A method for managing the supply and transport of fresh fish

Abstract

The application discloses a fresh fish meat supply transportation management method, and particularly relates to the supply transportation management field, and is used for solving the problems that the existing fish meat supply process cannot take into account the demand difference of different freshness grade products, the lag of market fluctuation response and the lack of dynamic feedback of transportation allocation; the freshness grade of fish meat products in each supply place is unified and integrated, and sentiment tendency analysis and keyword frequency analysis are carried out in combination with network public opinion data, and demand trend indexes of different supply places and product categories are calculated; then, mutation detection and historical sales volume correlation modeling are carried out, and a supply demand list containing fish meat freshness, supply place and expected demand volume is generated; transportation and cold chain resources are matched according to the supply demand list, a multi-objective planning model is established to generate a transportation allocation scheme, and model parameters are dynamically adjusted according to actual goods receiving damage rate results, so that closed-loop control of supply prediction, transportation optimization and quality feedback is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of supply transportation management, and more particularly, to a fresh fish supply transportation management method. BACKGROUND

[0002] With the rapid development of urban catering industry, chain supermarkets and online fresh food platforms, the supply chain of fish products presents diversification and high frequency characteristics. Different terminals have significant differences in the freshness level of fish products. High-end catering and sashimi processing places usually require the supply of fish at a very high freshness level, allowing only a very short transportation time limit; while ordinary household consumption or frozen processing enterprises can accept fish with lower freshness level to reduce transportation costs and inventory risks.

[0003] However, the existing supply transportation management system generally takes real-time order as the basis for allocation, lacks a differentiated allocation mechanism for market demand forecasting and freshness level, resulting in frequent problems such as high-freshness products being unsold and deteriorated, and low-freshness products occupying excessive cold chain resources. At the same time, changes in market demand are not only determined by historical sales, but also affected by network public opinion, regional preferences, seasonal activities and consumer emotional tendencies. However, traditional supply is based on static sales data for forecasting, making it difficult to timely perceive demand mutations for certain fish products and freshness levels in specific regions. In summary, the existing technology has not yet formed a fish supply transportation management method that can comprehensively analyze market public opinion, identify demand mutations, forecast supply, and dynamically match cold chain resources. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a fresh fish supply transportation management method to solve the problems raised in the background art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions:

[0006] A fresh fish supply transportation management method, comprising the following steps:

[0007] S1, integrating all supply sites of fish product categories, and classifying the fish product categories by freshness level;

[0008] S2, in the internet information of each supply site, collecting network discussion data related to each fish product category through keyword grabbing;

[0009] S3, performing sentiment analysis and keyword frequency analysis on the network discussion data, and calculating demand trend indicators for different fish product categories in each supply site;

[0010] S4, mutation detection is performed on the demand trend index time series data, historical sales data of the corresponding supply area is combined, and an association model is established to output a supply demand list containing fish freshness, supply area and expected demand;

[0011] S5, according to the supply demand list, the in-transit transport capacity and cold chain storage resources in the current transport network are matched, a multi-objective programming model is established, and a transport allocation scheme is generated;

[0012] S6, according to the actual cargo loss rate monitoring results of each supply area at the time of receiving goods, the parameters of the multi-objective programming model are dynamically adjusted.

[0013] In a preferred embodiment, the S1 integrates all fish product categories of all supply areas, and classifies the fish product categories according to freshness grades, which specifically includes:

[0014] A unified fish product classification system is established, including product categories and corresponding freshness grades;

[0015] The product categories are divided based on fish species and processing technology, and the freshness grades are divided according to the storage time of fish;

[0016] A supply area fish product information database is constructed to record the fish product categories and corresponding freshness grade ranges provided by each supply area, and product category data integration is performed regularly;

[0017] The supply area is the final target sales area of the supply chain flow;

[0018] The product category data integration generates standardized product category descriptions by eliminating brand description differences of the same products from different supply areas.

[0019] In a preferred embodiment, the S2 in each supply area internet information, the network discussion data related to each fish product category is collected by keyword grabbing, which specifically includes:

[0020] A multi-level keyword dictionary is constructed, including a basic keyword layer and an extended keyword layer;

[0021] The basic keyword layer is composed of product category descriptions, and the extended keyword layer is derived from product category descriptions and freshness grades;

[0022] Directional grabbing of public text data of social media platforms, forum communities and news portals in each supply area, using regular expression matching algorithm to extract network discussion data containing any keywords in the keyword dictionary from the original text, and storing them according to time stamp and supply area identification.

