A clothing recommendation system based on real-time social network services

Abstract

A real-time social network service-based clothing recommendation system is disclosed. The real-time social network service-based clothing recommendation system according to an embodiment of the present invention comprises: a social sentiment analysis unit (110) that collects unstructured text data including social media (SNS) posts and comments of users in a specific region and time zone and quantitatively calculates a sentimental temperature correction value to correct the official temperature; a temperature calculation unit (120) that receives data related to the current or predicted official temperature from an official weather information provider and calculates the user's actual temperature by applying the user's sentimental temperature correction value to the official temperature; and a clothing recommendation unit (130) that searches a database of clothing photos (UGC) tagged with GPS and time information uploaded by users, selects clothing data worn in an environment closest to the user's actual temperature, ranks suitability, and provides recommended clothing along with images to the user terminal.

Description

Technology Field

[0001] The present invention relates to a real-time social network service-based clothing recommendation system, and more specifically, to a system that can provide integrated weather information and clothing recommendation services, analyzes unstructured text data collected from social media using natural language processing (NLP) techniques to quantify the user's actual perceived sentiment, and recommends clothing after correcting the perceived temperature by reflecting this in official weather data. Background Technology

[0002] With the recent proliferation of mobile devices, various types of clothing recommendation services have emerged. These services have evolved into structures that suggest clothing suitable for users based primarily on structured data such as weather information, seasonal information, age group, gender, and fashion trends. Conventional clothing recommendation technologies rely on official weather data, such as temperature, precipitation, and wind speed, and adopt relatively simple rule-based or statistical-based recommendation methods. While this ensures a certain level of objectivity, it has limitations in that it fails to adequately reflect the environment experienced by actual users.

[0003] In particular, the weather information utilized in conventional technology mostly involves applying official temperature or wind chill indicators provided by credible institutions such as the Korea Meteorological Administration. Since these official indicators are calculated based on average environmental conditions, there is a problem of discrepancy with the subjective wind chill felt by individual users. Even though the perception of cold or heat can vary depending on an individual's physical condition, activity level, psychological state, and surrounding environment under the same temperature conditions, conventional technology applies uniform standards without reflecting these variable factors.

[0004] Furthermore, while conventional clothing recommendation services may partially reflect personalization factors such as user preferences, past purchase history, and click patterns, they have a structure that does not consider unstructured factors such as the user's emotional state or social atmosphere at that time. For example, even under the same temperature and weather conditions, the clothing preferred by a user may differ depending on whether they are feeling psychologically depressed or fatigued or have high energy levels; however, existing systems are unable to reflect these emotional differences.

[0005] Furthermore, conventional technology fails to fully utilize collective experience data regarding outfits actually chosen and worn by a large number of users. While some services refer to user reviews or ratings, this data is limited to subjective evaluations or collected primarily based on satisfaction with the product itself; it suffers from the problem of failing to reflect the actual context of wearing—specifically, which outfits were suitable for certain temperatures or perceived environments.

[0006] In modern society, where social network services have become part of daily life, users frequently express their feelings about the weather, discomfort from heat or cold, and perceptions of humidity in the form of natural language text; however, conventional clothing recommendation technologies have not fully recognized the potential for utilizing such social data. In particular, unstructured text data, such as social media posts and comments, is currently not being incorporated into recommendation algorithms simply because it is unstructured, despite its excellent real-time and immediacy.

[0007] Furthermore, even when existing technologies compensate for perceived temperature, they often rely simply on personal settings or manual input. A structure that requires users to directly input settings such as "sensitive to cold" or "sensitive to heat" places an additional burden on the user and has limitations in that these settings fail to reflect emotional states that change over time. As a result, the problem of recommendation results deviating from actual situations occurs repeatedly.

[0008] Furthermore, conventional clothing recommendation systems suffer from the problem of opaque basis for their recommendations. Users lack understanding as to why a specific outfit was recommended, and trust in the service tends to decline if the recommendation results do not align with their actual experiences. This can be attributed to the fact that recommendation logic relies solely on official data and limited personal information, failing to reflect the shared experiences of a large number of users in real time.

[0009] While some research in existing technologies attempts to utilize crowd-based data, these methods often remain limited to surveys or post-evaluations, resulting in limitations such as poor real-time performance and difficulty in reflecting subtle regional and time-based variations. In particular, despite the fact that weather and perceived environmental conditions can fluctuate rapidly on an hourly and regional basis, conventional technologies suffer from structural constraints that prevent them from precisely reflecting these dynamic changes.

[0010] As such, conventional clothing recommendation technologies rely on structured information centered on official weather data, failing to reflect individual subjective feelings and social consensus, and lacking the perspective of sentiment analysis and feeling correction utilizing unstructured social data. Consequently, issues regarding limited recommendation accuracy and insufficient user satisfaction have been continuously raised.

[0011] Therefore, to solve these problems arising from conventional technology, the present invention aims to overcome the limitations of conventional technology by analyzing unstructured text data collected from real-time social network services to quantify users' emotional perceptions and reflecting this in official weather data to more accurately calculate the actual perceived temperature, and then providing clothing recommendations based on actual wearing data from multiple users. Prior art literature

[0012] Korean Patent Publication No. 10-2565310 (Registration Date: August 4, 2023) The problem to be solved

[0013] This invention aims to solve the problem where conventional clothing recommendation technologies rely solely on structured data such as official weather information, seasons, trends, and age, failing to reflect the user's actual perceived environment and emotional state. In particular, it seeks to improve upon the limitations of failing to account for variations in region, time of day, social atmosphere, and the subjective perceptions of multiple users, even under identical temperature conditions. Accordingly, the objective is to provide a system that analyzes unstructured text data generated in real-time from social network services to calculate an emotional perception correction value, and based on this, accurately recommends clothing that matches the actual perceived temperature. means of solving the problem

[0014] A real-time social network service-based clothing recommendation system according to one aspect of the present invention for achieving such objectives may comprise: a social sentiment analysis unit that collects unstructured text data including social media (SNS) posts and comments of users in a specific region and time zone, extracts positive and negative perceived sentiment keywords related to temperature and humidity within the text using natural language processing (NLP) techniques, and quantitatively calculates a perceived temperature correction value to correct the official perceived temperature based on the frequency and intensity of the extracted sentiment keywords; a perceived temperature calculation unit that receives data related to the current or predicted official perceived temperature from an official weather information provider and calculates the user's actual perceived temperature by applying the user's perceived temperature correction value to the official perceived temperature; and a clothing recommendation unit that searches a database of clothing photos (UGC) tagged with GPS and time information uploaded by users, selects clothing data worn in an environment closest to the user's actual perceived temperature, ranks suitability, and provides recommended clothing along with images to a user terminal.

[0015] In one embodiment of the present invention, the social sentiment analysis unit (110) may be configured to include: an unstructured text data collection module (111) that collects various forms of text data including posts, comments, and hashtags posted on a user's social network service; a sentiment keyword extraction module (112) that identifies and extracts sentiment expressions directly related to the perceived temperature and humidity from the text data obtained through the unstructured text data collection module (111); and a sentiment correction value calculation module (113) that classifies the sentiment keywords extracted from the sentiment keyword extraction module (112) into three categories of 'cold', 'hot', and 'humid', and derives quantitative temperature correction values ​​and humidity correction values ​​through the frequency of occurrence for each category.

