Data-driven riding path intelligent recommendation method and system

Abstract

The invention discloses a data-driven riding path intelligent recommendation method and system, and relates to the technical field of intelligent planning, and the method comprises the steps: receiving riding demand information inputted by a target user, constructing a hierarchical reasoning framework based on a hypertree structure, and decomposing the hierarchical reasoning framework into a plurality of subtasks; multi-dimensional constraint conditions are introduced, independent riding path recommendation is performed on each sub-task in combination with an external knowledge base, and a plurality of independent recommended riding paths are obtained; and father-child node adaptive adjustment is carried out on the independent recommended riding path in combination with a hierarchical reasoning framework, father node association fusion is carried out, and a target recommended riding path is determined. The technical problems that an existing riding route planning method cannot comprehensively consider multi-dimensional constraint conditions and user personalized requirements and route recommendation is not accurate and flexible enough are solved, and the purposes of dynamically adapting to user requirements and environment changes through a data-driven hierarchical reasoning framework and multi-dimensional optimization are achieved. And the accuracy and flexibility of path recommendation are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent planning technology, specifically to a data-driven intelligent recommendation method and system for cycling routes. Background Technology

[0002] With the acceleration of urbanization and people's pursuit of a healthy lifestyle, cycling, as a green and healthy mode of transportation, is gaining popularity among more and more users. However, traditional cycling route planning methods often rely solely on simple map data and distance calculations, failing to fully consider multi-dimensional factors during the cycling process, such as real-time traffic conditions, weather changes, terrain conditions, and users' personalized preferences. This results in recommended cycling routes that may not be safe, comfortable, or efficient enough, making it difficult to achieve accurate route recommendations and meet the diverse needs of users. Summary of the Invention

[0003] This application provides a data-driven intelligent cycling route recommendation method and system to address the technical problem that existing cycling route planning methods cannot comprehensively consider multi-dimensional constraints and personalized user needs, resulting in insufficient accuracy and flexibility in route recommendations.

[0004] The first aspect of this application provides a data-driven intelligent cycling route recommendation method, the method comprising: receiving cycling demand information input by a target user; constructing a hierarchical reasoning framework based on a hypertree structure to decompose the cycling demand information into multiple sub-tasks; introducing multi-dimensional constraints and combining an external knowledge base to independently recommend cycling routes for each of the multiple sub-tasks, thereby obtaining multiple independently recommended cycling routes; performing parent-child node adaptive adjustment on the multiple independently recommended cycling routes using the hierarchical reasoning framework, thereby obtaining multiple adjusted independently recommended cycling routes; and performing parent node association fusion on the multiple adjusted independently recommended cycling routes to determine the target recommended cycling route.

[0005] A second aspect of this application provides a data-driven intelligent cycling route recommendation system, comprising: a cycling demand information decomposition module, used to receive cycling demand information input by a target user, construct a hierarchical reasoning framework based on a hypertree structure, and decompose the cycling demand information into multiple sub-tasks; an independent cycling route recommendation module, used to introduce multi-dimensional constraints, combine with an external knowledge base to recommend independent cycling routes for each of the multiple sub-tasks, and obtain multiple independent recommended cycling routes; a parent-child node adaptive adjustment module, used to combine the hierarchical reasoning framework to adaptively adjust the parent-child nodes of the multiple independent recommended cycling routes, and obtain multiple adjusted independent recommended cycling routes; and a parent node association and fusion module, used to perform parent node association and fusion on the multiple adjusted independent recommended cycling routes to determine the target recommended cycling route.

[0006] One or more technical solutions provided in this application have at least the following technical effects or advantages: This application provides a data-driven intelligent cycling route recommendation method and system, which relates to the field of intelligent planning technology. It constructs a hierarchical reasoning framework using a hypertree structure, decomposing user cycling needs into multiple sub-tasks. Multi-dimensional constraints and an external knowledge base are introduced to recommend independent routes for each sub-task. Then, through the fusion of parent nodes, the target recommended cycling route is determined. This solves the technical problem that existing cycling route planning methods cannot comprehensively consider multi-dimensional constraints and personalized user needs, resulting in insufficient accuracy and flexibility in route recommendations. It achieves the technical effect of dynamically adapting to user needs and environmental changes through a data-driven hierarchical reasoning framework and multi-dimensional optimization, thereby improving the accuracy and flexibility of route recommendations. Attached Figure Description

[0007] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0008] Figure 1 A schematic flowchart of a data-driven intelligent cycling route recommendation method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of a data-driven intelligent cycling route recommendation system provided in an embodiment of this application.

[0009] Figure labeling: Cycling demand information decomposition module 11, independent cycling route recommendation module 12, parent-child node adaptive adjustment module 13, parent node association and fusion module 14. Detailed Implementation

[0010] This application provides a data-driven intelligent cycling route recommendation method and system to address the technical problem that existing cycling route planning methods cannot comprehensively consider multi-dimensional constraints and personalized user needs, resulting in insufficient accuracy and flexibility in route recommendations.

