Location-based running course recommendation and sharing system

Abstract

The present invention relates to Location-Based Services (LBS) and mobile computing technologies, and more specifically, to a location-based running course recommendation and sharing system and method that comprehensively analyzes a user's current location, personal preferences, and real-time environmental data to recommend a running course optimized for an individual, manages related activity records, and supports information sharing among users.

Description

Technology Field

[0001] The present invention relates to Location-Based Services (LBS) and mobile computing technologies, and more specifically, to a location-based running course recommendation and sharing system and method that comprehensively analyzes a user's current location, personal preferences, and real-time environmental data to recommend a running course optimized for an individual, manages related activity records, and supports information sharing among users. Background Technology

[0002] With the recent increase in interest in health and the widespread use of wearable devices, the number of people enjoying running has grown significantly. Along with this, numerous mobile application services have emerged that recommend nearby running routes based on the user's location.

[0003] Most conventional running course recommendation systems have been limited to presenting routes frequently used by other users, highly rated ones, or pre-established lists of popular routes, centered on the user's current location. While this approach can provide users with basic route information, it harbors various problems because the data serving as the basis for recommendations is static and fragmentary.

[0004] Conventional technology has failed to comprehensively consider the physical characteristics or environmental factors of the route. For example, even if the overall distance of the route is suitable, in reality, the sidewalks are frequently interrupted or the road surface conditions are poor, often hindering the flow of running and increasing the risk of injury. Furthermore, it has failed to reflect comfort factors such as the amount of shade or the urban heat island effect depending on the season or time of day, thereby exposing users to the risk of heat-related illnesses, particularly during the day in the summer.

[0005] Conventional technology had limitations in reflecting real-time safety conditions at the time of recommendation. Although the route is fixed, the surrounding environment is constantly changing. Unpredictable risk factors, such as crowd congestion due to sudden events or protests, route control due to construction or accidents, or the appearance of wild animals, can seriously threaten user safety, but conventional systems could not reflect these real-time unexpected situations in their recommendation logic.

[0006] The level of personalization was insufficient. There were issues that caused negative user experiences, such as repeatedly recommending the same and stale routes by failing to consider the user's past driving history, or interrupting driving records midway by not taking into account technical constraints like the remaining battery level of the user's smart device.

[0007] Therefore, there is an increasing technical demand for advanced learning course recommendation systems that go beyond simply listing popular routes to comprehensively evaluate the multidimensional quality and real-time safety of routes, and dynamically filter and provide them according to the user's current situation. The problem to be solved

[0008] The present invention has been devised to solve the problems of the prior art described above, and its purpose is to provide a system and method for recommending a personalized running course that provides optimal safety and satisfaction to the user by comprehensively considering the physical characteristics of the path, environmental comfort, real-time safety conditions, and the user's personal circumstances.

[0009] Another objective of the present invention is to provide a system and method that maximizes the reliability of recommendation results through a stepwise evaluation logic that separates and evaluates positive and negative factors of a path, and particularly prioritizes filtering out critical risk factors that may compromise user safety.

[0010] Another objective of the present invention is to provide a system and method that self-develops the accuracy of recommendations over time by continuously learning feedback and driving data from actual users and optimizing evaluation criteria and thresholds on the system's own.

[0011] The problems of the present invention are not limited to those mentioned above, and other problems not mentioned will be clearly understood by those skilled in the art from the description below. means of solving the problem

[0013] According to an embodiment of the present invention for solving the above problem

[0014] A user interface module (100) that receives a running course recommendation request from the outside;

[0015] A location receiving unit (200) that generates the user's current location data in response to a request from the above user interface module (100);

[0016] A database unit (800) storing multiple candidate path data; and

[0017] A course recommendation module (300) that determines a final recommended path by performing multi-stage evaluation and filtering on candidate paths stored in the database unit (800) based on a recommendation request received from the user interface module (100) and current location data received from the location receiving unit (200);

[0018] Includes,

[0019] The above course recommendation module (300) is,

[0020] A step of evaluating the objective quality and safety of the path through multiple judgment criteria; and

[0021] Including the step of filtering the above evaluation results and adjusting the rankings based on the user's real-time situation and past history,

[0022] It is characterized by recommending user-customized running courses,

[0023] The above course recommendation module (300) is,

[0024] It includes a route comprehensive quality evaluation module that evaluates the overall quality of the route, and

[0025] The above-mentioned comprehensive path quality evaluation module is,

[0026] Step 1, which calculates a basic quality score of the path based on pedestrian surface continuity data representing the physical continuity of the path and path safety grade data representing the safety of the external environment of the path; and

[0027] It includes a second step of determining a final quality grade by deducting a score equal to a preset penalty ratio from the basic quality score calculated in the first step when the micro-terrain roughness data indicating the degree of road surface defects of the above path and the thermal stress index data indicating the degree of thermal burden of the above path each exceed a preset first standard value.

[0028] The above micro-topographic roughness data is,

[0029] It is generated by analyzing the frequency of vertical impact occurrences in accelerometer sensor data collected from multiple user terminals, and

[0030] The above heat stress index data is characterized by being generated by combining real-time temperature data, satellite surface temperature data, and GIS green space ratio data, and

[0031] The above route safety rating data is,

[0032] It is generated by the path safety evaluation module included in the above course recommendation module (300), and

[0033] The above path safety evaluation module is,

[0034] A first filtering step that prioritizes determining whether either crowd variability data indicating the possibility of sudden crowd density around a path or animal encounter probability data indicating the probability of encountering wild or stray animals exceeds a preset second threshold, and if exceeded, immediately determines the path as 'unsafe' regardless of other conditions and terminates the evaluation; and

[0035] Limited to routes that have passed the above-mentioned first stage, it includes a second evaluation stage that calculates a basic safety rating by synthesizing night lighting consistency data indicating the stability of night lighting and road surface friction estimate data indicating the risk of slipping on the road surface according to weather conditions, and

[0036] The above crowd volatility data is,

[0037] By monitoring the frequency of appearance of specific keywords (assemblies, events, demonstrations) around the above path in real time through a social media API, and determining that the second threshold is exceeded when the frequency of appearance exceeds a preset number of times per unit of time,

[0038] It is characterized by ensuring user safety by detecting in advance unpredictable real-time multi-density risks that cannot be identified solely by fixed geographical information, and by fundamentally blocking the recommendation of such routes.

[0039] The above path safety evaluation module is,

[0040] After the above-mentioned two-stage evaluation, it further includes a third correction stage in which, based on emergency escape accessibility data analyzing the distance from all points on the above-mentioned route to the nearest safety base (hospital, police station), if the accessibility is equal to or greater than a pre-set third threshold, a correction score is added to the above-mentioned basic safety grade to determine the final route safety grade.

[0041] This is characterized by increasing the reliability of recommendations by upgrading the safety rating of routes capable of responding to emergencies among routes with the same basic safety level.

[0042] The first threshold and penalty rate of the above-mentioned comprehensive path quality evaluation module are,

[0043] A machine learning model periodically analyzes subjective satisfaction feedback and injury experience data collected from multiple users through a community module (700), and

[0044] By dynamically updating the micro-terrain roughness data value or the thermal stress index data value corresponding to the inflection point where the incidence rate of negative experiences in the above feedback data increases statistically significantly to the above first reference value,

[0045] The system can be characterized by optimizing its own evaluation criteria through actual user experience data to improve the accuracy of the evaluation over time. Effects of the invention

[0047] The location-based running course recommendation and sharing system according to the present invention provides the following effects.

[0048] User satisfaction is improved through a multidimensional quality evaluation of the path. The present invention comprehensively evaluates specific quality indicators that were not considered in conventional technology, such as the continuity of the walking surface, the micro-terrain roughness of the road surface, and the thermal stress index. Through this, the quality of the running experience can be fundamentally improved by providing users with a high-quality path that allows for comfortable and immersive running with a low risk of injury, rather than just a path for running.

[0049] The present invention maximizes user safety by filtering real-time risk factors. Through a 'critical risk filtering' logic, the invention detects unpredictable real-time risks in advance, such as the possibility of crowd density or wildlife sightings at the time of recommendation, and fundamentally excludes the corresponding route from recommendations. This protects users from the risk of potential accidents and has the effect of significantly increasing user trust in the system.

[0050] Negative user experiences are prevented through practical personalized filtering. This invention uses practical personalization factors, such as the user's device battery level or past driving history, as filtering conditions at the final stage of the recommendation process. By doing so, negative experiences—such as interrupted exercise recording or receiving the same route recommendations every time—are blocked in advance, and long-term engagement with the service can be enhanced by presenting users with genuinely useful and novel routes.

[0051] The invention ensures continuous performance improvement of the system through data-driven self-learning. This invention includes a mechanism that dynamically optimizes evaluation criteria by learning from actual user feedback and driving data. This enables sustainable technological advancement, allowing the system to adapt to changing environments and user patterns over time rather than remaining stuck in fixed logic, thereby autonomously improving the accuracy of recommendations.

[0052] The effects according to the present invention are not limited to those exemplified above, and a wider variety of effects are included within the present invention. Brief explanation of the drawing

[0054] Figure 1 illustrates an overall relationship diagram according to the present invention. FIG. 2 illustrates an organic operation flowchart between all components according to the present invention. Figure 3 illustrates a flowchart of a comprehensive path quality evaluation module according to the present invention. Figure 4 illustrates a flowchart of a path safety evaluation module according to the present invention. Figure 5 illustrates a flowchart of a path difficulty determination module according to the present invention. Figure 6 illustrates a flowchart of a personalized recommendation filtering module according to the present invention. Specific details for implementing the invention

[0055] Hereinafter, various embodiments are described in more detail with reference to the attached drawings. The embodiments described in this specification may be modified in various ways. Specific embodiments may be depicted in the drawings and described in detail in the detailed description. However, specific embodiments disclosed in the attached drawings are intended only to facilitate understanding of various embodiments. Accordingly, the technical concept is not limited by specific embodiments disclosed in the attached drawings, and it should be understood that it includes all equivalents or substitutions that fall within the spirit and scope of the invention.

