Application of AI high risk detection algorithms based on emotional, cognitive and behavioral indicators

Abstract

The present invention relates to a system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, comprising: a dialogue analyzer that calculates emotional, cognitive, and behavioral indicators on a sentence-by-sentence basis from a user's voice or text input; a weight integration unit that calculates an integrated risk index by applying weights to each indicator; a stabilization unit that mitigates noise caused by short-term emotional fluctuations by applying an exponential moving average (EMA); and a risk grade classifier that classifies risk grades based on the integrated risk index and generates corresponding response scenarios. According to one embodiment of the present invention as described above, by applying an exponential moving average (EMA) to correct for temporal variability, temporary emotional response stabilization and inconsistent speech are eliminated, and only the changing trends of emotional, cognitive, and behavioral patterns that appear continuously over time are reflected to classify into “normal-low risk-moderate risk-high risk.” By performing step-by-step psychological interventions such as empathetic feedback, suggestion of coping strategies, and linkage to emergency counseling through an automatic response logic for each grade, an automatic response system can be implemented that detects the psychological risk level of a user in real time and supports step-by-step intervention in various mental health management environments encompassing education, welfare, and medical fields, as well as special occupational groups such as the military, police, and fire department.

Description

Technology Field

[0001] The present invention relates to a system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, and more specifically, to a system for applying an algorithm to precisely calculate the level of psychological risk (suicide, self-harm, and other crisis signs) by analyzing user speech data in an AI-based counseling or conversation environment. Background Technology

[0002] The digital mental healthcare market is growing rapidly, and the need for mental health services is increasing. As awareness of mental health issues such as stress, anxiety, and depression rises, accessibility and convenience through digital platforms are emerging as key topics in the digital mental healthcare market.

[0003] Furthermore, in order to improve psychological counseling services utilizing AI technology, there is a growing need to enhance the quality of psychological analysis to increase the accuracy and reliability of individual psychological tests, to create a customized psychological counseling environment that analyzes the causes of clients' emotional anxiety symptoms and discovers healing curricula effective for psychological stability, and to increase accessibility to initial psychological counseling.

[0004] However, conventional risk detection technology has the following structural limitations.

[0005] 1) Problems with detection methods based on simple keyword frequency (e.g., “I want to die,” “it is hard”) that fail to distinguish contextual meaning or speech intent, resulting in high rates of false positives and false negatives,

[0006] 2) Problems arising from the failure to integrally reflect key psychological factors such as emotional expression, cognitive distortion, and behavioral intention, resulting in a discrepancy between actual risk and psychological context, and

[0007] 3) A problem in which temporary emotional outbursts or verbal exaggerations are incorrectly judged as danger signals, leading to frequent misclassification of non-risk users as high-risk users.

[0008] Accordingly, the applicant proposes a system that implements an automated response system capable of detecting the psychological risk of users in real time and supporting step-by-step interventions in various mental health management environments, including education, welfare, and medical fields as well as special occupational groups such as the military, police, and fire department, by applying an Exponential Moving Average (EMA) to correct for temporal variability, thereby eliminating temporary emotional response stabilization and inconsistent speech, and classifying into “Normal-Low Risk-Moderate Risk-High Risk” by reflecting only the changing trends of emotional, cognitive, and behavioral patterns that appear continuously over time, and performing step-by-step psychological interventions such as empathy feedback, suggestion of coping strategies, and linkage to emergency counseling through an automated response logic for each grade. Prior art literature

[0009] Republic of Korea Published Patent No. 2024-14154 (Publication Date: Feb. 1, 2024) Republic of Korea Published Patent No. 2023-116523 (Publication Date: Aug. 4, 2023) The problem to be solved

[0010] The objective of the present invention is to provide an automated response system that supports early detection of high-risk groups and counseling intervention by calculating an ECA Risk Index based on psychological validity through a multidimensional analysis algorithm based on Emotion, Cognition, and Action indicators extracted from user conversations, and classifying it in real-time. means of solving the problem

[0011] The present invention, for achieving these technical challenges, is a system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, characterized by comprising: a dialogue analyzer that calculates emotional, cognitive, and behavioral indicators on a sentence-by-sentence basis from a user's voice or text input; a weight integration unit that calculates an integrated risk index by applying weights to each indicator; a stabilization unit that mitigates noise caused by short-term emotional fluctuations by applying an exponential moving average (EMA); and a risk grade classifier that classifies risk grades based on the integrated risk index and generates corresponding response scenarios.

