Supply chain multimodal transport intelligent data system and method

Abstract

The invention provides a supply chain multimodal transport intelligent data system and method, and belongs to the technical field of supply chain management. The invention provides a supply chain multimodal transport intelligent data system. The system comprises a unified data base module, a core calculation engine module, an intelligent decision engine module, a hierarchical alarm response module and a closed-loop optimization module. The unified data base module is used for accessing multi-source heterogeneous data and cleaning, standardizing and fusing the multi-source heterogeneous data to form a global unified data view; the core calculation engine module is used for performing data processing and feature extraction on the global unified data view; the intelligent decision engine module is used for carrying out prediction analysis on the extracted feature data and generating a prediction result and a risk score; the grading alarm response module is used for comparing the risk score with a preset threshold value, triggering an alarm and executing grading response; and the closed-loop optimization module is used for returning feedback data generated by execution and intervention to the unified data base module.

Description

A Smart Data System and Method for Multimodal Transport in Supply Chains Technical Field

[0001] This application relates to the field of supply chain management technology, and in particular to an intelligent data system and method for supply chain multimodal transport. Background Technology

[0002] Multimodal transport, as a core logistics model that integrates various modes of transportation such as road, rail, waterway, and air, and connects numerous participants such as shippers, carriers, ports, and consignees, is a key path to improve supply chain efficiency and reduce logistics costs. It has become the core of the development of modern logistics industry. With the increasing complexity of multimodal transport scenarios and the diversification of participating entities, the need for intelligent management to break down data barriers, optimize decision-making efficiency, and strengthen risk control is becoming increasingly urgent.

[0003] Currently, the industry mainly relies on three types of technical solutions for multimodal transport management: First, decentralized management systems, where each participant operates its own management system independently, and data exchange relies on traditional EDI point-to-point connections, emails, or manual entry, resulting in poor integration and real-time performance; second, rudimentary tracking platforms, which can only cover their own or a few partners' transport resources, with lagging data updates and an inability to achieve real-time visualization of the entire chain; and third, passive recording systems, which focus on post-event recording and statistics, lacking proactive monitoring and intelligent decision support. However, these existing solutions have significant shortcomings: poor data collaboration, with system heterogeneity and lack of standards leading to data fragmentation and insufficient supply chain transparency; low operational efficiency, relying on manual communication and repetitive data entry, which easily leads to connection delays and operational errors; lack of intelligent decision-making, making it difficult to achieve multi-objective optimization by comprehensively considering dynamic factors, resulting in unscientific resource allocation; and lagging risk management, lacking predictive and early warning capabilities, and only able to passively respond to risks.

[0004] Therefore, there is an urgent need for a supply chain multimodal transport intelligent data system to improve operational efficiency through process automation, achieve real-time transparency and visualization of the entire supply chain, provide intelligent decision support to reduce costs and increase efficiency, and establish a proactive risk warning mechanism to enhance supply chain resilience. Summary of the Invention

[0005] In view of this, this application provides a supply chain multimodal transport intelligent data system and method to improve operational efficiency through process automation, achieve real-time transparency and visualization of the entire supply chain, provide intelligent decision support to reduce costs and increase efficiency, and establish a proactive risk warning mechanism to enhance supply chain resilience.

