CN-121561200-B - Intelligent analysis recommendation method for civil aviation logistics data

CN121561200BCN 121561200 BCN121561200 BCN 121561200BCN-121561200-B

Abstract

The application discloses an intelligent analysis recommendation method of civil aviation logistics data, which relates to the technical field of data content management and comprises the steps of determining a causal chain and a scene label based on depth recommendation content, constructing a causal scene map, setting a dynamic authority mapping engine, establishing a scene and authority mapping rule base, mapping the scene label to a temporary authority level, determining the temporary authority level and a target object based on the current scene label and the causal scene map, issuing and binding a temporary authority token to the target object based on the temporary authority level, calculating the matching degree of the depth recommendation content and an original comprehensive factor of the target object, determining compensation recommendation content, forming a final recommendation content by the multilevel recommendation content and the compensation recommendation content, effectively identifying different attention directions of different users to the same content, and realizing multi-dimensional authority control and accurate deduplication.

Inventors

QIU BIN
YANG FAN
ZHANG WEI
YANG PEIYING
WEI SHEN
YANG MAN

Assignees

中国民用航空局信息中心

Dates

Publication Date: 20260508
Application Date: 20260121

Claims (9)

1. An intelligent analysis recommendation method for civil aviation logistics data is characterized by comprising the following steps: s100, receiving a plurality of recommended contents returned by a server side and a data tag set of each recommended content, and primarily removing duplication according to the data tag set to obtain primary recommended contents; s200, constructing a multi-level column system, setting user permission levels, and generating column permission labels, performing deep analysis and screening on the primary recommended content according to the column permission labels to obtain deep recommended content, and determining and displaying final recommended content by combining an unbounded hierarchical management and control mechanism; the unbounded hierarchical control mechanism comprises a causal link and a scene label based on depth recommended content, a dynamic authority mapping engine, a scene and authority mapping rule base and a temporary authority level, wherein the causal link and the scene label are determined and a causal scene map is constructed; Determining a temporary permission level and a target object based on the current scene tag and the causal scene map, issuing and binding a temporary permission token to the target object based on the temporary permission level, calculating the matching degree of the depth recommended content and the original comprehensive factors of the target object, and determining the compensation recommended content; s300, acquiring intervention times and intervention effects of a target object, calculating individual deviation degree to generate an optimal strategy under fairness constraint, and updating the step S200; the method comprises the steps of recording the number of times of authority adjustment of a target object and recording the number of times of intervention, wherein the intervention effect is used for measuring the change degree of user behaviors after the intervention and comparing the index measurement before and after the intervention, calculating the average intervention effect of all the target objects, respectively calculating the individual deviation degree according to the intervention effect of a single target object, and obtaining a user group with the individual deviation degree as a negative number, setting fairness constraint, compensating the deficiency of a weak group through compensation content, and realizing the balance of fairness and efficiency; The method comprises the steps of extracting key entities and causal relations between the key entities based on depth recommendation content, connecting the key entities with arrows to form a causal link, enabling a causal scene graph to represent the key entities by nodes and to represent the causal relations by edges, finding out user roles affected by the scene based on the causal scene graph, checking whether the current authority level of the user roles is lower than the scene required level, and marking the user roles as target objects if the current authority level is lower than the scene required level; The data tag set comprises a basic tag, an emotion tendency tag, a deep reading tag, an interaction attribute tag and a timeliness tag, wherein the target object refers to a target person with authority which is not consistent with the current recommended content viewing level; In the process of calculating the matching degree, checking the matching degree value of each feature vector, screening out feature vectors with the matching degree value lower than 80%, marking the feature vectors as demand features, and searching corresponding contents from a content library according to the demand features to serve as compensation recommended contents.
2. The intelligent analysis recommendation method for civil aviation logistics data according to claim 1, wherein the step of calculating the matching degree of the depth recommended content and the original comprehensive factors of the target object and determining the compensation recommended content comprises the steps of converting the original comprehensive factors of a user into feature vectors based on extracted features in the depth recommended content, calculating the matching degree of the content feature vectors and the user feature vectors by using a vector similarity algorithm, wherein the value range is 0, 1.
3. The intelligent analysis recommendation method of civil aviation logistics data according to claim 1, wherein the multi-level column system classifies and organizes contents according to different levels and comprises a plurality of levels of columns, each level of columns has a specific theme and coverage area to form a hierarchical structure, the levels of the multi-level column system are divided according to the theme and attribute of the contents, the user permission level means that different users have different access and operation permissions in a content management system, and the depth and width of the recommended contents are determined according to the different user permission levels.
4. The intelligent analysis recommendation method for civil aviation logistics data according to claim 1, wherein the column permission label is an identifier combining a column level with a user permission level and is used for identifying access and operation permissions of a user under a specific column, the format and the structure of the label are determined, the format and the structure comprise column level and permission level information, the column level and the user permission level are associated to form a permission control strategy, and corresponding column permission labels are generated for each recommended content according to the association relation and actual requirements.
5. The intelligent analysis recommendation method of civil aviation logistics data according to claim 1, wherein the deep analysis screening means that the preliminary recommended content is subjected to deep screening according to the generated column authority labels, and specifically comprises the steps of matching the recommended content with the column authority labels of users, screening out content which accords with the user authority level and the column preference, and further performing de-duplication processing in the screened content.
6. The intelligent analysis recommendation method for civil aviation logistics data of claim 1, wherein in step S200, the deep analysis screening further comprises: S210, splitting the preliminary recommended content, identifying component elements, and setting independent identifiers and attributes for each component element; S220, dynamically forming new recommended content based on the identification and the attribute, and establishing an intelligent deduplication caching mechanism for storing hash values of content components and combination modes thereof which are checked by a user, and identifying and filtering repeated component combination content; The identification is a unique identifier of each component for locating the component within the system, and the attribute is data content describing the characteristics of the component.
7. The intelligent analysis recommendation method of civil aviation logistics data according to claim 6, wherein the intelligent deduplication caching mechanism is characterized in that a caching structure is utilized and is used for storing hash values of content components and combination modes thereof which have been checked by a user, when new content is recommended, the hash values of the current content component combination are calculated first, whether the same hash values exist in the cache or not is checked, the key value pair is adopted for storage, wherein the key is a user ID, and the values are in a collection form and comprise the checked component identifications and the hash values in the combination modes.
8. The intelligent analysis recommendation method for civil aviation logistics data of claim 5, wherein the deep analysis screening further comprises: S230, establishing a cross-column content association mechanism, calculating content similarity according to the identifiers and the attributes of different columns, and constructing a cross-column content association network according to a similarity result to obtain association rules of the content among different columns; And S240, defining authority levels of different user groups, determining the accessible content range of users of each level, marking authority of each recommended content, generating a fusion recommended strategy, and further determining deep recommended content.
9. The intelligent analysis recommendation method of civil aviation logistics data according to claim 8, wherein the cross-column content association mechanism is used for establishing association relations between contents by analyzing and calculating correlation of contents among different columns, the content association network is used for displaying internal relations and mutual influences among different contents in a graph structure mode, each node represents one content, edges represent association among the contents, weights of the edges represent association strength, and the fusion recommendation strategy is used for mixing the contents with different authority levels according to preset mixing proportions in a recommendation list according to authority levels of users.