[0023] In a preferred embodiment, in S3, sentiment polarity analysis and keyword frequency analysis are performed on the network discussion data, and the demand trend indicators of each supply area for different fish and meat product categories are calculated, which specifically include:

[0024] The sentiment polarity score and product relevance score of the network discussion data corresponding to each supply area are output by the pre-trained language model;

[0025] A keyword frequency statistical window is established to accumulate the occurrence frequency and relevance score of the corresponding keywords of each product category within a fixed monitoring period, and the demand trend indicators are calculated by weighted calculation combined with the sentiment polarity score, and the smoothed demand trend indicator sequence of each supply area is generated.

[0026] In a preferred embodiment, in S4, mutation detection is performed on the demand trend indicator time series data, and a correlation model is established combined with the historical sales data of the corresponding supply area to output a supply demand list containing fish freshness, supply area and expected demand amount, which specifically includes:

[0027] A mutation threshold of the demand trend indicator is preset, and when the change rate of the demand trend indicator in a continuous set of monitoring periods exceeds the mutation threshold, it is marked as a potential mutation point;

[0028] Significance test is performed on each potential mutation point, and the test process includes trend consistency and duration evaluation, and the potential mutation point that passes the test is confirmed as an effective mutation event;

[0029] Combined with the historical sales data of each supply area, an association model between effective mutation events and product category sales changes is established, and the product category sales of the supply area in the future period is predicted through the association model, the product category sales is converted into the expected demand amount of different fish freshness, and a supply demand list is established.

[0030] In a preferred embodiment, the establishment of the association model between the effective mutation events and the product category sales changes specifically includes:

[0031] The mutation intensity and duration in the effective mutation event record are taken as independent variables, and the historical sales data of the corresponding monitoring period are taken as dependent variables;

[0032] The product category is taken as the basis for data grouping, and regression equations for different product categories are established respectively;

[0033] The historical sales data of the same period is introduced as a seasonal adjustment term, and the market promotion activity data is introduced as an auxiliary variable;

[0034] The stepwise regression method is used to screen the variable combination that has an impact on sales changes, and the least squares method is used to solve the coefficients of each variable to construct a complete multiple regression prediction model.

[0035] In a preferred embodiment, in the S5, according to the supply demand list, matching the in-transit transport capacity resources and cold chain storage resources in the current transport network, a multi-objective programming model is established to generate a transport deployment scheme, specifically comprising:

[0036] The supply demand list is parsed to obtain the location, freshness level and expected demand quantity of each supply site, and real-time data of in-transit transport capacity resources and cold chain storage resource data are synchronously obtained;

[0037] A set of objective functions is defined, including a transport total cost minimization function and a freshness maintenance maximization function, and a multi-objective programming model is established;

[0038] A set of constraint conditions is constructed, including transport capacity constraints, time window constraints, freshness maintenance constraints, and availability constraints of in-transit transport capacity resources and cold chain storage resources;

[0039] An optimal solution search algorithm is used to solve the multi-objective programming model to generate a non-dominated solution set;

[0040] The non-dominated solution set is comprehensively scored and calculated by a preset target weight, and the solution with the highest comprehensive score is selected as the final transport deployment scheme.

[0041] In a preferred embodiment, in the S6, according to the actual damage rate monitoring results of each supply site at the time of receiving goods, the parameters of the multi-objective programming model are dynamically adjusted, specifically comprising:

[0042] The damage rate recorded at the time of receiving goods at each supply site is collected, and the damage rate is calculated by detecting the proportion of the freshness level of fish meat being lower than the freshness level in the supply demand list;

[0043] When the damage rate exceeds a set acceptable threshold, the weight coefficient of the freshness maintenance maximization objective in the multi-objective programming model is adjusted;

[0044] The transport paths with the highest damage rate are identified, and the time window constraints and freshness maintenance constraints of the corresponding transport paths in the multi-objective programming model are tightened.

[0045] The technical effects and advantages of the fresh fish supply transport management method of the present application are:

[0046] By establishing a unified fish product classification system and freshness grade standard, the comparability and standardized management of fish products from different supply sources are realized; combined with sentiment tendency analysis and keyword frequency analysis of network public opinion data, the market demand trend and potential changes in consumer preferences can be captured in a timely manner, improving the real-time and accuracy of demand forecasting; through mutation detection and sales volume correlation modeling, sudden increases or decreases in market demand can be identified in advance, enabling proactive adjustment of supply plans; in the transportation link, based on a multi-objective planning model, transportation cost and freshness maintenance effect are optimized simultaneously, making transportation deployment more in line with the specific requirements of different freshness grade products; at the same time, a damage rate feedback mechanism is introduced, which can dynamically correct model parameters according to the actual receipt quality, forming a closed-loop adaptive optimization system.