[0006] Specifically, this application is implemented through the following technical solution: The first aspect of this application provides a supply chain multimodal transport intelligent data system, which includes a unified data foundation module, a core computing engine module, an intelligent decision engine module, a hierarchical alarm response module, and a closed-loop optimization module. The unified data foundation module is used to access multi-source heterogeneous data and perform cleaning, standardization, and fusion to form a globally unified data view. The unified data foundation module uses a multi-factor fuzzy matching algorithm for cross-system data association and combines an adaptive threshold judgment mechanism to aggregate heterogeneous data. The multi-factors include time matching degree, spatial matching degree, cargo attribute vector similarity, business logic matching degree, and transportation mode matching degree. The core computing engine module is communicatively connected to the unified data foundation module and is used to perform data processing and feature extraction on the globally unified data view. The intelligent decision engine module is communicatively connected to the core computing engine module and is used to process the extracted feature data... The system performs predictive analysis to generate prediction results and risk scores. The tiered alarm response module, connected to the intelligent decision engine module, compares the risk scores with dynamic risk thresholds, triggers alarms, and executes tiered responses. The dynamic risk threshold is based on a logarithmic positive distribution threshold model, adjusted using multiple factors. The closed-loop optimization module, connected to both the tiered alarm response module and the unified data foundation module, transmits feedback data generated from execution and intervention back to the unified data foundation module. The adaptive threshold determination mechanism includes: determining three evaluation dimensions—data source reliability, data transmission timeliness, and data integrity—assigning preset weights to each dimension, and performing weighted calculations to obtain a data quality confidence index. A fixed base threshold is set, and a correlation between the data quality confidence index and the threshold offset is established based on a hyperbolic tangent function. The data quality confidence index is based on 0.5; the greater the deviation from the base, the smaller the change in threshold offset, with the maximum offset controlled at 0.Within 0.5, an adaptive threshold is calculated by superimposing a fixed base threshold and a threshold offset. The comprehensive score of association confidence output by the multi-factor fuzzy matching algorithm is compared with the adaptive threshold. When the comprehensive score of association confidence reaches or exceeds the adaptive threshold, the corresponding data is determined to belong to the same business entity, and a data fusion operation is performed. The calculation model of the dynamic risk threshold is as follows: the base threshold is determined according to the log-normal distribution law, and the base threshold is constructed by preset mean and standard deviation parameters. Six types of influencing factors are selected: seasonal characteristics, weather conditions, resource utilization tension, historical risk handling accuracy, transportation cost level, and remaining transportation time pressure. Corresponding quantitative adjustment rules are set for each type of influencing factor. The base threshold is dynamically corrected by multiplying the quantitative results of each influencing factor to obtain real-time adapted transportation. The second aspect of this application provides a supply chain multimodal transport intelligent data method, which is applied to the supply chain multimodal transport intelligent data system provided in any of the claims of the first aspect of this application. The method includes: accessing multi-source heterogeneous data, cleaning, standardizing, and fusing it to form a globally unified data view; using a multi-factor fuzzy matching algorithm to perform cross-system data association, and combining it with an adaptive threshold determination mechanism to aggregate heterogeneous data; performing data processing and feature extraction on the globally unified data view; performing predictive analysis on the extracted feature data to generate prediction results and risk scores; comparing the risk scores with dynamic risk thresholds, triggering alarms and executing tiered responses; and adjusting the generation of the globally unified data view based on feedback data generated from execution and intervention.

[0007] The intelligent data system and method for supply chain multimodal transport provided in this application, firstly, breaks down data silos in traditional multimodal transport by implementing standardized data access capabilities (compatible with multi-source data such as API, EDI, and IoT devices), a built-in data cleaning and fusion engine, and a unified logistics data model. This achieves unified aggregation and standardized governance of data across the entire supply chain, laying the foundation for subsequent data collaboration. Secondly, the linkage between the core computing engine module and the intelligent decision-making engine module (the former processes and extracts feature data, while the latter generates results based on predictive analysis of feature data), combined with the "risk score comparison + tiered response" mechanism of the tiered alarm response module, enables early prediction of risks such as transportation delays and equipment failures. Furthermore, the tiered alarm response module automatically pushes early warning information and intervention guidance, transforming traditional passive risk management into proactive early warning and intervention, thus improving supply chain risk control capabilities. Thirdly, the entire system is divided into loosely coupled independent modules (microservice architecture). These modules link together through standardized communication interfaces and can also output standardized services externally via an API gateway, significantly reducing system integration complexity, improving the system's adaptability to new partners and business needs, and enabling flexible expansion. Overall, this application, through modular architecture design and functional definition, not only ensures a closed loop for the entire process of data access, processing, decision-making, response, and optimization, but also achieves deep integration of "data collaboration" and "intelligent decision-making" through the professional division of labor among various modules. This effectively solves the pain points of data fragmentation, inefficient decision-making, delayed risk management, and poor scalability in traditional multimodal transport management, ultimately achieving the effects of full supply chain visualization, improved operational efficiency, and proactive risk management. Attached Figure Description

[0008] Figure 1 is a schematic diagram of the intelligent data system for supply chain multimodal transport provided in Embodiment 1 of this application; Figure 2 is a flowchart of the intelligent data method for supply chain multimodal transport provided in Embodiment 2 of this application. Detailed Implementation