[0012] Preferably, the conversation analyzer comprises: an emotion analysis module that detects emotion words from text converted from the user's voice and calculates emotion intensity and polarity scores to produce an emotion index; a cognitive analysis module that detects negative beliefs and self-deprecating expressions and detects cognitive distortion patterns to produce a cognitive index; and a behavior analysis module that detects sentences containing intentions of avoidance, disconnection, self-harm, or abandonment to produce a behavior index.

[0013] In addition, the weight integration unit is characterized by including: a risk calculation module that calculates risk based on emotional change trends, behavioral risk scores, and time-series behavioral patterns authorized by the conversation analyzer; and a weight adjustment module that dynamically adjusts weights for the calculated risk according to user type.

[0014] In addition, the stabilization unit is characterized by including a stabilization module that emphasizes a continuity pattern to mitigate abrupt emotional fluctuations caused by short-term emotional fluctuations by applying an Exponential Moving Average (EMA); and a risk index management module that controls the application of the Exponential Moving Average (EMA) by the stabilization module.

[0015] In addition, the risk class classifier is characterized by including a risk class classification module that classifies the exponential moving average (EMA) based on risk into three grades: Alert, Warning, and Critical; and a scenario matching module that matches pre-set response scenarios for each classified grade. Effects of the invention

[0016] According to the present invention as described above, through a multidimensional analysis algorithm based on Emotion, Cognition, and Action indicators extracted from a user's conversation, an ECA Risk Index based on psychological validity is calculated and graded in real time, thereby eliminating temporary emotional response stabilization and inconsistent utterances, and allowing for classification into “Normal-Low Risk-Moderate Risk-High Risk” by reflecting only the changing trends of emotion, cognition, and behavior patterns that appear continuously over time.

[0017] According to the present invention, by enabling step-by-step psychological intervention, such as empathetic feedback, suggestion of coping strategies, and linkage to emergency counseling through grade-based automatic response logic, it is possible to implement an automatic response system that detects the psychological risk level of a user in real time and supports step-by-step intervention in various mental health management environments encompassing education, welfare, and medical fields, as well as special occupational groups such as the military, police, and fire department. Brief explanation of the drawing

[0018] FIG. 1 is a configuration diagram illustrating an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 2 is a configuration diagram illustrating a conversation analyzer of an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 3 is a configuration diagram illustrating a weight integration unit of an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 4 is a configuration diagram illustrating a stabilization unit of an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 5 is a configuration diagram illustrating a risk grade classifier of an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 6 is a configuration diagram illustrating an additional configuration of an artificial intelligence high-risk detection algorithm application system based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 7 is a flowchart illustrating a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 8 is a flowchart illustrating the detailed process of step S702 of a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 9 is a flowchart illustrating the detailed process of step S704 of a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 10 is a flowchart illustrating the detailed process of step S706 of a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. FIG. 11 is a flowchart illustrating the detailed process of step S708 of a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention. Specific details for implementing the invention

[0019] The specific features and advantages of the present invention will become more apparent from the following detailed description based on the accompanying drawings. Prior to this, terms and words used in this specification and claims should be interpreted in a meaning and concept consistent with the technical spirit of the present invention, based on the principle that the inventor may appropriately define the concept of the terms to best describe his invention. Furthermore, it should be noted that specific descriptions of known functions and configurations related to the present invention have been omitted where it is determined that such descriptions might unnecessarily obscure the essence of the invention.

[0020] As illustrated in FIG. 1, an artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention includes a conversation analyzer (100) that calculates emotional, cognitive, and behavioral indicators in sentence units from a user's voice or text input, a weighted integration unit (200) that calculates an integrated risk index by applying weights for each indicator, a stabilization unit (300) that mitigates noise caused by short-term emotional fluctuations by applying an exponential moving average (EMA), and a risk classifier (400) that classifies risk grades based on an integrated risk index including alert, warning, and critical, and generates response scenarios accordingly.

[0021] Although specific details will be omitted below, the system (S) applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention can secure much higher precision and clinical validity than simple keyword detection by integrating psychological grounds (emotional, cognitive, and behavioral theories, Beck's CBT Model, etc.) into an AI analysis framework.

[0023] Hereinafter, with reference to FIG. 2, the conversation analyzer (100) of the artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is described as follows.

[0024] A conversation analyzer (100) comprises an emotion analyzer (110) that detects emotional words from text converted from the user's voice and calculates emotional intensity and polarity scores to produce an emotional change trend (emotional indicator), a cognition analyzer (120) that detects negative beliefs and self-deprecating expressions and detects cognitive distortion patterns (dichotomy, overgeneralization, etc.) to produce a cognitive distortion score (cognitive indicator), and an action analyzer (130) that detects sentences containing intentions of avoidance, disconnection, self-harm, or abandonment to produce a behavioral risk score (behavioral indicator) and records time-series behavioral patterns from the detected sentences.