[0016] In one embodiment of the present invention, the perceived temperature calculation unit (120) may be configured to include: an official weather information collection module (121) that collects a plurality of weather-related data including temperature, perceived temperature, humidity, and wind speed from a weather information provider; a perceived temperature correction value application module (122) that applies a perceptual temperature correction value derived from a perceived temperature correction value calculation module (113) to the official perceived temperature-related data received from the official weather information collection module (121); and an actual perceived temperature calculation module (123) that receives weather data corrected through the perceived temperature correction value application module (122) and finally calculates the perceived temperature that the user is likely to actually feel.

[0017] In one embodiment of the present invention, the outfit recommendation unit (130) may be configured to include: a database search module (131) that stores outfit-related information uploaded by users and searches for outfit photos (UGC) tagged with GPS and time information uploaded by users and transmits them to an outfit data selection module (132); an outfit data selection module (132) that selects an outfit worn in an environment closest to the user's actual perceived temperature calculated by an actual perceived temperature calculation module (123) from among the candidate outfit data set secured by the database search module (131); a suitability ranking module (133) that determines the recommendation priority to be provided to the user for a plurality of outfit candidate data selected through the outfit data selection module (132); and a recommended outfit providing module (134) that visually provides the final recommended outfit to the user terminal based on the priority result calculated by the suitability ranking module (133).

[0018] In one embodiment of the present invention, the clothing recommendation unit (130) can analyze and store information regarding the type, material, and number of layers of clothing included in clothing photos uploaded by users using artificial intelligence, and recommend a composite clothing set that matches the information based on the user's actual perceived temperature data.

[0019] The present invention may also provide a real-time social network service-based clothing recommendation method using the real-time social network service-based clothing recommendation system. A real-time social network service-based clothing recommendation method according to one aspect of the present invention is,

[0020] In one embodiment of the present invention, the configuration may include a social data collection step (S110); a perceived temperature correction value calculation step (S120); an actual perceived temperature calculation step (S130); an outfit data selection step (S140); a suitability ranking step (S150); and a recommended outfit provision step (S160). Effects of the invention

[0021] According to the present invention, by analyzing posts and comments from multiple users collected from a social network service using natural language processing techniques to quantify perceived sentiment, it is possible to more precisely calculate the actual perceived environment, which is difficult to reflect using only official weather data. Through this, it is possible to derive the actual perceived temperature that reflects subtle differences in perception by region and time of day, and to provide recommendations based on clothing data actually worn by multiple users in environments similar to that perceived temperature. As a result, the reliability and realism of recommendations are improved, and user satisfaction and service utilization are simultaneously increased. Brief explanation of the drawing

[0022] FIG. 1 is a block diagram showing a real-time social network service-based clothing recommendation system according to one embodiment of the present invention. FIG. 2 is a flowchart illustrating a real-time social network service-based clothing recommendation method according to an embodiment of the present invention. Figure 3 is a table showing the perceived temperature and emotion correction-based comments used in a real-time social network service-based clothing recommendation method according to an embodiment of the present invention, classified by key differentiation elements. Figure 4 is a table showing specific weather elements and adverse weather preparedness comments used in a real-time social network service-based clothing recommendation method according to an embodiment of the present invention. FIG. 5 is a table showing the appropriateness of clothing and UGC-based feedback comments used in a real-time social network service-based clothing recommendation method according to an embodiment of the present invention. FIG. 6 is a table showing travel-specific and rule-emphasizing comments used in a real-time social network service-based outfit recommendation method according to an embodiment of the present invention. Specific details for implementing the invention

[0023] Preferred embodiments of the present invention will be described in detail below with reference to the drawings. Prior to this, terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, but should be interpreted in a meaning and concept consistent with the technical spirit of the present invention.

[0024] Throughout this specification, when it is stated that one component is located "on" another component, this includes not only cases where one component is in contact with another component, but also cases where another component exists between the two components. Throughout this specification, when it is stated that a part "includes" a component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0025] FIG. 1 shows a block diagram illustrating a real-time social network service-based outfit recommendation system according to one embodiment of the present invention, and FIG. 2 shows a flowchart illustrating a real-time social network service-based outfit recommendation method according to one embodiment of the present invention.

[0026] Referring to these drawings, the real-time social network service-based clothing recommendation system (100) according to the present embodiment is equipped with a social sentiment analysis unit (110) that performs specific roles, a perceived temperature calculation unit (120), and a clothing recommendation unit (130), so as to quantify the emotional state of the user on the social network service as well as the season, weather, fashion information, trends, age, and tendencies, calculate a perceived sentiment correction value, and then recommend and provide the most suitable clothing among the clothing data uploaded by multiple users based on the perceived sentiment correction value, thereby ultimately providing a service that can innovatively improve the accuracy of clothing recommendations and user satisfaction.

[0027] Hereinafter, each component constituting the real-time social network service-based clothing recommendation system (100) according to the present embodiment will be described in detail with reference to FIGS. 1 to 6.

[0028] Detailed explanation of the social sentiment analysis unit (110)

[0029] The social sentiment analysis unit (110) is a core component for quantitatively analyzing the sentiments related to temperature and humidity perceived by users based on large-scale unstructured text data generated in a real-time social network service environment. The social sentiment analysis unit (110) is designed to go beyond simply collecting text and to derive perceived information closely related to the actual environment by selectively processing user expressions concentrated in specific regions and time periods. Through this, it performs the role of technically absorbing subjective perceived elements that cannot be reflected by official weather data.

[0030] The social sentiment analysis unit (110) has a structure that takes the flow of collective sentiment as the subject of analysis rather than the individual sentiments of users. By comprehensively analyzing multiple posts and comments posted in the same region and at similar times, it is configured to mitigate deviations caused by temporary emotional expressions or personal tendencies and to derive a perceived sentiment closer to social consensus. This collective sentiment-based analysis structure serves as a technical basis for improving the reliability of perceived temperature correction.

[0031] The social sentiment analysis unit (110) operates by organically combining natural language processing-based preprocessing, semantic analysis, and sentiment classification processes. It is designed to minimize semantic distortion by considering slang, colloquial expressions, and emphasis expressions included in unstructured text, and is configured to enable context-centered analysis rather than a simple keyword matching method. Through this, expressions containing subtle nuances, such as "cold," "too hot," and "muggy," can be systematically interpreted.

[0032] In addition, the social sentiment analysis unit (110) structures the analysis results into units of perceived sentiment keywords and provides them for use in subsequent modules. The sentiment analysis results are not simply classified as positive or negative, but are organized into a data form that includes the perceived direction and intensity associated with temperature and humidity. This structure functions as an essential preprocessing step for quantitatively calculating the perceived temperature correction value.

[0033] Consequently, the social sentiment analysis unit (110) serves as a central hub that converts unstructured data generated from social network services into meaningful information that can be directly used to calculate the perceived temperature. Through this, sentiment-based perceived elements that were excluded in conventional technology can be systematically reflected, and a foundation is provided to fundamentally improve the accuracy and realism of clothing recommendations.

[0034] Detailed description of the unstructured text data collection module (111)

[0035] The unstructured text data collection module (111) is configured to form input data for the social sentiment analysis unit (110) and is configured to collect various forms of text data, such as posts, comments, and hashtags posted on a social network service. The unstructured text data collection module (111) has a structure that can secure data in real-time or near-real-time by utilizing a public API, a crawling interface, a streaming data channel, etc.

[0036] The unstructured text data collection module (111) is designed to filter data based on specific regions and time zones, rather than simply collecting random data. By comprehensively considering location information included in posts, upload time, user-set region information, etc., it selectively collects only data directly related to the current environment for clothing recommendations. This increases the environmental suitability of the data to be analyzed and minimizes unnecessary noise data.