[0009] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0010] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used herein are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0011] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0012] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0013] Figure 1 is a schematic diagram of the intelligent data system for supply chain multimodal transport provided in Embodiment 1 of this application. Referring to Figure 1, the intelligent data system for supply chain multimodal transport provided in this embodiment includes a unified data foundation module, a core computing engine module, an intelligent decision engine module, a hierarchical alarm response module, and a closed-loop optimization module. The unified data foundation module is used to access multi-source heterogeneous data and perform cleaning, standardization, and fusion to form a globally unified data view. The unified data foundation module uses a multi-factor fuzzy matching algorithm for cross-system data association and combines an adaptive threshold judgment mechanism to aggregate heterogeneous data. The multi-factors include time matching degree, spatial matching degree, cargo attribute vector similarity, business logic matching degree, and transportation mode matching degree. The core computing engine module is connected to the unified data... The base module is connected for communication and is used to process data and extract features from the global unified data view. The intelligent decision engine module is connected to the core computing engine module and is used to perform predictive analysis on the extracted feature data to generate prediction results and risk scores. The hierarchical alarm response module is connected to the intelligent decision engine module and is used to compare the risk score with a dynamic risk threshold, trigger an alarm, and execute a hierarchical response. The dynamic risk threshold is based on a logarithmic positive distribution threshold model and is adjusted using multiple factors. The closed-loop optimization module is connected to both the hierarchical alarm response module and the unified data base module and is used to optimize the execution and intervention results. Feedback data is transmitted back to the unified data foundation module. The adaptive threshold determination mechanism includes: determining three evaluation dimensions—data source reliability, data transmission timeliness, and data integrity—assigning preset weights to each dimension, and then performing a weighted calculation to obtain a data quality confidence index; setting a fixed base threshold and establishing a correlation between the data quality confidence index and the threshold offset based on the hyperbolic tangent function; using 0.5 as a benchmark for the data quality confidence index, the greater the deviation from the benchmark, the smaller the change in the threshold offset, with the maximum offset controlled within 0.05; and calculating the adaptive threshold by superimposing the fixed base threshold and the threshold offset; and integrating the correlation confidence scores output by the multi-factor fuzzy matching algorithm. The score is compared with the adaptive threshold. When the comprehensive score of the association confidence reaches or exceeds the adaptive threshold, it is determined that the corresponding data belongs to the same business entity and a data fusion operation is performed. The calculation model of the dynamic risk threshold is as follows: the basic threshold is determined by the log-normal distribution law, and the basic threshold is constructed by preset mean and standard deviation parameters; six types of influencing factors are selected, namely seasonal characteristics, weather conditions, resource utilization tension, historical risk handling accuracy, transportation cost level, and remaining transportation time pressure, and corresponding quantitative adjustment rules are set for each type of influencing factor; the basic threshold is dynamically corrected by multiplying the quantitative results of each influencing factor to obtain a dynamic risk threshold that is adapted to the transportation scenario in real time.

[0014] Specifically, the intelligent data system for multimodal transport in the supply chain consists of five core modules: a unified data foundation module, a core computing engine module, an intelligent decision-making engine module, a hierarchical alarm response module, and a closed-loop optimization module. These modules are interconnected to form a cohesive whole. Specifically, the core computing engine module directly communicates with the unified data foundation module, receiving processed data from it; the intelligent decision-making engine module directly communicates with the core computing engine module, acquiring its processed feature data; the hierarchical alarm response module directly communicates with the intelligent decision-making engine module, receiving its generated prediction results and risk scores; and the closed-loop optimization module uses bidirectional communication, connecting with the hierarchical alarm response module to collect feedback data from execution and intervention, and connecting with the unified data foundation module to send feedback data back to it, forming a complete data closed loop.

[0015] Furthermore, the unified data foundation module serves as the system's data infrastructure support module. Its core function is to enable the access of multi-source heterogeneous data (such as external platform data, IoT device data, enterprise system data, and manually entered data), and through cleaning, standardization, and fusion processes, it breaks down data silos and forms a globally unified and reliable data view, providing high-quality data input for subsequent modules.

[0016] The core computing engine module is the core data processing module of the system. Its core function is to perform real-time processing and feature extraction on the global unified data view output by the unified data foundation module, and transform the integrated data into feature data that meets the needs of intelligent decision-making, so as to provide data support for subsequent predictive analysis.

[0017] The intelligent decision engine module is the core module of the system's intelligent analysis. Its core function is to perform predictive analysis based on the feature data extracted by the core computing engine module and through the built-in machine learning model to generate specific prediction results (such as the estimated time of arrival (ETA) of goods) and quantified risk scores (such as the severity scores of risks such as delays and cargo damage).

[0018] The tiered alarm response module is the system's risk handling and response module. Its core function is to compare the risk score output by the intelligent decision engine module with the preset threshold in real time. When the risk reaches or exceeds the threshold, an alarm is triggered, and the corresponding response operation is executed according to the preset tiered mechanism to ensure timely risk handling.

[0019] The closed-loop optimization module is the system's iterative optimization module. Its core function is to collect feedback data (such as manual intervention results and anomaly handling records) generated by the hierarchical alarm response module during the intervention process, and send it back to the unified data base module to provide a basis for data updates and model optimization, thereby promoting the continuous iterative upgrade of the entire system.