[0047] Overall, the fish supply management has changed from static scheduling to dynamic intelligentization, effectively reducing cold chain transportation loss, improving the timeliness and market matching degree of high freshness products, and providing an efficient and scalable technical path for digital management of aquatic supply chains. BRIEF DESCRIPTION OF DRAWINGS

[0048] Figure 1 A schematic diagram of the supply and transportation management method of fresh fish meat according to the present application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0050] Embodiment 1

[0051] Figure 1 A supply and transportation management method of fresh fish meat according to the present application is given, which includes the following steps:

[0052] S1, integrating all fish product categories of supply sources, and classifying the fish product categories by freshness grade;

[0053] S2, in the internet information of each supply source, collecting network discussion data related to each fish product category through keyword grabbing;

[0054] S3, performing sentiment tendency analysis and keyword frequency analysis on the network discussion data, and calculating demand trend indicators of each supply source for different fish product categories;

[0055] S4, mutation detection is performed on the demand trend index time series data, historical sales data of the corresponding supply place is combined, and a correlation model is established to output a supply demand list containing fish freshness, supply place and expected demand;

[0056] S5, according to the supply demand list, the in-transit transport capacity resources and cold chain storage resources in the current transport network are matched, a multi-objective programming model is established to generate a transport deployment scheme;

[0057] S6, according to the actual cargo loss rate monitoring results of each supply place at the time of receiving goods, the parameters of the multi-objective programming model are dynamically adjusted.

[0058] In S1, all fish product categories of the supply places are integrated, and the fish product categories are classified according to freshness level.

[0059] According to the main processing methods and consumption purposes of fish in the supply chain, fish products are systematically sorted and divided into several main categories, such as raw food, cold fresh, frozen and processed. Each major category is further refined into specific product categories, such as raw food including salmon sashimi and tuna sashimi, cold fresh including sea bass fillets and yellow croaker whole segments, frozen including cod blocks and pomfret segments, and processed including dried fish, fish floss and fish balls. For each product category, freshness level standards are established, and freshness levels are divided into fresh, cold storage and frozen according to the storage time and temperature control conditions of fish from capture to consumption. Fresh usually refers to fish caught within 24 hours and stored in a cool environment, cold storage refers to products stored in an environment of 0-4 degrees Celsius within 48-72 hours, and frozen refers to products that have been quickly frozen and stored for a long time. When the corresponding freshness level time is reached, the freshness level is automatically updated (such as fresh level stored for more than 24 hours, automatically transferred to cold storage and updated freshness level). After the classification system is established, fish product standard directory is constructed with fish species, processing technology and freshness level as basic elements to provide a unified reference framework for subsequent data integration. During the development of the entire classification system, hierarchical numbering is used for identification, such as combining category code, process code and freshness code to generate a unique identification code to ensure consistent identification standards among different supply places during information integration, thereby eliminating naming differences and duplicate records.

[0060] Each supply source is an information collection unit, and detailed data of the fish meat products provided by each supply source is collected one by one. The collected contents include basic information such as product name, classification number, processing technology, storage condition, delivery cycle and average storage time, and the main sales areas and target consumer types corresponding to the products, such as catering wholesale, supermarket retail or e-commerce platform distribution. For the differentiated characteristics of the same products in different supply sources, an association table is established to reflect the corresponding relationship between product categories and fish freshness. Each record in the database is bound with a supply source identifier and a data update time to ensure the timeliness and traceability of the information. After the database is established, data aggregation and cleaning tasks are automatically performed at fixed intervals (for example, once a week) to update the newly added, changed and invalid product information, so as to ensure that the fish meat product information of different supply sources remains up-to-date.

[0061] The same fish meat product information from different supply sources is standardized and fused. Due to the inconsistencies in brand description, naming habits, origin identification or packaging instructions of different supply sources, a brand description difference elimination process is needed. This process first extracts key description words in the product name, such as "fresh", "quick frozen", "original cut", "selected", etc., and maps them to the corresponding standardized terms in the standard term library. Then, according to the processing technology and storage condition fields, multi-dimensional comparison is performed to determine whether they belong to the same standard product category. For example, "original cut salmon fillet" from A and "sashimi grade salmon fillet" from B are unified and merged into "sashimi grade salmon" after matching the process field and freshness level. For the brand word part, such as "X fishing ground" and "Y brand", etc., it is kept as an attached field and does not participate in classification identification calculation. In the integration process, for entries with slight differences in description but the same product, the final merging result is confirmed through a combination of fuzzy matching algorithm and manual review, and a standardized product description table is generated.