[0037] The unstructured text data collection module (111) is characterized by securing text containing various forms of expression in its original form. By collecting text without excluding spelling errors, repetitive characters, emoticons, interjections, etc. from the analysis, the actual user's perceived emotions can be reflected more richly. This collection method provides a favorable foundation for precisely analyzing perceived intensity and emotional expression in the subsequent natural language processing stage.

[0038] In addition, the unstructured text data collection module (111) manages data timestamps to track changes in sentiment over time. Since it is possible to analyze how sentiment expressions change over time in the same region, changes in sentiment caused by sudden temperature changes or deterioration of weather conditions can also be quickly reflected. This serves as a factor that compensates for the problem of delayed reflection of weather information that occurred in conventional technology.

[0039] In this way, the unstructured text data collection module (111) plays a role in directly connecting the real-time and on-site nature of the social network service environment to the tactile analysis system. Through this, the reliability and timeliness of the input data used for tactile temperature correction can be simultaneously secured, and a foundation is formed to improve the responsiveness and accuracy of the overall outfit recommendation system.

[0040] Detailed explanation of the perceived sentiment keyword extraction module (112)

[0041] The perceived sentiment keyword extraction module (112) is configured to identify and extract sentiment expressions directly related to perceived temperature and humidity from text data obtained through the unstructured text data collection module (111). The perceived sentiment keyword extraction module (112) is designed to perform not only word-level analysis but also sentence-level semantic interpretation based on natural language processing technology.

[0042] The sensory sentiment keyword extraction module (112) is configured to recognize not only explicit expressions such as "cold" or "hot," but also expressions that indirectly indicate the state of feeling, such as "took out the padding," "turned on the air conditioner," and "sweated," as sensory sentiment keywords. This allows for a wide range of user expressions to be reflected and reduces the possibility of missing sensory analysis.

[0043] In addition, the perceived emotional keyword extraction module (112) also analyzes the intensity of the emotional expression. It is designed so that the perceived intensity is evaluated differently even for the same keyword by considering elements such as emphatic adverbs, repetitive expressions, and exclamation marks. For example, by distinguishing between simple expressions of discomfort and extreme expressions of discomfort, the precision of calculating the perceived temperature correction value is increased.

[0044] The perceived sentiment keyword extraction module (112) converts the extracted keywords into a structured data form and transmits them to a subsequent module. Each keyword includes attribute information regarding whether it is a temperature-related sentiment or a humidity-related sentiment, and its frequency of occurrence and relative intensity are recorded together. This data structure serves as a technical basis that enables quantitative calculation during the process of calculating the perceived temperature correction value.

[0045] Consequently, the perceived sentiment keyword extraction module (112) performs the core function of converting unstructured text into structured data that can be used for perceived temperature correction. Through this, social sentiment information that could not be utilized in conventional technology can be systematically reflected, and the analysis accuracy of the entire system is substantially improved to enable clothing recommendations closely related to the actual user's perception.

[0046] Detailed description of the perceived temperature correction value calculation module (113)

[0047] The perceived temperature correction value calculation module (113) performs the role of converting the temperature and humidity deviations actually felt by the user into quantitative correction values ​​based on the perceived emotional keywords transmitted from the perceived emotional keyword extraction module (112). The perceived temperature correction value calculation module (113) is configured to calculate the correction value by considering both the frequency of keyword appearance and the emotional intensity, rather than simply the presence or absence of keywords. Through this, it is possible to distinguish between cases where emotional expression appears temporarily and cases where it is continuously repeated.

[0048] The perceived temperature correction value calculation module (113) classifies the extracted perceived emotional keywords into three categories: 'cold', 'hot', and 'humidity'. Each category is set as an emotional axis directly related to the perceived temperature or humidity, and even when multiple categories appear simultaneously within the same text data, they are calculated independently. This classification structure enables a more realistic reflection of complex perceived environments.

[0049] The perceived temperature correction value calculation module (113) quantifies the relative frequency and intensity of keywords for each category to derive a temperature correction value and a humidity correction value. For example, if keywords related to 'hot' are repeated with high frequency and are evaluated as having high intensity, a temperature correction value that is added to the official perceived temperature is calculated. Conversely, if expressions related to 'cold' are dominant, a correction value that is subtracted is calculated.

[0050] The perceived temperature correction value calculation module (113) includes a normalization process for multiple user data so that extreme expressions of specific individuals do not excessively affect the correction value. By statistically processing sentiment data collected in the same region and time zone on a group basis, distortion caused by outliers or temporary emotional expressions is minimized. This ensures the stability and reproducibility of the correction value.

[0051] In this way, the perceived temperature correction value calculation module (113) converts unstructured emotional data into numerical environmental correction parameters, thereby providing a link that can combine social perception, which could not be reflected in conventional technology, with official weather information. This serves as a core foundation that enables the perceived temperature calculation unit (120) to derive a more realistic perceived temperature.

[0052] Detailed explanation of the perceived temperature calculation unit (120)

[0053] The perceived temperature calculation unit (120) performs the function of calculating the perceived temperature that the user is likely to actually feel by applying the perceived temperature correction value derived from the perceived temperature correction value calculation module (113) to the official weather information. The perceived temperature calculation unit (120) is designed to perform a more precise calculation of the perceived temperature through a weighted application structure based on environmental conditions, rather than a simple merging of official data and emotional data.

[0054] The perceived temperature calculation unit (120) utilizes data such as current or predicted perceived temperature, actual temperature, and humidity obtained through the official weather information collection module (121) as input values. By applying a sentiment-based temperature correction value and a humidity correction value, different perceived results can be derived depending on the region and time of day, even under the same official temperature conditions. This is a feature that is clearly distinguished from the existing perceived temperature calculation method based on average values.

[0055] The perceived temperature calculation unit (120) can adjust the application range of the perceived temperature correction value according to environmental conditions. For example, when a specific weather event such as rain, strong wind, or heat wave occurs, it can increase the weight of official weather data, and during normal times, it can operate by relatively increasing the weight of the perceived temperature correction value. This structure allows the reliability of the perceived temperature calculation to be maintained flexibly according to the environment.

[0056] In addition, the perceived temperature calculation unit (120) may not provide the calculated actual perceived temperature as a single numerical value, but may convert it into a range value or grade value suitable for clothing recommendation and transmit it to the subsequent clothing recommendation unit (130). This allows the recommendation algorithm to not react sensitively to specific numerical errors and to utilize criteria suitable for actual wearing judgment.

[0057] Consequently, the perceived temperature calculation unit (120) combines official weather information and social sentiment data to structurally resolve the perceived discrepancy problem that occurred in conventional technology. Through this, it is possible to stably provide a perceived temperature that matches the environment the user actually feels, and it plays a pivotal role in fundamentally improving the accuracy of clothing recommendations.

[0058] Detailed description of the official weather information collection module (121)

[0059] The official weather information collection module (121) is configured to provide basic input data for the perceived temperature calculation unit (120) and is configured to collect various weather-related data, such as temperature, perceived temperature, humidity, and wind speed, from a reliable weather information provider. The official weather information collection module (121) reliably secures real-time or predicted data through API integration methods, data feed reception methods, etc.

[0060] The official weather information collection module (121) is designed not only to collect current weather conditions but also to collect short-term forecasts and time-based prediction data. This enables the calculation of a suitable wind chill temperature for a specific time period, even when a user requests clothing recommendations based on their planned time of going out or activity. This overcomes the limitations of conventional technology, which was limited to information at the current time.

