Psychological tendency-based matching recommendation system
The psychological tendency-based matching system addresses the limitations of conventional methods by structuring multidimensional data from psychological tests to calculate long-term relationship suitability, enhancing precision and reproducibility through a 3D model that reflects user behavior and interaction history.
Patent Information
- Application Number
- KR1020250200162
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2025-12-16
- Publication Date
- 2026-11-06
- Estimated Expiration
- 2045-12-16
AI Technical Summary
Conventional matching services rely excessively on superficial information, failing to reflect essential factors that determine relationship quality, such as psychological stability and value alignment, leading to reduced precision and reproducibility in relationship recommendations, and lack the capability to dynamically adjust candidate pools based on user behavior and interaction history.
A psychological tendency-based matching recommendation system that structures multidimensional tendency data from psychological test items, integrating user profiles and partner characteristics to calculate long-term relationship suitability using a 3D matching model, reflecting interaction history and dynamically adjusting candidate pools.
Predicts relationship compatibility by quantifying user values and tendencies in 11 dimensions, reducing duplicate recommendations and user fatigue, and improving the precision and reproducibility of relationship recommendations.
Smart Images

Figure 112025142070263-PAT00001_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to a psychological tendency-based matching recommendation system, and more specifically, to a psychological tendency-based matching recommendation system that structures multidimensional tendency data derived from a number of psychological test items to recommend a customized person based on a user's psychological tendency, values, and the possibility of maintaining a relationship, and integrates this with the user profile and the characteristics of a preferred partner to quantitatively calculate long-term relationship suitability. Background Technology
[0003] Conventional matching services or recommendation systems primarily adopt condition-based filtering methods that select candidates based on external, quantitative information such as age, occupation, region, and hobbies. Additionally, some platforms utilize taste-based recommendation techniques that search for users similar to the conditions when users directly set their preferred attributes. However, because these methods rely excessively on the match of superficial information, they have limitations in that it is difficult to reflect essential factors that determine the quality of a relationship, such as the continuity of the actual relationship, psychological stability, and alignment of values.
[0004] To address these issues, some services introduce simple personality type tests or preference surveys; however, due to the limited number of items and simplistic measurement structures, it is difficult to fully grasp in-depth dispositional factors such as an individual's values, emotional foundations, and worldview. In particular, conventional technologies often rely on simple comparisons of psychological test results or judgments based solely on unidimensional similarity, which presents structural limitations in performing sophisticated matching by multidimensionally combining user profile information, preferences for partners, and psychological disposition data.
[0005] Furthermore, because conventional recommendation engines lack structured data models to utilize psychological test data within their algorithms, they struggle to perform sophisticated operations that calculate relational fit by analyzing inter-dimensional interactions after segmenting user tendencies based on psychological items. Consequently, relying on a simple score summation method that fails to consider various personality factors has led to problems with reduced precision and reproducibility of recommendation results.
[0006] Meanwhile, there are several practical issues within the recommendation process. For instance, in condition-based recommendation systems, the same user is frequently recommended repeatedly, or entities with whom the user has previously had negative interactions are presented again as candidates. Such repetitive recommendations significantly increase user fatigue and dissatisfaction with the matching service. Furthermore, conventional systems lack the capability to dynamically adjust candidate pools based on user selection history, mutual acceptance / rejection history, and changes in interest; consequently, it is difficult to provide intelligent recommendations that reflect user preferences or relationship tendencies that change over time.
[0007] Furthermore, in the case of matching non-romantic relationships such as those in the workplace, school, clubs, and mentoring, systems must reflect role complementarity, harmony of work tendencies, and suitability of communication styles by considering the purpose and characteristics of the relationship; however, existing technologies fail to adequately consider these areas. As a result, recommendations based on simple preference matching have limitations in predicting the formation of long-term relationships, teamwork suitability, and collaboration stability.
[0008] Under these circumstances, there is a need for technology capable of quantitatively predicting relationship compatibility by subdividing an individual's values and tendencies based on high-resolution psychological test data, structuring them into multidimensional scores, and integrating them into a three-dimensional structure of user profile, partner preference, and psychological tendency. Furthermore, to resolve issues such as duplicate recommendations, the failure to reflect past interaction history, and increased user fatigue that occur during the recommendation process, a hybrid matching structure is required that dynamically reflects user behavior logs and selection history, generates an automatically screened candidate pool, and allows for final adjustment by a human.
[0009] Meanwhile, the aforementioned background technology is technical information that the inventor possessed for the derivation of the present invention or acquired during the process of deriving the present invention, and it cannot necessarily be considered publicly known technology disclosed to the general public prior to the filing of the present invention. The problem to be solved
[0011] The present invention has been devised to solve the aforementioned problems and aims to provide a psychological tendency-based matching recommendation system that structures multidimensional tendency data derived from multiple psychological test items to recommend a customized person based on the user's psychological tendency, values, and relationship maintenance potential, and integrates this with the user profile and the characteristics of the preferred partner to quantitatively calculate long-term relationship suitability.
[0012] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem
[0014] A psychological tendency-based matching recommendation system according to one embodiment of the present invention structures multidimensional tendency data derived from a number of psychological test items to recommend a customized person based on a user's psychological tendency, values, and the possibility of maintaining a relationship, and integrates this with the user profile and the characteristics of the preferred partner to quantitatively calculate the long-term relationship suitability.
[0015] The above psychological tendency-based matching recommendation system is,
[0016] A data collection unit that collects data on multiple psychological test items responded to by the user, profile data including the user's age, occupation, region, and activity information, and preference degree data indicating the characteristics of the counterpart the user prefers or wishes to avoid;
[0017] A value and tendency calculation unit that generates a user-specific tendency profile by calculating multiple value dimensions and psychological tendency dimensions based on psychological test response data collected from the above data collection unit;
[0018] A 3D matching model generation unit that generates a multidimensional matching model composed of three axes of profile fit, preference fit, and psychological fit using the above propensity profile, the above profile information, and the above preference criteria as inputs, and calculates a matching score corresponding to each axis;
[0019] A matching fit calculation unit that calculates a matching score by quantifying the degree of relationship fit between users based on the results of a 3D matching model; and
[0020] It includes a candidate group derivation unit that automatically selects an optimal candidate group using the matching score and the analysis results of the history-based candidate exclusion unit mentioned above.
[0021] The above matching suitability calculation unit is,
[0022] It is characterized by quantifying the relationship fit between users based on the results of the multidimensional matching model by reflecting interaction history including past acceptance / rejection history between users, message exchange history, and candidate exposure history, and calculating a final matching score that reflects the removal of duplicate recommendations.
[0023] The above-mentioned value orientation calculation unit is,
[0024] Each of the multiple psychological test items is configured to correspond to at least one of 11 value dimensions, consisting of the social energy value dimension, emotional stability value dimension, logical judgment value dimension, emotional harmony value dimension, optimism and trust value dimension, financial soundness value dimension, life norms value dimension, initiative value dimension, cooperation and consideration for others value dimension, self-awareness value dimension, and openness and modernity value dimension, by referring to predefined item mapping information; and a user-specific value profile consisting of 11 value dimensions is generated by aggregating the psychological test response values for each value dimension to calculate a user-specific value score.