[0062] In S2, in the internet information of each supply source, network discussion data related to each fish meat product category is collected through keyword grabbing.

[0063] According to the fish product classification system and freshness grade standard established in the early stage, the core description words of each product category are extracted as the composition of the basic keyword layer. For each basic keyword, combined with its corresponding freshness grade description, the extended keyword layer is generated. The construction of the extended keyword layer is oriented to the actual consumption context, and the synonyms, colloquial expressions and derivative modifiers are included in the dictionary. To cover the non-standardized descriptions commonly appeared by users in network discussions. In order to ensure the comprehensiveness and scalability of the dictionary, the combination of manual review and corpus statistics is used in the process of generating vocabulary, and the usage frequency and semantic relevance of new words are screened, and invalid or noise words are deleted. Finally, a multi-level keyword dictionary including basic layer and extended layer is formed, in which the basic layer reflects the core semantics of product category, and the extended layer reflects the actual usage expressions under different freshness grades and consumption contexts.

[0064] The administrative division or the name of the main city of each supply place is taken as the identification of the grabbing area, and the data collection process is executed in turn. The collection range covers the public discussion content of mainstream social media platforms, regional forum communities, food review areas and news portals. The collection cycle is set to fixed period every day or every week to ensure the continuity and timeliness of the data. During the grabbing process, the basic information such as page title, text, publishing time and source domain name of each web page content is extracted and converted into text data format. Then, the multi-level keyword dictionary mentioned above is called, and the regular expression matching algorithm is used to scan the text segment by segment to identify whether it contains any keyword in the dictionary. The regular matching rule is based on the complete word form and common variants of the keyword, such as setting the matching mode for "fresh salmon" to identify variants such as "ultra-fresh salmon" and "salmon ultra-fresh". For the text segment containing the keyword, the publishing time of the web page where it is located is extracted as the timestamp, and the geographical identification of the data source is recorded to form a multi-dimensional structured record of "supply place-time-text content". In the matching process, if a text contains multiple keywords at the same time, it is labeled with keyword grouping to ensure that the subsequent statistical analysis can distinguish the discussion content of different product categories. All the extracted network discussion data are uniformly classified and stored in the data warehouse.

[0065] In S3, the network discussion data is analyzed for sentiment and keyword frequency, and the demand trend indicators for different fish and meat product categories in each supply place are calculated.

[0066] The network discussion text data after keyword matching and timestamp labeling is used as the model input corpus. To ensure the relevance of the analysis, the input text is cleaned and standardized, including removing emoticons, website links, advertising phrases, and repetitive content, and only retaining natural language passages related to fish and meat product descriptions, consumer experience, taste evaluation, transportation experience, and other semantic content. The cleaned text is grouped by supply source label to ensure that the corpus of each supply source reflects the true consumer sentiment characteristics of the region. Then, a pre-trained language model is called, which is based on a general semantic understanding framework and fine-tuned in the food category evaluation corpus, and has the ability to distinguish the sentiment orientation of words such as "fresh", "fishy", "whitish", "cold chain", "delivered on the same day", etc. The model outputs two indicators when analyzing: sentiment polarity score and product relevance score. The sentiment polarity score reflects the positive or negative degree of the evaluation tendency in the text, for example, "meat is firm" and "sashimi tastes good" correspond to high positive scores, while "smells fishy" and "color is dark" correspond to negative scores; The product relevance score measures the semantic association strength of the text with the target fish and meat product category, and is used to filter out irrelevant comments. The score results of each text are recorded with supply source, product category, and timestamp as indexes. If the same comment involves multiple fish and meat categories, it is processed by sentence segmentation according to semantic segments and calculated separately, and finally summarized as a table of supply source level public opinion score distribution data.

[0067] A fixed monitoring period is set to cover the main active period of network discussion data. Within each window, the cumulative frequency of keywords for each supply and corresponding fish product category is counted in that time period. Keyword frequency statistics not only include the number of basic keyword layer appearances, but also include the number of derived word appearances in the extended keyword layer. For synonyms or near-synonymous expressions, the same word mapping table is used for combined statistics to prevent dispersion errors caused by vocabulary diversity. Then, the keyword frequency in the time window is combined with the product relevance score to improve the semantic representativeness of the statistical results, where the product relevance score is obtained by the similarity of the semantic vector output by the pre-trained language model. Further, a comprehensive weighting operation is performed in combination with the sentiment polarity score, and the sentiment polarity score is calculated by the proportion of positive public opinion and the proportion of negative public opinion. Finally, the trend value of the strength of consumer demand is obtained. To avoid extreme data interference, the trend value is smoothed by using a moving average or a local weighted regression method to generate a continuous demand trend index sequence. The weight setting is determined based on the statistical regression results of the historical data. Specifically, in the model training phase, the historical sales data of multiple supply places are selected as the true demand reference, and the correlation between each index and the actual sales is calculated. The sequence takes time as the horizontal axis and the trend value as the vertical axis, which can directly reflect the market attention and demand change trend of fish and meat products in different time periods in each supply place. When the trend index shows a significant upward trend in consecutive multiple periods, it is identified as a demand growth trend; when the trend index continues to decline or the fluctuation weakens, it is identified as a demand decline trend.