[0025] Specifically, the emotion analysis module (110) produces an emotion change trend (emotion indicator) including emotion words, emotion intensity scores, and the direction of emotion change.

[0026] At this time, the emotion analysis module (110) calculates an emotion score using an emotion dictionary and a deep learning-based emotion classifier, for example, if the detected word is “tired”, the emotion score is calculated as “sadness +0.7, fatigue +0.6”, and if the detected word is “I want to disappear”, the emotion score is calculated as “despair +0.9”, and the final result emotion index E = 0.8 (depression, despair level) is calculated.

[0027] Additionally, the cognitive analysis module (120) analyzes cognitive distortion patterns (Cognition Indicators) including negative beliefs, self-deprecating expressions, and cognitive distortions (overgeneralization and dichotomous thinking).

[0028] At this time, if the word detected by the cognitive analysis module (120) is “I want to disappear,” the cognitive distortion pattern is calculated as “dichotomous thinking,” and if it is “I have no meaning,” the cognitive distortion pattern is calculated as “belief of worthlessness,” and the cognitive score C = 0.65 is calculated according to the frequency and intensity of such negative cognitive expressions.

[0029] And, the behavior analysis module (130) analyzes time-series behavioral patterns (action indicators) including avoidance, disconnection, self-harm, and helplessness.

[0030] At this time, if the sentence detected by the behavior analysis module (130) is “I want to disappear,” it is classified as “passive self-harm / avoidance intention” and calculated as a behavior score A = 0.7, and if the same pattern continues in the last 3 utterances, the behavior score is weighted and reinforced ( ) can.

[0032] Hereinafter, with reference to FIG. 3, the weight integration unit (200) of the artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is described as follows.

[0033] The weight integration unit (200) is configured to include a risk calculation module (210) that calculates risk based on emotional change trends, behavioral risk scores, and time-series behavioral patterns authorized by the conversation analyzer (100), a weight adjustment module (220) that dynamically adjusts the weights of the calculated risk according to user type, and an update module (230) that updates each indicator in real time.

[0034] Specifically, the risk calculation module (210) [calculates] the trend of emotional change (emotion indicator). Defined as, and cognitive distortion patterns (Cognition Indicators) Defined as, time-series behavioral patterns (Action Indicators) Defined as such, and each indicator is summed to calculate the risk as shown in [Equation 1] below.

[0035] [Mathematical Formula 1]

[0036]

[0037] For example, emotional indicators, cognitive indicators, and behavioral indicators In this case, the risk is It is calculated as.

[0038] And, the weight adjustment module (220) is the weight applied to the risk calculation module (210). , , and It is configured to dynamically adjust according to a pre-set user type (e.g., youth, soldier, civilian) and conversation context. At this time, the weight adjustment module (220) is driven by an adaptive weight adjustment algorithm that adjusts each weight we, wc, waw_e, w_c, w_awe, wc, wa in real time according to the user type and conversation context.

[0039] As mentioned earlier, emotional indicators, cognitive indicators, and behavioral indicators In this case, the risk calculated by the risk calculation module (210) is 0.735, and when the weight adjustment module (220) applies the exponential moving average (EMA) to it, It is calculated as, and the risk to which the Exponential Moving Average (EMA) is applied is It is adjusted (judged as a persistent risk signal).

[0041] Hereinafter, with reference to FIG. 4, the stabilization unit (300) of the artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is described as follows.

[0042] The stabilization unit (300) is configured to include a stabilization module (310) that emphasizes a continuity pattern to mitigate rapid emotional fluctuations caused by short-term emotional fluctuations by applying an exponential moving average (EMA), and a risk index management module (320) that controls the application of the exponential moving average (EMA) of the stabilization module (310) to emphasize the continuity pattern.

[0043] Specifically, the stabilization module (310) is configured to mitigate the rapid rise in Risk value caused by short-term emotional outbursts and to correct the Risk index to emphasize continuous risk patterns, but to correct the volatility caused by short-term emotional outbursts, it can apply a time-weighted EMA as shown in [Equation 2] below to reinforce only continuous risk patterns.

[0044] [Mathematical Formula 2]

[0045]

[0047] Hereinafter, with reference to FIG. 5, the risk classifier (400) of the artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is described as follows.

[0048] The risk class classifier (400) is configured to include a risk class classification module (410) that classifies the exponential moving average (EMA) according to risk into three grades: Alert, Warning, and Critical, and a scenario matching module (420) that matches pre-set response scenarios for each classified grade.