[0025] The above 3D matching model generation unit is,
[0026] It is characterized by forming a three-dimensional matching structure that includes a profile dimension reflecting the above-mentioned user-specific value profile and the above-mentioned profile data, an ideal type dimension reflecting the user's preference data, and a psychological tendency dimension reflecting the above-mentioned value score and psychological tendency profile, and calculating a matching score between users based on the suitability of each dimension.
[0027] The above-mentioned value orientation calculation unit is,
[0028] After identifying the value dimension corresponding to each psychological test item, for multiple item responses corresponding to the same value dimension,
[0029] Calculate a response tendency index representing the characteristics of the user's response distribution by integrating extreme response tendencies and central response tendencies, calculate a self-balance index representing the degree of balance between the user's egocentrism and sociability by analyzing the difference in responses between self-related items and social value-related items, and calculate a behavioral consistency index representing the degree of agreement between the item's intention and response behavior by analyzing the consistency between the item's psychological context and the user's response method.
[0030] It is characterized by generating intermediate scores for each value dimension by reflecting the above response tendency index, self-balance index, and behavioral consistency index in the value score calculation process, and then normalizing the above intermediate scores according to a standard scale to calculate user value scores for each value dimension.
[0031] The above-mentioned value orientation calculation unit is,
[0032] For each item included in the item group corresponding to each value dimension, the response value selected by the user is converted to a range from 1 to 5 points; then, the number of items in which 5 points or 1 point was selected among all items belonging to the said item group is aggregated and divided by the total number of items to calculate the extreme selection ratio, and the number of items in which 3 points were selected is divided by the total number of items to calculate the middle selection ratio; the value obtained by subtracting the middle selection ratio from the extreme selection ratio is set as the default value of the response tendency index; and the response tendency index is calculated by applying a function defined to indicate an extreme response tendency when the default value is positive and a neutral response tendency when it is negative.
[0033] After calculating a first average value by averaging the response values of items included in the self-related item group and a second average value by averaging the response values of items included in the social value-related item group, a difference value representing the degree of egocentrism is calculated by subtracting the second average value from the first average value, and it is defined that the greater the absolute value of the difference value, the more severe the bias toward the self or society; and by normalizing the difference value by dividing it by a predefined maximum bias value, a self-balance index having a range between 0 and 1 is calculated.
[0034] The method is characterized by predefining pairs of items with the same psychological purpose among the items constituting each item group of value dimensions, calculating a concordance ratio by comparing whether the response values of each item pair are in the same direction (indicating whether both are high scores or both are low scores) and dividing the number of matching item pairs by the total number of item pairs, calculating a separate concordance ratio for item pairs composed of reverse items by checking whether the response values are in reverse direction, setting the average value of the two concordance ratios as the first degree of consistency within the item group, additionally arranging the response values of the entire item group sequentially to calculate a variability value by summing all differences between adjacent responses, calculating an inverse normalization value between 0 and 1 by dividing the variability value by a predefined maximum variability value, and calculating a final behavioral consistency index by weighted averaging the inverse normalization value and the first degree of consistency. Effects of the invention
[0036] According to one aspect of the present invention described above, it provides a structure that is essentially different from a simple preference-matching-based matching system centered on external information or a keyword-based search method, and provides the effect of predicting the possibility of forming relationships between users in an advanced manner by utilizing a 3D matching model composed of profile fit, ideal type fit, and psychological tendency fit after quantifying the user's values and tendencies in 11 dimensions based on a total of 110 in-depth psychological test data. Brief explanation of the drawing
[0038] FIG. 1 is a diagram illustrating the schematic configuration of a psychological tendency-based matching recommendation system according to one embodiment of the present invention. Specific details for implementing the invention
[0039] The following detailed description of the invention refers to the accompanying drawings, which illustrate specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that various embodiments of the invention are different but need not be mutually exclusive. For example, specific shapes, structures, and characteristics described herein may be implemented in other embodiments without departing from the spirit and scope of the invention in relation to one embodiment. It should also be understood that the location or arrangement of individual components within each disclosed embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not intended to be limiting, and the scope of the invention is limited only by the appended claims, including all equivalents to those claimed therein, provided appropriately described. Similar reference numerals in the drawings refer to the same or similar functions across various aspects.
[0040] In this specification, terms such as “comprising” or “having” are intended to indicate 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. Where one component is described as being “connected” or “combined” with another component, it may be directly connected to or connected to that other component, but there may be other in between.
[0041] It should be understood that components may exist. On the other hand, when it is stated that one component is "directly connected" or "directly coupled" to another component, it should be understood that no other component exists in between.
[0042] 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.
[0043] 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.
[0044] 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.
[0045] 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.
[0046] Hereinafter, preferred embodiments of the present invention will be described in more detail with reference to the drawings.
[0047] FIG. 1 is a diagram illustrating the schematic configuration of a psychological tendency-based matching recommendation system according to one embodiment of the present invention.
[0048] The purpose is to structure multidimensional tendency data derived from multiple psychological test items and integrate it with user profiles and preferred partner characteristics to quantitatively calculate long-term relationship suitability, in order to recommend customized partners based on the user's psychological tendencies, values, and relationship maintenance potential.
[0049] To this end, the psychological tendency-based matching recommendation system according to the present invention can be implemented in a management server (hereinafter referred to as the server).
[0050] The server may be a self-contained server for the service according to the present invention, a cloud server, or a peer-to-peer (P2P) set of distributed nodes.
[0051] The server can perform one or more of the computation, storage, reference, input / output, and control functions of a general computer, and may include an artificial neural network described below based on input data.
[0052] The server may include a processor and memory. The processor may perform a smart interior service according to the present invention and may include devices capable of performing such service. The processor may execute a program or control the server. Program code executed by the processor may be stored in memory. The memory may store relevant information for performing the service according to the present invention or a program for implementing a method. The memory may be volatile memory or non-volatile memory.
[0053] The server can use a network to send data to an external device or receive data from an external device.
[0054] The server can train artificial neural networks and can also utilize artificial neural networks once training is complete. The processor can train or execute artificial neural networks stored in memory, and memory can store artificial neural networks once training is complete. The electronic device that trains the artificial neural network and the electronic device that uses it may be the same, but they may also be separate.
[0055] The artificial intelligence-related functions 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.
[0056] 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.
[0057] 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
[0058] A neural network through operations between the computation results of the previous layer and multiple weights
[0059] It performs operations. Multiple weights possessed by 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 from 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 aforementioned examples.
[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] The terminal can be implemented in various forms. For example, the terminal described in this specification may be a mobile terminal such as a smartphone, tablet PC, PDA (Personal Digital Assistant), PMP (Portable Multimedia Player), or MP3 player, as well as a fixed terminal such as a smart TV or desktop computer.
[0062] Specifically, the psychological tendency-based matching recommendation system according to the present invention includes a data collection unit, a value tendency calculation unit, a three-dimensional matching model generation unit, a matching suitability calculation unit, and a candidate group derivation unit.
[0063] The data collection unit of the present invention is structured to include a plurality of collection modules and a preprocessing engine to secure information on the user's psychological characteristics and external conditions from various angles, and is a core component that determines the accuracy of value calculation, psychological tendency analysis, and matching model generation performed in subsequent processing steps. The data collection unit consists of a psychological test response collection unit, a profile information collection unit, and a preference criteria collection unit, and each component is designed to include verification functions and data structuring functions to stably process real-time data input, ensuring that the overall flow is maintained consistently.