[0020] In practical implementation, the unified data foundation module accesses multi-source heterogeneous data, including data from external platforms, IoT devices, enterprise systems, and manually entered data. It cleans and standardizes this multi-source heterogeneous data, then fuses it using waybill numbers or container numbers as unique identifiers to form a globally unified data view. The core computing engine module obtains the globally unified data view output by the unified data foundation module through a communication connection, performs real-time KPI calculations, business event identification, data aggregation and normalization, and extracts standardized feature data. The intelligent decision engine module receives the feature data output by the core computing engine module through a communication connection, uses its built-in machine learning model for predictive analysis, and generates the estimated arrival time of the goods. The system collects prediction results such as ETA and quantifies risk scores. The tiered alarm response module obtains the risk scores output by the intelligent decision engine module via communication, compares them in real-time with preset risk thresholds, and triggers corresponding alarms when thresholds are reached or exceeded. Alarm information is distributed via SMS, app push notifications, emails, etc., and tiered response operations are executed. If no action is taken within the time limit, the alarm level is automatically escalated. The closed-loop optimization module collects feedback data such as manual intervention results and anomaly handling records generated by the tiered alarm response module via communication. After standardizing the feedback data, it is transmitted back via communication with the unified data foundation module to update the global unified data view and optimize subsequent data processing and model analysis.

[0021] The working process of each module will be introduced in turn.

[0022] Optionally, the unified data base module includes a data access unit, which supports API gateway access to external platform data, IoT protocol access to IoT device data, customized connector access to enterprise system data, and standardized form reception of manually entered data.

[0023] Specifically, for external platform data, a standardized access channel is established through an API gateway to receive data from third-party platforms; for IoT device data, IoT protocols (such as MQTT) are used to establish communication with device sensors to collect data generated by the devices in real time; for enterprise system data, customized connectors (such as JDBC) are configured to interface with customers' ERP, TMS, WMS and other systems to extract data such as orders, inventory and transportation plans; for manually entered data, standardized web or mobile forms are provided so that drivers, on-site operators and others can report abnormal events, receipt certificates and other data through the forms.

[0024] Furthermore, after integrating heterogeneous data from multiple sources, data quality rules are written for verification. For example, these rules check data integrity (non-empty), validity (e.g., latitude and longitude within a reasonable range), and consistency (e.g., correct waybill number format). Invalid data will be marked, corrected, or filtered out.

[0025] Optionally, the unified data foundation module includes a data standardization unit, which is used to unify data formats and units of measurement, establish a master data model, map data fields from different sources to standard fields, and associate them to form a complete data view corresponding to the business entity.

[0026] Specifically, the verified multi-source heterogeneous data undergoes format unification processing, with timestamps standardized to ISO8601 format and units of measurement such as weight and temperature standardized to preset standard units such as kilograms and degrees Celsius. Furthermore, a master data model centered on the waybill is established, clearly defining standard fields within the model. Corresponding fields from different data sources are mapped to standard fields in the master data model; for example, the "order ID" from the ERP system, the "shipping order number" from the TMS system, and the "tracking number" from GPS data are all associated with a unified "waybill number" field. Based on these mapped standard fields, the dispersed multi-source data is integrated to form a complete data view corresponding to the cargo transportation business entity.

[0027] For example, by using the waybill number or container number as a unique identifier, data from different sources such as order information, vehicle location, and port plans can be linked together to form a complete "cargo journey" record.

[0028] Optionally, the unified data base module includes a data context building unit, which is used to add business context to the raw data after standardization and fusion processing, match the original GPS coordinates of the device with the geographic information system, supplement the road segment name and city information corresponding to the original data, compare the transportation status with the preset transportation plan timetable, and mark the real-time transportation status.

[0029] Specifically, the original GPS coordinates of the devices are extracted from the standardized and fused data, and matched with the coordinates of the Geographic Information System (GIS). Based on the matching results, the corresponding road segment names, cities, and other geographic information are added. The preset transportation schedule is retrieved, and the real-time collected transportation status data is compared with the planned nodes in the schedule item by item, marking the real-time transportation status such as "on time," "delayed," and "ahead of schedule." The supplemented geographic information and the marked transportation status are integrated into the original data to complete the addition of the business context.

[0030] Optionally, the unified data base module uses a multi-factor fuzzy matching algorithm to associate data across systems and combines it with an adaptive threshold determination mechanism to aggregate heterogeneous data; the multi-factors include time matching degree, spatial matching degree, cargo attribute vector similarity, business logic matching degree, and transportation mode matching degree.

[0031] Specifically, the confidence score model of the multi-factor fuzzy matching algorithm is: Score = 1 - √[w1×(1-match1)² + w2×(1-match2)² + w3×(1-match3)² + w4×(1-match4)² + w5×(1-match5)²]; where Score is the confidence score; match1 is the temporal matching degree; match2 is the spatial matching degree; match3 is the similarity of cargo attribute vectors; match4 is the business logic matching degree; match5 is the transportation mode matching degree; and w1-w5 are the corresponding weights.

[0032] It should be noted that w1-w5 are determined according to the entropy weight method. In one possible implementation, w1=0.28, w2=0.25, w3=0.22, w4=0.15, and w5=0.10.