[0056] Terms including ordinal numbers, such as first, second, etc., may be used to describe various components, but these components are not limited by the aforementioned terms. The aforementioned terms are used solely for the purpose of distinguishing one component from another.

[0057] Functions related to artificial intelligence according to the present disclosure are operated through a processor and memory. The processor may be composed of one or more processors. In this case, the one or more processors may be general-purpose processors such as CPUs, APs, and DSPs (Digital Signal Processors), graphics-dedicated processors such as GPUs and VPUs (Vision Processing Units), or artificial intelligence-dedicated processors such as NPUs. The one or more processors control the processing of input data according to predefined operation rules or artificial intelligence models stored in memory. Alternatively, if the one or more processors are artificial intelligence-dedicated processors, the artificial intelligence-dedicated processors may be designed with a hardware structure specialized for processing a specific artificial intelligence model.

[0058] The predefined rules of operation or artificial intelligence models are characterized by being created through learning. Here, being created through learning means that a basic artificial intelligence model is trained using a number of training data by a learning algorithm, thereby creating predefined rules of operation or artificial intelligence models configured to perform desired characteristics (or objectives). Such learning may be performed on the device itself where the artificial intelligence according to the present disclosure is executed, or it may be performed through a separate server and / or system. Examples of learning algorithms include supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but are not limited to the examples described above.

[0059] An artificial intelligence model can be composed of multiple neural network layers. Each of the multiple neural network layers has multiple nodes and weight values, and performs neural network operations through calculations between the results of previous layers and the multiple weights. The multiple weights possessed by the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, multiple weights can be updated so that the loss value or cost value obtained by the artificial intelligence model during the learning process is reduced or minimized. Additionally, to minimize the loss value or cost value, multiple weights can be updated in a direction that minimizes the gradient associated with the loss value or cost value. Artificial neural networks may include deep neural networks (DNNs), such as Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), Recurrent Neural Networks (RNNs), Restricted Boltzmann Machines (RBMs), Deep Belief Networks (DBNs), Bidirectional Recurrent Deep Neural Networks (BRDNNs), or Deep Q-Networks, but are not limited to the examples mentioned above.

[0060] A network is a network that serves as a transmission path for web pages; it may be a closed network such as a LAN (Local Area Network) or WAN (Wide Area Network), but it is desirable for it to be an open network such as the Internet. The Internet refers to a global open computer network structure that provides the TCP / IP protocol and various services existing at its upper layers, namely HTTP (HyperText Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System), SMTP (Simple Mail Transfer Protocol), SNMP (Simple Network Management Protocol), NFS (Network File Service), and NIS (Network Information Service).

[0061] Terminals can be implemented in various forms. For example, the terminals described in this specification may include mobile terminals such as smartphones, tablet PCs, PDAs, portable multimedia players, and MP3 players, as well as fixed terminals such as smart TVs and desktop computers.

[0062] In this specification, terms such as “comprising” or “having” are intended to specify the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof. When a component is described as being “connected” or “connected” to another component, it should be understood that it may be directly connected to or connected to that other component, or that there may be other components in between. On the other hand, when a component is described as being “directly connected” or “directly connected” to another component, it should be understood that there are no other components in between.

[0063] Meanwhile, a "module" or "part" for a component as used in this specification performs at least one function or operation. Furthermore, a "module" or "part" may perform a function or operation by hardware, software, or a combination of hardware and software. Additionally, a plurality of "modules" or a plurality of "parts," excluding a "module" or "part" that must be performed on specific hardware or on at least one processor, may be integrated into at least one module. A singular expression includes a plural expression unless the context clearly indicates otherwise.

[0064] In addition, power, power transmission, and control therefor for the following assembly configurations and embodiments, including "by control," follow conventional technology including terminals, applications, hardware control modules, etc., so they are omitted to avoid redundancy.

[0065] In addition, the operation embodiments and configurations described in a general manner without being explained in detail below follow the prior art and are omitted in order to focus on describing the purpose of the present invention and the resulting effects.

[0066] Furthermore, in describing the present invention, if it is determined that a detailed description of related known functions or configurations may unnecessarily obscure the essence of the invention, such detailed description is abbreviated or omitted.

[0067] The present invention relates to a location-based running course recommendation and sharing system that recommends personalized running courses based on a user's location information, manages related activity records, and allows sharing courses and records with other users.

[0068] The system may include a user interface module (100), a location receiving unit (200), a course recommendation module (300), a surrounding information provision module (400), a sensor unit (500), an activity recording module (600), a community module (700), and a database unit (800). Both the location receiving unit (200) and the sensor unit (500) may include GPS functions, but their roles are clearly distinguished in the present invention. The location receiving unit (200) is primarily used to determine the 'current location' as input data for the recommendation logic, whereas the sensor unit (500) is used to measure continuous trajectory and motion data occurring during the user's 'running activity'. That is, since the former is intended for 'obtaining static coordinates' and the latter for 'recording dynamic activities', they operate independently according to their respective purposes without functional overlap.

[0069] The above user interface module (100) is defined as a software module that provides and controls a visual and tactile interface for interaction between a user and a system. The above user interface module (100) performs the function of allowing the user to input standard information necessary for course recommendation through actions such as text input, button selection, and map manipulation, and the standard information is configured to include a target distance, desired difficulty level, and specific running themes such as 'riverside course', 'forest path course', and 'city night view course'. If a single route has multiple theme characteristics (e.g., a forest path created along a riverside), the above course recommendation module (300) assigns multiple theme tags ('riverside course', 'forest path course') to the route. When a user selects a specific theme to search, all routes with the corresponding theme tags are included in the search results to prevent omissions in the search. In addition, it performs the function of receiving final recommended route data from the course recommendation module (300) and visualizing and displaying it on a map as multi-layered information including not only a simple route line, but also color differentiation according to the 'overall route quality grade', warning icons generated by the 'route safety evaluation module', and information badges assigned by the 'personalized recommendation filtering module'. At this time, the user interface module (100) includes logic that dynamically controls the level of detail of the information displayed according to the scale level of the map. For example, at a scale level that displays a wide area, only the overall color (quality grade) of the route is displayed, and only when the user zooms in on the map and switches to a scale level that displays a detailed area is detailed information such as warning icons or information badges rendered. This prevents information overcrowding and allows the user to perceive information step by step, thereby increasing visibility.The above user interface module (100) transmits course recommendation criteria information entered by the user to the course recommendation module (300), and conversely receives the final processed visualization data from the course recommendation module (300), surrounding information provision module (400), etc.

[0070] The above location receiving unit (200) is defined as a hardware unit that receives GPS satellite signals to obtain the user's real-time geographical coordinates. The above location receiving unit (200) generates location data including the user's current latitude and longitude coordinates through a built-in GPS receiver. The location data performs the function of generating the data once to determine the current location at the time of a course recommendation request, and continuously at a set frequency (e.g., once per second) for real-time route tracking while the user is recording running activities. Additionally, the above location receiving unit (200) includes a function to dynamically adjust the data generation frequency in conjunction with accelerometer sensor data to optimize the user's smartphone battery consumption. Specifically, if the user's movement speed is detected to be below a preset threshold (e.g., 2 km / h) and the user is determined to be in a stationary state, the GPS data generation frequency is automatically reduced from once per second to once per 10 seconds, and returns to the original frequency when movement is detected again. The above location receiving unit (200) transmits the generated location data to the course recommendation module (300) and the activity recording module (600).

[0071] The above course recommendation module (300) is defined as a central processing software module that performs the core recommendation logic of the present invention. The course recommendation module (300) receives course recommendation criteria information from the user interface module (100) and current location information from the location receiving unit (200), and based on this, searches for a group of course candidates stored in the database unit (800). Internally, it performs complex operations to evaluate and filter each candidate path by sequentially or in parallel calling the 'Route Comprehensive Quality Evaluation Module', 'Route Safety Evaluation Module', 'Route Difficulty Determination Module', and 'Personalized Recommendation Filtering Module'. The method of calling the modules is determined by data dependency. For example, since the 'Route Safety Grade', which is the result of the 'Route Safety Evaluation Module', is used as an input value for the 'Route Comprehensive Quality Evaluation Module', these two modules are called sequentially. On the other hand, the 'Route Comprehensive Quality Evaluation Module' and the 'Route Difficulty Determination Module', which do not have data dependency on each other, can be called and processed simultaneously in parallel to increase the computational efficiency of the system. In addition, it includes a function to actively reflect environmental data in the recommendation results, such as automatically lowering the recommendation ranking of an outdoor course or generating a guide recommending indoor running to the user when real-time air quality data received via an external API is rated as 'very bad' according to the Ministry of Environment standards. Specifically, the 'lowering' is performed by applying differential penalties based on the air quality grade to the final score calculated by the 'Route Comprehensive Quality Evaluation Module'. For example, if the air quality is 'bad', 10% is deducted from the final score, and if it is 'very bad', 30% is deducted to recalculate the final recommendation ranking. The course recommendation module (300) receives data from the user interface module (100), the location receiving unit (200), and the database unit (800), and finally transmits the processed recommended route list to the user interface module (100).