[0068] In S4, mutation detection is performed on the demand trend index time series data, and a correlation model is established based on the historical sales data of the corresponding supply place to output a supply demand list containing fish freshness, supply place, and expected demand quantity.

[0069] A time series trend data set is established for each supply place and its corresponding fish product category, with a fixed monitoring period as a unit to record the demand trend index values in consecutive multiple periods. To identify sudden changes in demand, a mutation threshold for the demand trend index is preset, which can be determined by the standard deviation statistics of historical samples, for example, taking the value of the mean of the historical trend index change rate plus twice the standard deviation as the mutation threshold. In the running process, when the change rate of the trend index exceeds the threshold continuously for three consecutive monitoring periods, it is marked as a potential mutation point. For each potential mutation point, a significance test is performed, including trend consistency and duration assessment. Trend consistency is used to determine whether the trend change is continuous and consistent in direction, and duration assessment is used to exclude short-term abnormal fluctuations, for example, requiring the trend change direction to be maintained for more than two periods and the average change amplitude to be more than 1.2 times the threshold, to confirm an effective mutation event.

[0070] In establishing the correlation model between demand mutation and historical sales, effective mutation events that pass the test are selected as input features, with mutation intensity and duration as independent variables. Mutation intensity refers to the change in demand trend indicators before and after the mutation in a unit of time, for example, an increase of 0.25 in three consecutive periods, which can be defined as a mutation intensity of 0.25. The duration refers to the number of monitoring periods that the mutation state is maintained. The above independent variables are paired with the historical sales data of the corresponding period to form a multi-dimensional data table. Based on product categories, regression equations are established for different fish and meat categories to ensure that the model independently models the sales change characteristics of high freshness sashimi, cold fresh processing or frozen storage products, preventing feature interference between different categories.

[0071] In the model establishment process, to improve the prediction accuracy, historical sales data of the same period is introduced as a seasonal adjustment term. Historical sales data of the same period reflects the periodic influence of a specific time period (such as holidays, fishing seasons or high temperature seasons) on sales. The seasonal adjustment term corrects the trend deviation caused by seasonal fluctuations by adding the average sales of the previous year or adjacent time period in the model. For example, there are stable regular differences in fish and meat sales before and after the Spring Festival or during the summer high temperature period. By adding this term, the periodic factor interference can be eliminated. Market promotion activity data is integrated as an auxiliary variable. Promotion activity data includes promotion frequency, discount rate and online publicity exposure, etc. The data is derived from sales platform records or supply activity records. Promotion activities often directly stimulate short-term sales changes, so when they are included as auxiliary variables in the model, they can explain part of the exogenous reasons for the sudden increase in sales, improving the explanatory power and stability of the regression equation.

[0072] After the variables are selected, stepwise regression method is used to select the variable combination that is significantly related to sales change. By testing the contribution of each variable to the model fitting degree at each step, the most explanatory set of independent variables is gradually retained. Finally, the least squares method is used to solve the regression coefficients of each variable to form a complete multiple regression prediction model. After the model is established, the mutation intensity, duration and auxiliary variables of the effective mutation events are input into the model, and the predicted sales value of the product category in the future period is output. Then, according to the correspondence between product category sales and freshness level, the predicted sales is divided into high freshness, medium freshness and frozen levels to form the expected demand of fish and meat freshness in each supply area. By integrating the expected demand of each supply area, a supply demand list is generated, providing accurate quantitative basis for subsequent transportation allocation and cold chain resource allocation, realizing the whole process data-driven closed loop from market public opinion change to supply plan generation.

[0073] In S5, according to the supply demand list, the in-transit transport capacity resources and cold chain storage resources in the current transportation network are matched, and a multi-objective planning model is established to generate a transportation allocation scheme.