[0033] The formula for calculating the time matching degree is: Match1=max(0,1-|Δt| / window)×exp(-|Δt|² / (2×σ²)); where match1 is the time matching degree; Δt is the time difference; window is the time window threshold (3h for highways, 6h for railways, 12h for waterways, and 2h for air travel); σ=window / 3 (Gaussian decay coefficient).

[0034] The formula for calculating spatial matching degree is: Match2=2 / (1+exp(d / scale))-1; where match2 is the spatial matching degree; d is the distance, calculated by Vincenty geodesy; and scale is the distance scaling factor.

[0035] d = a × Δσ + b × (sinσ × cos2σm + cosσ × sin2σm × cosΔα); where d is the distance; a is the major radius of the Earth ellipsoid; b is the minor radius of the Earth ellipsoid; Δσ is the difference in spherical angles between the two spatial points on the Earth ellipsoid; σ is the average of the spherical angles corresponding to the two spatial points; σm is the intermediate spherical angle parameter defined in the Vincenty algorithm; and Δα is the difference in normalized latitudes between the two spatial points.

[0036] The formula for calculating the similarity of cargo attribute vectors is: Match3=(V1·V2) / (‖V1‖×‖V2‖)+λ×Jaccard(Set1,Set2); V=[type code, weight, volume, density, packaging type, temperature control requirements, hazard class, value class, special treatment, insurance requirements, customs classification, loading and unloading difficulty]; where match3 is the similarity of cargo attribute vectors; V is the cargo attribute vector; V1 and V2 are the attribute vectors of the two sets of goods; λ is the weight coefficient, usually taken as 0.3; Jaccard(Set1,Set2) is the set similarity.

[0037] Both the business logic matching degree and the transportation mode matching degree are calculated based on preset business rules and transportation model features, which will not be described here.

[0038] Optionally, the adaptive threshold determination mechanism includes: determining three evaluation dimensions—data source reliability, data transmission timeliness, and data integrity—assigning preset weights to each dimension, and then performing a weighted calculation to obtain a data quality confidence index; setting a fixed base threshold and establishing a correlation between the data quality confidence index and the threshold offset based on the hyperbolic tangent function; using 0.5 as a benchmark for the data quality confidence index, the greater the deviation from the benchmark, the smaller the change in the threshold offset, with the maximum offset controlled within 0.05; and calculating the adaptive threshold by superimposing the fixed base threshold and the threshold offset; comparing the comprehensive association confidence score output by the multi-factor fuzzy matching algorithm with the adaptive threshold; and when the comprehensive association confidence score reaches or exceeds the adaptive threshold, determining that the corresponding data belongs to the same business entity and performing a data fusion operation.

[0039] For example, in one embodiment, the adaptive threshold is calculated as follows: Threshold=0.80+0.05×tanh((CI-0.5) / 0.2); where Threshold is the adaptive threshold and CI is the data quality confidence index.

[0040] Optionally, the core computing engine module is built on a stream processing framework and is used to process data and extract features from the globally unified data view, including: real-time calculation of key performance indicators (KPIs) for the entire transportation chain, including average vehicle speed, current transportation mileage, estimated arrival time of goods, and average temperature of cold chain transportation; identification of business events formed by combinations of multiple preset types of independent events; aggregation and normalization of single-dimensional, discrete granular data to form a comprehensive data view covering the entire chain; and extraction and targeted output of feature data that meets the input requirements of the intelligent decision engine module.

[0041] Specifically, the system reads data from a unified global data view in real time, calculates key performance indicators for the entire transportation chain, calculates average vehicle speed and current transportation mileage based on GPS trajectory data, estimates estimated arrival time of goods by combining transportation routes and real-time road conditions, and collects sensor data from cold chain equipment to calculate average temperature during cold chain transportation. It identifies complex patterns (business events) composed of multiple simple events (independent events). For example, it identifies the combination of "vehicle staying at point A for more than 2 hours (independent event 1)" and "temperature sensor alarm (independent event 2)" as a business event of "suspected abnormal cargo loading / unloading or equipment failure." Furthermore, it performs time- or spatial-dimensional aggregation operations on single-dimensional, discrete, fine-grained data (such as single-point GPS data or single-time temperature collection data), and then normalizes it according to a unified standard to form a comprehensive data view covering the entire transportation chain. For example, it aggregates the location information of all vehicles on the same route to display the overall traffic conditions of that route. Following the input specifications of the intelligent decision engine module, it extracts feature data from the processed data and transmits it to the intelligent decision engine module.

[0042] Optionally, the intelligent decision engine module has a pre-trained machine learning model, which includes a classification model, a regression model, and a time series prediction model. This model is used to predict and analyze the extracted feature data, and generate prediction results and risk scores. Specifically, it generates a specific prediction result of the expected arrival time of the goods through the regression model and the time series prediction model; and calculates a risk score by combining the classification model with multi-dimensional feature data. The risk score is represented in a quantitative form of 0-100 points and is used to quantitatively assess the severity of risks during transportation.