[0072] The above surrounding information provision module (400) is defined as a software module that provides additional information about the surroundings of a route to enhance the convenience and usefulness of the recommended route. The above surrounding information provision module (400) receives final recommended route data from the above course recommendation module (300) and subsequently performs the function of selecting information according to the flow of the user's running experience, rather than simply listing surrounding facilities, by linking with an external map API such as OpenStreetMap. Here, the 'midpoint' is determined by an algorithm that prioritizes searching for a point that is 50% of the total estimated time required for the route, rather than simply a point that divides the total length of the route in half, and if there are no suitable amenities within the radius of that point, expands the search range to the 40% and 60% points. This is intended to more accurately predict the point at which the user's fatigue or physiological needs occur based on time rather than distance. For example, it performs the function of preferentially extracting and displaying information on 'parking lots' within a radius of 500m from the starting point of the route, information on 'drinking fountains' or 'restrooms' within a radius of 100m from the middle point of the route, and information on 'cafes' or 'restaurants' within a radius of 500m from the end point of the route on a map. The surrounding information providing module (400) receives route data from the course recommendation module (300), receives surrounding facility data from an external API, and finally transmits the processed information to the user interface module (100).

[0073] The sensor unit (500) is defined as a set of sensors embedded in a user's smartphone or a wearable device connected thereto, and as a hardware unit that measures the user's physical data. The sensor unit (500) includes a GPS, an accelerometer, a gyroscope, and a heart rate sensor, and these sensors are activated during the user's running activity to continuously measure each physical data and perform the function of converting it into a digital signal. The sensor unit (500) transmits the measured raw sensor data to the activity recording module (600) in real time.

[0074] The activity recording module (600) is defined as a software module that converts and records the user's running activity into quantitative data. The activity recording module (600) receives raw sensor data in real time from the sensor unit (500) and combines and calculates the received data to generate specific exercise metrics. For example, it performs calculations such as calculating the 'total distance traveled' by accumulating changes in GPS coordinates over time and calculating the 'average pace' by dividing it by the total time taken, and converts all generated metric data into a structured JSON data structure along with the user's identification information and activity time information. The activity recording module (600) receives data from the sensor unit (500) and transmits the activity record in JSON format, which is finally generated, to the database unit (800) for storage.

[0075] The community module (700) is defined as a social function software module that enriches the system's data through information sharing and interaction among users. The community module (700) provides an interface that allows users to share their activity records or favorite courses with other users, and other users can press 'Like' on the shared courses or leave specific feedback comments such as "This section is particularly dimly lit at night." This interaction data does not stop at simple social activities but performs a function that directly contributes to the system's self-learning mechanism. To quantify the collected text comments, the community module (700) includes a natural language processing (NLP)-based keyword extraction and sentiment analysis engine. For example, when a comment stating "the lighting is dim" is detected, the keyword "lighting" and a sentiment index of "negative" are extracted, and this is reflected in temporarily lowering the reliability of the "nighttime lighting consistency" data for that route. If this quantified feedback data accumulates above a certain threshold, it automatically triggers a re-evaluation of the route. The community module (700) reads activity records and course information from the database unit (800), displays them to the user, and transmits data such as shares, 'likes', and comments generated by the user back to the database unit (800) for storage.

[0076] The above database unit (800) is defined as a storage unit that stores and manages all static and dynamic data of the system of the present invention. The database unit (800) systematically stores user account information, course candidate groups in Geographic Information System (GIS) data format, road network data converted into a graph structure, user activity records in JSON format, and all data generated by community activities, and performs the function of quickly querying, adding, modifying, and deleting data according to the request of each module. The database unit (800) transmits and receives data bidirectionally with all software modules within the system and performs the role of maintaining data consistency and integrity.

[0077] As an example of an organic embodiment between the components, first, the user requests a 'high difficulty, 5km mountain course' through the user interface module (100). Subsequently, the location receiving unit (200) identifies the user's current location and transmits it to the course recommendation module (300). Next, the course recommendation module (300) queries five candidate routes from the database unit (800) based on the current location and request conditions. Subsequently, the course recommendation module (300) calls the 'route safety evaluation module' for each of the five routes. At this time, the database unit (800) checks that one route has a "wild boar sighting" comment posted via the community module (700), and determines that route as 'unsafe' to exclude it from the candidates. At this time, the course recommendation module (300) does not immediately exclude a route based on a single comment, but includes logic to evaluate the reliability of the comment. Specifically, the time at which the comment was written (timeliness), other users' 'like' or 'trust' reactions to the comment (empathy), and the commenter's past feedback reliability score are comprehensively considered. For example, a higher risk weight is assigned to a comment from one hour ago rather than a comment from one year ago, and to a comment with high empathy; the comment is judged as 'unsafe' only when this weight exceeds a preset threshold. Next, the 'Route Difficulty Judgment Module' and the 'Route Comprehensive Quality Evaluation Module' are executed for the remaining four routes to calculate evaluation scores. Subsequently, the 'Personalized Recommendation Filtering Module' is called to check that the battery of the smartphone linked to the user's sensor unit (500) is 30%, and the route with the highest evaluation score is excluded from the recommendation list because it is determined that 35% of the battery is required to complete the expected course. Finally, three routes are confirmed as the recommendation list, and the course recommendation module (300) transmits this data to the user interface module (100) and the surrounding information provision module (400).Next, the user interface module (100) displays three routes on the map in different colors according to their grade, and the surrounding information provision module (400) displays parking lot information near the starting point of each route as an icon. Finally, when the user selects a route and starts driving, the sensor unit (500) measures data and the activity record module (600) records it in real time and sends a command to save it to the database unit (800) after the drive ends.

[0079] The comprehensive route quality evaluation module of the present invention performs an evaluation aimed at assessing the overall quality of a route through multidimensional criteria rather than a single dimension, in order to provide the user with an optimal running experience. To this end, it processes commands using four core judgment criteria. Walking surface continuity is defined as an indicator representing the degree to which the flow is not interrupted by sidewalks, crosswalks, stairs, etc., throughout the entire running route. It is established with the purpose of ensuring that the user can maintain a consistent pace and engage in immersive running without unnecessarily stopping or stepping onto dangerous roads, and this generates data that directly affects the exercise efficiency and satisfaction of running. Micro-terrain roughness is defined as an indicator representing the degree of slight unevenness of the road surface caused by protrusions in sidewalk blocks, cracks in asphalt, or poor paving. It performs the function of reducing fatigue by preemptively avoiding the risk of minor injuries that may occur during running, such as ankle sprains or falls, and by providing the user with a stable landing sensation. The thermal stress index is defined as an indicator representing the degree of heat burden received by the user during running, by comprehensively considering factors such as the shade ratio around the route, radiant heat from the road surface (like asphalt), and the density of surrounding buildings. In particular, it processes calculations to minimize the risk of heat-related illnesses, such as heatstroke and dehydration, during daytime or summer running, and supports the continuation of exercise in a more comfortable environment. The route safety rating is a result calculated through a separate route safety evaluation module and is defined as a comprehensive grade representing the level of safety secured from external environmental factors, such as nighttime lighting, road surface friction, and the potential for crowd density. It provides data that ensures users can focus solely on driving in an environment where safety from external threats, such as crime and accidents, is secured.

[0080] The above module does not adopt a method of simply averaging the four judgment criteria arithmetically, but rather executes logic to resolve the problem where a single fatal flaw is diluted by several other advantages, thereby potentially exposing the user to risk. Accordingly, the present invention employs a stepwise comprehensive evaluation method that performs calculations by weighting and subtracting negative factors after a basic quality evaluation. As the technical basis for this method, a first-stage basic quality score calculation procedure is performed to evaluate the fundamental value of the route by first combining pedestrian surface continuity—a positive indicator directly related to driving satisfaction—and the route safety rating, a prerequisite for safety. Subsequently, a second-stage negative factor application procedure is performed, applying micro-terrain roughness and thermal stress index—negative indicators that increase driving risk—as penalty factors. This sends a signal to the user clearly warning of potential danger by significantly lowering the overall evaluation rating of the route if a specific risk factor exceeds a threshold. This stepwise approach performs calculations that enhance the reliability of the final evaluation result by clearly distinguishing the nature of each judgment criterion and reflecting it in the evaluation logic.

[0081] Data for each judgment criterion used by the above module for evaluation is obtained through specific methods. Pedestrian surface continuity data is obtained by linking OpenStreetMap 'footway' tag data with public GIS data from the National Geographic Information Institute to extract the number of crosswalks, discontinuous sections, and stairs within a 1km route. A normalized value of 0.9 or higher is assigned for pedestrian surface continuity when there are 0 to 1 discontinuous section within a 1km route, and 0.5 or lower when there are 5 or more; for example, a pedestrian surface continuity value of 0.8 is assigned to a 1km route where there are 2 crosswalks and 1 discontinuity. Micro-terrain roughness data utilizes accelerometer sensor data collected anonymously from the smartphones of other users of this system. Roughness indices for each section are generated by analyzing the frequency of vertical impacts exceeding a specific threshold during driving. For example, if the average number of vertical impacts during a 100m section is less than 1, the micro-terrain roughness is normalized to 0.1 or less; if it is 10 or more, it is normalized to 0.8 or more. If an average of 7 impacts are detected in a section, a micro-terrain roughness value of 0.65 is assigned to that section. Heat stress index data is calculated by receiving real-time temperature data from the Korea Meteorological Administration API, combining it with regional surface temperature data analyzed via satellite thermal infrared data, and additionally reflecting green space ratio data for parks and street trees obtained through GIS data. For example, under conditions of a real-time temperature of 32℃ at 2 PM on a midsummer afternoon, a forest path with a shade ratio of 70% or more is calculated to have a heat stress index of 0.2, while an asphalt road with almost no shade is calculated to have a heat stress index of 0.9. Route safety rating data is obtained by directly inputting the final output rating of the route safety evaluation module included in this system.

