things is provided as an embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, but not all of the embodiments.

[0020] Reference Figure 1 , as one embodiment of the present invention, provides a watch inventory management system based on the Internet of Things, including a sensing module, a transmission module, a processing module, a control module and a decision module: The perception module is used to collect user terminal data and watch inventory data based on the Internet of Things; The transmission module is used to transmit data remotely using the MQTT protocol, encrypt user privacy and supply chain data through HTTPS, deploy a local server for the watch inventory management network system, and automatically identify and pre-process abnormal data; The processing module is used to preprocess and extract features from the data perceived by the perception module based on the Internet of Things. Preprocessing includes data cleaning and data completion, and feature extraction includes extracting time features and device features through a pre-trained neural network. The control module is used to perform k-means classification on the data obtained by the perception module, perform hierarchical forecasting based on the watch demand forecasting model, add long-term and short-term influencing factors and demand influencing factors, and adjust the inventory quantity; The decision-making module is used to set up forecast dashboards, adjust safety stock parameters, predict cost changes in real time, and output cost-service balance curves.

[0021] In this embodiment, the perception module includes: User terminal data and watch inventory data are collected based on the Internet of Things. User terminal data includes user usage scenarios, user watch replacement cycle, dial usage preferences and user behavior data. User behavior data includes dial switching frequency, smart function activation time, maintenance work order data and watch spare parts replacement cycle. Watch inventory data includes watch spare parts electronic tags and watch spare parts environment. Watch spare parts electronic tags include spare parts location, spare parts precision unit and spare parts inventory quantity. Watch spare parts include movement, dial and strap. Watch spare parts environment includes warehouse temperature, warehouse humidity and light intensity.

[0022] In this embodiment, the transmission module includes: Use the MQTT protocol for remote data transmission, encrypt user privacy and supply chain data through HTTPS, deploy a local server for the watch inventory management network system, and automatically identify and pre-process abnormal data in the transmission chain of the local server of the watch inventory management network system. Processing abnormal data includes filtering sensor noise and discarding garbled data packets.

[0023] In this embodiment, the processing module includes: Preprocess and extract features of the data perceived by the perception module based on the Internet of Things; Preprocessing includes data cleaning and data completion. Data cleaning involves removing demand records for discontinued models and outliers that exceed the preset factory settings for watch usage data, such as battery consumption > 5 times per day. Data completion involves filling missing values ​​using a sliding window mean method with a sliding window size of 7 days. Feature extraction involves extracting time features and device features through pre-trained neural networks. Time features include holiday identifiers, promotional event date labels, and peak factors of the seasonal curve for watch sales. For example, the demand for diving watches is highest in August in summer. Device features include watch factory timestamps, watch usage time, watch repair times, watch repair cycles, and user watch usage frequency.

[0024] In this embodiment, the control module includes a forecasting unit and an inventory control unit: The prediction unit includes: Based on the data obtained by the perception module, k-means classification is performed to obtain Class A, Class B and Class C watches. Class A includes first-value watches and first-turnover watches, Class B includes second-value watches and second-turnover watches, and Class C includes third-value watches and third-turnover watches. The data obtained by the perception module is subjected to k-means classification for the second time to obtain Class A, Class B and Class C watches. Class A represents first-flow watches, Class B represents second-flow watches, and Class C represents third-flow watches. The data obtained by the perception module is subjected to k-means classification for the third time to obtain detailed labels of watches. The detailed labels of watches include watch price range, watch brand, watch type and watch inventory range. Watch types include mechanical watches, quartz watches and smart watches. The classified Class A, Class B, Class C, Class A, Class B, Class C and watch detailed labels are stored as index nodes in the watch initial information storage queue in the background of the watch inventory management system.

[0025] In this embodiment, the watch demand forecasting model is optimized based on the watch initial information storage queue of each watch and the data processed by the processing module. The optimization process includes: Initialize the watch demand forecasting model. The watch demand forecasting model adopts a hierarchical forecasting structure. The first layer forecasts the demand for Class A, Class B and Class C watches. The second layer forecasts the demand for Class A, Class B and Class C watches based on the demand for Class A, Class B and Class C watches. The third layer conducts in-depth demand forecasting of detailed watch labels based on the forecast output of the second layer. In addition, long-term and short-term impact factors and continuous and interrupted demand impact factors are added to the watch demand forecasting model.