[0011] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0012] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0013] Example 1, as Figure 1 As shown, this application provides a data-driven intelligent recommendation method for cycling routes, which includes: P10: Receive cycling demand information input by the target user, construct a hierarchical reasoning framework based on a supertree structure, and decompose the cycling demand information into multiple sub-tasks.

[0014] The cycling demand information includes the starting point, destination, budget, time constraints, and user preferences. Multiple sub-tasks include specific decision-making tasks related to route selection, mode of transportation selection, weather impact, and road condition analysis. The hypertree structure includes multiple parent nodes and child nodes; each parent node represents a sub-task, and each child node represents a specific breakdown of that task. Parent nodes are connected to multiple child nodes through edges, forming a hierarchical reasoning relationship.

[0015] It should be understood that the first step is to receive the cycling needs information input by the target user, which forms the basis for personalized cycling route recommendations. Cycling needs information includes key elements such as start and end points, budget, time constraints, and user preferences. Specifically, the start and end points are the user's starting and ending locations, the budget involves cost limits during the ride, the time constraint refers to the user's desired ride duration, and user preferences encompass the user's personalized requirements for the cycling route, such as whether they prefer scenic routes or need to pass through specific locations.

[0016] To efficiently process these complex cycling demand information, this application constructs a hierarchical reasoning framework based on a hypertree structure. A hypertree structure is a data organization method that decomposes complex tasks into multiple sub-tasks through a tree-like hierarchical relationship, and further breaks them down into more specific smaller tasks, thereby achieving hierarchical reasoning and decision-making. In this structure, each parent node represents a sub-task, such as specific decision-making tasks like route selection, mode of transportation selection, weather impact, and road condition analysis, while each child node represents a specific breakdown task of that sub-task. Parent nodes are connected to multiple child nodes through edges, forming a clear hierarchical reasoning relationship. This structure enables the system to refine and optimize the cycling route decision-making process layer by layer.

[0017] Taking the route selection subtask as an example, the parent node "Route Selection" connects to multiple child nodes, such as the shortest distance path, the safest path, and the most scenic path. The system analyzes and makes decisions for each child node based on the user's input requirements, combined with external data such as map data and real-time traffic information. For instance, if the user prefers scenic routes, the system prioritizes the "most scenic path" child node and further analyzes the specific road segments within that path to ensure the route selection aligns with the user's preferences.

[0018] Meanwhile, external data sources, such as real-time traffic information and weather forecasts, are incorporated to optimize each sub-task. For example, in route selection, the system incorporates real-time traffic flow data to avoid recommending congested routes to users. In weather impact assessment, the system determines whether there will be rain or other adverse weather conditions based on the weather forecast and recommends more suitable routes or provides appropriate suggestions, such as avoiding areas with heavy rain or adjusting departure times. In road condition analysis, the system integrates real-time road condition data to ensure that recommended routes do not pass through closed roads or construction areas, thereby avoiding inconvenience during cycling.

[0019] Using the methods described above, the hypertree structure can not only break down tasks, but also perform fine-grained optimization on each subtask, so that the final recommended cycling route meets the user's specific needs and can make real-time adjustments based on changes in the external environment.

[0020] P20: Introduce multi-dimensional constraints and combine them with an external knowledge base to recommend independent cycling routes for the multiple sub-tasks, thereby obtaining multiple independent recommended cycling routes.

[0021] Furthermore, step P20 in this embodiment of the application also includes: P21: Traverse and obtain the features of the multiple subtasks; P22: Retrieve the external knowledge base to perform cycling path matching on the features of the multiple subtasks respectively, and obtain a set of independent recommended cycling paths for multiple subtasks; P23: Iteratively filter the set of independent recommended cycling paths for multiple subtasks to determine multiple independent recommended cycling paths.

[0022] Optionally, multi-dimensional constraints can be further introduced, and external knowledge bases such as maps, real-time traffic information, and weather data can be combined to recommend independent cycling routes for each sub-task, thereby optimizing the decision-making of each sub-task and obtaining multiple independent recommended cycling routes.

[0023] First, we iterate through and acquire features from multiple subtasks. The feature data for each subtask comes from user input and external data sources. These features form the basis for subsequent path matching and filtering, and specifically include path length features, traffic condition features, weather impact features, terrain features, user preference features, and surrounding facility features. Specifically, path length features represent the straight-line distance or actual cycling distance from the start to the end point, used to assess the length of the path; traffic condition features include road congestion, average speed, and the number of traffic lights, used to assess the efficiency of the path; weather impact features include precipitation probability, wind speed, and temperature, used to assess the potential impact of weather on cycling; terrain features include slope and altitude changes, used to assess the difficulty and physical exertion of cycling; user preference features include whether users prefer scenic routes or need to pass through specific locations, used to meet users' personalized needs; and surrounding facility features include rest stops, supply points, and hospitals along the route, used to assess the convenience and safety of the path. By traversing each subtask, extracting all relevant features, and storing these features as a subtask feature set, we can ensure that the features of each subtask are comprehensive and accurate, providing a foundation for subsequent path matching.

