Internet-of-things-based pension safety monitoring method and system

Abstract

The invention belongs to the field of old-age care safety monitoring, and discloses an Internet of Things-based old-age care safety monitoring method and system, and the method comprises the steps: carrying out the correlation analysis of an intervener, an intervention strategy, a medicine-taking conflict behavior mode and an intervention result in a historical intervention record, and constructing a mapping relation between an intervention success rate and a situation, determining and recommending suitable interveners and intervention strategies in combination with the available state of the interveners and the current conflict behavior mode which are acquired in real time; the problem that in the prior art, personalized interveners and intervention strategies cannot be intelligently recommended according to the medicine taking conflict behavior mode of the old people and the available state of the interveners is solved, and therefore intervention effectiveness and medicine taking compliance are improved.

Description

Technical Field

[0001] The present application relates to the field of elderly care safety monitoring, and specifically to a method and system for elderly care safety monitoring based on the Internet of Things. Background Art

[0002] IoT-based medication monitoring systems have become widely used in home-based elderly care settings. Their core is a smart pillbox with multiple medicine compartments that automatically unlocks according to schedule and uses sensors to detect opening and medication removal. Data is uploaded to a server via a wireless network. The system then generates medication adherence reports, sends voice reminders, and sends alerts to family members if medication is not removed within the prescribed timeframe. However, this mechanism suffers from unclear responsibilities: when multiple people receive alerts simultaneously, responses can be inconsistent or delayed, leading to ineffective interventions.

[0003] To improve response coordination, the existing system incorporates a task assignment mechanism: alerts are pushed to a pre-set list of contacts, with the time of receipt recorded. If no one responds after a certain timeout, the alert is automatically pushed to the next contact. Once someone confirms the action, the system marks the task as "assigned" to prevent duplication or omission. While this approach clarifies accountability, it presents the challenge of insufficient verification of intervention effectiveness. Elderly individuals may feign medication intake due to emotional or cognitive issues (e.g., holding medication without swallowing or vomiting it afterward). The system only confirms that the task has been taken over, but cannot verify the actual effectiveness of medication, creating a monitoring blind spot.

[0004] To address this issue, some systems have integrated intervention verification modules: the interventionist initiates "on-site medication verification" through a mobile app, using a camera to record the medication process. The system then uses gesture recognition to analyze swallowing movements and record an intervention observation log. Once uploaded, the verification data is linked to the compliance record to provide feedback on the effectiveness. However, the system can only determine the success or failure of a single intervention and lacks the ability to analyze behavioral patterns: Elderly resistance behaviors often exhibit regular patterns (such as specific time periods, medication flavors, or resistance induced by the interventionist). Existing technologies are unable to identify triggers and patterns, resulting in a lack of strategies for caregivers when faced with repeated resistance, which increases the burden on caregivers.

[0005] To this end, the advanced system has added a behavioral pattern analysis module: it aggregates intervention verification records (emotions, intervention descriptions, results) and environmental sensor data (millimeter wave radar, speakers, lighting), and constructs a personalized "medication behavior map" through trend analysis. For example, it identifies patterns such as "high resistance to taking antihypertensive drugs at 3 pm and low activity in the morning" or "child C's intervention success rate is significantly lower than child A." However, such systems still have a key flaw: they can only explain "why" resistance occurs, and lack actionable intervention strategy recommendations. For example, after identifying that "the bitter taste of the drug causes resistance," it is unable to automatically generate specific plans such as "mixing with yogurt to take," "playing favorite music," or "designating a caregiver to assist," forcing caregivers to explore strategies on their own, limiting the actual value of the system in improving compliance.

[0006] Therefore, there is an urgent need for a solution that can dynamically recommend personalized intervention strategies based on real-time behavioral patterns and intervener status to break through the existing technical bottleneck. Summary of the Invention

[0007] The purpose of this application is to provide a method and system for elderly care safety monitoring based on the Internet of Things, which solves the problem in the existing technology that it is impossible to intelligently recommend personalized interveners and intervention strategies based on the elderly's medication resistance behavior patterns and the available status of interveners, thereby improving the effectiveness of intervention and medication compliance.

[0008] In a first aspect, the present application provides an Internet of Things-based elderly care safety monitoring method for monitoring elderly people's medication to improve elderly people's compliance with medication safety. The method comprises the following steps: A1. Obtain intervention records; these records include the identity of the intervener, intervention strategy, the elderly individual's medication resistance behavior patterns, and intervention outcomes. A2. Based on the intervention records, analyze the correlation between the interventionist, intervention strategy, medication resistance behavior pattern, and intervention success rate; A3. When the elderly person's medication resistance behavior is monitored, the available status information of each preset interventionist and the current medication resistance behavior pattern are obtained; A4. Determine a recommended intervener and a recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship; A5. Send intervention prompt information including the recommended intervention strategy to the recommended interventionist.