[0074] The supply demand list is parsed to obtain the location, freshness level and expected demand quantity of each supply source, and real-time data of in-transit transport resources and cold chain storage resources are synchronously obtained;

[0075] According to the expected demand quantity of fish and meat products of different supply sources and different freshness levels in the supply demand list and the corresponding transport path information, dual objectives of transport optimization are determined. The first objective is a total transport cost minimization function, which is a weighted linear function defined as minimizing the total expenditure of the entire transport process under the condition of meeting all order demands. The total transport cost includes vehicle fuel cost, highway toll, cold chain refrigeration energy cost, manual loading and unloading cost, and storage transfer cost, etc. Each cost item is determined according to historical operation data and real-time market prices, and a single transport path is taken as the basic calculation unit. The second objective is a freshness retention maximization function, which is an exponential decay function defined as maximizing the preservation degree of the freshness level of fish and meat products when they arrive at the destination during the entire transport process. The evaluation of freshness retention is based on transport time, temperature control stability and original product level. For high freshness fish and meat products, the extension of transport time and temperature fluctuations will cause the freshness to decrease, and the freshness retention function calculates the maximum value through the penalty term of transport time and temperature deviation.

[0076] The transport capacity constraint is defined, and in any transport task, the loadable volume and load weight of each transport vehicle must not exceed its rated upper limit. For example, the maximum load volume of a cold chain vehicle is 12 cubic meters or the maximum load weight is 8 tons, when the volume or weight of fish and meat products exceeds any limit, the transport task needs to be automatically split. The time window constraint is used to ensure that the expected demand can be transported within the time limit of the delivery time window. When the planned transport time exceeds the time limit, the scheme is determined as an infeasible solution. The freshness retention constraint is used to limit the fluctuation range of temperature and humidity during transport to prevent quality decline due to cold chain abnormalities, for example, requiring the temperature fluctuation to be within ±1 degree Celsius throughout the journey. In addition, the in-transit transport resource constraint is used to ensure that the actual state of the current in-transit vehicle is considered when scheduling transport, preventing repeated allocation. This constraint selects only vehicles and routes that are in a dispatchable state by querying the real-time transport database. The cold chain storage resource availability constraint is used to ensure that each supply source or transit station has sufficient cold storage capacity and refrigeration equipment resources during the current scheduling period. If the storage space is insufficient or the cold storage is in a maintenance state, the node is automatically excluded as a candidate storage point. The above constraint conditions ensure that the model has executability and reliability in actual operation, avoiding the generation of scheduling schemes that are theoretically feasible but operationally infeasible.

[0077] The optimal solution search algorithm is used to solve the multi-objective programming model. The algorithm is based on the non-dominated solution set generation mechanism, that is, by simultaneously optimizing the total transportation cost and freshness maintenance, a set of mutually trade-off feasible solutions is obtained. Each solution corresponds to a set of transportation task allocation scheme, path selection and resource allocation mode. In order to select the optimal scheme from these non-dominated solutions, a preset target weight is introduced for comprehensive score calculation. The setting of the target weight is determined according to the enterprise operation strategy or seasonal market priority, for example, in the fish supply peak period, the freshness maintenance target weight is given as 0.6 and the transportation cost weight is given as 0.4; in the off-season with high cost pressure, the weights are set as 0.5 and 0.5. The score calculation is obtained by weighting and summing the standardized target values of each solution, and the solution with the highest score is selected as the final transportation deployment scheme. If the score difference is small, then the time utilization rate and cold chain resource occupation rate indexes are further compared to determine the uniqueness of the final scheme.

[0078] In S6, according to the actual damage rate monitoring results of each supply site at the time of receiving goods, the parameters of the multi-objective programming model are dynamically adjusted.

[0079] After each transportation task is completed, the receiving end detects and classifies the freshness grade of the arrived fish products according to the acceptance record. The detection method is performed according to the pre-established standard freshness grade system, and the actual freshness grade of the arrived fish is determined by combining time, temperature tracking and texture determination. Then, the freshness grade of the arrived fish is compared with the freshness grade corresponding to the batch of fish in the supply demand list. When the detection result is lower than the target grade in the list, it is regarded as a damage event. The calculation of the damage rate takes the total weight of the arrived fish as the basis, and calculates the proportion of the freshness grade decrease. For example, a supply site transports a total of 500 kg of fish to the target city, and 50 kg of the actual detection grade is lower than the predetermined grade, so the damage rate is 10%. The calculation result is directly stored in the transportation data record table, which is used for subsequent model parameter correction and path evaluation. In order to exclude the influence of single batch abnormal situation, the damage rate calculation uses the moving average method, and the average value is calculated with three transportation cycles as the statistical window to ensure the stability and representativeness of the data.