[0064] The psychological test response collection unit is configured to systematically record responses selected by the user for a total of 110 psychological test items. The response values corresponding to each item are stored with multidimensional data fields in the form of item number, item group identifier, selected score, response time, and response processing status. During the response collection process, an input status monitoring function operates to verify whether the user has completed the response. By recording the time difference between item transitions, it enables the user's response concentration or response patterns to be assessed in subsequent analysis stages. Additionally, components such as repeated items, reverse items, and conditional items inserted to measure the same psychological element are identified and stored with separate tags; these responses can be utilized for response consistency analysis or behavioral consistency analysis during the value calculation stage.
[0065] After response data is collected, an input validation procedure is performed to ensure data integrity. For example, if a response time is abnormally short, indicating that a user made a random selection without reading the item content, a warning tag is assigned to the response to reduce its weight during the subsequent value calculation process. Additionally, if the same value is consistently selected within a specific item group, the monotonicity of the response pattern is recorded; this serves as an important criterion when calculating the response tendency index in subsequent stages. The psychological test response collection module is configured to detect user input values in real time and perform exception handling in the event of input errors or abnormal patterns.
[0066] The profile information collection unit is configured to acquire data regarding the user's physical characteristics and living environment, receiving objective and repetitive data such as age, occupation, residential area, work type, daily routine, hobbies, and interests. This module generates a basic profile based on the information entered by the user during initial registration and includes an update function to save changes as needed. Since the data types of the input profile vary by item, categorical data is converted into code values, numerical data is normalized and stored, and unstructured text data is converted into structured data by applying predefined keyword-based encoding rules.
[0067] The profile information collection unit verifies the validity of input values by automatically detecting, for example, if an unallowed age range or region code is entered and requesting the user to re-enter it, and by blocking the completion of input if some items are missing, thereby ensuring the stability of the entire data. This verification function is intended to prevent incorrect profile information from being reflected in the matching model and resulting in inaccurate matching results, and the collected profile information is used as a reference value for calculating profile fit in the 3D matching model generation unit.
[0068] The preference criteria collection unit is a module configured to allow users to specify the characteristics of a partner they consider important when forming a relationship, collecting data such as the expected age range, acceptable geographical range, personality preferences, and interest alignment. Multiple selections are possible for each item, and the selected values are structured and stored as preference level data. Additionally, since users may strongly wish to avoid specific characteristics, avoidance factors are recorded separately so that they can be reflected as negative fit factors during the process of constructing the ideal partner dimension.
[0069] The preference criteria collection unit provides a score input function to classify user-selected preference items by importance, and this importance is reflected as a weight in the matching model. For example, if a user places great importance on similarity in specific values, the weight of that item increases, strongly influencing the judgment of suitability in the ideal type dimension. Furthermore, if the preference criteria entered by the user conflict internally—for instance, allowing a very broad geographical range while simultaneously strongly requiring a specific lifestyle pattern—the module detects such inconsistencies and performs a verification function to suggest corrections to the user.
[0070] The data collection unit includes a function to preprocess all input values obtained through the above module so that subsequent analysis algorithms can process them efficiently. Psychological test response data is converted into a structured table classified into items such as item group identifiers, response scores, response times, and response pattern information, and profile information is stored as encoded and normalized values for each item. Preference criteria data is divided into preference values and avoidance values, and each item is assigned an importance value and sorted so that it can be used in the 3D matching model generation unit.
[0071] During the data structuring process, the collection unit links all values to a single user identifier to form a single user data object. It then maintains an indexing system that allows the algorithm module to retrieve only specific items as needed for weight calculation or fitness analysis. This data structuring function is a key element for preventing calculation errors caused by differences in data formats between modules and for ensuring the computational speed and stability of the entire system.
[0072] The data collection unit includes error detection and quality correction functions to verify the quality of collected data, intended to prevent the subsequent values calculation unit from performing calculations based on data with low reliability. In psychological test responses, it detects abrupt fluctuations in responses between items, differing responses upon item repetition, and abnormal response time patterns; in such cases, it assigns reliability tags to the relevant data to reduce its weight in subsequent calculations. During the profile information input process, it guides the user to correct missing values, out-of-range figures, or impossible combinations; additionally, it provides a correction procedure if contradictory choices are detected when entering preference criteria.
[0073] Such error detection and quality correction functions play a role in ensuring the accuracy of the finally generated value scores, psychological tendency scores, and matching suitability by preemptively blocking various input errors that may occur in the previous stage.
[0074] The value orientation calculation unit is a core component that quantifies the user's values and psychological orientation by subdividing them into 11 dimensions based on psychological test response data transmitted from the data collection unit, and calculates the values that form the basis of the entire matching algorithm of the present invention. This component consists of an item group classification engine, a raw score calculation module, a correction factor application module, an index-based weighting processing module, and a normalization engine, and each step operates in a sequential yet interconnected manner to evaluate the value characteristics of each individual user in a multifaceted way.
[0075] The item group classification engine identifies the value dimension corresponding to each of the 110 items recorded in the data collection unit and maps each item to one or more of the 11 value dimensions. Each item includes metadata such as an item identifier, a psychological purpose tag, and an item group number, and the classification engine refers to this metadata to accurately classify the item into the corresponding dimension among the 11 dimensions: the social energy value dimension, the emotional stability value dimension, the logical judgment value dimension, the emotional harmony value dimension, the optimism and trust value dimension, the financial soundness value dimension, the life norm value dimension, the initiative value dimension, the cooperation and consideration for others value dimension, the self-awareness value dimension, and the openness and modernity value dimension.
[0076] In this classification process, since the original purpose of an item may overlap with two or more value dimensions, item-specific weight distribution rules may be applied; these rules are executed automatically by referencing a predefined table within the item group classification engine. By performing this classification process beforehand, it ensures that the influence between dimensions based on the purpose of the items is accurately reflected in the subsequent calculation stage.
[0077] The raw score calculation module converts the response values for each item into integer scores ranging from 1 to 5 and calculates dimension-specific raw scores based on the response values of items belonging to the same value dimension. In this process, rather than simply summing or averaging the scores, a weighted summation method is used that applies importance coefficients assigned to each item to reflect differences in importance among items. For example, for items designated as core items in a specific dimension, the importance coefficient is set to be greater than 1, while for items with relatively low influence, the importance coefficient is set to be less than 1 and reflected in the raw score.
[0078] In addition, for repeated or reverse items, the raw score calculation module corrects the scores by applying predefined reverse transformation rules, and if duplicate items exist, it creates a separate score field to determine response consistency and stores it for use in subsequent steps.
[0079] The correction factor application module is a component designed to reflect users' psychological characteristics, which are difficult to adequately reflect in the response values themselves, into the value score; it generates an intermediate score for each dimension by applying the response tendency index, self-balance index, and behavioral consistency index to the raw score.
[0080] First, the response tendency index is an indicator that assesses response bias by analyzing the ratio of extreme and median responses within a given dimension group; for users with many extreme values, weights are applied to enhance score clarity, while for users exhibiting response patterns centered on median values, the sensitivity of scores by dimension is adjusted.