[0024] Subsequently, an external knowledge base is retrieved to perform cycling route matching for multiple sub-task features. This external knowledge base includes map data, real-time traffic information, weather data, and surrounding facility data, which provide relevant background information and dynamic data support for each sub-task. Based on this information, multiple candidate cycling routes can be generated for each sub-task. For example, route length matching calculates all possible routes from the start to the end point based on map data and filters routes that meet the user's time constraints based on route length features; traffic condition matching assesses the traffic efficiency of each route based on real-time traffic information, eliminating routes with severe congestion; weather impact matching evaluates the weather conditions of each route based on weather data, prioritizing routes with better weather conditions; terrain matching assesses the slope and elevation changes of each route based on terrain data, eliminating routes with overly complex terrain; user preference matching selects routes that match user preferences, such as scenic routes or routes passing through specific locations; and surrounding facility matching evaluates the convenience and safety of each route based on surrounding facility data, prioritizing routes with rest stops and supply points along the way. Through the matching process described above, a set of candidate independent recommended cycling routes is generated for each subtask. These sets contain multiple cycling routes that match the characteristics of the current subtask, providing a wide range of choices for subsequent filtering.

[0025] Finally, the set of candidate independent cycling routes for multiple sub-tasks is iteratively filtered to determine multiple independent recommended cycling routes. The filtering criteria include multiple dimensions such as route time efficiency, economy, comfort, and safety. For example, routes that can be completed within the specified time can be prioritized based on the user's input time constraint; routes with lower costs can be selected based on the user's budget; routes with rich scenery and good road conditions can be prioritized if the user prefers a comfortable cycling experience; in addition, the system will avoid recommending routes with high safety risks, such as areas with frequent traffic accidents or road sections under construction. Through these iterative filtering processes, the recommendation results are gradually optimized to ensure that the final recommended route for each sub-task can meet the user's needs to the greatest extent.

[0026] Ultimately, an independent recommended path is provided for each subtask, corresponding to different subtask decisions. For example, the route selection subtask might recommend a fast and flat route; the transportation mode selection subtask might recommend the mode of transportation that best suits the user's budget and time requirements; the weather impact subtask might recommend a route that avoids severe weather; and the traffic condition analysis subtask would recommend the smoothest and safest route. Through this process, not only can personalized route recommendations be provided based on user needs, but also changes in the external environment can be responded to in real time, ensuring the optimality of the recommended path under multi-dimensional constraints.

[0027] Furthermore, step P22 in the embodiments of this application also includes: P22-1: Match multiple cycling path reasoning units of the external knowledge base based on the features of the multiple sub-tasks; P22-2: Use the multiple cycling path reasoning units to perform path reasoning on the features of the multiple sub-tasks respectively to obtain a set of candidate independent recommended cycling paths for the multiple sub-tasks.

[0028] In one possible embodiment of this application, based on the features of multiple sub-tasks, multiple cycling route reasoning units in an external knowledge base are matched to achieve path reasoning for each sub-task and obtain a set of candidate independent recommended cycling routes.

[0029] First, the features of each subtask are matched with multiple cycling route reasoning units in an external knowledge base. These units are pre-built to handle different types of cycling route problems. They can be neural network models based on different inputs and outputs, each optimized for specific cycling route features. For example, some models might focus on traffic condition features, while others might focus on weather-related features. By analyzing the features of each subtask and selecting the best-matching cycling route reasoning unit, we ensure that each subtask receives targeted processing.

[0030] Next, multiple cycling path reasoning units are used to perform path reasoning on the features of multiple sub-tasks. Each cycling path reasoning unit can be regarded as an independent neural network model, responsible for processing specific types of input data and generating corresponding path recommendations. In this process, the system inputs the features of each sub-task into the corresponding reasoning unit, which infers the path based on its pre-set training model. For example, the reasoning unit for traffic condition features considers real-time traffic data to generate a path that avoids congested sections; the reasoning unit for weather impact features considers precipitation probability and wind force to generate a path with better weather conditions. In this way, the output of each reasoning unit generates a set of independent candidate recommended paths, which contains multiple cycling paths recommended for that sub-task.

[0031] Furthermore, step P23 in this embodiment of the application also includes: P23-1: Perform intra-set path similarity identification on the multiple sub-task candidate independent recommended cycling path sets respectively, and take the sub-task candidate independent recommended cycling path corresponding to the maximum average similarity with other sub-task candidate independent recommended cycling paths in each sub-task candidate independent recommended cycling path set as multiple central sub-task candidate independent recommended cycling paths; P23-2: Based on the multiple central sub-task candidate independent recommended cycling paths, perform center iterative drift on the multiple sub-task candidate independent recommended cycling path sets to determine multiple independent recommended cycling paths.