[0049] Specifically, the risk classification module (410) classifies the exponential moving average (EMA) based on risk into three grades: Alert, Warning, and Critical, wherein the risk index Alert is Set to, and the risk index risk (Warning) is It is set to, and the high-risk (Critical) risk index is Set to.

[0050] for example When -> -> Determined as "Critical".

[0051] And, the scenario matching module (420) automatically matches pre-set response scenarios (e.g., connecting to a counselor, providing safety plan guidance, reporting notification, etc.) according to the classified grade.

[0052] For example, in the case of “high risk,” the AI ​​response can provide a response such as, “I’m so glad you told me now. Is there anyone nearby who can be with you?”

[0053] Meanwhile, [Table 1] below is an example scenario of the operation of the scenario matching module (420).

[0054] step Input sentence Analysis results Risk index Risk level Automatic response 1 "I'm having such a hard time these days." E=0.6, C=0.3, A=0.1 0.43 Alert Maintain empathetic conversation 2 "I really want to quit everything." E=0.8, C=0.6, A=0.6 0.69 Warning Stabilization conversation transition 3 I'd rather it just disappear. E=0.9, C=0.7, A=0.8 0.82 Critical Activation of safety plan scenario

[0056] Hereinafter, with reference to FIG. 6, the additional configuration of the artificial intelligence high-risk detection algorithm application system (S) based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is as follows.

[0057] First, the input unit (500) is configured to include a voice input module (510) for collecting the user's voice, a text input module (520) for collecting chat and conversation text, and a video input module (530) for collecting non-verbal signals including the user's facial expressions and posture.

[0058] Additionally, the preprocessing / conversion unit (600) is configured to include a STT (Speech-to-Text) conversion module (610) that converts speech collected by the input unit (500) into text, a sentence splitting module (620) that separates the text collected by the input unit (500) into sentence units, and a morphological / semantic analysis module (630) that extracts key words and contextual meanings from the separated sentences.

[0059] For example, when a user says, “I’m so tired these days, and I just want to disappear,” the STT conversion module (610) converts the voice input into text, the sentence splitting module (620) separates the text into sentences, and the morphological / semantic analysis module (630) extracts key words (“tired”, “want to disappear”) through morphological analysis.

[0060] And, the output unit (700) is configured to include an LLM-based conversation engine (710) that generates a pre-set customized response according to the risk level, an administrator dashboard (720) that visualizes and displays the risk index, level, and conversation summary on the screen, and a notification module (730) that transmits user information classified as high risk by the risk level classifier (400) to the administrator.

[0061] The notification module (730) immediately sends an alert to the administrator dashboard (720) and generates a notification to connect with a counselor when specific user information is classified as high risk, and automatically returns to the caution state by reducing the EMA-based score when the risk level stabilizes downward.

[0063] Hereinafter, with reference to FIG. 7, a method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to an embodiment of the present invention is described as follows.

[0064] First, a conversation analyzer (100) calculates sentiment, cognition, and behavior-related indicators in sentence units from the user's voice or text input (S702).

[0065] Next, the weight integration unit (200) calculates the integrated risk index by applying weights to each indicator (S704).

[0066] Subsequently, the stabilization unit (300) applies an exponential moving average (EMA) to mitigate noise caused by short-term emotional fluctuations (S706).

[0067] Then, the risk class classifier (400) classifies the risk level based on the integrated risk index and generates a response scenario accordingly (S708).

[0069] Hereinafter, with reference to FIG. 8, the detailed process of step S702 of the method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is as follows.

[0070] First, a conversation analyzer (100) detects emotional words from text converted from the user's voice and calculates emotional intensity and polarity scores to produce an emotional index (S802).

[0071] Next, the conversation analyzer (100) detects negative beliefs and self-deprecating expressions, and detects cognitive distortion patterns (dichotomy, overgeneralization, etc.) to calculate cognitive indicators (S804).

[0072] And, the conversation analyzer (100) detects sentences containing the intention to avoid, disconnect, self-harm, or give up and calculates a behavioral indicator (S806).

[0074] Hereinafter, with reference to FIG. 9, the detailed process of step S704 of the method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is as follows.

[0075] After step S702, the weight integration unit (200) calculates the risk based on the emotional change trend, the behavioral risk score, and the time-series behavioral pattern (S902).

[0076] Next, the weight integration unit (200) dynamically adjusts the weights of the calculated risk according to user type (S904).

[0077] Then, the weight integration unit (200) updates each indicator in real time (S906).

[0079] Hereinafter, with reference to FIG. 10, the detailed process of step S706 of the method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is as follows.