[0081] Next, the Self-Balance Index reflects the difference in average response values between the self-related and social-related question groups to generate a correction value that considers the balance of egocentrism or social orientation. A high score on this index has a direct impact on scores for specific value dimensions, and is significantly reflected in scores for dimensions such as cooperation and consideration, and emotional harmony.
[0082] The behavioral consistency index is an indicator that evaluates the alignment between the psychological purpose of an item and response patterns. It is calculated based on the consistency of response directions between item pairs with the same purpose within an item group. Because this index reflects how consistently users interpreted the psychological meaning of an item and responded, it is used as a reliability adjustment variable in the score calculation process; in the case of dimensions with low behavioral consistency, it affects the median score by reducing a portion of the raw score.
[0083] The index-based weighting processing module is configured to calculate the final score by additionally reflecting inter-dimensional correlations and item group structure to the intermediate score, which is the result of integrating the previously calculated raw score and correction value. This module refers to the cross-weighting matrix between values; when specific dimensions have a strong correlation with each other, it assigns complementary weights to the intermediate score of that dimension, and for dimensions exhibiting opposing characteristics, it applies adjustment weights to balance the scores.
[0084] For example, since the Initiative dimension and the Logical Judgment dimension can exhibit a high correlation together, a cross-correction value is added when the scores of both dimensions are simultaneously high or low; conversely, because the Sentimental Consonance dimension and the Self-Awareness dimension may show opposing tendencies in some aspects, a balance correction value is applied to mitigate extreme biases. This reduces bias in values and improves the consistency of individual user profiles.
[0085] The normalization engine is configured to eliminate scale deviations that may occur when the median score of each value dimension is relatively higher or lower compared to other dimensions; it normalizes the scores of all dimensions into an interval of 0 to 1 so that they can be expressed within the same range. During the normalization process, the normalization ratio is calculated based on the maximum and minimum possible scores of each dimension, and the calculated values are stored as the final scores for each of the 11 value dimensions.
[0086] The normalized score is used as a reference score to calculate the degree of matching fit along with the scores of the profile dimension and the ideal type dimension in the subsequent 3D matching model generation unit, and prevents calculation errors that may occur during the process of integrating scores of various dimensions by ensuring scale uniformity.
[0087] The response tendency index calculation unit is configured to select question groups corresponding to each value dimension from the response values of 110 psychological test items transmitted from the data collection unit, and to quantify the characteristics of the response distribution appearing in those question groups; this serves as key information for determining whether a user's response pattern is extreme or neutral. This unit consists of a response distribution calculation module, a response pattern analysis module, a correction rule application module, and an index generation module, and each module interacts with the others to generate a value that quantitatively expresses the response tendency for each user.
[0088] The response distribution calculation module is configured to convert the response values of all items corresponding to a specific value dimension into a range of 1 to 5 points, and then calculate frequency information on how often each option score is selected within the item group.
[0089] The module calculates the frequency of 1-point selection, 2-point selection, 3-point selection, 4-point selection, and 5-point selection, respectively, based on the total number of items included in the item group, and each frequency is stored as a ratio value relative to the total number of items. This ratio information is used as basic information to determine not only the average trend of responses but also whether extreme values are selected.
[0090] In particular, responses corresponding to 1 point and 5 points are defined as extreme choices, and the value obtained by summing the number of times 1 point or 5 point is selected in the corresponding dimension item group and dividing it by the total number of items is stored as the extreme choice ratio. On the other hand, 3-point responses are defined as neutral choices, and the value obtained by dividing the number of items in which 3 points are selected by the total number of items is stored as the middle choice ratio.
[0091] The response distribution data generated in this manner is utilized as a key input value in the subsequent response pattern analysis module to interpret the directionality of the user's response behavior.
[0092] The response pattern analysis module performs an analysis function to determine whether the user's response behavior has an extreme tendency or a neutral tendency, based on the extreme choice ratios and median choice ratios calculated by the response distribution calculation module.
[0093] First, if the proportion of extreme choices exceeds a certain threshold value, for example, if 40 percent or more of the item group are answered with extreme choices, the user's response pattern is interpreted as showing a strong preference or strong rejection tendency toward the corresponding value dimension. On the other hand, if the proportion of intermediate choices exceeds a certain threshold value, for example, if 50 percent or more of the total items are answered with 3 points, it can be judged that the user's attitude toward the corresponding value dimension is not clearly revealed or exhibits a neutral tendency.
[0094] Based on these analysis results, this module generates a tendency classification value indicating the extreme or neutrality of user responses, and this classification value is reflected in the calculation of the response tendency index in the subsequent correction rule application module.
[0095] The correction rule application module is a component that adjusts the sensitivity of the response trend index by combining additional correction elements with the trend classification value calculated based on the response pattern analysis results.
[0096] This module evaluates whether a user's response pattern is reliable by utilizing additional information such as response time-based concentration information, response consistency information upon item repetition, and whether there are sudden changes in responses within an item group.
[0097] For example, if there are many items with abnormally fast response times, it is highly likely that responses of that dimension were not sufficiently considered; therefore, a correction is performed to reduce the influence of extreme or median selection rates by a certain percentage.
[0098] Conversely, if a consistent response pattern is identified in repeated or reversed items, the user's response to that dimension is deemed to have relatively high reliability, and a correction rule is applied to reinforce the influence of the extreme choice rate or adjust the reduction in the median choice rate.
[0099] In addition, if the response values within the item group fluctuate rapidly, for example, if a pattern of alternately selecting 1 point and 5 points is repeated, the correction module detects this response variability and takes measures to reduce the tendency score of the corresponding dimension by a certain percentage, as this may indicate that the response was made impulsively or spontaneously.
[0100] The response tendency index generation module is configured to calculate the final response tendency index by synthesizing the response distribution, response pattern analysis results, and the application of correction rules. Fundamentally, the response tendency index is calculated based on the value obtained by subtracting the median selection rate from the extreme selection rate; a positive value indicates a tendency to respond to extremes regarding the corresponding value dimension, while a negative value indicates a tendency toward neutral or indecisive responses.
[0101] However, in this invention, instead of using the simple subtraction value as the index, the final index is calculated after finely adjusting the influence of extremes and neutrality by reflecting the reliability correction value generated by the correction rule application module. For example, when response concentration is high and consistent responses are observed in repeated items, weights are applied to the index calculation formula to strengthen the influence of extreme values; conversely, when response variability is high, the influence of neutrality is strengthened to prevent the index from being excessively skewed toward extreme values.
[0102] The response tendency index generated in this manner has a range of positive or negative values with 0 as the reference value, and is finally input into the intermediate score calculation process of the value tendency calculation unit to finely adjust the score for each value dimension.
[0103] The self-balance index calculation unit is configured to quantitatively evaluate the balance between the self-related item group and the social value-related item group in a user's response, and plays a key role in determining whether an individual places greater weight on self-orientation or social orientation in relationship formation or social interaction. This unit consists of an item group classification module, an item group representative value calculation module, a balance difference analysis module, a bias correction module, and a self-balance index generation module, and generates a stable indicator by performing sophisticated calculations step by step.
[0104] The item group classification module separates self-related items and social-related items from the 110 items transmitted from the data collection unit and forms each item group.