[0032] Specifically, the set of candidate independent recommended cycling routes for multiple sub-tasks is further optimized to determine the final independent recommended cycling route.

[0033] First, path similarity is identified within each of the multiple candidate independent recommended cycling path sets for each subtask. By calculating the similarity between paths, the central path in each subtask candidate path set is identified. For example, based on factors such as path geography, traffic conditions, and path length, algorithms such as cosine similarity or Euclidean distance are used to measure the similarity between paths, calculating the similarity between paths within each subtask candidate independent recommended cycling path set.

[0034] After calculating the similarity between paths, for each path in the set of candidate independent recommended cycling paths for each subtask, the mean similarity between that path and other paths is calculated. This mean reflects the representativeness of the path within the entire set. Subsequently, the path with the highest mean similarity in the set of candidate independent recommended cycling paths for each subtask is selected as the central candidate independent recommended cycling path for that subtask. These central paths have the highest representativeness in the set and can better reflect the feature requirements of that subtask.

[0035] After completing path similarity identification and center path selection within the set, an iterative center drift is performed on the set of multiple candidate independent recommended cycling paths for various sub-tasks. This process can be implemented using the mean drift algorithm, with the candidate independent recommended cycling paths for the central sub-tasks serving as the initial center points, representing the initial path sets and guiding subsequent iterative optimization. Next, mean drift iterations are performed on each set of candidate independent recommended cycling paths for each sub-task. In each iteration, the similarity between each path and the current center point is calculated, and the center point is updated based on the similarity. The update rule can be based on a weighted average, where paths with higher similarity contribute more to the center point update. By calculating the similarity weight between each path and the current center point, the position of the center point is updated according to the similarity weight, making it closer to paths with higher similarity. After each iteration, the change in the center point needs to be checked. If the change in the center point is less than a preset threshold or reaches a preset number of iterations, the iteration is considered converged, and the iteration process stops. After multiple iterations, several independent recommended cycling routes were finally obtained. These routes represent the optimal routes under the constraints of multiple sub-tasks, ensuring that the routes are not only optimal in their respective sub-tasks, but also provide consistent recommendations that meet user needs as a whole.

[0036] P30: By combining a hierarchical reasoning framework, the parent-child nodes of the multiple independent recommended cycling routes are adaptively adjusted to obtain multiple adjusted independent recommended cycling routes.

[0037] Furthermore, step P30 in this embodiment of the application also includes: P31: Extract multiple parent nodes corresponding to the multiple independent recommended cycling routes based on the hierarchical reasoning framework; P32: Obtain multiple child node sets of the multiple parent nodes; P33: Calculate the path adaptability score of the multiple independent recommended cycling routes based on the multiple child node sets to obtain multiple child node path adaptability score sets; P34: Calculate the mean of the multiple child node path adaptability score sets to obtain the mean of the multiple child node path adaptability scores; P35: Adaptively adjust the multiple independent recommended cycling routes based on the mean of the multiple child node path adaptability scores to obtain multiple adjusted independent recommended cycling routes.

[0038] It should be understood that by combining a hierarchical reasoning framework to adaptively adjust the parent-child nodes of multiple independent recommended cycling routes, the route recommendation is optimized to ensure that the recommendation results more accurately meet the user's needs.

[0039] First, multiple parent nodes corresponding to the various independent recommended cycling routes are extracted based on a hierarchical reasoning framework. Parent nodes represent relatively macro-level decision-making tasks, such as route selection, mode of transportation selection, and weather influence. These parent nodes define the goals and constraints of each task; therefore, by extracting parent nodes, the position and role of each route in the task can be determined, providing a framework for subsequent route adjustments.

[0040] Next, we obtain the set of multiple child nodes associated with each parent node. Child nodes are specific breakdowns of the parent node's tasks, typically representing more granular path selection objectives. For example, under a path selection parent node, child nodes might include choosing the shortest path, avoiding traffic congestion, and considering aesthetic appeal. Each parent node can have multiple child nodes, each representing a different characteristic or objective of the task. By obtaining the set of child nodes, we can gain a deeper understanding of the multidimensional requirements contained within each parent node, providing detailed optimization basis for path recommendation.

[0041] Subsequently, path adaptability scores are applied to multiple independent recommended cycling routes based on multiple sets of child nodes. This process is achieved by evaluating the adaptability of each child node path across specific dimensions. Path adaptability score is a quantitative metric for assessing the degree to which a path matches the sub-task objective. For example, for a route selection task, the adaptability score might consider the route's time consumption and whether it avoids congested sections; for a weather-related task, the adaptability score would consider whether the route avoids areas affected by severe weather. By applying adaptability scores to all candidate routes, a set of adaptability scores for each child node path can be generated.