[0061] The official weather information collection module (121) selectively collects detailed weather data based on regional information. It is configured to obtain weather information that matches the actual activity range of the recommended user as closely as possible by utilizing various location standards, such as administrative district units and GPS coordinate units. This minimizes errors caused by regional average values.

[0062] Additionally, the official weather information collection module (121) organizes the collected weather data into a standardized data structure and transmits it to the wind chill temperature calculation unit (120). It includes a normalization process so that even data received from different weather information providers can be processed in a consistent format, and enables stable service provision through a correction logic in the event of data omission or error.

[0063] In this way, the official weather information collection module (121) performs the role of providing objective criteria for calculating the perceived temperature. By combining this official weather data with a sentiment-based correction value, the present invention overcomes the problem of dependence on averaged weather information that was present in conventional technology and establishes a technical foundation that enables clothing recommendations that correspond to actual perception.

[0064] Detailed description of the perceived temperature correction value application module (122)

[0065] The perceived temperature correction value application module (122) performs the function of applying the perceived temperature correction value derived from the perceived temperature correction value calculation module (113) to the official perceived temperature data received from the official weather information collection module (121). The perceived temperature correction value application module (122) is designed with a structure that applies the correction target item and the correction intensity separately, rather than a simple numerical addition or subtraction method. Through this, the influence of temperature and humidity elements on the perceived temperature can be reflected independently.

[0066] The perceived temperature correction value application module (122) dynamically adjusts the application weight by considering the reliability of the perceived emotional correction value. It operates by increasing the reflection weight of the emotional correction value when a sufficient amount of social sentiment data is secured in a specific region and time period, and relatively maintaining the weight of official weather information when the data is limited. This structure contributes to ensuring the stability of the perceived temperature calculation result.

[0067] The perceived temperature correction value application module (122) includes buffering logic to prevent excessive correction from occurring even in situations of rapid weather change. For example, in situations where the official temperature changes rapidly, the range of application of the emotional correction value is limited to prevent the perceived temperature from being unrealistically distorted. This prevents the problem of overcorrection that may occur when applying emotional information in conventional technology.

[0068] In addition, the perceived temperature correction value application module (122) has a cumulative correction structure to reflect the trend of perceived change over time. By applying the correction in a way that smoothly transitions the direction and intensity of the correction by comparing the correction result of the previous time period with the sentiment data of the current time period, the phenomenon of the perceived temperature changing rapidly is mitigated. Through this, a more natural recommendation standard can be provided to the user.

[0069] Consequently, the perceived temperature correction value application module (122) functions as an intermediate processing layer that effectively combines official weather data and sentiment-based perceived data. This structurally resolves the discrepancy between the official perceived temperature and the actual perceived temperature that occurred in conventional technology, and provides a foundation for the actual perceived temperature calculation module (123) to derive realistic results.

[0070] Detailed explanation of the actual perceived temperature calculation module (123)

[0071] The actual perceived temperature calculation module (123) receives weather data corrected through the perceived temperature correction value application module (122) and performs the role of finally calculating the perceived temperature that the user is likely to actually feel. The actual perceived temperature calculation module (123) is designed not to simply output numerical calculation results, but to derive a perceived standard optimized for clothing recommendations.

[0072] The actual perceived temperature calculation module (123) can not only calculate the corrected perceived temperature value in the form of an absolute value, but also convert it into a certain range or a perceived grade. This prevents recommendation instability caused by minute numerical differences during the clothing recommendation process and provides a useful criterion for making a practical judgment on wearing. This structure alleviates the recommendation consistency problem that occurred in conventional technology.

[0073] The actual perceived temperature calculation module (123) can also calculate the perceived temperature considering the user's activity time. By reflecting predicted weather data and sentiment correction values ​​for not only the current time but also the near-future time, it can calculate a perceived temperature suitable for the planned time of going out or the duration of the activity. Through this, it provides practical recommendation criteria rather than one-off recommendations.

[0074] Additionally, the actual perceived temperature calculation module (123) transmits the calculation result along with metadata to the clothing recommendation unit (130). The metadata includes regional information, time zone information, and the degree of correction applied, enabling the recommendation algorithm to understand the context of the perceived temperature calculation. This serves as a factor that increases the explainability of the recommendation result.

[0075] In this way, the actual perceived temperature calculation module (123) provides a final judgment result by synthesizing official weather information and social sentiment data, thereby fundamentally reducing recommendation errors caused by perceived discrepancies in conventional technology. Through this, it plays a key role in providing users with a more reliable standard for clothing recommendations.

[0076] Detailed explanation of the clothing recommendation section (130)

[0077] The clothing recommendation unit (130) performs the function of selecting and providing the most suitable clothing to the user based on the perceived temperature information derived from the actual perceived temperature calculation module (123). The clothing recommendation unit (130) is configured to provide highly realistic recommendation results by utilizing actual user wearing data, rather than simply providing recommendations based on theoretical standards.

[0078] The clothing recommendation unit (130) searches a database for clothing photo data tagged with GPS and time information uploaded by users. In this process, by prioritizing the selection of clothing data taken in a region and time zone similar to the current or planned activity environment of the recommended user, the consistency between the perceived temperature and the actual wearing environment is improved. This improves the abstract recommendation problem that occurred in conventional technology.

[0079] The clothing recommendation unit (130) performs a suitability evaluation on the selected clothing data. It calculates a similarity based on the difference between the perceived temperature at the time the clothing was applied and the current user's actual perceived temperature, and determines the recommendation priority based on this. This suitability-based structure is differentiated from simple popularity-based or trend-based recommendations.

[0080] Additionally, the clothing recommendation unit (130) can utilize information on the type, material, and number of layers of clothing included in the clothing photo through artificial intelligence analysis. This enables recommendations for complex clothing combinations rather than single clothing, and allows for precise recommendations that take into account warmth and breathability based on the perceived temperature. This is an advanced recommendation method that was not provided in conventional technology.

[0081] As a result, the clothing recommendation unit (130) combines a realistic standard of actual perceived temperature with actual wearing cases of multiple users to provide a highly reliable clothing recommendation to the user. This overcomes the limitations of uniform and abstract recommendations in conventional technology and provides the effect of substantially improving user satisfaction and service utilization.

[0082] Detailed description of the database search module (131)

[0083] The database search module (131) performs the function of quickly and precisely searching for necessary candidate data from a database storing clothing-related information uploaded by users, so that the clothing recommendation unit (130) can perform recommendations based on actual wear. The database search module (131) is not merely at the level of viewing image files, but has a structure that configures search conditions based on GPS information and upload time information tagged in clothing photos (UGC), and selectively loads data sets that match the region and time zone conditions transmitted from the actual perceived temperature calculation module (123). Through this, the problem of generalized recommendations, which commonly occurred in conventional technology—that is, the problem of repeating the same recommendation if it belongs to the same seasonal category—is reduced, and a candidate group reflecting the wearing context of a specific region and specific time zone can be secured.

[0084] The database search module (131) manages location metadata included in the uploaded data in a multi-layered manner to increase the precision of the location-based search. For example, it stores and manages auxiliary information such as regional codes mapped to administrative district units, area identifiers based on living areas, and areas of stay estimated from the user's movement path, in addition to the GPS coordinates themselves, thereby mitigating the problem of candidate data being omitted due to a single coordinate error during the search. Furthermore, since there are sections within the same city with slightly different perceived environments, such as coastal areas, downtown areas, and inland areas, the database search module (131) applies a search policy that variably expands or contracts the regional range to stably secure wearing cases within a range sufficiently similar to the recommended environment.