[0105] The self-related item group includes questions regarding an individual's intrinsic motivation, self-evaluation, self-esteem, and self-processing of emotions, while the social-related item group consists of questions evaluating interpersonal relationships, cooperation, consideration for others, and awareness of social norms.
[0106] The classification module automatically constructs item groups by referencing tag information included in item metadata, and if the same item possesses both self- and social characteristics simultaneously, it applies a distribution ratio per item to ensure that the item's score is partially reflected in both item groups.
[0107] The item group representative value calculation module is a step that calculates representative values based on the response values for each of the self-related item group and the society-related item group.
[0108] The average value is calculated by applying weights to the response values of all items included in each item group, and if an importance coefficient exists for each item, the coefficient is multiplied and reflected in the process of calculating the representative value.
[0109] For example, in the self-related item group, the importance coefficient is set to be greater than 1 for core items that directly assess an individual's level of self-awareness, while a relatively lower importance coefficient is applied for items asking about general tendencies. The first mean value (self-related representative value) and the second mean value (social-related representative value) calculated in this manner are used in the subsequent balance analysis stage.
[0110] The balance difference analysis module calculates the difference between the first mean value and the second mean value to analyze the direction of the user's egocentrism or social orientation.
[0111] If the difference is positive, it means the user has strong self-oriented characteristics, and if it is negative, it means the user has relatively strong social-oriented characteristics.
[0112] In addition, if the absolute value of the difference is above a certain standard, for example, 0.5 or higher, it is evaluated as having a distinct bias, and this is also used to judge the stability or strength of the user's values.
[0113] This module does not stop at simply calculating these differences but also includes additional functions to analyze the distribution of difference values and response characteristics, and can re-evaluate specific item groups or perform exception handling when the variability of the difference values is high.
[0114] The bias correction module is a function that determines whether a user's response pattern is distorted based on the results of the balance difference analysis and reflects the result in the calculation of the self-balance index.
[0115] For example, if the response tendency index shows a high tendency toward extreme responses, a correction is performed to reduce the difference value by a certain percentage because self- or social biases may be exaggerated compared to reality.
[0116] Conversely, if the behavioral consistency index is high and it is confirmed that the response aligns with the item's intent, adjustments can be made to reflect the influence of the bias difference value more strongly.
[0117] In addition, if only items with excessively short or excessively long response times are clustered within a specific range, it implies that the level of concentration was inconsistent during the response process; therefore, additional reliability correction is performed for that item group.
[0118] The self-balance index generation module calculates the self-balance index by dividing the difference between the first average value and the second average value by the maximum bias value and normalizing it to a value between 0 and 1.
[0119] The normalization process is intended to ensure the uniformity of scores across dimensions and serves to limit the exponent from increasing beyond an upper limit, even if the bias value is large.
[0120] The final calculated self-balance index serves as an indicator representing the balance between the user's self-orientation and social orientation. It is utilized in the intermediate score calculation process of the value orientation calculation unit and acts as a crucial correction factor, particularly in the analysis of interpersonal relationship-related tendencies such as cooperation and consideration, emotional harmony, and self-awareness.
[0121] The behavioral consistency index calculation unit is configured to quantitatively analyze how consistently the psychological purpose or evaluation intention of each item is combined with the user's response behavior when the user responds to psychological test items. Unlike simple response value analysis, this is an advanced computational module that cross-analyzes the psychological structure of the item and the response pattern, and plays a very important role in the reliability and in-depth analysis of the 11 value dimensions proposed in this invention.
[0122] This output unit consists of an item pair mapping module, a co-direction analysis module, a reverse item analysis module, a response flow variability analysis module, a consistency integration module, and a behavioral consistency index generation module.
[0123] The item pair mapping module predefines pairs of items configured to assess the same psychological objective and stores each item pair in a database to be used as a standard for future consistency analysis. For example, if there are multiple items assessing emotion regulation ability, these items are classified into a group of items with the same objective, and the item pair mapping module generates a list of specific item pairs to maintain a correspondence table.
[0124] Reverse items are mapped in the same way; however, since the criteria for determining the direction of the response are opposite for reverse items, tags are assigned to apply separate comparison rules during response analysis.
[0125] The co-direction analysis module compares and analyzes whether the user's response values have the same directionality for the item pairs configured in the item pair mapping module. For example, if Item A and Item B are a pair of items measuring a positive tendency, if 4 points or 5 points are selected for both items, they are judged to have the same directionality, and conversely, if one is selected as 2 points or 1 point and the other as 4 points or 5 points, they are judged to have inconsistent directionality.
[0126] For each pair of items, whether the direction matches is recorded, and the proportion of items with matching directions among all pairs of items is calculated and stored as a primary agreement value. This value is used as an indicator to show whether the user maintains consistency in the process of interpreting and responding to the relevant psychological objective.
[0127] The reverse item analysis module is a function that verifies whether the response direction is opposite for item pairs designated as reverse items. For example, if Item A evaluates a specific behavior positively and Item B is a reverse item that evaluates the same behavior negatively, and a high score is selected for Item A and a low score for Item B, the direction is determined to be consistent.
[0128] In this manner, the response values of each inverse item pair are compared to determine whether there is a reverse match, and the ratio of matched item pairs to the total number of inverse item pairs is calculated and stored as a separate value for the inverse item match.
[0129] The response flow variability analysis module is a function that analyzes how stable or variable the flow of user responses is within the same question group. This module arranges the response values within the question group in sequential order and calculates the variability value by summing the absolute differences between each response. A lower variability value indicates that user responses have remained stable, while a higher value signifies greater variability in the responses.
[0130] The variability value is normalized by dividing it by a predefined maximum variability value, and a higher normalized variability value is interpreted as indicating that the response flow is unstable. This value is used as a weighting factor in the subsequent alignment integration step.
[0131] The consistency integration module calculates a final consistency score by combining the two consistency values derived from the co-direction analysis and reverse item analysis results with the response flow variability analysis results. In this process, co-direction consistency and reverse item consistency are assigned high weights as they indicate high consistency, while variability is given an inverse weight (1 minus the variability value) as it signifies response instability.
[0132] These three values are integrated using a weighted average method by applying different weighting factors according to the importance of each dimension, and the integrated consistency value is input into the behavioral consistency index generation module.
[0133] The behavioral consistency index generation module generates the final behavioral consistency index after performing the necessary normalization procedures so that the value calculated by the consistency integration module ranges from 0 to 1. A consistency value closer to 1 indicates that the user consistently interpreted the psychological purpose of the item and responded stably, while a value closer to 0 indicates that the response was unstable or the user failed to consistently interpret the purpose of the item.
[0134] The final behavioral consistency index acts as a reliability adjustment variable when calculating scores for each value dimension in the value orientation calculation unit, and is used to enhance score accuracy, particularly when combined with the response tendency index or self-balance index.
[0135] The 3D matching model generation unit is a core component that calculates the degree of fit between users by simultaneously reflecting user-specific value scores, profile data, and preference criteria data, calculated through the data collection unit and the value tendency calculation unit, into a matching model composed of multiple axes. This unit consists of a profile fit calculation module, an ideal type fit analysis module, a psychological tendency fit calculation module, a dimension alignment correction module, and an integrated matching model generation engine, and these modules are organically combined to complete a multidimensional matching structure.