[0043] Specifically, the intelligent decision-making engine module receives feature data (such as delay duration, current road congestion index, weather severity, and vehicle historical performance) output by the core computing engine module, and calls the internally pre-trained regression model and time series prediction model; it inputs the feature data into the regression model and time series prediction model, and generates a specific prediction result of the estimated arrival time of the goods through model inference calculation; it calls the internally pre-trained classification model, and inputs multi-dimensional feature data into the classification model; the classification model processes the input multi-dimensional feature data and calculates a risk score in a 0-100 quantification form.

[0044] Optionally, the calculation model for the dynamic risk threshold is as follows: a basic threshold is determined based on a log-normal distribution, and the basic threshold is constructed using preset mean and standard deviation parameters; six types of influencing factors are selected, including seasonal characteristics, weather conditions, resource utilization tension, historical risk handling accuracy, transportation cost level, and remaining transportation time pressure, and corresponding quantitative adjustment rules are set for each type of influencing factor; the basic threshold is dynamically corrected by multiplying the quantitative results of each influencing factor to obtain a dynamic risk threshold that adapts to the transportation scenario in real time.

[0045] Among them, seasonal characteristics are related to the adjustment of peak and off-peak transportation patterns; weather conditions are based on key meteorological indicators such as precipitation, wind speed, and visibility to quantify the degree of impact; resource utilization tension is calculated in conjunction with the actual occupancy rate of current transportation capacity and warehousing resources; historical risk handling accuracy is referenced to the past early warning response effects of similar scenarios; transportation cost level is related to the fluctuation of the current market freight rate index; and remaining transportation timeliness pressure is determined based on the ratio of remaining transportation time to the total planned time.

[0046] Specifically, the formula for calculating the dynamic risk threshold is: DynamicThreshold=BaseThreshold×∏{i=1}^{6}Fi; BaseThreshold=exp(μ+σ×Z); where DynamicThreshold is the dynamic risk threshold; BaseThreshold is the base threshold; ∏{i=1}^{6}Fi is the product of the influencing factors; μ is the location parameter (mean) of the log-normal distribution; σ is the scale parameter (standard deviation) of the log-normal distribution; and Z is a random variable that follows a standard normal distribution.

[0047] The influencing factor of seasonal characteristics is: Fseason=1+0.4×sin(2π×(doy-peak) / 365)+0.2×sin(4π×(doy-peak) / 365); where Fseason is the influencing factor of seasonal characteristics; doy is the year sequence date; and peak is the peak date of the season.

[0048] The weather condition influencing factor is: Fweather=1-∑{j=1}^{5}wj×Sj / max(Sj); where Fweather is the weather condition influencing factor; Sj is the severity of 5 types of weather (precipitation, wind speed, visibility, temperature, road conditions); and wj is the weight of weather element.

[0049] The influencing factor of resource utilization stress is: Resource = 1 + 0.3 × (1 - U / Umax)^2; where, Resource is the influencing factor of resource utilization stress; U is the current resource utilization rate; and Umax is the maximum availability rate.

[0050] The impact factor of historical risk handling accuracy is: Fhistory=1+α×(AR-target); where Fhistory is the impact factor of historical risk handling accuracy; AR is the historical accuracy; target=0.85, α=0.5.

[0051] The influencing factor of transportation cost level is: Fcost = 1 + β × ln(C / baseline); where Fcost is the influencing factor of transportation cost level; C is the current freight rate index; baseline is the base freight rate; and β is the cost sensitivity coefficient.

[0052] The influencing factor of remaining transportation time pressure is: Ftime=1+γ×(remaining / total); where Ftime is the influencing factor of remaining transportation time pressure; γ is the time sensitivity coefficient; remaininging is the remaining transportation time; and total is the planned total transportation time.

[0053] Optionally, the hierarchical alarm response module includes an alarm distribution unit, which distributes alarm information through one or more of the following methods: SMS, App push, and email. If an alarm is not processed within a timeout period after being triggered, the alarm level is automatically escalated.

[0054] Specifically, upon receiving triggered alarm information, the system selects one or more methods—SMS, App push, or email—to distribute the alarm content to the corresponding responsible person; it sets timeout monitoring for distributed alarms to track their processing status in real time; if no processing feedback is received within a preset time after an alarm is triggered, the alarm level is automatically upgraded by one level; and according to the upgraded alarm level, the alarm information is redistributed again via one or more methods—SMS, App push, or email—to ensure that alarms receive timely attention and processing.

[0055] Optionally, the feedback data includes the results of manual intervention after alarm response, transportation anomaly handling records, cargo receipt confirmation information, and risk disposal effect data. The closed-loop optimization module standardizes the feedback data according to a unified format and then sends it back to the data storage unit and data processing unit of the unified data base module to update the global unified data view and provide an update basis for data standardization, fusion processing, and context construction. At the same time, it serves as iterative training data for the machine learning model built into the intelligent decision engine module.