Description

Intelligent analysis recommendation method for civil aviation logistics data Technical Field The invention relates to the technical field of data content management, in particular to an intelligent analysis recommendation method for civil aviation logistics data. Background With the digital transformation of the aviation logistics industry, the algorithm can accurately analyze user attributes and behavior data in the aspect of content recommendation, and content such as videos and articles can be pushed to users in a personalized mode, so that content browsing experience of the users is greatly improved. Meanwhile, the aviation logistics industry is actively exploring the application of big data to optimize the whole-flow business management. The aviation logistics public information service platform of the civil aviation bureau of China enriches the data dimension through the business data processing and integration technology based on the data weight, and remarkably improves the quality and accuracy of the data. However, in a multi-display area interface such as an APP home page, the problem of repeatability of recommended content is increasingly prominent, affecting user satisfaction. How to combine the intelligent technology of content recommendation with big data management of aviation logistics, avoid recommending content repetition, realize individuation and accurate service becomes important demand. The Chinese patent application number 202310862674.1 provides a recommended content management method and device, an electronic device and a storage medium, wherein the recommended content management method and device comprises the steps of performing de-duplication processing according to a data tag of each recommended content in a plurality of recommended contents and according to the data tag of each recommended content in the plurality of recommended contents, and exposing at least one recommended content remained after the de-duplication processing to a corresponding display area. In the prior art, the duplication removal effect on data information is realized only by carrying out duplication removal processing on tag contents of different data, but the tag contents only roughly show the general direction of the data, the difference characteristics of the information contents are ignored, the recommended contents correspond to display areas one by one, the contents are displayed in solid state, and the nuances of the contents under different columns or authority levels cannot be accurately distinguished. Disclosure of Invention The intelligent analysis recommendation method for civil aviation logistics data solves the problems that in the prior art, duplication removal is carried out only by depending on tag content, the difference of information content and personalized requirements are ignored, and the technical effects of multidimensional authority control and accurate duplication removal are achieved. The application provides an intelligent analysis recommendation method for civil aviation logistics data, which comprises the following steps: s100, receiving a plurality of recommended contents returned by a server side and a data tag set of each recommended content, and primarily removing duplication according to the data tag set to obtain primary recommended contents; s200, constructing a multi-level column system, setting user permission levels, and generating column permission labels, performing deep analysis and screening on the primary recommended content according to the column permission labels to obtain deep recommended content, and determining and displaying final recommended content by combining an unbounded hierarchical management and control mechanism; the unbounded hierarchical control mechanism comprises a causal link and a scene label based on depth recommended content, a dynamic authority mapping engine, a scene and authority mapping rule base and a temporary authority level, wherein the causal link and the scene label are determined and a causal scene map is constructed; Determining a temporary permission level and a target object based on the current scene tag and the causal scene map, issuing and binding a temporary permission token to the target object based on the temporary permission level, calculating the matching degree of the depth recommended content and the original comprehensive factors of the target object, and determining the compensation recommended content; s300, acquiring the intervention times and the intervention effects of the target object, calculating the individual deviation degree to generate an optimal strategy under fairness constraint, and updating the step S200. Further, determining a target object, including finding out a user role affected by a scene based on a causal scene graph, checking whether a current authority level of the user role is lower than a scene required level, and if so, marking the user role as the target object; The data tag set com