[0136] This generation unit utilizes a multi-layered integrated algorithm that separates external characteristics, preference criteria, and psychological tendencies into independent dimensions, calculates the suitability of each dimension individually, and reflects the interrelationships between them, rather than relying on a simple attribute matching comparison method. Through this, unlike existing systems that filter matching candidates based solely on profile information, it provides the ability to quantitatively analyze users' inner tendencies and values to select candidates with a higher potential for maintaining a relationship.
[0137] The profile fit calculation module is a module that quantifies the similarity of living environments and conditions between two users based on external data such as age, occupation, residential area, and lifestyle patterns.
[0138] First, this module compares the profile items of User A and User B item by item to calculate the match or difference value for each item. For example, in the case of age, the absolute difference is calculated and then scored by applying a predefined allowable age difference range; for region, a score is calculated based on whether they are in the same region or commuting distance; and for occupation group, a similarity score is calculated based on whether they belong to the same or similar occupational group.
[0139] Importance weights are assigned to each profile item because their impact on matching varies. For example, information such as daily routines or activity patterns can act as important factors in forming relationships, so it is set to a high weight, while hobby information may be set to a relatively low weight.
[0140] The individual item scores calculated in this manner are integrated using a weighted average method and are finally stored as a profile fit score. This score is expressed within the range of 0 to 1, and the closer it is to 1, the better the basic living environment and external conditions match between users.
[0141] The ideal partner compatibility analysis module is configured to analyze, in a cross-sectional manner, how well User B's profile matches the characteristics of a partner preferred by User A, and conversely, how well User A's profile corresponds to the characteristics preferred by User B.
[0142] This module first retrieves the user's preference criteria data stored in the data collection unit and normalizes the preference and avoidance values set by the user on an item-by-item basis. The normalized preference values are adjusted based on weights to consider the priority or importance of each item. Subsequently, the user's profile data is compared with the preference criteria data to calculate the match or deviation value.
[0143] For example, when the age range preferred by user A is 30 to 35, if user B's age falls within this range, a weighted score is assigned, and if it does not fall within it, points are deducted based on the deviation value.
[0144] Other items, such as occupational group, region, and similarity of values, are processed in the same way.
[0145] For avoidance items, weights are applied in the opposite direction, and candidates meeting the conditions are automatically excluded if their ideal type fit score decreases or falls below a certain threshold.
[0146] Ideal type fit is calculated by considering both bidirectional preference values of A and B, then normalized to the same scale and expressed as a value between 0 and 1. This is a factor that significantly improves user satisfaction compared to existing matching systems that reflect only unidirectional conditions.
[0147] The psychological orientation suitability calculation module performs the function of evaluating psychological similarity or complementarity between users based on 11 value dimension scores calculated by the value orientation calculation unit.
[0148] This module arranges the value scores of User A and User B in the same dimensional order, and then calculates the difference in scores by dimension to determine similarity or complementarity.
[0149] In this case, instead of a simple difference calculation method, an advanced calculation is performed using dimensional correlation matrices and complementarity matrices.
[0150] For example, if the initiative and collaboration dimensions are complementary, a difference in scores within an appropriate range is reflected as a positive matching factor; conversely, if the difference is excessively large, points are deducted from the fit score.
[0151] In addition, the response tendency index, self-balance index, and behavioral consistency index are used as weighting adjustment variables in the calculation of psychological fit, and are reflected in a way that automatically reduces the influence of a specific dimension if the score confidence of that dimension is low.
[0152] Unlike existing matching systems that compare only simple score differences, this complex calculation structure enables a sophisticated assessment of the alignment of inner tendencies between users by analyzing psychological structural elements at multiple levels.
[0153] The dimension alignment correction module does not simply sum profile fit, ideal type fit, and psychological tendency fit, but rather performs the function of dynamically adjusting weights by reflecting the relationship in which each dimension influences one another.
[0154] First, if the user profile demonstrates high fit, additional weights are applied in combination with psychological fit to increase the likelihood of maintaining the actual relationship.
[0155] Conversely, in cases where profile fit is high but ideal type fit is significantly low, adjustments are made to amplify the influence of the ideal type dimension in the overall model calculation to ensure that results faithful to the relative criteria actually desired by the user are derived.
[0156] In addition, when specific value dimensions have a strong correlation with one another—for example, when the emotional stability dimension and the emotional harmony dimension, or the logical judgment dimension and the initiative dimension, both record high scores—a mutual reinforcement effect is applied; conversely, when the score deviation between dimensions is excessive, a balance correction rule is applied.
[0157] The integrated matching model generation engine is configured to generate a 3-axis matching model by combining the fit scores for each dimension, and calculates an overall matching score between User A and User B.
[0158] The integration engine first combines the three-dimensional fit scores into a single vector and then calculates the final matching score by applying dimension-specific importance weights. These importance weights can be set differently depending on the user type, relationship purpose, or specific environmental situation, and can be adjusted by the system administrator or updated by an automatic learning module.
[0159] The final matching score is normalized to a range of 0 to 1, and a higher value indicates a higher likelihood of forming a long-term relationship between users. This score is used as a key input value in the matching suitability calculation unit and the candidate group derivation unit, and serves as the basis for the system's automatic recommendation algorithm.
[0160] The matching fit calculation unit receives the profile fit score, ideal type fit score, and psychological tendency fit score calculated by the 3D matching model generation unit as input, and calculates the final matching score by integrating the overall fit between users into a single indicator. This calculation unit consists of a dimension importance adjustment module, a history-based candidate exclusion module, a fit adjustment engine, a normalization and stabilization module, and a final matching score calculation engine. Each module is interconnected to perform detailed score calculations that reflect the user's past matching behavior, preference criteria, psychological stability, etc.
[0161] The dimension importance adjustment module performs the function of reflecting the importance of each dimension in the score by variably adjusting it according to the situation, rather than simply averaging the scores of the three dimensions: profile fit, ideal type fit, and psychological tendency fit.
[0162] First, this module dynamically sets the weights assigned to each dimension based on user-specific preference criteria and usage purposes. For example, for users who prioritize consideration for others, the weight of the Psychological Fit dimension is increased, while for users who strongly prefer specific age groups or regional conditions, the weight of the Profile Fit dimension is raised.
[0163] In addition, a default weight table is set based on the overall system's training data or operational policies, and this can be adjusted according to specific age groups, relationship purposes, or user types.
[0164] Through this adjustment process, scores with weights applied to each dimension are integrated using a weighted summation method and stored as a dimension-integrated score, which is used as the base value for calculating the final matching score.
[0165] The history-based candidate exclusion module queries the history of past matching attempts or interactions between User A and User B, and performs the function of automatically excluding the candidate or adjusting the score if there is a negative impact on the matching result.
[0166] This module refers to the following historical data.
[0167] - Records of users consistently rejecting the same candidate in the past
[0168] - Records of the other party repeatedly rejecting the user
[0169] - History of conversations ending early in past matches
[0170] - Block processing or reporting history
[0171] - List of candidates already recently recommended
[0172] If any of these histories are repeated multiple times, the module takes measures to apply penalty factors to the corresponding match or, if it falls below a certain standard, completely exclude it from the matching target.
[0173] Conversely, if there is a positive history, for example, if the other party has previously shown interest or there are positive cooperation indicators, a correction may be made to slightly increase the matching suitability of the candidate.