[0042] Next, the mean of the adaptive scores for multiple child node paths is calculated to obtain the average adaptive score for multiple child node paths. This mean reflects the overall performance of each path across multiple child node tasks. For example, if a path provides good adaptability across multiple child nodes, its average score will be high. This mean serves as the adaptive score for the path at the global level, providing guidance for subsequent path adjustments.

[0043] Finally, adaptive adjustments are made to multiple independent recommended cycling routes based on the average path adaptability scores of multiple child nodes. This optimizes route selection, making it more aligned with user needs under the constraints of each sub-task. If a route has a low adaptability score, some features of the route can be adjusted, such as changing the route selection, cycling time, or mode of transportation, to improve its adaptability. This process ensures that the recommended routes meet user expectations at both the global and local levels. After adaptive adjustments, multiple adjusted independent recommended cycling routes are obtained. These routes are the optimal paths optimized under the adaptive adjustments of parent and child nodes. Each route not only meets the needs of a single sub-task but also provides the best overall cycling experience.

[0044] Furthermore, step P35 in this embodiment of the application also includes: P35-1: Determine whether the average adaptive score of the multiple sub-node paths is greater than or equal to a preset threshold. If not, divide the average adaptive score of the multiple sub-node paths by the sum of the average adaptive scores of the multiple sub-node paths to obtain multiple adaptive adjustment coefficients. P35-2: Calculate the product of the multiple adaptive adjustment coefficients and the preset adjustment bandwidth to obtain multiple adaptive adjustment bandwidths. P35-3: Randomly adjust the multiple independent recommended cycling paths based on the multiple adaptive adjustment bandwidths to obtain multiple stage-adjusted independent recommended cycling paths. P35-4: When the average adaptive score of the sub-node paths corresponding to the multiple stage-adjusted independent recommended cycling paths is greater than or equal to the preset threshold, the multiple stage-adjusted independent recommended cycling paths are used as multiple adjusted independent recommended cycling paths.

[0045] Specifically, multiple independently recommended cycling routes will be further adjusted and optimized to ensure that the final recommended route better meets user needs and various constraints.

[0046] First, the system determines whether the average adaptability score of multiple child paths is greater than or equal to a preset threshold. The average adaptability score reflects the overall performance of each path across various sub-tasks, while the threshold represents the system's minimum acceptance standard for path recommendations. If the average adaptability score of a path does not reach the preset threshold, it means that the path performs poorly in certain sub-tasks and requires further optimization. Therefore, if the average adaptability score is below the threshold, the system enters an adjustment phase. Multiple adaptability adjustment coefficients are calculated by dividing the average adaptability score of each path by the sum of the average adaptability scores of all paths. These adjustment coefficients provide a basis for subsequent path adjustments.

[0047] Next, these adaptive adjustment coefficients are multiplied by a preset adjustment bandwidth to obtain multiple adaptive adjustment bandwidths. The adjustment bandwidth is a crucial parameter in the adjustment process, determining the magnitude of adjustment for each path. The preset adjustment bandwidth defines the range and flexibility of path adjustment. The product of the adjustment coefficients and the bandwidth helps the system determine the degree of adjustment for each path, ensuring that the adjustment results do not deviate from the expected range.

[0048] Subsequently, multiple independent recommended cycling routes are randomly adjusted based on multiple adaptive adjustment bandwidths. This random adjustment step can be viewed as an exploratory process, achieved by randomly perturbing within the adaptive adjustment bandwidth range of each sub-node. Specifically, based on the adaptive adjustment bandwidth of each sub-node, the independent recommended cycling routes are fine-tuned, such as adjusting certain segments of the route, changing the start or end point, etc. This random adjustment can explore more possibilities in the route space, making route recommendations more flexible and diverse, and finding better route solutions.

[0049] After phased adjustments, the average adaptability score of the child nodes corresponding to each path is re-evaluated. If the average adaptability score of these paths is greater than or equal to a preset threshold, it indicates that the path has met all sub-task requirements, and the adjusted path can provide a high-quality cycling experience under various constraints. At this point, the system uses these adjusted paths as the final independent recommended cycling paths and ends the path adjustment process. These paths have undergone a process of parent-child node adaptability adjustment and random adjustment to ensure that each path not only meets the user's basic needs but can also flexibly respond to environmental changes and dynamic constraints.

[0050] P40: Perform parent node association and fusion on the multiple independently recommended cycling routes to determine the target recommended cycling route.

[0051] Furthermore, step P40 in this embodiment of the application also includes: P41: Obtain the weights of multiple subtasks corresponding to multiple parent nodes; P42: Based on the weights of the multiple subtasks, perform association and fusion on the multiple independently recommended cycling routes to obtain the target recommended cycling route.