[0082] The final determination result derived by the 'Route Comprehensive Quality Evaluation Module' is utilized within the system. Based on the final evaluation result, a four-level grade—'Excellent', 'Recommended', 'Average', and 'Caution'—is assigned to the route, and a signal is transmitted to intuitively display this to the user through route colors on the map or separate icons. When a user searches for a route, the result list is sorted in order of highest quality grade calculated by this module, providing data that allows the user to prioritize access to high-quality routes without the need for separate filtering. Users process commands to select and search only for routes that satisfy their desired minimum quality level through filter options such as "Show only routes with a Recommended grade or higher."

[0083] In a scenario where a user in Seoul searches for a 5km running course at 3:00 PM in early August, the system performs a quality evaluation on 'Yangjaecheon Course A', one of the candidate routes. First, the system collects data for each judgment criterion regarding 'Yangjaecheon Course A', obtaining data with walking surface continuity at 0.9, micro-terrain roughness at 0.2, heat stress index at 0.3, and a route safety rating of 'Excellent'. Next, a base score is calculated based on the positive indicators of walking surface continuity and route safety rating, and the system converts the 'Excellent' route safety rating to 90 points out of 100. For walking surface continuity (converted from 0.9 to 90 points) and route safety rating (90 points), a weighted average is calculated by applying the importance weights (e.g., running satisfaction 40%, safety 60%) pre-set in the system, and the base quality score is calculated as = (90 points * 0.4) + (90 points * 0.6) = 90 points. Next, penalties are calculated for the negative indicators, micro-terrain roughness and thermal stress index. Since micro-terrain roughness (0.2) falls within the 'Good' range, the penalty is set to 5%, and since thermal stress index (0.3) also falls within the 'Comfortable' range, the penalty is set to 5%, resulting in a total penalty ratio of 10%. Subsequently, the total penalty ratio is subtracted from the basic quality score to calculate the final score, which is calculated as Final Score = 90 points * (1 - 0.10) = 81 points. The system determines that the final score of 81 points falls within the 'Recommended' grade range (e.g., 80 points to 94 points). Finally, the system assigns a 'Recommended' grade to 'Yangjaecheon Course A' and sends a command to display this on the user interface with a green path and a 'Recommended' icon.

[0084] If user sensor data for a specific route is insufficient, exception handling is performed to prevent gaps in the evaluation by using the average micro-terrain roughness value of the area or default values ​​based on the road paving material type in the OSM data as a substitute. Data collected from users is preprocessed using an outlier removal algorithm, and a mechanism is activated to gradually increase the reliability of the data when consistent data is collected from multiple users. When a user submits feedback regarding the evaluation results of a specific route, that feedback is used to temporarily increase the penalty weight for the relevant index of that route, and a command is sent to permanently reflect the changes in the evaluation database after verification by the operations team.

[0085] In the scenario of a pedestrian path next to a bicycle path along the Han River, the prior art presents the path as the "best recommended path" by considering only high user ratings and excellent path continuity, whereas the present invention evaluates the pedestrian surface continuity (0.95) and fine terrain roughness (0.1) as excellent, but calculates the heat stress index to be extremely high at 0.9 (danger). According to the "step-by-step comprehensive evaluation method" of the present invention, a high heat stress index value applies a high penalty of 40% or more to the total score, and as a result, the path is determined to be rated as "caution" and a signal is transmitted to the user along with a warning that "the risk of heat stress is very high during the day."

[0086] [Table 1]

[0087]

[0088] (Base Score = (SC100 * 0.4) + (RSI Score * 0.6), RSI: Excellent=100, Good=90, Average=70, Risky=20)

[0089] The penalty is summed according to the judgment rule tables of μR and HSD.

[0090] The following shows the result data of simulation examples for 10 virtual paths with various characteristics. No. 1 'Downtown Tree-lined Street' received a 'Normal' rating with a final score of 61.5 points, No. 2 'Riverside Bicycle Path' received a 'Caution' rating with a final score of 51.15 points, No. 3 'University Campus' received a 'Normal' rating with a final score of 68.4 points, No. 4 'National Park Forest Path' received a 'Caution' rating with a final score of 49.4 points, No. 5 'New Town Lake Park' received a 'Recommended' rating with a final score of 86.4 points, No. 6 'Old Downtown Alley' received a rating excluded from evaluation with a final score of 16 points, No. 7 'Within Apartment Complex' received a 'Normal' rating with a final score of 69.6 points, No. 8 'Coastal Promenade' received a 'Normal' rating with a final score of 72 points, No. 9 'Gangnam Teheran-ro' received a 'Caution' rating with a final score of 34.1 points, and No. 10 'Olympic Park' received a 'Recommended' rating with a final score of 87.3 points, respectively. Calculate.

[0091] [Table 1-1]

[0092]

[0093] A judgment rule is applied to set the penalty ratio to 5% if the normalized value of the micro-terrain roughness data is in the range of 0.0 to 0.2 and to 30% if it is 0.7 or higher, and to set the penalty ratio to 5% if the normalized value of the heat stress index data is in the range of 0.0 to 0.3 and to 40% if it is 0.7 or higher.

[0094] The judgment rules and penalty rates used in this module are not values ​​arbitrarily set by the applicant, but are established based on objective grounds and technical necessity. Each threshold is established based on the risk thresholds for heat-related illnesses presented in research data from the fields of sports science and public health, while the penalty rate for micro-terrain roughness is established based on the correlation between ankle joint fatigue, injury probability, and road surface vibration reported in the field of rehabilitation medicine. Since running is an activity that requires a combination of physical safety and environmental comfort, a logical procedure that strongly reflects risk factors as penalties provides an essential technical component to ensure user safety.

[0095] Unlike conventional technology, the present invention achieves clear technical effects by adopting specific judgment criteria and logical rules. By quantitatively evaluating potential risk factors such as 'micro-terrain roughness' and 'thermal stress index' and strongly reflecting them as penalties, it executes commands to effectively filter out risk paths that the user was unaware of. The 'step-by-step comprehensive evaluation method' logically combines data of different characteristics to prevent evaluation distortion, which ultimately increases the reliability of the recommended paths provided to the user and yields the effect of improving overall system satisfaction. Objective and multidimensional path evaluation provides a robust technical foundation that enables the provision of highly personalized recommendation services in the future by combining with the user's personal characteristics.

[0096] The 'Basic Quality Score' is defined as an initial score calculated by evaluating only the positive aspects of a route, and the 'Penalty Ratio' is defined as the percentage deducted from the Basic Quality Score due to the negative aspects of a route. The 'Threshold,' 'Basis,' 'Penalty Ratio,' etc. mentioned in this specification are not fixed constants; they are set based on public research data during the initial launch of the system, and thereafter, a machine learning model periodically analyzes millions of actual user driving data points and feedback accumulated through the system to automatically transmit commands for dynamic optimization tailored to regional and seasonal characteristics.

[0097] To set penalty thresholds for the ‘heat stress index (HSD)’ and ‘micro-terrain roughness (μR)’ of the present invention, a simulation was performed to investigate the ‘subjective satisfaction’ (out of 10 points) and ‘experience of injury / discomfort’ (Y / N) after driving by a group of actual users (50 men and women in their 20s to 40s) on 10 virtual paths with various characteristics.

[0098] [Table 1-2]

[0099]

[0100] The simulation results confirmed a strong correlation in which user satisfaction drops sharply to 6.0 points or lower and the rate of injury or discomfort experiences increases significantly to 28% or higher when the Heat Stress Index (HSD) or Micro-Terrain Roughness (μR) exceeds 0.7. This suggests that these values ​​represent significant inflection points that negatively impact the user's actual experience. Therefore, based on this empirical data, the present invention established a threshold for imposing the highest penalty in the range where HSD and μR exceed 0.7. In cases where 'satisfaction' and 'injury experience rate' show conflicting results in the simulation (e.g., unpaved forest path), the penalty logic of the present invention prioritizes user safety. Accordingly, when determining the penalty ratio, a higher weight is assigned to the more negative result between the two indicators (in this case, the 'injury experience rate') to calculate the final penalty. This is based on the technical philosophy that ensuring objective safety is a more important role of the recommendation system than user subjective satisfaction.

[0101] To verify the effect of the penalty threshold setting of the present invention on risk path filtering, a simulation was performed to compare the changes in recommendation rankings between the 'no penalty applied' (simple average) method and the 'penalty applied by the present invention' method.

[0102] [Table 1-3]

[0103]

[0104] As a result of the verification simulation above, the simple average method resulted in a problem where routes were recommended in the upper ranks (2nd and 5th) if their base scores were high, even if their HSD or μR values ​​were high. On the other hand, when the penalty threshold of the present invention was applied, the recommendation rankings of routes such as 'shadeless riverside', 'old sidewalk blocks', and 'unpaved forest path', where the HSD or μR exceeded 0.7, dropped significantly by 5, 4, and 3 steps, respectively, moving to the lower ranks. This demonstrates that the threshold and penalty logic set in the present invention have a technical effect of effectively identifying routes that appear good for tracking but contain actual risks and excluding them from the recommendation priority.