[0056] Specifically, the system collects feedback data such as the results of manual intervention after alarm response, records of transportation anomaly handling, cargo receipt confirmation information, and risk management effectiveness data; it standardizes the collected feedback data according to a unified format to ensure that the data format matches the requirements of the unified data foundation module; it then transmits the standardized feedback data back to the data storage and data processing units of the unified data foundation module; the feedback data is used to update the global unified data view of the unified data foundation module, providing an update basis for data standardization, fusion processing, and context construction; and it is also used synchronously as iterative training data for the machine learning model built into the intelligent decision engine module to support model optimization.

[0057] The system provided in this embodiment, firstly, breaks down data silos in traditional multimodal transport by implementing standardized data access capabilities (compatible with multi-source data such as API, EDI, and IoT devices), a built-in data cleaning and fusion engine, and a unified logistics data model. This achieves unified aggregation and standardized governance of data across the entire chain, laying the foundation for subsequent data collaboration. Secondly, the linkage between the core computing engine module and the intelligent decision-making engine module (the former processes and extracts feature data, while the latter generates results based on predictive analysis of feature data), combined with the "risk score comparison + tiered response" mechanism of the tiered alarm response module, enables early prediction of risks such as transportation delays and equipment failures. Furthermore, the tiered alarm response module automatically pushes early warning information and intervention guidance, transforming traditional passive risk management into proactive early warning and intervention, thus enhancing supply chain risk control capabilities. Thirdly, the entire system is divided into loosely coupled independent modules (microservice architecture). These modules link together through standardized communication interfaces and can also output standardized services externally via an API gateway, significantly reducing system integration complexity, improving the system's adaptability to new partners and business needs, and enabling flexible expansion. Overall, this application, through modular architecture design and functional definition, not only ensures a closed loop for the entire process of data access, processing, decision-making, response, and optimization, but also achieves deep integration of "data collaboration" and "intelligent decision-making" through the professional division of labor among various modules. This effectively solves the pain points of data fragmentation, inefficient decision-making, delayed risk management, and poor scalability in traditional multimodal transport management, ultimately achieving the effects of full supply chain visualization, improved operational efficiency, and proactive risk management.

[0058] Corresponding to the aforementioned embodiment of a supply chain multimodal transport intelligent data system, this application also provides an embodiment of a supply chain multimodal transport intelligent data method.

[0059] Figure 2 is a flowchart of the intelligent data method for supply chain multimodal transport provided in Embodiment 2 of this application. Referring to Figure 2, the intelligent data method for supply chain multimodal transport provided in this embodiment is applied to the intelligent data system for supply chain multimodal transport as described in any of the first aspects of this application; the method includes: S201, accessing multi-source heterogeneous data, and cleaning, standardizing and merging it to form a globally unified data view; using a multi-factor fuzzy matching algorithm to perform cross-system data association, and combining an adaptive threshold determination mechanism to aggregate heterogeneous data.

[0060] S202. Perform data processing and feature extraction on the global unified data view.

[0061] S203. Perform predictive analysis on the extracted feature data to generate prediction results and risk scores.

[0062] S204. Compare the risk score with the dynamic risk threshold, trigger an alarm and execute a graded response.

[0063] S205. Adjust the generation of the global unified data view based on feedback data generated from execution and intervention.

[0064] The method in this embodiment can be used to execute the steps of the system embodiment shown in Figure 1. The specific implementation principle and process are similar and will not be described again here.

[0065] The specific implementation process of the functions and roles of each unit in the above device can be found in the implementation process of the corresponding steps in the above method, and will not be repeated here.

[0066] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0067] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A supply chain multimodal transport intelligent data system, characterized in that, The intelligent data system for multimodal transport in the supply chain includes a unified data foundation module, a core computing engine module, an intelligent decision-making engine module, a hierarchical alarm response module, and a closed-loop optimization module. The unified data foundation module is used to access multi-source heterogeneous data, and to clean, standardize, and integrate it to form a globally unified data view. The unified data foundation module uses a multi-factor fuzzy matching algorithm for cross-system data association, combined with an adaptive threshold determination mechanism to aggregate heterogeneous data. The multi-factors include time matching degree, spatial matching degree, cargo attribute vector similarity, business logic matching degree, and transportation mode matching degree. The core computing engine module is communicatively connected to the unified data base module and is used to perform data processing and feature extraction on the global unified data view. The intelligent decision engine module is communicatively connected to the core computing engine module and is used to perform predictive analysis on the extracted feature data to generate prediction results and risk scores. The hierarchical alarm response module is communicatively connected to the intelligent decision engine module and is used to compare the risk score with the dynamic risk threshold, trigger an alarm, and execute a hierarchical response. The dynamic risk threshold is based on a log-positive distribution threshold model and is adjusted by multiple factors. The closed-loop optimization module is communicatively connected to both the hierarchical alarm response module and the unified data base module and is used to transmit the feedback data generated by the execution and intervention back to the unified data base module. The adaptive threshold determination mechanism includes: determining three evaluation dimensions: data source reliability, data transmission timeliness, and data integrity; assigning preset weights to each dimension and performing weighted calculations to obtain a data quality confidence index. A fixed base threshold is set, and a correlation between the data quality confidence index and the threshold offset is established based on the hyperbolic tangent function. The data quality confidence index is based on 0.5; the further it deviates from the base, the smaller the change in the threshold offset. The maximum offset is controlled within 0.