[0052] Optionally, the system first obtains the weights of multiple subtasks corresponding to multiple parent nodes. Each parent node represents a different decision-making level, while each subtask has a different level of importance. For example, the route selection subtask might be more important than the weather impact subtask in certain situations, or certain constraints, such as budget or time limits, might have a greater impact on the final route recommendation. By assigning appropriate weights, the system can quantify the influence of each subtask on the final route recommendation. The determination of subtask weights can be based on various factors, such as the priority of user needs and the degree of impact of the subtask on the cycling experience. For example, if the user is particularly concerned about route safety, the weights of the transportation mode selection and road condition analysis subtasks might be higher; if the user values ​​scenery more, the weight of the route selection subtask would be higher. These weights affect the decision priority of different subtasks, thus weighting them during route fusion to ensure that the relative importance of each subtask is fully considered during the fusion process.

[0053] Subsequently, based on these sub-task weights, multiple adjusted independent recommended cycling routes are correlated and fused. The purpose of correlation fusion is to merge the route recommendation results of different sub-tasks to obtain the final target recommended route. Simple weighting methods may not fully express the complex interrelationships between sub-tasks; therefore, this application employs a neural network model for correlation fusion. Specifically, the neural network model uses sub-task weights as analysis weight network layers and embeds these weights into the route fusion process. Through the neural network model, the system can learn the relative contribution of each sub-task to the final route recommendation and fuse them based on this learning result, thereby achieving more accurate and reasonable route selection. For example, a neural network model specifically for route fusion is constructed. The input of this model is multiple adjusted independent recommended cycling routes and their corresponding sub-task weights, and the output is the target recommended cycling route after comprehensively considering all sub-task weights. In the neural network model, a weight network layer is designed, and the weight parameters of this layer correspond to the weights of each sub-task. These weight parameters are adjusted according to data during model training to ensure that the model can accurately reflect the importance of each sub-task. For each adjusted independent recommended cycling route, its key features, such as route length, safety level, and scenery rating, are extracted. These features serve as input to the neural network, participating in the model's computation along with the sub-task weights. Through forward propagation, the features that adjust each independently recommended cycling route are combined with their corresponding sub-task weights for comprehensive calculation. The output is a target recommended cycling route that takes all sub-task weights into account.

[0054] This process not only considers the independent contribution of each subtask but also captures the complex relationships between subtasks through the nonlinear characteristics of neural networks. By leveraging a large amount of historical cycling data and user feedback, the system can automatically adjust and optimize route selection during path fusion based on user needs, external environmental constraints, and the priority of each subtask. This ensures that the recommended route satisfies all constraints while best meeting the user's overall expectations, providing personalized and efficient cycling route selection.

[0055] In summary, the embodiments of this application have at least the following technical effects: This application provides highly personalized cycling route recommendations by receiving user cycling demand information and combining it with user preferences, meeting the diverse needs of different users. By introducing multi-dimensional constraints, such as traffic, weather, and terrain, the application comprehensively optimizes cycling routes to ensure superior performance in terms of safety and comfort. By combining real-time data, the application dynamically adjusts the recommended routes, improving their flexibility and robustness. Through a hierarchical reasoning framework and neural network model, the application efficiently integrates the recommendation results from multiple sub-tasks to determine the optimal cycling route, enhancing its safety, comfort, and efficiency, thus meeting diverse user needs and improving the cycling experience.

[0056] It achieves the technical effect of improving the accuracy and flexibility of path recommendation by dynamically adapting to user needs and environmental changes through a data-driven hierarchical reasoning framework and multi-dimensional optimization.

[0057] Example 2, based on the same inventive concept as the data-driven intelligent cycling route recommendation method in the foregoing examples, such as... Figure 2 As shown, this application provides a data-driven intelligent cycling route recommendation system. The system and method embodiments in this application are based on the same inventive concept. The system includes: The cycling demand information decomposition module 11 is used to receive cycling demand information input by the target user, and construct a hierarchical reasoning framework based on a hypertree structure to decompose the cycling demand information into multiple sub-tasks.

[0058] The independent cycling route recommendation module 12 is used to introduce multi-dimensional constraints and combine them with an external knowledge base to recommend independent cycling routes for the multiple sub-tasks, thereby obtaining multiple independent recommended cycling routes.

[0059] The parent-child node adaptive adjustment module 13 is used to perform parent-child node adaptive adjustment on the multiple independent recommended cycling routes in combination with the hierarchical reasoning framework to obtain multiple adjusted independent recommended cycling routes.

[0060] The parent node association and fusion module 14 is used to perform parent node association and fusion on the multiple independently recommended cycling routes to determine the target recommended cycling route.

[0061] Furthermore, in the cycling demand information decomposition module 11: The cycling demand information includes the starting point, destination, budget, time constraints, and user preference information; multiple sub-tasks include specific decision-making tasks corresponding to route selection, mode of transportation selection, weather impact, and road condition analysis.

[0062] Furthermore, in the cycling demand information decomposition module 11: The hypertree structure includes multiple parent nodes and child nodes; each parent node represents a subtask, and each child node represents a specific decomposition task of that task. The parent nodes are connected to multiple child nodes through edges to form a hierarchical reasoning relationship.