[0174] The history-based candidate exclusion module prevents users from experiencing fatigue due to repetitive candidate recommendations during the matching process and plays a role in continuously providing a new pool of candidates.
[0175] The fit adjustment engine combines dimension integration scores and history-based candidate exclusion results to generate a final adjustment value that more realistically reflects the potential for relationship formation between users.
[0176] The fitness adjustment engine additionally considers the mutual interest index, relationship sustainability index, and complementarity correction value.
[0177] The mutual interest index is a value that evaluates the degree to which mutually preferred attributes align, verifying in both directions how well A's preference criteria fit B's profile and how well B's preference criteria align with A's profile.
[0178] The higher the mutual interest, the more the adjustment engine operates to increase the final score.
[0179] The relationship sustainability index is a measure that predicts the likelihood of maintaining a relationship based on combined values such as value similarity, emotional stability, and emotional processing stability, derived from psychological dispositions. While existing matching systems focus solely on simple preference matching, this invention is designed to reflect the potential for long-term relationship sustainability based on psychological structure analysis.
[0180] The complementarity adjustment value indicates that when two users exhibit antagonistic or complementary tendencies, this can have either a positive or negative effect on relationship formation. For example, if one user is highly assertive and the other is highly emotionally receptive, it is judged as positive complementarity and the score is increased; conversely, if a user with an excessively strong logical judgment tendency meets a user with an excessively low emotional harmony dimension, it is judged as negative complementarity and the score is decreased.
[0181] This engine combines each element using a weighted average method to convert the emotional, psychological, and physical combination levels among users into a single continuous score.
[0182] The normalization and stabilization module is configured to normalize the scores calculated by the goodness-of-fit engine to a range between 0 and 1, enabling comparison between candidates under various conditions. This module includes a volatility mitigation function along with normalization.
[0183] First, the normalization function performs a linear transformation based on the minimum and maximum values of the scores to distribute them uniformly within a certain range. Additionally, a stabilization filter is applied to prevent the excessive reflection of specific user tendencies when score variability is excessive; this serves to mitigate the impact of dimensions where the amount of change increases rapidly.
[0184] Through this process, the normalized goodness-of-fit scores are configured to ensure stable comparisons between candidate groups during system operation.
[0185] The final matching score calculation engine generates a final matching score based on the normalized goodness-of-fit score, and this score is a key criterion value used in the candidate group derivation unit.
[0186] The engine first applies a weighting table reflecting the user's preference criteria and importance to ultimately adjust the influence of specific items on the overall score. If the degree of alignment with the relative conditions set by the user is high, the score increases by a weight; conversely, if specific conditions are not met, a weighted deduction occurs.
[0187] The final matching score is expressed as a real value between 0 and 1, and the closer it is to 1, the higher the probability of a match between users. Since this score is not based simply on the matching of external conditions but comprehensively reflects preference criteria, psychological tendencies, relationship sustainability, complementary tendencies, and past history, user satisfaction can be dramatically improved compared to existing matching systems.
[0188] The candidate pool derivation unit is configured to automatically select final recommendation candidates based on the user-to-user matching scores calculated by the matching suitability calculation unit, and to generate an optimized matching list in a way that minimizes user fatigue.
[0189] This configuration consists of a candidate selection engine, a history-based duplicate filtering module, a matching stability correction module, a candidate stratification module, a user interaction reflection module, and a final candidate generation engine; each module is organically interconnected to provide a candidate set suitable for the user while avoiding repetitive recommendations.
[0190] The candidate selection engine sequentially analyzes the matching scores between all users transmitted from the matching suitability calculation unit and extracts only users with scores above a certain standard as a candidate group.
[0191] The engine first sets all users excluding the current user from the entire user set as comparison targets and sequentially queries the matching scores between each target user. At this time, if the matching score has a predefined threshold value, for example, 0.6 or higher, it is filtered as a potential candidate group in the first stage.
[0192] The threshold value is not a fixed value but is dynamically adjusted by considering the system operation period, user regional distribution, and the distribution of matching scores across the entire data. For example, if the number of users in a specific region is low, the threshold value can be automatically lowered to ensure an appropriate number of candidates are secured.
[0193] The candidate selection engine is designed to prevent over- or under-candidate situations in this way, thereby preventing the presentation of too many or too few matching candidates to the user.
[0194] The history-based duplicate filtering module performs the function of reducing matching fatigue by querying the user's past matching history and automatically excluding candidates that the user has already contacted or that have been repeatedly suggested.
[0195] This module reflects the following historical information.
[0196] - Candidates that were previously recommended but rejected by the user
[0197] - Candidates who have previously rejected the user
[0198] - Negative history such as unilateral or mutual blocking, reporting, etc.
[0199] - Candidates already recommended within the last period
[0200] - Candidates whose conversations ended quickly in past matches
[0201] If such history values accumulate above a certain threshold, the candidate is automatically excluded from the candidate pool. Conversely, candidates with positive interactions—for example, those who have exchanged interest indicators at least once—can be retained in the pool by applying a certain weight rather than being excluded.
[0202] This module combines matching suitability and history-based information to perform user experience-based optimization beyond simple duplicate filtering.
[0203] The matching stability correction module is a module that readjusts the candidate pool by deducting points or removing candidates with unstable matching potential from the users selected by the candidate selection engine, so that only users with a high probability of forming actual relationships remain.
[0204] Psychological stability correction
[0205] Candidates with excessive variability in dimensions such as emotional stability and emotional harmony are judged to have a low likelihood of maintaining a long-term relationship, and their scores are adjusted.
[0206] Complementarity Imbalance Correction
[0207] When one user's scores are very high and the other user's scores are extremely low in terms of initiative, logical judgment, emotional harmony, and self-awareness, stability adjustment is applied because the likelihood of conflict within the relationship increases.
[0208] Lifestyle Collision Correction
[0209] In terms of profile suitability, if lifestyle patterns, work types, or activity times are opposite, it is determined that maintaining a long-term relationship is difficult, and a penalty factor is applied.
[0210] This module reflects actual relationship sustainability, which is difficult to judge solely by simple matching scores, ensuring that the candidate pool consists of more realistic and sustainable users.
[0211] The candidate hierarchy module is a function that sets priorities for candidate groups by classifying them into multiple layers based on matching quality, rather than organizing the final candidate group into a single set.
[0212] Hierarchization is typically carried out based on the following criteria.
[0213] - Top Candidates: Matching score 0.8 or higher, Psychological fit 0.7 or higher
[0214] - Median Candidates: Matching score 0.6 or higher, less than 0.8
[0215] - Preliminary Candidates: Candidates who do not meet the baseline but satisfy some of the preference criteria
[0216] This stratification is utilized by the candidate derivation unit when determining the recommendation order in subsequent stages of the sentiment matching manager or system.
[0217] The hierarchical structure is not a fixed hierarchy rule, and system operators or automated learning engines can automatically adjust it by reflecting regional distribution, the number of users, matching performance over a specific period, etc.
[0218] The user interaction reflection module is a feature that analyzes the user's previous response patterns during the candidate generation process to customize the recommendation structure for the user.
[0219] This module considers the following elements.