[0080] In the process of judging and adjusting the model parameters, the acceptable threshold of the damage rate is set in advance. The threshold is determined according to the product type and the requirements of the target market, for example, the threshold for the fish meat level of the sword is set to 5%, the threshold for the cold fresh fish meat is set to 8%, and the threshold for the frozen fish meat is set to 12%. When the average damage rate exceeds the corresponding threshold, it is determined that the current transportation scheme cannot effectively maintain the freshness. At this time, the weight coefficient of the freshness maintenance maximization target in the multi-objective planning model is adjusted to strengthen the optimization priority of the freshness dimension in the model. If the weight of the transportation total cost and the freshness maintenance target in the original model is set to 0.5:0.5, when the damage rate exceeds the threshold, the weight of the freshness maintenance target is adjusted to 0.65, and the weight of the transportation total cost is adjusted to 0.35 accordingly; when the damage rate exceeds twice the threshold, it is further increased to 0.75:0.25. The adjustment is performed in a hierarchical form, and each update takes effect in the next round of model solving, forming a dynamic adaptive target function weight system, so that the subsequent scheduling process prioritizes product quality.

[0081] While adjusting the target weight, the transportation path with the highest damage rate is identified and optimized. The identification process takes the last three transportation records as samples, sorts them from high to low according to the damage rate, and selects the top 10% of the paths as the focus of correction. For these paths, the time window constraint and freshness maintenance constraint in the multi-objective planning model are tightened. The tightening of the time window constraint is performed according to the proportion of the original upper limit, for example, the original maximum transportation time is 10 hours, which is tightened to 8 hours; if the original is 6 hours, it is tightened to 5 hours. The tightening of the freshness maintenance constraint is reflected in the strictness of the temperature fluctuation range and the humidity control range, for example, the original temperature fluctuation range is ±2℃, which is tightened to ±1℃; the original humidity fluctuation range is ±5%, which is tightened to ±3%. The model after parameter adjustment is solved again, and the new transportation scheme generated will prioritize the time and temperature control requirements of high-freshness-grade fish products. Through the above steps, a closed-loop association between the damage rate data and the model constraints is established, enabling the model to have dynamic learning and continuous correction capabilities, ensuring that fish products maintain stable freshness quality in multi-supply and multi-path cold chain transportation.

[0082] The above formulas are dimensionless and calculated by taking their numerical values. The formula is obtained by collecting a large amount of data to simulate the most recent real situation. The preset parameters and threshold values in the formula are set by a person skilled in the art according to the actual situation.

[0083] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center through a wired (for example, infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0084] Those of ordinary skill in the art can realize that the modules and algorithm steps of the examples described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0085] Those of ordinary skill in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device, and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0086] In several embodiments provided in the present application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed ones can be indirect coupling or communication connection through some interfaces, devices, or modules, which can be electrical, mechanical, or other forms.

[0087] The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules, and may be located in one place or distributed on multiple network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment.

[0088] In addition, the functional modules in each embodiment of the present application can be integrated into one processing module, or each module can exist physically alone, or two or more modules can be integrated into one module.

[0089] The functions, if realized in the form of software function modules and sold or used as independent products, can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program code storage media.

[0090] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0091] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application, should be included in the protection scope of the present application.

Claims

1. A method for managing the supply and transport of fresh fish meat, characterized by, The method comprises the following steps: S1, integrating fish product categories of all supply places and classifying the fish product categories according to freshness grades; S2, collecting network discussion data related to each fish product category through keyword crawling in the internet information of each supply place; S3, performing sentiment tendency analysis and keyword frequency analysis on the network discussion data, and calculating demand trend indexes of different fish product categories in each supply place respectively; S4, detecting mutations in the time series data of the demand trend indexes, combining historical sales data of the corresponding supply place, and establishing an association model to output a supply demand list containing fish freshness, supply place and expected demand amount; S5, matching in-transit transportation resources and cold chain storage resources in the current transportation network according to the supply demand list, establishing a multi-objective programming model to generate a transportation deployment scheme; S6, dynamically adjusting parameters of the multi-objective programming model according to the actual damage rate monitoring results of each supply place when receiving goods.

2. The method for managing the supply and transport of fresh fish meat according to claim 1, characterized by, In the S1, the fish product categories of all supply places are integrated, and the fish product categories are classified according to freshness grades, which specifically comprises: A unified fish product classification system is established, including product categories and corresponding freshness grades; The product categories are divided based on fish species and processing technology, and the freshness grades are divided according to the storage time of fish; A supply place fish product information database is constructed to record the fish product categories provided by each supply place and the corresponding freshness grade range, and product category data integration is performed regularly; The supply place is the final target sales area of the supply chain; The product category data integration generates a standardized product category description by eliminating brand description differences of the same products from different supply places.