[0026] In this embodiment, the class A, class B, and class C labels are defined as optimization targets for first-level prediction. The prediction process includes: The mathematical expression of the first-layer prediction objective function is: ; in, For the cycle, is the forecast demand for the t+1th period, is the forecast demand in the tth period, is the actual detection requirement of the t-th cycle, is the smoothing coefficient; Through two smoothing processes, the forecast data trend is obtained. The mathematical expression of the first-level forecast quadratic objective function is: ; ; in, is the trend value of the t-th period, The trend value of the t-1th period, is the trend smoothing coefficient; The mathematical expression of the final prediction output is: ; in, Predict the output value for the first layer.

[0027] In this embodiment, based on the first-level prediction output values, demand forecasts are conducted for Class A, Class B, and Class C watches. Based on the first-level prediction process, the classification ratios and demand characteristics in the first-level prediction process are changed based on the ratio of the classification ratios of Class A, Class B, and Class C to Class A, Class B, and Class C. Furthermore, the smoothing coefficient and trend smoothing coefficient are manually adjusted to conduct the second-level predictions for Class A, Class B, and Class C watches. Based on the second-level prediction output value, demand forecast is performed on the watch type corresponding to the watch detailed label. According to the second-level prediction process, based on the ratio of the watch type corresponding to the watch detailed label to the classification ratio of Class A, Class B, and Class C, the classification ratio and demand characteristics in the second-level prediction process are changed, and the smoothing coefficient and trend smoothing coefficient are manually adjusted to perform the third-level prediction of the watch type corresponding to the watch detailed label.

[0028] In this embodiment, in the forecasting process of the watch demand forecasting model, short-term and long-term impact factors are added, and the smoothing coefficient and trend smoothing coefficient are adjusted to distinguish the weights of short-term and long-term impacts. The differentiation logic includes: Set the smoothing coefficient threshold and trend smoothing coefficient threshold, set the short-term demand fluctuation threshold and long-term demand fluctuation threshold; When the data is adjusted for short-term demand forecasting, set the smoothing coefficient greater than the smoothing coefficient threshold and the trend smoothing coefficient less than the trend smoothing coefficient threshold; this allows for quick response to market changes and more sensitive forecasting.

[0029] When the data is adjusted for long-term demand forecasts, set the smoothing coefficient to be less than the smoothing coefficient threshold and the trend smoothing coefficient to be greater than the trend smoothing coefficient threshold; this can smooth out short-term fluctuations and pay more attention to trend stability and long-term changes.

[0030] Factors influencing the need to incorporate continuous and interrupted demands include: During certain time periods, demand may experience seasonal or holiday fluctuations, and such demand should be modeled separately to capture the disruption and recovery of demand.

[0031] The Lagrangian residual term is added to the prediction model. The weight of the Lagrangian residual term is the weighted large number probability distribution of time characteristics and equipment characteristics to consider the impact of special events on demand.

[0032] Obtain end-user growth targets, adjust inventory quantities based on the forecast data of the watch demand forecasting model, monitor watch out-of-stock situations in real time, and predict future demand by increasing the feature weight of watch inventory quantities in the watch demand forecasting model.

[0033] In this embodiment, the decision module includes: Set up a forecast dashboard to display sales trends by category, forecast error heatmaps, and inventory health; Inventory health includes categorized display of turnover rate and inventory percentage views.

[0034] Adjust safety stock parameters, predict cost changes in the next three months in real time, and output the cost-service balance curve.