[0063] Furthermore, the independent cycling route recommendation module 12 is also used to perform the following steps: The process involves iterating through and obtaining features of multiple subtasks; retrieving external knowledge bases to perform cycling path matching on the features of each subtask to obtain a set of independent recommended cycling paths for each subtask; and iteratively filtering the set of independent recommended cycling paths for each subtask to determine multiple independent recommended cycling paths.

[0064] Furthermore, the independent cycling route recommendation module 12 is also used to perform the following steps: For each of the multiple sub-task candidate independent recommended cycling path sets, path similarity within the set is identified. The sub-task candidate independent recommended cycling path corresponding to the maximum average similarity with other sub-task candidate independent recommended cycling paths in each sub-task candidate independent recommended cycling path set is taken as multiple central sub-task candidate independent recommended cycling paths. Based on the multiple central sub-task candidate independent recommended cycling paths, a center iterative drift is performed on the multiple sub-task candidate independent recommended cycling path sets to determine multiple independent recommended cycling paths.

[0065] Furthermore, the independent cycling route recommendation module 12 is also used to perform the following steps: Multiple cycling path reasoning units are matched with the external knowledge base based on the features of the multiple sub-tasks; the multiple cycling path reasoning units are used to perform path reasoning on the features of the multiple sub-tasks respectively to obtain a set of independent recommended cycling paths for the multiple sub-tasks.

[0066] Furthermore, the parent-child node adaptive adjustment module 13 is also used to perform the following steps: Based on the hierarchical reasoning framework, extract multiple parent nodes corresponding to the multiple independent recommended cycling routes; obtain multiple child node sets of the multiple parent nodes; perform path adaptability scoring on the multiple independent recommended cycling routes based on the multiple child node sets to obtain multiple child node path adaptability score sets; calculate the mean of the multiple child node path adaptability score sets to obtain the mean of the multiple child node path adaptability scores; and perform adaptive adjustment on the multiple independent recommended cycling routes based on the mean of the multiple child node path adaptability scores to obtain multiple adjusted independent recommended cycling routes.

[0067] Furthermore, the parent-child node adaptive adjustment module 13 is also used to perform the following steps: Determine whether the average adaptive score of the multiple sub-node paths is greater than or equal to a preset threshold. If not, divide the average adaptive score of each of the multiple sub-node paths by the sum of the average adaptive scores of the multiple sub-node paths to obtain multiple adaptive adjustment coefficients. Calculate the product of the multiple adaptive adjustment coefficients and a preset adjustment bandwidth to obtain multiple adaptive adjustment bandwidths. Randomly adjust the multiple independent recommended cycling paths based on the multiple adaptive adjustment bandwidths to obtain multiple stage-adjusted independent recommended cycling paths. When the average adaptive score of the sub-node paths corresponding to the multiple stage-adjusted independent recommended cycling paths is greater than or equal to the preset threshold, the multiple stage-adjusted independent recommended cycling paths are used as multiple adjusted independent recommended cycling paths.

[0068] Furthermore, the parent node association and fusion module 14 is also used to perform the following steps: Obtain the weights of multiple subtasks corresponding to multiple parent nodes; based on the weights of the multiple subtasks, perform association and fusion on the multiple independently recommended cycling routes to obtain the target recommended cycling route.

[0069] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0070] The above description is only 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 protection scope of this application.

[0071] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application intends to include such modifications and variations.

Claims

1. A data-driven intelligent cycling route recommendation method, characterized in that, The method includes: Receive cycling demand information input by the target user, and build a hierarchical reasoning framework based on a supertree structure to decompose the cycling demand information into multiple sub-tasks; By introducing multi-dimensional constraints and combining an external knowledge base, independent cycling route recommendations are made for the multiple sub-tasks, resulting in multiple independent recommended cycling routes. By combining a hierarchical reasoning framework, the parent-child nodes of the multiple independent recommended cycling routes are adaptively adjusted to obtain multiple adjusted independent recommended cycling routes; The multiple independently recommended cycling routes are associated and fused with their parent nodes to determine the target recommended cycling route.

2. The data-driven intelligent cycling route recommendation method as described in claim 1, characterized in that, The cycling demand information includes the starting point, destination, budget, time constraints, and user preference information; The sub-tasks include specific decision-making tasks related to route selection, mode of transportation selection, weather impact, and road condition analysis.

3. The data-driven intelligent cycling route recommendation method as described in claim 1, characterized in that, The supertree structure includes multiple parent nodes and child nodes; Each parent node represents a subtask, and each child node represents a specific breakdown of that task. Parent nodes are connected to multiple child nodes through edges, forming a hierarchical reasoning relationship.