[0085] The database search module (131) is designed to search by setting a time window centered on the recommended target time, rather than matching a single time for time information. This takes into account that the perception of morning and night may differ even on the same day, and that the perception of change may occur rapidly during seasonal transitions even within the same time period. The database search module (131) manages multiple time fields, such as the time of shooting, the time of device creation, and the time of posting, in addition to the time of upload, to estimate the actual time of wearing as closely as possible, and reduces recommendation errors caused by time discrepancies by forming a search candidate group based on the estimated result.

[0086] The database search module (131) can increase the precision of the search by querying text information linked to images and structured attribute information together. If hashtags, comments, place names, mentions related to clothing, and expressions of physical sensation included in user posts are stored in the database, the database search module (131) can include these additional texts in the search conditions to filter candidate data that reflects the intention to wear. In addition, if artificial intelligence analysis results such as the type of clothing, material, and number of layers mentioned in claim 5 are stored in advance, the database search module (131) can quickly search for representative layer combinations that are common in a specific temperature range and transmit them to the clothing data selection module (132).

[0087] In this way, the database search module (131) has a search structure that minimizes response delay to respond to a real-time recommendation environment. Since search speed and accuracy are simultaneously required in an environment where large-scale UGC data is continuously accumulated, the database search module (131) is based on an indexing structure centered on location and time keys, and secures timeliness and real-time accuracy by maintaining candidate data for frequently referenced regions and time zones in a cache form or by applying a policy that prioritizes searching for the latest uploaded data. As a result, the database search module (131) goes beyond the structure of conventional technology that only performs recommendations based on structured data, and stably provides a data base that can utilize actual wearing cases for real-time recommendations.

[0088] Detailed description of the clothing data selection module (132)

[0089] The clothing data selection module (132) performs the function of precisely selecting clothing worn in an environment closest to the user's actual perceived temperature calculated by the actual perceived temperature calculation module (123) from among the set of candidate clothing data obtained by the database search module (131). The clothing data selection module (132) operates not through simple filtering, but by estimating the environmental context at the time the candidate data was generated and calculating the similarity with the current recommended target environment. Through this, it provides a selection criterion that can solve the problem in conventional technology where the wearing results vary depending on humidity, wind, and regional characteristics even if the temperature is the same.

[0090] The clothing data selection module (132) can utilize weather history data linked with location and time metadata included in the uploaded data to configure the environmental parameters of the candidate data. That is, the official weather conditions at the time each candidate clothing photo was taken or uploaded, or the wind chill value corresponding to that time, are stored in conjunction, and through this, it is possible to objectively compare in what kind of wind chill environment the candidate clothing was selected. This structure enables much more detailed selection than the method that relied simply on seasonal tags such as "winter outfit," and allows for differentiation of recommended candidates by distinguishing between above-freezing and below-freezing temperatures, and between dry and humid days, even within the same winter.

[0091] The clothing data selection module (132) can directly reflect the sentiment-based correction results, which are a feature of the present invention, into the selection process. That is, by evaluating whether the current sentiment-based correction value or social atmosphere at the time the candidate data was generated is similar to the current sentiment-based atmosphere, it is possible to distinguish between days when people actually feel colder and days when they feel less cold, even if the official temperature is similar. Such sentiment-matching-based selection is an area that was impossible in conventional technology and enables the refinement of candidate clothing in a way that reduces the discomfort felt by actual users.

[0092] Additionally, the clothing data selection module (132) can exclude data unsuitable for recommendation by including quality and validity verification of the image itself. For example, images that are blurry to the point where a person's clothing cannot be identified, images that appear to be taken indoors, specific promotional images, and data that is highly likely to cause bias due to repeated uploads by the same user can be excluded from the recommendation candidate group. These selection criteria are essential for maintaining the reliability of recommendations based on actual wearing cases and have the effect of mitigating the bias problem associated with simple popularity-based recommendations in conventional technology.

[0093] Furthermore, the clothing data selection module (132) can be expanded to perform selection at the level of complex clothing combinations. When artificial intelligence analysis results regarding clothing type, material, and number of layers are stored for candidate images, the clothing data selection module (132) prioritizes selecting combinations that provide similar warmth or breathability in the current perceived temperature range. Through this, beyond simply recommending a "coat," it is possible to construct a practical wearing set that includes underwear, tops, outerwear, and accessories, and the selection performance can be enhanced in a way that solves the abstract recommendation problem, which had low user satisfaction in conventional technology.

[0094] Detailed description of the suitability ranking module (133)

[0095] The suitability ranking module (133) performs the function of determining the recommendation priority to be provided to the user for multiple candidate outfit data selected through the outfit data selection module (132). The suitability ranking module (133) does not simply rank based on a single criterion, but forms a multidimensional evaluation structure centered on the similarity between the actual perceived temperature and the environment in which the candidate data was generated. Through this, the recommendation results are designed to comprehensively reflect actual wearability without being biased toward specific factors.

[0096] The suitability ranking module (133) first utilizes the difference between the perceived temperature value derived from the actual perceived temperature calculation module (123) and the perceived temperature at the time each candidate outfit data was generated as a key evaluation indicator. A higher base score is assigned as the difference in perceived temperature decreases, thereby ensuring that outfits worn under conditions most similar to the current environment are prioritized for exposure. This evaluation method enables much more precise ranking calculation than the conventional method of simply matching temperature ranges.

[0097] The suitability ranking module (133) considers not only the perceived temperature but also the consistency with the perceived humidity and the emotional atmosphere. For example, even if the official temperature is similar, the suitability for wearing may differ between an environment where the social sentiment analysis result is "muggy" and an environment where it is "cool," so the similarity of the emotional perception correction value is reflected as an additional evaluation factor. Through this, recommendation errors caused by perceived inconsistencies that occurred in conventional technology can be reduced.

[0098] Additionally, the suitability ranking module (133) may include an evaluation of the composition information of the candidate outfit. The type, material, and number of layers of the clothing identified through artificial intelligence analysis determine whether they provide appropriate warmth or breathability in the current perceived temperature range and assign weights. For example, since the suitability may differ between wearing a single outer garment and wearing multiple layers even within the same perceived temperature range, this difference is reflected in the ranking calculation.

[0099] In this way, the suitability ranking module (133) determines the recommendation priority by comprehensively reflecting environmental similarity, emotional matching, and clothing composition suitability. Through this, the recommendation structure provided in the conventional technology in order of simple popularity or trend is improved, and the reliability and practicality of the recommendation are simultaneously improved by ensuring that outfits optimized for the actual experience environment are displayed at the top.

[0100] Detailed description of the recommended outfit provision module (134)

[0101] The recommended outfit providing module (134) performs the function of visually providing the final recommended outfit to the user terminal based on the priority result calculated by the suitability ranking module (133). The recommended outfit providing module (134) is configured to provide recommendations based on images of actual users wearing the outfit, rather than a simple text listing method, so that the user can intuitively predict the wearing result. This is clearly distinguished from the recommendation method in the prior art, where only abstract descriptions were provided.

[0102] The recommended outfit providing module (134) organizes multiple outfits selected in the top rankings into a list or card format and outputs them to the user terminal. Each recommended item may be provided with an image of the outfit, along with brief information regarding the perceived temperature range and the wearing environment in which the outfit is judged to be suitable. Through this, the user can understand the context of the recommendation results and more easily make a choice that suits their situation.

[0103] The recommended outfit providing module (134) can flexibly adjust the display method of recommendation information according to the screen size, resolution, and interface environment of the user terminal. It is configured so that the recommended outfit is clearly recognized in various terminal environments such as smartphones, tablets, and wearable devices, and an image-centered or summary information-centered output method can be selectively applied. This acts as a factor that increases usability in an actual service environment.