[0080] After step S704, the stabilization unit (300) applies an exponential moving average (EMA) to emphasize a continuous pattern so that rapid emotional fluctuations caused by short-term emotional fluctuations are mitigated (S1002).

[0081] And, the stabilization unit (300) controls the application of the exponential moving average (EMA) (S1004).

[0083] Hereinafter, with reference to FIG. 11, the detailed process of step S708 of the method for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators according to one embodiment of the present invention is as follows.

[0084] After step S706, the risk class classifier (400) classifies the exponential moving average (EMA) according to risk into three grades: Alert, Warning, and Critical (S1102).

[0085] Then, the risk class classifier (400) matches pre-set response scenarios according to the classified grades (S1104).

[0086] As such, according to one embodiment of the present invention as described above, by applying an exponential moving average (EMA) to correct for temporal variability, temporary emotional response stabilization and inconsistent speech are eliminated, and only the changing trends of emotional, cognitive, and behavioral patterns that appear continuously over time are reflected to classify into “normal-low risk-moderate risk-high risk.”

[0087] Furthermore, by performing step-by-step psychological interventions—such as empathetic feedback, suggestions for coping strategies, and referrals for emergency counseling—through grade-based automated response logic, it is possible to implement an automated response system that detects users' psychological risk levels in real time and supports step-by-step interventions in diverse mental health management environments encompassing education, welfare, and medical fields, as well as special occupational groups such as the military, police, and fire department.

[0088] Although preferred embodiments illustrating the technical concept of the present invention have been described and illustrated above, the present invention is not limited to the configuration and operation as illustrated and described, and those skilled in the art will understand that numerous changes and modifications can be made to the present invention without departing from the scope of the technical concept. Accordingly, all such appropriate changes and modifications and equivalents should be considered to be within the scope of the present invention. Explanation of the symbols

[0089] S: System applying an AI high-risk detection algorithm based on emotional, cognitive, and behavioral indicators 100: Conversation Analyzer 110: Sentiment Analysis Module 120: Cognitive Analysis Module 130: Behavioral Analysis Module 200: Weight Integration Unit 210: Risk Calculation Module 220: Weight Adjustment Module 230: Update Module 300: Stabilization unit 310: Stabilization module 320: Risk Index Management Module 400: Risk Classifier 410: Risk Classification Module 420: Scenario Matching Module 500: Input section 510: Voice input module 520: Text input module 530: Image input module 600: Preprocessing / Conversion Unit 610: STT Conversion Module 620: Sentence Partitioning Module 630: Morphological / Semantic Analysis Module 700: Output section 710: LLM-based dialogue engine 720: Admin Dashboard 730: Notification Module

Claims

Claim 1 A system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, characterized by comprising: a conversation analyzer that calculates emotion, cognitive, and behavioral indicators on a sentence-by-sentence basis from a user's voice or text input; a weight integration unit that calculates an integrated risk index by applying weights for each indicator; a stabilization unit that mitigates noise caused by short-term emotional fluctuations by applying an exponential moving average (EMA); and a risk grade classifier that classifies risk grades based on the integrated risk index and generates corresponding response scenarios, wherein the stabilization unit includes: a stabilization module that emphasizes continuous patterns to mitigate abrupt emotional fluctuations caused by short-term emotional fluctuations by applying an exponential moving average (EMA); and a risk index management module that controls the application of the exponential moving average (EMA) by the stabilization module. Claim 2 A system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, characterized in that, in claim 1, the conversation analyzer comprises: an emotional analysis module that detects emotional words from text converted from the user's voice and calculates emotional intensity and polarity scores to produce an emotional indicator; a cognitive analysis module that detects negative beliefs and self-deprecating expressions and detects cognitive distortion patterns to produce a cognitive indicator; and a behavioral analysis module that detects sentences containing intentions of avoidance, disconnection, self-harm, or abandonment to produce a behavioral indicator. Claim 3 A system for applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, characterized in that, in claim 1, the weight integration unit comprises: a risk calculation module that calculates risk based on an emotional change trend, a behavioral risk score, and a time-series behavioral pattern authorized by the conversation analyzer; and a weight adjustment module that dynamically adjusts weights for the calculated risk according to user type. Claim 4 delete Claim 5 A system applying an artificial intelligence high-risk detection algorithm based on emotional, cognitive, and behavioral indicators, characterized in that, in claim 1, the risk classifier comprises: a risk class classification module that classifies the exponential moving average (EMA) according to risk into three grades: Alert, Warning, and Critical; and a scenario matching module that matches pre-set response scenarios for each classified grade.

Images