[0035] The present invention innovatively integrates user behavior and inventory environment data to construct an associative mapping of "user demand-product characteristics-inventory status", and provides differentiated storage management for precision spare parts such as movements and straps, avoiding inventory losses due to environmental anomalies. It fills the gap in inventory quality control of traditional solutions, constructs a three-level dynamic prediction system, realizes refined classification management, introduces long-term and short-term influencing factors and continuous and interrupted demand influencing factors into the prediction model, and realizes a paradigm upgrade from "inventory management" to "demand insight". This solution realizes the full-process intelligence of "demand perception-intelligent prediction-dynamic regulation-decision optimization" through the user-product-inventory full-link data closed loop, transforming inventory management from a "cost center" to a "value center", not only solving the problems of inventory backlog and out-of-stock, but also feeding back product design and supply chain planning through user behavior data, providing the watch industry with an integrated empowerment path from the supply chain execution layer to the strategic layer.

[0036] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The storage medium may be implemented by any type of volatile or non-volatile storage device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0037] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A watch inventory management system based on the Internet of Things, characterized in that: It includes perception module, transmission module, processing module, control module and decision module: The sensing module is used to collect user terminal data and watch inventory data based on the Internet of Things; The transmission module is used to use the MQTT protocol for remote data transmission, encrypt it through HTTPS and deploy a local server, and automatically identify and pre-process abnormal data; The processing module is used to preprocess and extract features from the data perceived by the perception module based on the Internet of Things; The control module is used to classify the data acquired by the perception module, perform hierarchical forecasting based on the watch demand forecasting model and adjust the inventory quantity; The decision module is used to set up a forecast dashboard, adjust safety stock parameters, predict cost changes in real time and output a cost-service balance curve.

2. The watch inventory management system based on the Internet of Things according to claim 1, characterized in that: The perception module includes: User terminal data and watch inventory data are collected based on the Internet of Things. The user terminal data includes user usage scenarios, user watch replacement cycle, dial usage preferences and user behavior data. The user behavior data includes dial switching frequency, smart function activation time, maintenance work order data and watch spare parts replacement cycle. The watch inventory data includes watch spare parts electronic tags and watch spare parts environment. The watch spare parts electronic tags include spare parts location, spare parts precision unit and spare parts inventory quantity. Watch spare parts include movement, dial and strap. The watch spare parts environment includes warehouse temperature, warehouse humidity and light intensity.

3. The watch inventory management system based on the Internet of Things according to claim 1, characterized in that: The transmission module includes: The MQTT protocol is used for remote data transmission, user privacy and supply chain data are encrypted through HTTPS, and a local server of the watch inventory management network system is deployed to automatically identify and pre-process abnormal data in the transmission chain of the local server of the watch inventory management network system. Processing abnormal data includes filtering sensor noise and discarding garbled data packets.

4. The watch inventory management system based on the Internet of Things according to claim 2, characterized in that: The processing module includes: Preprocess and extract features of the data perceived by the perception module based on the Internet of Things; Preprocessing includes data cleaning and data completion. Data cleaning involves removing demand records for discontinued models and outliers that exceed the preset factory settings for watch usage data, such as battery consumption >5 times per day. Data completion involves filling missing values ​​using the sliding window mean method. Feature extraction includes extracting time features and device features through a pre-trained neural network. The time features include holiday identifiers, promotional activity date labels, and peak factors of the watch sales seasonality curve. The device features include watch factory timestamps, watch usage time, watch repair times, watch repair cycles, and user watch usage frequency.

5. The watch inventory management system based on the Internet of Things according to claim 4, characterized in that: The control module includes a forecasting unit and an inventory control unit: The prediction unit includes: Based on the data obtained by the perception module, k-means classification is performed to obtain Class A, Class B and Class C watches, where Class A includes first-value watches and first-turnover watches, Class B includes second-value watches and second-turnover watches, and Class C includes third-value watches and third-turnover watches. The data obtained by the perception module is subjected to k-means classification for a second time to obtain Class A, Class B and Class C watches, where Class A represents first-flowing watches, Class B represents second-flowing watches, and Class C represents third-flowing watches. The data obtained by the perception module is subjected to k-means classification for a third time to obtain detailed labels for watches, where the detailed labels for watches include price range, brand, type and inventory range of watches, where the types of watches include mechanical watches, quartz watches and smart watches. The classified Class A, Class B, Class C, Class A, Class B, Class C and detailed labels for watches are stored as index nodes in the watch initial information storage queue in the background of the watch inventory management system.