[0104] Additionally, the recommended outfit providing module (134) can be designed to utilize user response data for subsequent learning and improvement. The user's act of selecting or saving a specific recommended outfit, or ignoring the recommendation results, can be used as feedback data to improve recommendation accuracy in the future. Through this, the recommended outfit providing module (134) functions as part of a cyclic structure that continuously improves recommendation quality, going beyond a one-time output function.

[0105] Consequently, the recommended outfit provision module (134) serves as the final point of contact that delivers the results derived through the analysis, selection, and ranking processes to the user in the most understandable form. Through this, the limitations of conventional technology, such as a lack of trust in recommendation results and low utilization, are overcome, and the effect of providing a practical outfit recommendation service that users can actually refer to and utilize is achieved.

[0106] Below, a real-time social network service-based clothing recommendation method (S100) using a real-time social network service-based clothing recommendation system (100) is described in detail.

[0107] The real-time social network service-based clothing recommendation method (S100) according to the present embodiment comprises a social data collection step (S110), a perceived temperature correction value calculation step (S120), an actual perceived temperature calculation step (S130), a clothing data selection step (S140), a suitability ranking step (S150), and a recommended clothing provision step (S160).

[0108] Detailed explanation of the social data collection step (S110)

[0109] The social data collection step (S110) is a step for securing unstructured text data that serves as the basis for calculating the emotional perception correction value in the real-time social network service-based outfit recommendation method (S100). By setting SNS posts and comments written by users based on specific regions and time zones as the collection targets, the social data collection step (S110) has a structure that secures emotional expressions directly connected to the actual recommendation target environment, rather than merely collecting general emotional expressions. This establishes an input basis that can complement the reality where conventional technology relied only on the average characteristics of official weather data.

[0110] The social data collection step (S110) acquires unstructured text data using public APIs, search interfaces, streaming channels, or equivalent data collection means provided by social network services to ensure the real-time and on-site nature of the data. In the social data collection step (S110), not only the body of the post but also comments, hashtags, short exclamations, repetitive characters, and emoticons can be treated as valid signals for estimating the intensity and direction of the perceived expression, and the data is configured to be collected in its original form with minimal loss. This collection structure enables the precise capture of subtle perceptual differences during the subsequent natural language processing.

[0111] The social data collection step (S110) performs location-based filtering to enhance the regional consistency of the collection target. If location tags or GPS information explicitly included in the post exist, they are utilized directly; if location information is not specified, the collection scope can be adjusted by utilizing place names mentioned in the post, user profile regions, or regional information estimated from the context of post creation. This method is advantageous for reflecting the local perceived environment of a specific region and contributes to reducing recommendation errors that occurred due to regional averaging in conventional technology.

[0112] The social data collection step (S110) sets a collection window based on the time of post creation to ensure time zone consistency and assigns priority to data generated within a time range similar to the time of the recommendation request. Since the perceived atmosphere can differ significantly between morning and night even on the same day, and changes in perception can occur rapidly in a short period, the social data collection step (S110) is configured to secure highly up-to-date data by flexibly adjusting the time criteria. This can mitigate the time lag issues that occurred in conventional technology due to weather forecast update cycles or delays in data reflection.

[0113] The social data collection step (S110) may include management operations to reduce unnecessary noise in order to maintain the quality and reliability of the collected data. For example, since promotional phrases, automatically generated sentences, and event promotional text unrelated to tangible feelings can impede the accuracy of subsequent analysis, the social data collection step (S110) may filter out abnormal data based on text repetition patterns, abnormal account activity, and the excessive appearance of identical phrases. This management structure ensures that the objective of the present invention, which is to reflect tangible sentiment based on multiple users, is stably achieved.

[0114] Detailed explanation of the perceived temperature correction value calculation step (S120)

[0115] The perceived temperature correction value calculation step (S120) is a step for quantifying perceived sentiment from unstructured text data obtained in the social data collection step (S110) and calculating a sentiment perceived temperature correction value to correct the official perceived temperature. The perceived temperature correction value calculation step (S120) functions as a core step for converting the perceived responses actually expressed by users into numerical correction parameters in order to solve the problem where conventional technology performed recommendations that were out of touch with user perception by utilizing only structured data of official weather information.

[0116] In the step of calculating the perceived temperature correction value (S120), positive or negative perceived sentiment keywords related to temperature and humidity are extracted from the text using natural language processing techniques. At this time, not only direct expressions such as "cold" or "hot" but also indirect perceived expressions such as "took out padding," "sweating," "muggy," and "hands are cold" can be interpreted based on context and reflected as perceived sentiment keywords. Through this, various linguistic perceived signals that were missed in conventional technology are absorbed, thereby increasing the precision of the correction value calculation.

[0117] The perceived temperature correction value calculation step (S120) classifies the extracted perceived sentiment keywords into 'cold', 'hot', and 'humidity', and analyzes the frequency of occurrence and sentiment intensity for each category. The frequency of occurrence can be used as an indicator of the degree of collective empathy, and the sentiment intensity can be quantified through linguistic reinforcement elements such as emphatic adverbs, repetitive expressions, and exclamation marks. The combination of these two indicators allows for a more accurate reflection of realistic perceived differences than simply the presence or absence of keywords.

[0118] The step for calculating the perceived temperature correction value (S120) may include normalization and outlier reduction processing to ensure stability based on collective data. To prevent exaggerated expressions or temporary complaints from specific users from distorting the overall correction value, the calculation logic may be configured to mitigate excessive values ​​based on the data distribution of the same region and time zone, while emphasizing common perceived patterns that appear repeatedly. This ensures that the correction value reflects the actual flow of environmental perception rather than being swayed by one-off emotions.

[0119] In this way, the emotional temperature correction value derived through the perceived temperature correction value calculation step (S120) is applied to official weather data and used as a standard for calculating the actual perceived temperature. Consequently, the perceived temperature correction value calculation step (S120) provides a key technical effect that resolves the discrepancy between official data and user perception in conventional technology and enables clothing recommendations that reflect differences in perception by region and time of day.

[0120] Detailed explanation of the actual perceived temperature calculation step (S130)

[0121] The actual perceived temperature calculation step (S130) is a step that calculates the perceived temperature likely to be actually felt by the user by reflecting the emotional perceived temperature correction value derived in the perceived temperature correction value calculation step (S120) into official weather information. The actual perceived temperature calculation step (S130) functions as a key step to resolve the problem where conventional technology performs recommendations based only on average indicators of official weather data, which deviates from the user's actual perception. In particular, based on the premise that perceived responses vary depending on the region and time of day even under the same official temperature conditions, social sentiment-based correction values ​​are systematically incorporated into the perceived temperature calculation process.

[0122] In the actual perceived temperature calculation step (S130), current or predicted official perceived temperature data obtained through the official weather information collection module (121) is used as an input value. Here, the official perceived temperature data does not refer only to a single temperature value, but can be collected in a form that includes additional information such as humidity, wind speed, and precipitation that affect the perceived temperature. The actual perceived temperature calculation step (S130) maintains the objectivity of these official weather elements as a basic axis, while also applying the results of social sentiment analysis to reduce the perceived temperature discrepancy, and the calculation logic is configured in this way.

[0123] The actual perceived temperature calculation step (S130) can process the effects of the temperature correction value and the humidity correction value on the perceived temperature separately by reflecting the application results of the perceived temperature correction value application module (122). For example, if the frequency of occurrence of the 'humid' category is high and the perceived muggy feeling is strong, the humidity correction value may be reflected in a direction that raises the perceived temperature, and if the occurrence of the 'cold' category is dominant, it may be reflected in a direction that lowers the perceived temperature. Such a separate application structure reflects a more realistic perceived response than the simple temperature standard recommendation of the conventional technology.