[0106] The route safety evaluation module of the present invention performs an evaluation aimed at classifying the objective safety level of a route by multi-facetedly analyzing various external environmental risk factors that may occur during running activities. To this end, it processes commands using five key judgment criteria by distinguishing the nature of potential risks. Nighttime lighting consistency is defined as the degree to which lighting maintains an illuminance level above a certain threshold without interruption throughout the entire route when driving at night or during the early morning hours. It is established with the purpose of ensuring actual personal safety by enabling users to clearly identify obstacles on the road surface to prevent falls, alleviating psychological anxiety regarding crimes that may occur in dark areas. The road surface friction estimate is defined as an inverse indicator representing the degree of risk that the road surface may become slippery due to external factors such as rain, snow, fallen leaves, and moss. It performs the function of preventing falls caused by slipping and resulting injuries such as fractures and bruises, and ensuring driving stability in response to changes in weather conditions. Crowd variability is defined as the possibility that crowds may suddenly congregate or traffic may be blocked on the route due to unforeseen events such as rallies, demonstrations, festivals, or construction. It processes calculations to prevent the driving flow from being forcibly interrupted and to protect users from risks such as collisions, loss of personal belongings, or safety accidents that may occur in crowded environments. The probability of encountering wild / stray animals is defined as the likelihood of encountering wild animals or stray dogs based on the ecological characteristics of the surrounding route. It calculates data to prevent users from experiencing fear due to sudden encounters with animals and to proactively prevent rare but possible attacks or collision accidents caused by animals. Emergency escape accessibility is defined as the distance from any point along the route to a safe base from which help can be requested, such as a hospital, police station, or 24-hour convenience store, in the event of an injury or emergency while driving.It provides data that enables rapid initial response in the event of an actual emergency, minimizes damage, and enhances users' psychological sense of security regarding remote routes.

[0107] The above module avoids a simple summation method that evaluates all safety standards with equal weight and executes logic to resolve the problem where a single critical risk factor can nullify all other advantages. Accordingly, the present invention adopts and performs calculations using a "step-by-step safety verification method that evaluates basic safety after critical risk filtering and adds mitigation factors." As the technical basis for the above method, a Stage 1 critical risk filtering procedure is performed, which prioritizes the review of crowd variability and the probability of animal encounters, which are difficult to predict and pose an immediate risk upon occurrence. If any of these exceed the "risk" threshold, the route is immediately determined to be "unsafe" without the need to evaluate other conditions, and the evaluation is terminated. Subsequently, for routes that passed Stage 1, a Stage 2 basic safety evaluation procedure is performed to assess the route's basic safety level by synthesizing nighttime lighting consistency and road surface friction estimates, which are constant safety factors. Finally, a Stage 3 safety correction procedure is performed, which reflects emergency escape accessibility—a mitigation factor that reduces damage in the event of an emergency—as an additional point in the final score.

[0108] Data for each judgment criterion used by the above module for evaluation is acquired through specific methods. Nighttime lighting consistency data is obtained by estimating the illuminance level for each section through the analysis of smart streetlight location and illuminance data from local government public data portals and EXIF ​​metadata from photos taken and uploaded by users at night. It is normalized to 0.95 if more than 95% of the route maintains illuminance above the average residential area level, and to 0.3 or lower if more than 30% of the route consists of dark sections. Road surface friction estimate data is obtained by receiving real-time precipitation and temperature data via the Korea Meteorological Administration API; if the temperature is 0℃ or lower, a 'freezing risk' flag is activated, and the slip risk is calculated by combining this with road surface type data from OSM. For example, an asphalt road on a clear day is calculated as 0.9, while a paved sidewalk immediately after rain in sub-zero weather is calculated as 0.2. Crowd variability data is obtained by monitoring specific keywords within a 500m radius of the route in real-time via social media APIs and generating a 'danger' signal if the frequency of such keywords increases sharply. The value is normally 0.1, but it is updated to 0.9 if 50 or more posts containing the keyword 'assembly' are detected within 10 minutes at a specific location. The probability of encountering wild or abandoned animals calculates the risk level by analyzing relevant text from local community apps and news data, as well as data on wildlife sighting warning zones provided by national parks and local governments. The value for urban areas is 0.05, but it is set to 0.8 if a residential route adjacent to a mountain is included in an official wild boar sighting warning zone. Emergency escape accessibility data calculates the shortest walking distance from every point on the route to the nearest safety point using a GIS database, and normalizes the result by taking the reciprocal of the maximum value among those distances. For example, if a safety point is within 200m of every point on the route, the value is calculated as 0.9, while if the farthest point is more than 1km away, it is calculated as 0.3 or lower.

[0109] The final determination result derived by the 'Route Safety Evaluation Module' is utilized within the system. The final 'Safety Grade' calculated by this module is directly transmitted as one of the core input data for the 'Route Comprehensive Quality Evaluation Module,' which processes commands to reflect this in the comprehensive route evaluation. If a user attempts night running on a route with a low night lighting consistency score, a signal is sent to display an active warning message as a pop-up, such as "This route has many dark sections at night and poses a safety risk." An interface is provided that allows users to directly set detailed safety-related conditions in the search filters, such as "Show only routes with the best night safety" or "Show only routes with low slip risk."

[0110] In a scenario where a user searches for a walking trail in a park on the outskirts of the city at 8 PM during the winter, the system performs a safety assessment. First, the system checks SNS APIs and animal sighting data to determine that crowd variability is 0.1 and the probability of encountering an animal is 0.2. Since both items are below the 'danger' threshold, the system processes a command to pass the filtering and proceed to the next step. Next, the system checks lighting data and weather data to calculate that nighttime lighting consistency is 0.5 and the road surface friction estimate is 0.4. According to the judgment rule table, 'nighttime lighting consistency' receives 30 points and 'road surface friction estimate' receives 20 points, and the base safety score is calculated as (30+20) / 2 = 25 points (converted to a maximum of 100 points). Next, the system calculates emergency escape accessibility through GIS data to determine that it is 0.8, and accordingly, adds 10 points as a correction score. Finally, the final safety score is calculated as 25 points + 10 points = 35 points, and the system determines that the final score of 35 points corresponds to a 'Caution' grade section. Lastly, the system assigns a 'Caution' grade to the route, transmits the grade to the 'Route Comprehensive Quality Evaluation Module,' and sends a command to the user to display a specific warning message, "This section is dark at night and there is a possibility of road icing."

[0111] If the real-time social media monitoring API fails to respond, exception handling is performed to treat the 'Crowd Volatility' index as 'Unknown' and conservatively downgrade the final safety rating of the corresponding route by one level. All real-time data has an expiration time; once this time elapses, a command is sent to automatically discard the data and re-request it to maintain the latest status. For locations where protests frequently occur or routes with repeated risk reports, a command is processed to add them to a blacklist by the administrator, automatically fixing their safety rating at the lowest level for a specific period.

[0112] In a scenario involving a walking trail known to local residents as a "quiet forest path," prior art presents it as a "recommended healing course" based on high ratings, whereas the present invention detects the keyword "wild boar" through news data analysis during the process of calculating the "animal encounter probability" and sets the probability value of the corresponding path above a "danger" threshold. In the "critical risk filtering" stage, this path is immediately determined to be "unsafe." Consequently, the path is not recommended to the user, or a signal is transmitted presenting a clear reason stating, "Not recommended due to the risk of wild animal sightings."

[0113] In the fatal risk filtering stage, if crowd variability or the probability of encountering animals exceeds 0.7, it is immediately judged as 'unsafe'. In the basic safety evaluation stage, 50 points are awarded if night lighting consistency is 0.8 or higher and 10 points if it is less than 0.5, and 50 points are awarded if the road surface friction estimate is 0.7 or higher and 5 points if it is less than 0.4. In the safety correction stage, a judgment rule is applied to add 10 points if emergency exit accessibility is 0.7 or higher.

[0114] The judgment rules used in this module are established based on objective grounds and technical necessity. Criteria classified as 'critical risk' share the common characteristic that the consequences are difficult to predict and, once they occur, can lead to serious casualties beyond the user's control; this follows objective standards consistent with the Ministry of the Interior and Safety's disaster safety guidelines. Since safety is determined by the most vulnerable parts, this invention possesses a technical necessity to perform the essential functions of a safety evaluation system.

[0115] [Table 2]

[0116]

[0117] This invention achieves the effect of maximizing the sophistication and realism of evaluation logic by separating risks into 'immediate / critical risks' and 'continuous / manageable risks' and evaluating them differentially. Going beyond a simple recommendation system, it achieves the technical effect of a 'risk management system' that proactively identifies potential risk factors and provides clear warnings to users. By combining real-time data with 'critical risk filtering' logic, it has the effect of actively responding to rapidly changing external situations to guarantee the safety of recommendations in real time.

[0118] 'Critical Risk Filtering' is defined as a top-priority safety verification step that takes precedence over all other conditions and halts the entire process if even a single condition is not satisfied, and 'Safety Correction' is defined as a procedure that incorporates additional positive factors capable of mitigating risks or reducing damage after the basic assessment is completed. The threshold values ​​presented in this specification are set based on public safety data as initial values, and after system operation, when data on 'risk reports' made by users accumulates, the machine learning model dynamically resets new threshold values ​​that show a high correlation with those reports, thereby operating with a learning mechanism that enables the system to autonomously strengthen safety standards in accordance with the actual environment.

[0119] To verify the validity of the ‘critical risk filtering’ threshold (0.7) of the present invention, data from 10 regions where actual ‘crowd volatility’ or ‘wild animal sighting’ safety accidents were reported over the past year were simulated to calculate ‘crowd volatility (CV)’ and ‘animal encounter probability (AEP)’ indices.

[0120] [Table 2-1]

[0121]

[0122] As a result of the simulation above, it was found that in 6 out of 7 cases (85.7%) where actual safety accidents occurred, the index calculated one hour prior to the accident exceeded 0.7. On the other hand, in the control group where no accident occurred and in cases of simple observation, the index did not exceed 0.7. This demonstrates that the threshold value of 0.7 is a significant value that has a high predictive correlation with the actual occurrence of risk. Therefore, setting 0.7 as the criterion for judging 'critical risk' in this invention is not an arbitrary value, but a reasonable setting intended to effectively pre-detect signs of risk based on actual accident data. Furthermore, to prevent excessive warnings during real-time operation, the path safety evaluation module determines the severity by additionally considering the 'persistence' and 'spreadability' of the risk. For example, it includes dynamic severity assessment logic that, even if the 'crowd volatility' index exceeds a threshold, if the frequency of the keyword decreases rapidly within 10 minutes, it is judged as 'temporary congestion' and the rating is lowered, and conversely, if the related keyword shows a pattern of geographical spread, it is judged as 'large-scale event' and the filtering is maintained.