[0220] - Candidate types that users have shown high interest in in the past
[0221] - Persistent preference for specific profiles or disposition dimensions
[0222] - Candidate types that users showed low interest in
[0223] - Patterns where the user repeatedly rejected
[0224] For example, if a user consistently shows a positive response to a partner with a high emotional compatibility dimension, it operates by giving weight to candidates with high emotional compatibility scores to increase their priority when generating future candidate pools.
[0225] Conversely, if a user repeatedly rejects candidates with specific profile criteria, those candidates will have their recommendation ranking lowered or be excluded.
[0226] This module goes beyond simple algorithm-based recommendations to enable user-centric optimization and performs the function of significantly improving recommendation satisfaction.
[0227] The final candidate generation engine is a function that generates a final list of recommended candidates based on candidate data selected and layered in the previous stage.
[0228] First, priorities are set based on the top candidates, and the median and reserve candidates are sorted as auxiliary candidates. Next, the final sorting order is adjusted by reflecting user preference criteria and mutual interest values. Finally, history-based exclusion conditions or dynamic filtering conditions are checked to exclude candidates.
[0229] The generated candidate pool is saved individually for each user and configured to be retrieved by the sentiment matching manager or the system recommendation module. The final generated candidate pool is displayed on the system's matching recommendation screen, allowing users to view detailed information about each candidate and make a selection.
[0230] The technology according to the present invention, as described above, may be implemented in the form of program instructions that can be executed through various computer components or implemented as an application, and may be recorded on a computer-readable recording medium. The computer-readable recording medium may include program instructions, data files, data structures, etc., either individually or in combination.
[0231] The program instructions recorded on the above-mentioned computer-readable recording medium are those specifically designed and configured for the present invention, but may also be those known and available to those skilled in the art of computer software.
[0232] Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions such as ROM, RAM, and flash memory.
[0233] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.
[0234] Although the invention has been described above with reference to embodiments, those skilled in the art will understand that various modifications and changes can be made to the invention without departing from the spirit and scope of the invention as set forth in the following claims.
Claims
Claim 1 In a psychological tendency-based matching recommendation system for recommending customized individuals based on a user's psychological tendency, values, and the possibility of maintaining a relationship, the system structures multidimensional tendency data derived from multiple psychological test items and integrates this with a user profile and the characteristics of a preferred partner to quantitatively calculate long-term relationship suitability. The psychological tendency-based matching recommendation system comprises: a data collection unit that collects data from multiple psychological test items responded to by a user, profile data including the user's age, occupation, region, and activity information, and preference degree data indicating the characteristics of a partner the user prefers or wishes to avoid; a values tendency calculation unit that generates a user-specific tendency profile by calculating multiple values dimensions and psychological tendency dimensions based on the psychological test response data collected from the data collection unit; a 3D matching model generation unit that generates a multidimensional matching model composed of three axes of profile suitability, preference suitability, and psychological suitability using the tendency profile, the profile data, and the preference degree data as inputs, and calculates a matching score corresponding to each axis; and a matching suitability calculation unit that calculates a matching score by quantifying the relationship suitability between users based on the results of the 3D matching model. The system includes a candidate group derivation unit that automatically selects an optimal candidate group using the matching score and the analysis results of the history-based candidate exclusion unit; the matching suitability calculation unit is characterized by reflecting interaction history including past acceptance / rejection history between users, message exchange status, and candidate exposure history, quantifying the relationship suitability between users based on the results of the multidimensional matching model, and calculating a final matching score that reflects the removal of duplicate recommendations; and the value orientation calculation unit is characterized by each of a plurality of psychological test items referring to predefined item mapping information, and each item [regarding] the social energy value dimension, emotional stability value dimension, logical judgment value dimension, emotional harmony value dimension, optimism and trust value dimension, financial soundness value dimension,The method is characterized by setting the value profile to correspond to at least one of 11 value dimensions, which consist of the dimension of lifestyle norms, the dimension of initiative, the dimension of cooperation and consideration for others, the dimension of self-awareness, and the dimension of openness and modernity; generating a user-specific value profile composed of 11 value dimensions by aggregating the psychological test response values for each value dimension to calculate a user-specific value score; the 3D matching model generation unit forms a 3D matching structure including a profile dimension reflecting the user-specific value profile and the profile data, an ideal type dimension reflecting the user's preference data, and a psychological tendency dimension reflecting the value score and psychological tendency profile; and calculating a matching score between users based on the suitability of each dimension; and the value tendency calculation unit identifies the value dimension corresponding to each psychological test item, and for multiple item responses corresponding to the same value dimension, calculates a response tendency index representing the characteristics of the user's response distribution by integrating extreme response tendencies and median response tendencies, and analyzes the difference in responses between self-related items and social value-related items to determine the user's egocentrism and A psychological tendency-based matching recommendation system characterized by calculating a self-balance index indicating the degree of social balance, calculating a behavioral consistency index indicating the degree of congruence between the item intention and the response behavior by analyzing the consistency between the psychological context of the item and the user's response method, generating an intermediate score for each value dimension by reflecting the above response tendency index, self-balance index, and behavioral consistency index in the value score calculation process, and then normalizing the above intermediate score according to a standard scale to calculate a user value score for each value dimension. Claim 2 delete Claim 3 delete Claim 4 delete Claim 5 delete Claim 6 In claim 1, the value orientation calculation unit converts the response value selected by the user for each item included in the item group corresponding to each value dimension into a range of 1 to 5 points, then calculates an extreme selection ratio by aggregating the number of items in which 5 points or 1 point was selected among all items belonging to the item group and dividing it by the total number of items, and calculates a middle selection ratio by dividing the number of items in which 3 points were selected by the total number of items, sets the value obtained by subtracting the middle selection ratio from the extreme selection ratio as the default value of the response orientation index, and calculates the response orientation index by applying a function defined such that if the default value is positive, it indicates an extreme response orientation, and if it is negative, it indicates a neutral response orientation; calculates a first average value by averaging the response values of items included in the self-related item group and a second average value by averaging the response values of items included in the social value-related item group, then calculates a difference value indicating the degree of egocentrism by subtracting the second average value from the first average value, wherein the greater the absolute value of the difference value, the more severe the bias toward the self or society. It is defined as such, and by normalizing the above difference value by dividing it by a predefined maximum bias value, a self-balancing index having a range between 0 and 1 is calculated; after predefining item pairs with the same psychological purpose among the items constituting the item group for each value dimension, the concordance ratio is calculated by comparing whether the response values of each item pair are in the same direction—indicating whether both are high or both are low scores—and dividing the number of matching item pairs by the total number of item pairs; for item pairs composed of inverse items, the concordance of the response values in the reverse direction is checked to calculate a separate concordance ratio, and then the average value of the two concordance ratios is set as the first-order consistency within the item group; additionally, the response values of the entire item group are arranged sequentially to calculate a variability value by summing all differences between adjacent responses, and after calculating an inverse normalized value between 0 and 1 by dividing the above variability value by a predefined maximum variability value,A psychological tendency-based matching recommendation system characterized by calculating a final behavioral consistency index by weighted averaging the above-mentioned denormalized value and the above-mentioned first-order consistency.
Citation Information
Patent Citations
Method for personality matching and recording medium thereof
KR100840215B1
Method for providing character analyzing service using questionnaire for psychotherapy in diffrent circumstance
KR101977559B1
Matching Process System And Method
US20140074824A1