05. An adaptive threshold is calculated by superimposing the fixed base threshold and the threshold offset. The comprehensive score of the association confidence output by the multi-factor fuzzy matching algorithm is compared with the adaptive threshold. When the comprehensive score of the association confidence reaches or exceeds the adaptive threshold, the corresponding data is determined to belong to the same business entity, and a data fusion operation is performed. The calculation model of the dynamic risk threshold is as follows: the base threshold is determined by the log-normal distribution law, and the base threshold is constructed by preset mean and standard deviation parameters. Six types of influencing factors are selected: seasonal characteristics, weather conditions, resource utilization tension, historical risk handling accuracy, transportation cost level, and remaining transportation time pressure. Corresponding quantitative adjustment rules are set for each type of influencing factor. The base threshold is dynamically corrected by multiplying the quantitative results of each influencing factor to obtain a dynamic risk threshold that is adapted to the transportation scenario in real time.

2. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The unified data base module includes a data access unit, which supports API gateway access to external platform data, IoT protocol access to IoT device data, customized connector access to enterprise system data, and standardized form reception of manually entered data.

3. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The unified data foundation module includes a data standardization unit, which is used to unify data formats and units of measurement, establish a master data model, map data fields from different sources to standard fields, and associate them to form a complete data view corresponding to the business entity.

4. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The unified data base module includes a data context construction unit, which is used to add business context to the raw data after standardization and fusion processing, match the original GPS coordinates of the device with the geographic information system, and supplement the road segment name and city information corresponding to the raw data. The transportation status is compared with the preset transportation schedule, and the real-time transportation status is marked.

5. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The core computing engine module is built on a stream processing framework and is used to process data and extract features from the globally unified data view. This includes: real-time calculation of key performance indicators (KPIs) for the entire transportation chain, including average vehicle speed, current transportation mileage, estimated arrival time of goods, and average temperature of cold chain transportation; identification of business events formed by combinations of multiple preset types of independent events; aggregation and normalization of single-dimensional, discrete granular data to form a comprehensive data view covering the entire chain; and extraction and targeted output of feature data that meets the input requirements of the intelligent decision engine module.

6. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The intelligent decision engine module has a pre-trained machine learning model, which includes a classification model, a regression model, and a time series prediction model. This model is used to predict and analyze the extracted feature data, and generate prediction results and risk scores. Specifically, it generates a specific prediction result of the estimated arrival time of the goods through the regression model and the time series prediction model; and calculates a risk score by combining the classification model with multi-dimensional feature data. The risk score is represented in a quantitative form of 0-100 points and is used to quantitatively assess the severity of risks during transportation.

7. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The hierarchical alarm response module includes an alarm distribution unit, which distributes alarm information through one or more methods such as SMS, App push, and email. If an alarm is not processed within a timeout period after being triggered, the alarm level is automatically upgraded.

8. The intelligent data system for supply chain multimodal transport according to claim 1, characterized in that, The feedback data includes the results of manual intervention after alarm response, transportation anomaly handling records, cargo receipt confirmation information, and risk disposal effect data. The closed-loop optimization module standardizes the feedback data according to a unified format and then sends it back to the data storage unit and data processing unit of the unified data base module to update the global unified data view and provide an update basis for data standardization, fusion processing, and context construction. At the same time, it serves as iterative training data for the machine learning model built into the intelligent decision engine module.

9. A smart data method for multimodal transport in a supply chain, characterized in that, The intelligent data method for supply chain multimodal transport is applied to the intelligent data system for supply chain multimodal transport as described in any one of claims 1-8; the method includes: accessing multi-source heterogeneous data, cleaning, standardizing, and merging it to form a globally unified data view; using a multi-factor fuzzy matching algorithm to perform cross-system data association, and combining an adaptive threshold determination mechanism to aggregate heterogeneous data; performing data processing and feature extraction on the globally unified data view; performing predictive analysis on the extracted feature data to generate prediction results and risk scores; comparing the risk scores with dynamic risk thresholds, triggering alarms and executing graded responses; and adjusting the generation of the globally unified data view based on feedback data generated from execution and intervention.

Images