[0123] To verify the effect of the 'critical risk filtering' logic of the present invention in fundamentally blocking risk path recommendations, a simulation was performed comparing the recommendation results of the 'no filtering applied' method and the 'filtering applied by the present invention' method.

[0124] [Table 2-2]

[0125]

[0126] As a result of the verification simulation above, when filtering was not applied, a serious problem occurred in which paths with actual risk signs (CV / AEP exceeding 0.7) were recommended to the user at the top level solely because of their high base safety. On the other hand, when the 'critical risk filtering' logic of the present invention was applied, the risky paths were all judged as 'unrecommended' regardless of their base safety and were excluded from the recommendation list at the source. This clearly demonstrates that the threshold-based filtering logic of the present invention plays a decisive role in protecting users from unpredictable critical risks.

[0128] The route difficulty determination module of the present invention performs an evaluation aimed at comprehensively assessing and grading the objective difficulty of a route so that users can select a route suitable for their exercise ability, training goals, and physical condition. To this end, it processes commands using three core judgment criteria based on measurable data rather than subjective feelings. Normalized elevation gain is defined as a value normalized by dividing the total cumulative elevation gain of the route by the total length of the route, and this functions as a key indicator representing the overall gradient of the route. It objectively predicts the total muscle strength and cardiorespiratory load required for a user to complete the route, thereby generating data that induces users who desire high-intensity interval training to select a high value, and users who desire recovery running to select a low value. Route substitutability is defined by the existence and ease of an alternative route to reach the target point when a specific section is unexpectedly blocked during the run, and it is quantified by applying concepts of graph theory. In addition to physical difficulty, it performs the function of incorporating the psychological pressure of having to complete the route in the event of an unexpected situation and the actual risk of isolation into the difficulty factors. The GPS signal attenuation rate is defined as the degree of positional error caused by GPS satellite signals being blocked or reflected in dense urban canyons or deep forests. It evaluates the 'technical difficulty' that impedes the accuracy of GPS-based pace measurement, distance calculation, and route guidance, and processes calculations that increase the difficulty by lowering the quality of training and raising the risk of route deviation, as it is difficult to accurately measure records on routes with unstable signals.

[0129] The above module performs calculations by adopting a 'multi-factor difficulty point summation method,' which determines the final difficulty grade by independently evaluating difficulty factors of different natures and summing them. As the technical basis for the above method, since 'normalized altitude ascent' represents physical load, 'path substitutability' represents psychological / situational load, and 'GPS signal attenuation rate' represents technical load, an independent evaluation procedure is performed to assign difficulty points by individually evaluating each characteristic. Subsequently, a point summation procedure is performed to calculate an overall difficulty level by summing the difficulty points calculated from each factor, encompassing cases where a single element is extremely difficult or multiple elements are complexly difficult.

[0130] Data for each judgment criterion used by the above module for evaluation is obtained through specific methods. Normalized elevation gain data is obtained by extracting the elevation of each coordinate point constituting the path based on publicly available elevation data from the U.S. Geological Survey. The cumulative elevation gain of the entire path is calculated and then divided by the total path length; for example, a walking trail in Namsan Park with a total length of 3 km and a cumulative elevation gain of 45 m has a value of 15 (45 m / 3 km), which is calculated to correspond to a normalized value of 0.95. Path substitutability data is obtained by converting OSM road network data into a graph structure and calculating the number of alternative paths based on the start and end points of the path. For example, an urban path in a grid-type road network is calculated as 0.9 because there are 5 or more alternative paths, while a single-path walking trail between a river and a cliff is calculated as 0.1 because there are 0 alternative paths. GPS signal attenuation rate data is obtained by utilizing 3D building modeling data to analyze building heights and densities around the path and calculating sky visibility. For example, an open park path is calculated as 0.05, while a sidewalk on Teheran-ro in Gangnam, which is densely packed with high-rise buildings, is calculated as 0.8.

[0131] The determination result finally derived by the 'Route Difficulty Determination Module' is utilized within the system. Based on the total sum of the final calculated difficulty points, the system assigns a four-level grade to the route: 'Beginner', 'Intermediate', 'Advanced', and 'Expert'. During a search, the system processes commands that allow the user to filter only routes suitable for their skill level, such as "Show only routes of Intermediate or lower." If the user sets an "Interval Training" goal in the system, it prioritizes recommending routes of 'Advanced' or higher with high 'Normalized Elevation Gain' points; if the user sets a "Long-Distance Recovery Running" goal, it sends a command to recommend 'Beginner' or 'Intermediate' routes with low 'GPS Signal Attenuation Rate' and 'Normalized Elevation Gain' points. For routes where the 'GPS Signal Attenuation Rate' points exceed a specific threshold, the system sends a signal to display a warning message stating, "The GPS signal on this route is unstable, so the record may be inaccurate," before the user begins running.

[0132] In a scenario where a user checks the difficulty of a perimeter trail (4km) around a hill in the city, the system performs a difficulty evaluation. First, the system analyzes data for each judgment criterion to calculate difficulty points. For normalized elevation gain, it calculates 40 difficulty points based on a cumulative elevation gain of 80m; for route substitutability, it calculates 20 difficulty points based on one alternative route; and for GPS signal attenuation rate, it calculates 10 difficulty points based on an attenuation rate of 0.4. Next, the system sums all difficulty points calculated from each factor to determine the total difficulty points as = 40 points + 20 points + 10 points = 70 points. Then, the system determines that the total difficulty points of 70 correspond to the 'Advanced' grade section. Finally, the system assigns the 'Advanced' grade to the route and sends a command to the user to display the result along with a detailed description stating, "This is a course for advanced users with many uphill sections and insufficient detours."

[0133] It processes commands to maintain data up-to-date by updating elevation data and road network data quarterly, and 3D building modeling data monthly only in areas with many new buildings. If multiple users leave feedback on a specific route, the system temporarily applies a correction factor to the difficulty points of that route and sends a command to modify the database after review by the operations team.

[0134] In a 5km route scenario that includes a flat riverside tunnel section, the prior art evaluates it as "very easy" by considering only elevation changes, whereas the present invention calculates the "normalized elevation gain" point close to 0 points, but calculates the "GPS signal attenuation rate" as very high at 0.95 due to the tunnel section, thereby obtaining a high difficulty point (18 points) in this category. Consequently, the total difficulty point is judged as "intermediate," and a signal is transmitted to the user providing specific information such as, "It is flat, but there are sections where the GPS signal is cut off, so it may be unsuitable for record measurement."

[0135] If the normalized elevation gain value is 2.0 or less, 0 to 10 difficulty points are assigned, and if it is 10.0 or more, 31 to 50 points are assigned. If the path substitutability value is 3 or more, 0 to 10 points are assigned, and if it is 0, 21 to 30 points are assigned. A judgment rule is applied to assign 0 to 5 points if the GPS signal attenuation rate value is 0.3 or less, and 16 to 20 points if it is 0.6 or more.

[0136] The judgment rules used in this module are established based on objective grounds and technical necessity. 'Normalized Altitude Ascent' is based on universal indicators used in exercise physiology to calculate training load, while 'Path Substitutability' and 'GPS Signal Attenuation Rate' utilize objective measures from network theory and satellite navigation systems, respectively. Since a successful running experience is determined not only by physical ability but also by psychological stability and the reliability of training data, comprehensively evaluating these multifaceted difficulty factors constitutes a technical necessity to provide truly useful information to the user. Furthermore, the present invention can calculate total difficulty points by personalizing the weights of each difficulty factor based on the user's profile and past activity records. For example, if 'Marathon Preparation' is set in the user's profile or if there are many long-distance runs with low pace variability in the past activity records, the weight assigned to 'Technical Difficulty (GPS Signal Attenuation Rate)' is increased by 1.5 times. Conversely, for users with extensive hiking or trail running records, the weight for 'physical difficulty (normalized elevation gain)' is increased by 1.2 times to provide customized difficulty information that matches the individual's training goals and characteristics.

[0137] [Table 3]

[0138]

[0139] This invention achieves the technical effect of overcoming the limitations of conventional fragmentary slope-based difficulty evaluation methods by separating, evaluating, and synthesizing 'physical,' 'psychological,' and 'technical' difficulty. By providing various difficulty factors, it establishes a foundation for recommending optimal routes tailored to the user's specific training objectives. Furthermore, by clearly informing the user in advance of technical pitfalls in a route that are difficult to know without actually traveling it, it has the effect of resolving the asymmetry between the user and route information.

[0140] 'Difficulty Points' are defined as values ​​converted into scores according to established rules for each difficulty judgment criterion, while 'Total Difficulty Points' are defined as the sum of all difficulty points, serving as the standard for determining the final difficulty grade. The point intervals in the 'Judgment Rule Table' and the threshold values ​​dividing the final grades are not fixed; initial values ​​are set through consultation with professional running coaches and groups of athletes. Following the system's launch, a machine learning model continuously analyzes the difference between the 'subjective difficulty feedback' data submitted by users and the 'objective difficulty grade' calculated by the system, automatically sending commands to dynamically correct the difficulty point calculation rules themselves.

[0141] To set the difficulty grade threshold according to the 'normalized altitude rise' of the present invention, a simulation was performed to analyze data from 10 representative routes extracted from the 'Strava' public dataset and the average exercise intensity (proportion of heart rate zone 4 or higher) of users who completed the routes.