6. The watch inventory management system based on the Internet of Things according to claim 5, characterized in that: The watch demand forecasting model is optimized based on the watch initial information storage queue of each watch and the data processed by the processing module. The optimization process includes: Initialize a watch demand forecasting model, which adopts a hierarchical forecasting structure. The first layer forecasts the Class A, Class B, and Class C watches. The second layer forecasts the demand for Class A, Class B, and Class C watches based on the demand for Class A, Class B, and Class C watches. The third layer forecasts the in-depth demand for detailed watch labels based on the forecast output of the second layer. In addition, long-term and short-term impact factors and continuous and interrupted demand impact factors are added to the watch demand forecasting model.

7. The watch inventory management system based on the Internet of Things according to claim 6, characterized in that: The class A, class B, and class C labels are defined as optimization targets for first-level prediction. The prediction process includes: The mathematical expression of the first-layer prediction objective function is: ； in, For the cycle, is the forecast demand for the t+1th period, is the forecast demand in the tth period, is the actual detection requirement of the t-th cycle, is the smoothing coefficient; Through two smoothing processes, the forecast data trend is obtained. The mathematical expression of the first-level forecast quadratic objective function is: ； ； in, is the trend value of the t-th period, The trend value of the t-1th period, is the trend smoothing coefficient; The mathematical expression of the final prediction output is: ； in, Predict the output value for the first layer.

8. The watch inventory management system based on the Internet of Things according to claim 7, characterized in that: Based on the output values ​​of the first-level forecast, demand forecasts are conducted for Class A, Class B, and Class C watches. According to the first-level forecast process, based on the ratio of the classification ratios of Class A, Class B, Class C to Class A, Class B, and Class C, the classification ratios and demand characteristics in the first-level forecast process are changed, and the smoothing coefficient and trend smoothing coefficient are manually adjusted to conduct the second-level forecast for Class A, Class B, and Class C watches. Based on the second-level prediction output value, demand forecast is performed on the watch type corresponding to the watch detailed label. According to the second-level prediction process, based on the ratio of the watch type corresponding to the watch detailed label to the classification ratio of Class A, Class B, and Class C, the classification ratio and demand characteristics in the second-level prediction process are changed, and the smoothing coefficient and trend smoothing coefficient are manually adjusted to perform the third-level prediction of the watch type corresponding to the watch detailed label.

9. The watch inventory management system based on the Internet of Things according to claim 8, characterized in that: In the forecasting process of the watch demand forecasting model, long-term and short-term impact factors are added, and the smoothing coefficient and trend smoothing coefficient are adjusted to distinguish the weights of short-term and long-term impacts. The differentiation logic includes: Set the smoothing coefficient threshold and trend smoothing coefficient threshold, set the short-term demand fluctuation threshold and long-term demand fluctuation threshold; When the data is adjusted to short-term demand forecast, the smoothing coefficient is set to be greater than the smoothing coefficient threshold and the trend smoothing coefficient is set to be less than the trend smoothing coefficient threshold; When the data is adjusted to long-term demand forecast, the smoothing coefficient is set to be less than the smoothing coefficient threshold and the trend smoothing coefficient is set to be greater than the trend smoothing coefficient threshold; the factors affecting continuous and interrupted demand are as follows: Adding a Lagrangian residual term to the prediction model, wherein the weight of the Lagrangian residual term is a weighted large number probability distribution of time characteristics and device characteristics; Obtaining end-user growth targets, adjusting inventory quantities based on forecast data from the watch demand forecasting model, monitoring watch shortages in real time, and forecasting future demand by increasing feature weights of watch inventory quantities in the watch demand forecasting model.

10. The watch inventory management system based on the Internet of Things according to claim 1, characterized in that: The decision module includes: Setting up a forecast dashboard that displays sales trends, forecast error heatmaps, and inventory health by category; The inventory health includes a categorized display of turnover rate and inventory ratio; Adjust safety stock parameters, predict cost changes in the next three months in real time, and output the cost-service balance curve.

Images