3. The method for managing the supply and transport of fresh fish meat according to claim 1, characterized by, In the S2, the network discussion data related to each fish product category is collected through keyword crawling in the internet information of each supply place, which specifically comprises: A multi-level keyword dictionary is constructed, including a basic keyword layer and an extended keyword layer; The basic keyword layer is composed of product category descriptions, and the extended keyword layer is derived from product category descriptions and freshness grades; Directional crawling is performed on the public text data of social media platforms, forum communities and news portals in each supply place, and regular expression matching algorithm is used to extract network discussion data containing any keyword in the keyword dictionary from the original text, and the network discussion data is classified and stored according to timestamp and supply place identifier.

4. The method for managing the supply and transport of fresh fish meat according to claim 1, characterized by, In the S3, the network discussion data is subjected to sentiment tendency analysis and keyword frequency analysis, and the demand trend indexes of different fish product categories in each supply place are calculated respectively, which specifically comprises: The pre-trained language model outputs the sentiment polarity score and product relevance score of the network discussion data corresponding to each supply place; A keyword frequency statistical window is established to accumulate the appearance frequency and relevance score of each product category corresponding keyword in a fixed monitoring period, and the demand trend index is calculated by weighted calculation combining the sentiment polarity score, and a smoothed demand trend index sequence of each supply place is generated.

5. The method for managing the supply and transport of fresh fish meat according to claim 1, characterized by, The S4 includes mutation detection of demand trend index time series data, combination of historical sales data of corresponding supply sites, establishment of an association model, and output of a supply-demand list containing fish freshness, supply sites, and expected demand, specifically including: A mutation threshold of the preset demand trend index is set, and when the change rate of the demand trend index exceeds the mutation threshold in a continuous setting period, it is marked as a potential mutation point; Significance test is performed on each potential mutation point, and the test process includes trend consistency and duration evaluation, and the potential mutation point that passes the test is confirmed as an effective mutation event; The historical sales data of each supply site is combined to establish an association model between the effective mutation event and the product category sales change, and the product category sales in the future period is predicted through the association model, and the product category sales is converted into the expected demand of different fish freshness and the supply-demand list is established.

6. The method for managing the supply and transport of fresh fish meat according to claim 5, wherein The association model between the effective mutation event and the product category sales change specifically includes: The mutation intensity and duration in the effective mutation event record are used as independent variables, and the historical sales data of the corresponding monitoring period are used as dependent variables; The product category is used as the data grouping basis, and a regression equation for different product categories is established respectively; The historical sales data of the same period is introduced as a seasonal adjustment item, and the market promotion activity data is introduced as an auxiliary variable; The stepwise regression method is used to screen the variable combination that has an impact on the sales change, the least squares method is used to solve the variable coefficients, and a complete multivariate regression prediction model is constructed.

7. The method for managing the supply and transport of fresh fish meat according to claim 1, wherein In the S5, according to the supply-demand list, the in-transit transport capacity resources and cold chain storage resources in the current transport network are matched, a multi-objective planning model is established, and a transportation deployment scheme is generated, specifically including: The supply-demand list is analyzed to obtain the location, freshness grade, and expected demand of each supply site, and the real-time data of the in-transit transport capacity resources and the cold chain storage resource data are synchronously obtained; A set of objective functions is defined, including a minimum transportation cost minimization function and a maximum freshness retention maximization function, and a multi-objective planning model is established; A set of constraint conditions is constructed, including transportation capacity constraints, time window constraints, freshness retention constraints, and availability constraints of in-transit transport capacity resources and cold chain storage resources; An optimal solution search algorithm is used to solve the multi-objective planning model to generate a non-dominated solution set; The non-dominated solution set is comprehensively scored by a preset target weight, and the solution with the highest comprehensive score is selected as the final transportation deployment scheme.

8. The method for managing the supply and transport of fresh fish meat according to claim 1, wherein In the S6, the parameters of the multi-objective planning model are dynamically adjusted according to the actual cargo loss rate monitoring results of each supply site, specifically including: The cargo loss rate recorded at the actual cargo receiving time of each supply site is collected, and the cargo loss rate is calculated by detecting the proportion of fish freshness grade lower than the freshness grade in the supply-demand list; When the cargo loss rate exceeds a set acceptable threshold, the weight coefficient of the maximum freshness retention maximization objective in the multi-objective planning model is adjusted; The transportation paths with the highest cargo loss rate are identified, and the time window constraints and freshness retention constraints of the corresponding transportation paths in the multi-objective planning model are tightened.

Images