[0142] [Table 3-1]

[0143]

[0144] As a result of the simulation above, a pattern was confirmed in which the proportion of users' average heart rates remaining in the high-intensity exercise zone (Zone 4) increased sharply from the 50% range to the 60% range, with the normalized altitude elevation value set at 10.0. In addition, there was a clear tendency for subjective difficulty to be perceived as 'intermediate' to 'advanced'. This suggests that a value of 10.0 is a practical threshold that distinguishes between 'challenging' difficulty and 'very difficult' difficulty at the level of general amateurs. Therefore, based on this analysis of actual exercise data, the present invention established an objective criterion for classifying paths with a value exceeding 10.0 as 'advanced'.

[0145] To verify how much more accurate information the multi-factor difficulty evaluation of the present invention provides compared to a single-factor (altitude) evaluation, a simulation was performed to compare the evaluation results for paths under specific situations.

[0146] [Table 3-2]

[0147]

[0148] The results of the verification simulation above clearly demonstrated the limitations of the altitude-based single-factor evaluation method, which incorrectly classified routes with high technical difficulty—such as "flat terrain including tunnels" or "flat terrain between buildings"—and routes with high psychological difficulty—such as "single-path coastal cliffs"—as "beginner." In contrast, the present invention comprehensively evaluated GPS attenuation rates and route substitutability to accurately determine all such routes as "intermediate." This proves that the multi-factor evaluation method of the present invention has a decisive effect in preventing the selection of incorrect routes unsuitable for training purposes by effectively informing the user of hidden difficulties that cannot be identified by altitude alone.

[0150] The personalized recommendation filtering module of the present invention performs an evaluation aimed at ultimately adjusting and filtering objectively superior routes evaluated by the system to match the unique circumstances and preferences of a specific user located "here and now." To this end, it processes commands using three key judgment criteria that reflect the user's internal and external circumstances in real time. Route novelty is defined as an indicator representing how little a recommended candidate route overlaps with the user's past driving history. It is established with the purpose of preventing psychological boredom and decreased motivation resulting from repeatedly driving the same route, and enhancing long-term service immersion by providing the user with the joy of new exploration. The battery level consideration range coefficient is defined as a suitability index representing the current battery level of the user's wearable device or smartphone, indicating whether GPS and sensor functions can be stably maintained during the expected route completion time. It performs the function of fundamentally blocking technical and safety issues, such as the interruption of exercise recording or the loss of emergency contact due to battery discharge during driving. The sudden event index is defined as the presence of real-time events currently occurring on or in very adjacent areas of the recommended candidate route that may interfere with driving or cause danger. It processes calculations that protect the user from unpredictable real-time risks occurring after the system evaluates the route, prevent wasted trips, and ensure the best driving experience.

[0151] The above module does not simply use a method of adding personalization elements to the score, but performs logic to resolve the problem that a route should not be recommended if the battery is low or a fire occurs on the route. Accordingly, the present invention adopts and performs calculations using a 'preference-based ranking rebalancing method after essential condition screening filtering.' As the technical basis for the above method, a first-stage essential condition screening filtering procedure is performed using a 'sudden event index' and a 'battery remaining capacity consideration range coefficient,' which are directly related to user safety and technical constraints, as 'pass / fail' gates. Routes that do not satisfy these two essential conditions are immediately excluded from the recommendation candidate group regardless of preference. Subsequently, a second-stage preference-based ranking rebalancing procedure is performed to dynamically rebalance the order within the recommendation list by applying a 'path novelty' index, which affects user motivation and satisfaction, only to safe routes that have passed the first stage.

[0152] Data for each judgment criterion used by the above module for evaluation is obtained through specific methods. Route novelty data is obtained by comparing the GPS route data of recommended candidate routes with the user's past three months of driving records database. The length of the overlapping segment between the two route data is calculated to determine the ratio of the overlapping segment to the total route length, and the novelty index is calculated as 1 - (overlap ratio). For example, if 1 km of a 5 km candidate route overlaps with a past driving route, the route novelty index becomes 0.8. To evaluate novelty more precisely, the above module applies weights that consider the location of the overlapping segment and the time of driving. Specifically, a higher overlap weight is assigned if the overlapping segment is located closer to the center than the starting or ending point of the route, and a higher time weight is assigned if it overlaps with a recent driving record within one week rather than a driving record from one year ago, in order to calculate the final overlap ratio. This is intended to more accurately reflect the degree of 'staleness' actually perceived by the user. Battery remaining capacity consideration range coefficient data is obtained by directly querying the real-time battery remaining capacity through the operating system API of the user's smartphone or connected wearable device. At the same time, the estimated completion time and the average battery consumption rate per hour are calculated based on the length of the route and the average gradient, and if (current battery remaining capacity - estimated consumption) is greater than or equal to the safety margin, it is determined as 'suitable (1)', and if it is less than, it is determined as 'unsuitable (0)'. For example, if the current battery is 40% and the estimated consumption is 20%, it is determined as 'suitable (1)', and if the current battery is 30%, it is determined as 'unsuitable (0)'. The sudden event index data is obtained by comprehensively monitoring the government public safety alert text message API, the traffic information center's real-time sudden event information API, and relevant keywords on social media in real time.If a related event is detected within a radius of 1 km of the candidate route, it is determined as ‘occurred (1)’, otherwise as ‘not occurred (0)’. For example, if a ‘large collision’ is reported at an intersection 500 m away from the candidate route, the sudden event index becomes ‘occurred (1)’.

[0153] The determination results ultimately derived by the 'Personalized Recommendation Filtering Module' are utilized within the system. This module receives the 'Candidate Path List' generated in the previous step as input and processes commands to generate and output a 'Personalized Recommendation List' to be displayed to the user. For paths excluded during the 'Essential Condition Selection Filtering' step, a signal is sent to the user explaining the clear reason, such as "Excluded from recommendations due to low battery." For paths whose rankings have changed due to 'Preference-Based Ranking Rebalancing,' data is provided to allow the user to intuitively recognize the reason for the ranking change by displaying a badge such as "A new path for you!"

[0154] In a scenario where a user is recommended three routes (A, B, and C) with a 'Recommended' rating, has a current battery of 35%, and has the 'New Route Search' option enabled, the system performs personalized filtering. First, Route A is judged as unsuitable for battery use and is immediately excluded from the recommendation list because its estimated battery consumption is 25%, and 35% - 25% = 10% is below the safety margin. Routes B and C pass the filtering as suitable for battery use and have no unexpected events. Next, 'Route Novelty' is calculated for the filtered Routes B and C, with Route B calculated as Novelty 0.2 and Route C as Novelty 1.0. Since the user has enabled the 'New Route Search' option, the recommendation ranking of Route C, which has a higher 'Route Novelty' index, is adjusted to be higher than Route B. Finally, the system sends a command to present Route C as the first choice and Route B as the second choice to the user, along with a notification message stating, "Route A has been excluded from recommendations due to low battery."

[0155] The command processes the average hourly battery consumption rate to improve prediction accuracy by continuously learning and updating personalized values ​​based on the user's device model, OS version, and usage habits. When analyzing social media keywords, it executes logic that minimizes false positives by assigning higher weight to information originating from trusted media outlets or public institution accounts and eliminating duplicate detections.

[0156] In a scenario where a large fire occurs at the park entrance at the moment a user searches for an objectively excellent urban park route, the prior art does not recognize the real-time situation and recommends the route as the top priority, whereas the present invention immediately detects 'fire' and 'road control' information through the public safety alert text API and traffic information API during the process of calculating the 'sudden event index'. In the 'essential condition selection filtering' step, the sudden event index of the route is determined to be 'occurred (1)', and a command is executed to immediately exclude it from the recommendation list regardless of the route's quality grade.

[0157] In the mandatory condition filtering step, if the battery level consideration range is 'unsuitable (0)' or the sudden event index is 'occurred (1)', it is immediately excluded from the recommendation. In the ranking readjustment step, a judgment rule is applied to give 10 ranking bonus points if the path novelty is 0.8 or higher, and 5 points if it is between 0.5 and 0.8.

[0158] The judgment rules used in this module are established based on objective grounds and technical necessity. 'Battery level,' 'spontaneous events,' and 'path novelty' are all based on objective data. Since even the best path is meaningless if the user's device is turned off or access to the path itself is impossible, the logical procedure of this invention, which prioritizes filtering these technical and situational constraints, possesses a technical necessity that completes the intrinsic value of the recommendation system.

[0159] [Table 4]

[0160]

[0161] This invention achieves the technical effect of providing the most valid information to users "right now" by adopting real-time unexpected event data as a filtering condition. By preemptively blocking the most critical negative experiences users may encounter, such as low battery or inability to access the site, it maximizes user trust in the system. By reflecting "path novelty" in ranking adjustments to satisfy users' desire for exploration and prevent boredom, it has the effect of inducing users to stay with the service for an extended period and engage actively.

[0162] 'Essential Condition Selection Filtering' is defined as a procedure that excludes unsuitable candidates at the source by first checking whether they satisfy the safety and technical constraints that serve as prerequisites for recommendations, and 'Preference-Based Ranking Reordering' is defined as a procedure that adjusts the recommendation order within the candidate pool that has passed the essential conditions by reflecting factors that influence the user's subjective satisfaction. The initial value of the battery safety margin is set based on the standards recommended by the operating systems of major smartphone manufacturers, and a command is automatically transmitted to optimize the safety margin by dynamically increasing it based on an analysis of actual discharge pattern data for each user's device model.

[0163] To verify the appropriateness of the ‘battery safety margin’ threshold (15%) of the present invention, 10 simulations were performed to measure the rate of degradation of GPS accuracy (horizontal accuracy, HDOP) according to battery level for 5 major smartphone models.