[0124] The actual perceived temperature calculation step (S130) may include an operation to adjust the strength of the application of correction values ​​by considering the reliability and distribution characteristics of the input data. In regions and time zones where social data is sufficiently secured, stable calculation is possible even if the reflection of sentiment correction values ​​is increased; however, in cases where data is insufficient or repetitive expressions by specific accounts appear excessively, buffering processing is required to prevent over-correction. In such situations, the actual perceived temperature calculation step (S130) suppresses rapid fluctuations in the perceived temperature through methods such as normalization, outlier reduction, and the application of gradual transitions, and provides consistent recommendation criteria to the user.

[0125] In addition, the actual perceived temperature calculation step (S130) organizes the calculated perceived temperature into a form suitable for clothing recommendations and transmits it to the subsequent step. Since transmitting the perceived temperature as a single numerical value can lead to excessive variations in recommendation results due to minute differences, the actual perceived temperature calculation step (S130) may convert and transmit it into a certain range value, a perceived grade value, or a perceived interval value. This mitigates the recommendation instability issues observed in conventional technology and provides perceived temperature information based on criteria that are easy for actual users to refer to.

[0126] Detailed explanation of the clothing data selection step (S140)

[0127] The clothing data selection step (S140) is a step of selecting clothing data suitable for recommendation from a database of clothing photos tagged with GPS and time information uploaded by users, based on the user's actual perceived temperature calculated in the actual perceived temperature calculation step (S130). To solve the problem where conventional technology performs recommendations based on product categories or seasonal labels and provides results detached from actual wearing situations, the clothing data selection step (S140) uses wearing examples verified in real environments as the central basis for recommendation. Through this, recommendations based on realistic wearing results can be provided to the user, rather than abstract advice.

[0128] In the clothing data selection step (S140), candidate clothing data is secured first through the database search module (131), and among the candidate data, data that satisfies location and time zone conditions similar to the current recommended environment is selected first. GPS tags are not processed merely as simple coordinates, but are mapped to administrative district units or living area units to evaluate the similarity of the regional range, and upload time information is filtered based on the same time zone or similar time window to refine it so that the perceived environment at the time of wearing is similar. This process alleviates the regional deviation problem that occurred due to weather average values ​​in conventional technology.

[0129] The clothing data selection step (S140) can improve selection precision by comparing the environmental context at the time the candidate clothing data was generated with the current perceived temperature. To this end, weather history or perceived temperature information corresponding to the time and location of candidate data generation can be stored in conjunction, and a method can be applied to prioritize the selection of candidates that are below a certain standard by calculating the difference from the actual perceived temperature calculated at the present. This method improves the simplification of conventional recommendations by enabling the distinction of situations where the actual perceived temperature varies significantly even within the same season.

[0130] In addition, the clothing data selection step (S140) can enhance perceived consistency by indirectly reflecting social sentiment-based correction results in the selection process. For example, even if the official temperature is similar, a wearing example filmed in an environment where the social sentiment analysis result is strongly 'hot' or 'humid' may be more suitable for the current environment; therefore, it is possible to prioritize and maintain candidates with high similarity by comparing the perceived correction characteristics at the time the candidate data was generated with the current correction characteristics. This forms a differentiating feature of selecting wearing examples based on sentiment matching, which was not provided by conventional technology.

[0131] The clothing data selection step (S140) may include validation to prevent the inclusion of data that impairs recommendation quality. Low-quality images where clothing is not identifiable, instances of wearing that are presumed to have been taken indoors and are unrelated to the external temperature, promotional posts, and data that causes bias through repeated posting of the same clothing may be excluded during the selection process. Accordingly, the clothing data selection step (S140) operates in a way that ensures the reliability and consistency of recommendations based on actual wear, moving away from the reality where conventional technology relied simply on trend information or general rules.

[0132] Detailed explanation of the suitability ranking step (S150)

[0133] The suitability ranking step (S150) is a step for determining the recommendation priority to be provided to the user regarding multiple candidate outfit data obtained through the outfit data selection step (S140). The suitability ranking step (S150) does not simply list candidate data, but has a structure that quantitatively compares and evaluates how suitable each candidate outfit is for the current user environment based on the actual perceived temperature. Through this, subjective or inconsistent recommendation problems that frequently occurred in conventional technology are improved.

[0134] In the suitability ranking step (S150), the difference between the perceived temperature value derived in the actual perceived temperature calculation step (S130) and the perceived temperature at the time each outfit data was generated, or the corresponding environmental information, is used as the core evaluation criterion. A higher suitability score is assigned as the difference in perceived temperature decreases, thereby placing outfits worn under conditions most similar to the current environment at the top of the priority list. This comparison method is differentiated from conventional methods that simply recommend outfits based on seasonal classification or temperature ranges.

[0135] The suitability ranking step (S150) can calculate a ranking by considering the results of humidity and emotional sentiment correction in addition to the perceived temperature. For example, even if the official temperature is similar, the suitability for wearing may differ between an environment where the social sentiment analysis result is strongly 'humid' and one where it is not; therefore, the emotional sentiment characteristics of the environment where the candidate outfit was generated are compared with those of the current environment and reflected in the ranking calculation. Through this, recommendation results with enhanced consistency in perception can be derived.

[0136] Additionally, the suitability ranking step (S150) may include an evaluation utilizing the clothing composition information of the candidate outfit. Weights are assigned by determining how well the information on the type, material, and number of layers of the clothing, identified through artificial intelligence analysis, matches the warmth or breathability required at the current perceived temperature. This evaluation structure complements the limitations of conventional technology that judged based solely on a single outer garment and adjusts the ranking in a direction that enhances actual wearing satisfaction.

[0137] In this way, the suitability ranking step (S150) calculates the recommendation priority by comprehensively reflecting environmental similarity, emotional alignment, and clothing composition suitability. Through this, the recommendation structure based on popularity or trends that occurred in conventional technology is improved, and the reliability and accuracy of recommendations are substantially enhanced by ensuring that outfits most suitable for the actual perceived environment are displayed at the top.

[0138] Detailed explanation of the recommended outfit provision step (S160)

[0139] The recommended outfit provision step (S160) is the final step for delivering the priority results calculated in the suitability ranking step (S150) to the user. The recommended outfit provision step (S160) is configured to provide intuitive information centered on images of outfits actually worn by users, rather than providing the analysis results as simple numbers or text. Through this, users can more easily understand the recommendation results and make a choice that suits their situation.

[0140] In the recommended outfit provision step (S160), multiple outfits determined to be in the top rank are displayed on the user terminal in the form of a list or cards. Each recommended item may be provided with an image of the outfit, along with a brief description of the perceived temperature range in which the outfit is deemed suitable and the wearing environment. This method of providing information contributes to mitigating the problem of low user trust caused by unclear grounds for recommendations in conventional technology.

[0141] The recommended outfit provision step (S160) can flexibly adjust the output method according to the environment of the user's device. The screen composition and information density are adjusted so that the recommended outfit is clearly recognized in various device environments, such as smartphones, tablets, and wearable devices, and image-centered or summary information-centered expressions can be selectively applied. This serves as a factor that enhances usability and accessibility when applied to actual services.