4. The data-driven intelligent cycling route recommendation method as described in claim 1, characterized in that, By introducing multi-dimensional constraints and combining them with an external knowledge base, independent cycling route recommendations are made for each of the multiple sub-tasks, resulting in multiple independently recommended cycling routes, including: Iterate through and obtain the features of multiple subtasks; External knowledge bases are retrieved to perform cycling route matching on the features of the multiple sub-tasks, resulting in a set of independent recommended cycling routes for multiple sub-tasks. The set of candidate independent recommended cycling routes for the multiple sub-tasks is iteratively filtered to determine multiple independent recommended cycling routes.

5. The data-driven intelligent cycling route recommendation method as described in claim 4, characterized in that, The set of candidate independent recommended cycling routes for the multiple sub-tasks is iteratively filtered to determine multiple independent recommended cycling routes, including: For each of the multiple sub-task candidate independent recommended cycling path sets, path similarity within the set is identified. The sub-task candidate independent recommended cycling path corresponding to the maximum average similarity with other sub-task candidate independent recommended cycling paths in each sub-task candidate independent recommended cycling path set is taken as multiple central sub-task candidate independent recommended cycling paths. Based on the multiple candidate independent recommended cycling routes for the multiple central sub-tasks, a center-based iterative drift is performed on the set of candidate independent recommended cycling routes for the multiple sub-tasks to determine multiple independent recommended cycling routes.

6. The data-driven intelligent cycling route recommendation method as described in claim 4, characterized in that, External knowledge bases are retrieved to perform cycling route matching on the features of the multiple sub-tasks, resulting in a set of independent recommended cycling routes for multiple sub-tasks, including: Multiple cycling path reasoning units in the external knowledge base are matched based on the features of the multiple sub-tasks; The multiple cycling path reasoning units are used to perform path reasoning on the features of the multiple sub-tasks respectively to obtain a set of candidate independent recommended cycling paths for the multiple sub-tasks.

7. The data-driven intelligent cycling route recommendation method as described in claim 1, characterized in that, By combining a hierarchical reasoning framework to adaptively adjust the parent-child nodes of the multiple independent recommended cycling routes, multiple adjusted independent recommended cycling routes are obtained, including: Based on the hierarchical reasoning framework, multiple parent nodes corresponding to the multiple independent recommended cycling routes are extracted; Obtain a set of multiple child nodes of the multiple parent nodes; Based on the multiple sets of child nodes, the multiple independent recommended cycling routes are scored for path adaptability, resulting in multiple sets of child node path adaptability scores. Calculate the mean of the multiple child node path adaptability score sets to obtain the mean of the multiple child node path adaptability scores; Based on the average path adaptability score of the multiple sub-nodes, the multiple independent recommended cycling routes are adaptively adjusted to obtain multiple adjusted independent recommended cycling routes.

8. The data-driven intelligent cycling route recommendation method as described in claim 7, characterized in that, Based on the average path adaptability scores of the multiple sub-nodes, the multiple independent recommended cycling routes are adaptively adjusted to obtain multiple adjusted independent recommended cycling routes, including: Determine whether the average path adaptability score of the multiple sub-nodes is greater than or equal to a preset threshold. If not, divide the average path adaptability score of the multiple sub-nodes by the sum of the average path adaptability scores of the multiple sub-nodes to obtain multiple adaptive adjustment coefficients. Calculate the product of multiple adaptive adjustment coefficients and preset adjustment bandwidths to obtain multiple adaptive adjustment bandwidths; Based on the multiple adaptive adjustment bandwidths, the multiple independent recommended cycling routes are randomly adjusted to obtain multiple stage-adjusted independent recommended cycling routes; When the average path adaptability score of the sub-nodes corresponding to the multiple stages of independently recommended cycling paths is greater than or equal to a preset threshold, the multiple stages of independently recommended cycling paths are treated as multiple independently recommended cycling paths.

9. The data-driven intelligent cycling route recommendation method as described in claim 3, characterized in that, The multiple independently recommended cycling routes are associated and fused with their parent nodes to determine the target recommended cycling route, including: Get the weights of multiple subtasks corresponding to multiple parent nodes; Based on the weights of the multiple sub-tasks, the multiple independently recommended cycling routes are correlated and fused to obtain the target recommended cycling route.

10. A data-driven intelligent cycling route recommendation system, characterized in that: The system includes: The cycling demand information decomposition module is used to receive cycling demand information input by the target user, and construct a hierarchical reasoning framework based on a hypertree structure to decompose the cycling demand information into multiple sub-tasks. The independent cycling route recommendation module is used to introduce multi-dimensional constraints and combine an external knowledge base to recommend independent cycling routes for the multiple sub-tasks, thereby obtaining multiple independent recommended cycling routes. The parent-child node adaptive adjustment module is used to combine the hierarchical reasoning framework to perform parent-child node adaptive adjustment on the multiple independent recommended cycling routes to obtain multiple adjusted independent recommended cycling routes; The parent node association and fusion module is used to perform parent node association and fusion on the multiple independently recommended cycling routes to determine the target recommended cycling route.

Images