[0164] [Table 4-1]

[0165]

[0166] As a result of the simulation above, it was confirmed that for most models, when the remaining battery level drops below 15%, the HDOP degradation rate, which indicates GPS accuracy, exceeds 15% and increases rapidly, frequently leading to actual 'record failures.' In particular, when the remaining battery level was 10% or lower, record failures occurred in all models. This demonstrates that a battery level of 15% is a practical critical point where stable sensor operation is difficult to guarantee due to factors such as switching to low-power mode. Therefore, setting the safety margin to 15% in this invention is a reasonable setting based on experimental data to prevent the fatal experience of losing the user's valuable exercise records.

[0167] To verify how effectively the 'essential condition filtering' of the present invention prevents negative user experiences, a simulation was performed comparing recommendation results when 'filtering is not applied' and when 'filtering of the present invention is applied'.

[0168] [Table 4-2]

[0169]

[0170] As a result of the verification simulation above, when filtering was not applied, recommendations (Scenarios A, B, and D) that caused critical negative experiences, such as "recording interruption" or "wasted trips," frequently occurred because the user's current situation (battery level, surrounding events) was not taken into account. On the other hand, when the "essential condition filtering" logic of the present invention was applied, all paths that could cause such negative experiences were completely blocked in advance, and optimal alternatives appropriate to the situation were presented. Furthermore, in cases where negative factors were absent, such as in Scenario C, it was confirmed that reflecting "path novelty" enhanced positive experiences that increased user satisfaction. This demonstrates that the filtering logic of the present invention goes beyond merely avoiding risks and directly contributes to managing and improving the user's overall experience.

[0172] User A, residing in Songpa-gu, Seoul, executes the system of the present invention at 10:00 AM on a weekend to receive recommendations for running courses around their current location, and inputs conditions through the user interface module (100) such as a target distance of approximately 5 km, a desired difficulty level of intermediate, a theme park or forest path, and a new route search option. When the request is received, the location receiving unit (200) identifies User A's current location and transmits it to the course recommendation module (300), and the personalized recommendation filtering module (400) confirms that the current battery level is 55% from User A's smartphone linked with the sensor unit (500).

[0173] The course recommendation module (300) queries the database unit (800) based on the received information to search for three candidate routes with a theme of 'park' or 'forest path' within a radius of 5 km within a 3 km radius of the current location, namely 'Candidate 1: Olympic Park Circular Course', 'Candidate 2: Seokchon Lake Circular Course', and 'Candidate 3: Jamsil Hangang Park Riverside Course'. For each candidate route, the course recommendation module (300) internally calls the 'Route Difficulty Determination Module' and the 'Route Comprehensive Quality Evaluation Module' sequentially. As a result of determining the route difficulty, 'Candidate 1' is determined to be 'Intermediate', and 'Candidate 2' and 'Candidate 3' are determined to be 'Beginner'. As a result of evaluating the route comprehensive quality, 'Candidate 1' is evaluated as 'Excellent', 'Candidate 2' as 'Recommended', and 'Candidate 3' as 'Average'.

[0174] The course recommendation module (300) calls the 'Route Safety Evaluation Module' for three candidate routes that passed the first evaluation to verify their safety in real time. The module checks SNS APIs, traffic information APIs, animal sighting databases, etc., and in this process, through the community module (700), it checks information such as "Traffic is very congested around the east side of Seokchon Lake this morning due to a marathon event" among recent comments stored in the database unit (800). Based on this information, the 'Route Safety Evaluation Module' determines that the 'crowd variability' index of Candidate 2: Seokchon Lake Circular Course is above the 'danger' threshold and determines the route as 'unsafe'. The course recommendation module (300) receives this determination result and processes a command to exclude Candidate 2 from the final recommendation list.

[0175] The course recommendation module (300) calls the 'personalized recommendation filtering module' for the remaining two routes. First, as a result of the battery suitability check, both 'Candidate 1' and 'Candidate 3' are determined to be 'suitable' because the estimated battery consumption is within the safety margin. Next, as a result of the sudden event check, both routes are determined to be 'suitable' because there are no real-time sudden events. Next, since the user has activated the 'new route search' option, the past activity records of User A stored in the database unit (800) are queried, and as a result, it is confirmed that User A has driven Candidate 3: Jamsil Hangang Park riverside course more than 5 times, but has never driven Candidate 1: Olympic Park loop course. Accordingly, the 'route novelty' is calculated as 1.0 for 'Candidate 1' and 0.1 for 'Candidate 3', and a ranking adjustment is performed to set the recommendation ranking of Candidate 1, which has an overwhelmingly high 'route novelty', as the highest and Candidate 3 as the next highest.

[0176] The course recommendation module (300) transmits the finally confirmed recommendation list and additional information to each module. The user interface module (100) displays the first-ranked 'Olympic Park Circular Course' in dark green, signifying an 'Excellent' grade, along with a badge that reads "A new route for you!", and the second-ranked 'Jamsil Hangang Park Riverside Course' in yellow, signifying an 'Average' grade. At the same time, a notice message is displayed at the bottom stating, "The Seokchon Lake Course has been excluded from recommendations because it is currently crowded due to an event." Additionally, the surrounding information provision module (400) displays parking lot information near the starting point of the Olympic Park Course and convenience store information at the midpoint of the Jamsil Hangang Park Course as icons on the map. When the user selects the 'Olympic Park Circular Course' and starts driving, the sensor unit (500) and the activity record module (600) record distance, pace, heart rate, etc., in real time. After finishing the drive, the activity record module (600) saves the final record to the database unit (800). Finally, the user interface module (100) displays a record saving completion screen and sends a command to induce the user to leave a review through the community module (700), and to store this review in the database unit (800) so that it can be used for other users in the future.

[0177] Although preferred embodiments of the present invention have been illustrated and described above, the present invention is not limited to the specific embodiments described above. Various modifications are possible by those skilled in the art without departing from the essence of the invention as claimed in the claims, and such modifications should not be understood individually from the technical spirit or perspective of the present invention. Explanation of the symbols

[0179] User interface module (100) Location receiving unit (200) Course recommendation module (300) surrounding information providing module (400) Sensor unit (500) Activity log module (600) Community Module (700) Database unit (800)

Claims

Claim 1 A user interface module (100) that receives a running course recommendation request from an external source; a location receiving unit (200) that generates the user's current location data in response to the request of the user interface module (100); a database unit (800) that stores multiple candidate path data; and a course recommendation module (300) that determines a final recommended path by performing multi-stage evaluation and filtering on candidate paths stored in the database unit (800) based on the recommendation request received from the user interface module (100) and the current location data received from the location receiving unit (200); wherein the course recommendation module (300) includes the step of evaluating the objective quality and safety of the path through multiple judgment criteria; A location-based running course recommendation and sharing system characterized by recommending a user-customized running course, including the step of filtering and adjusting rankings based on the evaluation results based on the user's real-time situation and past history, wherein the course recommendation module (300) includes a comprehensive path quality evaluation module that evaluates the comprehensive quality of a path, and the comprehensive path quality evaluation module includes a first step of calculating a basic quality score of a path based on walking surface continuity data representing the physical continuity of the path and path safety grade data representing the safety of the external environment of the path;The method includes a second step of determining a final quality grade by deducting a score equal to a pre-set penalty ratio from the basic quality score calculated in the first step when the micro-terrain roughness data indicating the degree of road surface defects of the path and the heat stress index data indicating the degree of heat burden of the path each exceed a pre-set first threshold; wherein the micro-terrain roughness data is generated by analyzing the frequency of vertical impact occurrence of accelerometer sensor data collected from a plurality of user terminals, and the heat stress index data is generated by combining real-time temperature data, satellite surface temperature data, and GIS green space ratio data; wherein the path safety grade data is generated by a path safety evaluation module included in the course recommendation module (300); and the path safety evaluation module includes a first filtering step of determining, with the highest priority, whether either crowd variability data indicating the possibility of sudden crowd density around the path or animal encounter probability data indicating the probability of encountering wild or stray animals exceeds a pre-set second threshold, and if exceeded, immediately determining the path as 'unsafe' regardless of other conditions and terminating the evaluation;And, limited to routes that have passed the above-mentioned first stage, it includes a second evaluation stage that calculates a basic safety rating by synthesizing night lighting consistency data, which indicates the stability of night lighting, and road surface friction estimate data, which indicates the risk of slipping on the road surface according to weather conditions; the above-mentioned crowd variability data is characterized by ensuring user safety by ensuring that unpredictable real-time crowd density risks that cannot be identified by fixed geographical information alone are detected in advance and that the recommendation of the corresponding route is fundamentally blocked, by monitoring the frequency of appearance of specific keywords (assemblies, events, demonstrations) around the above-mentioned route in real time via a social media API and determining that the second threshold is exceeded if the frequency of appearance exceeds a preset number of times per unit time; and the route safety evaluation module further includes a third correction stage that, after the above-mentioned second-stage evaluation, determines the final route safety rating by adding a correction score to the above-mentioned basic safety rating if the accessibility is above a preset third standard value, based on emergency escape accessibility data analyzing the distance from all points on the above-mentioned route to the nearest safety base (hospital, police station); thereby, among routes with the same basic safety level, when an emergency situation occurs A location-based running course recommendation and sharing system characterized by increasing the reliability of recommendations by upgrading the safety rating of a responsive route, wherein the first threshold and penalty rate of the route comprehensive quality evaluation module are characterized by a machine learning model periodically analyzing subjective satisfaction feedback and injury experience data collected from multiple users through a community module (700), and dynamically updating the micro-terrain roughness data value or thermal stress index data value corresponding to the inflection point where the occurrence rate of negative experiences in the feedback data increases statistically significantly to the first threshold, thereby allowing the system to optimize its own evaluation criteria through actual user experience data and improve the accuracy of the evaluation over time. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 delete

Images