[0142] Additionally, the recommended outfit provision step (S160) can utilize the results of user interactions to improve the quality of subsequent recommendations. Actions such as a user selecting or saving a specific recommended outfit, or ignoring a recommendation, can be used as feedback data to improve future recommendation accuracy, and this data contributes to supplementing the suitability evaluation criteria. Through this, the recommended outfit provision step (S160) functions as part of a continuous quality improvement structure rather than being limited to a one-time output.

[0143] Consequently, the recommended outfit provision step (S160) performs the role of delivering the recommendation results derived through the analysis, selection, and ranking processes to the user in the most understandable form. Through this, the problem where recommendation results were not practically utilized in conventional technology is resolved, and the effect of providing a practical outfit recommendation service that users can refer to in their daily lives is achieved.

[0144] As described above, according to the present invention, the problem of discrepancies between the actual user's perception and the recommendations caused by conventional clothing recommendation technology relying solely on official weather information to perform recommendations based on average environmental conditions can be effectively resolved. By analyzing user expressions generated in real-time on social network services, the degree of heat or cold that people actually feel under the same temperature conditions can be reflected, thereby overcoming the limitations of recommendations based on uniform standards.

[0145] Furthermore, by quantifying unstructured sentiment information—which was not considered in conventional technologies—through natural language processing techniques, subjective user experience factors can be systematically reflected. This provides a structure where perceived temperature correction based on social consensus is automatically performed without requiring users to input separate settings, thereby simultaneously improving the burden of user input and the problem of recommendation inaccuracies caused by setting errors.

[0146] This invention complements the limitations of conventional recommendation methods that relied on limited information at the individual level by utilizing collective experience data from multiple users. By comprehensively reflecting the sentiments actually expressed by multiple users in specific regions and time periods, it can more reliably reflect environmental characteristics that are difficult to capture based on a single user standard.

[0147] Furthermore, this invention can resolve the problem where existing outfit recommendation services operate separately from actual wearing cases. By performing recommendations based on user-uploaded outfit data tagged with GPS and time information, this invention can reflect wearing results verified in real-world environments rather than theoretical standards, thereby improving the realism and effectiveness of the recommendation results.

[0148] Furthermore, it can alleviate the problem of reduced user trust caused by the unclear basis of recommendation results in conventional technology. Since the recommendation results according to the present invention are based on the process of calculating actual perceived temperature and wear data from multiple users, the recommendation logic becomes clear, and psychological resistance to users accepting the results can be reduced.

[0149] This invention also has the effect of compensating for the time delay and regional averaging problems inherent in official weather data. Since social network services possess excellent real-time and on-site capabilities, they can rapidly reflect sudden weather changes or local differences in perceived weather conditions, thereby effectively resolving the time lag issues that occurred in conventional technologies.

[0150] Consequently, the present invention overcomes the structural limitations of conventional technology, which relies on structured data-based recommendations centered on official weather information, and presents a new recommendation paradigm that combines sentiment data with actual wear data. Through this, it substantially improves the accuracy and suitability of outfit recommendations and provides the effect of simultaneously increasing user satisfaction and service competitiveness. Explanation of the symbols

[0151] 100: Real-time social network service-based outfit recommendation system 110: Social Sentiment Analysis Department 111: Unstructured Text Data Collection Module 112: Perceived Sentiment Keyword Extraction Module 113: Perceived temperature correction value calculation module 120: Wind chill calculation section 121: Official Weather Information Collection Module 122: Perceived temperature correction value application module 123: Actual perceived temperature calculation module 130: Outfit Recommendation Section 131: Database Search Module 132: Clothing Data Selection Module 133: Fitness Ranking Module 134: Outfit Recommendation Module S100: Real-time social network service-based outfit recommendation method S110: Social data collection stage S120: Calculation step for perceived temperature correction value S130: Actual perceived temperature calculation stage S140: Clothing data screening stage S150: Fitness Ranking Stage S160: Recommended Outfit Provision Step

Claims

Claim 1 A social sentiment analysis unit (110) that collects unstructured text data including social media (SNS) posts and comments of users in a specific region and time zone, extracts positive and negative sentiment keywords related to temperature and humidity within the text using natural language processing (NLP) techniques, and quantitatively calculates a sentiment temperature correction value to correct the official sentiment temperature based on the frequency and intensity of the extracted sentiment keywords; and a sentiment temperature calculation unit (120) that receives data related to the current or predicted official sentiment temperature from an official weather information provider and calculates the user's actual sentiment temperature by applying the user's sentiment temperature correction value to the official sentiment temperature. The system includes an outfit recommendation unit (130) that searches a database of outfit photos (UGC) tagged with GPS and time information uploaded by users, selects outfit data worn in an environment closest to the user's actual perceived temperature, ranks suitability, and provides recommended outfits along with images to the user terminal; and the social sentiment analysis unit (110) includes an unstructured text data collection module (111) that collects various forms of text data including posts, comments, and hashtags posted on the user's social network service; and a perceived sentiment keyword extraction module (112) that identifies and extracts sentiment expressions directly related to perceived temperature and humidity from the text data obtained through the unstructured text data collection module (111). and includes a perceived temperature correction value calculation module (113) that classifies the perceived emotional keywords extracted from the perceived emotional keyword extraction module (112) into three categories of 'cold', 'hot', and 'humid', and derives quantitative temperature correction values ​​and humidity correction values ​​through the frequency of occurrence for each category; and the perceived temperature calculation unit (120) includes an official weather information collection module (121) that collects a number of weather-related data including temperature, perceived temperature, humidity, and wind speed from a weather information provider;A perceived temperature correction value application module (122) that applies a perceived temperature correction value derived from a perceived temperature correction value calculation module (113) to the official perceived temperature data received from the above official weather information collection module (121); The actual perceived temperature calculation module (123) receives weather data corrected through the perceived temperature correction value application module (122) and finally calculates the perceived temperature that the user is likely to actually feel; and the clothing recommendation unit (130) includes: a database search module (131) that stores clothing-related information uploaded by users and searches for clothing photos (UGC) tagged with GPS and time information uploaded by users and transmits them to a clothing data selection module (132); a clothing data selection module (132) that selects a clothing worn in an environment closest to the user's actual perceived temperature calculated by the actual perceived temperature calculation module (123) from a set of candidate clothing data secured by the database search module (131); and a suitability ranking assignment module (133) that determines the recommendation priority to be provided to the user for a plurality of candidate clothing data selected through the clothing data selection module (132). The method includes a recommended outfit providing module (134) that visually provides a final recommended outfit to a user terminal based on the priority result calculated by the suitability ranking module (133); a social data collection step (S110) for securing unstructured text data that serves as the basis for calculating an emotional perception correction value; a perceived temperature correction value calculation step (S120) for quantifying the perceived sentiment from the unstructured text data secured in the social data collection step (S110) and calculating an emotional perceived temperature correction value to correct the official perceived temperature; and an actual perceived temperature calculation step (S130) for calculating a perceived temperature that is likely to be actually felt by the user by reflecting the emotional perceived temperature correction value derived in the perceived temperature correction value calculation step (S120) in the official weather information.A real-time social network service-based clothing recommendation system characterized by being operated by a real-time social network service-based clothing recommendation method (S100), comprising: a clothing data selection step (S140) for selecting clothing data suitable for recommendation from a database of clothing photos tagged with GPS and time information uploaded by users, based on the user's actual perceived temperature calculated in the actual perceived temperature calculation step (S130); a suitability ranking step (S150) for determining the recommendation priority to be provided to the user for a plurality of clothing candidate data obtained through the clothing data selection step (S140); and a recommended clothing provision step (S160) for delivering the priority result calculated in the suitability ranking step (S150) to the user. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete

Images