[0009] In a second aspect, the present application provides an IoT-based elderly care safety monitoring system for monitoring elderly people's medication use to improve elderly people's medication compliance. The system includes an IoT-based medication monitoring system, a server, and multiple interventionist terminals. The medication monitoring system and the multiple interventionist terminals are both communicatively connected to the server. The medication monitoring system is used to monitor the medication status of the elderly, identify the elderly's medication resistance behavior and the corresponding medication resistance behavior pattern, and upload the identification results to the server; The server is used to store intervention records and to perform: Extracting intervention records stored locally; the intervention records include the identity of the intervener of the historical medication intervention event, the intervention strategy, the elderly person's medication resistance behavior pattern, and the intervention results; Based on the intervention records, the correlation between the interventionist, intervention strategy, medication resistance behavior pattern and intervention success rate was analyzed; When the medication monitoring system detects medication resistance behavior of the elderly, obtaining available status information and current medication resistance behavior pattern of each preset intervener; Determining a recommended intervener and a recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship; Sending intervention prompt information including the recommended intervention strategy to the intervener terminal of the recommended intervener.

[0010] Beneficial effects: The present application provides an Internet of Things-based elderly care safety monitoring method and system, which correlates and analyzes the interveners, intervention strategies, medication resistance behavior patterns and intervention results in historical intervention records, thereby constructing a mapping relationship between intervention success rate and context, and combining the real-time available status of the interveners and the current resistance behavior patterns to determine and recommend suitable interveners and intervention strategies; it solves the problem in the existing technology that it is impossible to intelligently recommend personalized interveners and intervention strategies based on the elderly's medication resistance behavior patterns and the available status of the interveners, thereby improving the effectiveness of the intervention and medication compliance. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a flowchart of the IoT-based elderly care safety monitoring method provided in an embodiment of the present application.

[0012] Figure 2 A schematic diagram of the structure of the Internet of Things-based elderly care safety monitoring system provided in an embodiment of the present application.

[0013] Explanation of the numbers: 1. Medication monitoring system; 2. Server; 3. Intervention terminal. DETAILED DESCRIPTION

[0014] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The components of the embodiments of the present application generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the application for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work fall within the scope of protection of the present application.

[0015] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and should not be understood as indicating or implying relative importance.

[0016] Please refer to Figure 1 In some embodiments of the present application, an IoT-based elderly care safety monitoring method is used to monitor elderly people's medication use to improve elderly people's medication compliance. The method includes the following steps: A1. Obtain intervention records; these records include the identity of the intervener, intervention strategy, the elderly individual's medication resistance behavior patterns, and intervention outcomes. A2. Based on the intervention records, analyze the correlation between the interventionist, intervention strategy, medication resistance behavior pattern, and intervention success rate; A3. When the elderly person's medication resistance behavior is monitored, the available status information of each preset interventionist and the current medication resistance behavior pattern are obtained; A4. Determine a recommended intervener and a recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship; A5. Send intervention prompt information including the recommended intervention strategy to the recommended interventionist.

[0017] Among them, intervention records refer to the historical intervention event data stored in the system, which can be implemented using a database, data warehouse or distributed file system, such as a table in a relational database or a document collection in a NoSQL database. It is mainly used to accumulate and manage past intervention experience and provide a data basis for subsequent analysis.

[0018] Among them, the intervener identity refers to the identification information of the person participating in the intervention, which can be implemented by user ID, name or role code, such as the child's ID and the caregiver's number. It is mainly used to distinguish different intervention subjects to facilitate the analysis of the intervention effects of different interveners in specific situations.

[0019] Among them, intervention strategies refer to the response measures taken against the elderly's resistance to medication. They can be implemented using preset text descriptions, behavioral instructions or multimedia content, such as mixing drugs with food, playing soothing music, and persuasion by family members. They are mainly used to guide the interventionists to take action and improve the targetedness and success rate of the intervention.

[0020] Among them, the medication resistance behavior pattern refers to the non-compliance behavior characteristics or patterns exhibited by the elderly during the medication process. It can be achieved by using behavioral labels, emotional state descriptions or a combination of physiological indicators, such as rejection of bitter-tasting medicines in the early morning, resistance to specific interventionists, and spitting out medicine immediately after taking it. It is mainly used to identify the causes and patterns that lead to non-compliance behavior and provide a basis for personalized intervention.

[0021] Among them, the association relationship refers to the data mapping or statistical connection between the intervener, intervention strategy, medication resistance behavior pattern and intervention success rate, which can be implemented using a decision tree model, regression model or rule set. For example, when the elderly show resistance to bitter-tasting medicines, the success rate of the mixed food strategy adopted by their children is higher. It is mainly used to reveal the probability of intervention success in different intervention scenarios and provide a decision-making basis for recommendations.

[0022] Among them, the available status information refers to the status of whether the preset intervener is currently able to perform the intervention task, which can be determined by online / offline status, busy / idle status or geographic location information, such as through status updates of mobile phone applications and activity data of smart wearable devices. It is mainly used to ensure that the recommended intervener can be contacted at the current moment and can actually perform the intervention to avoid invalid recommendations.

[0023] The core innovation of this application lies in correlating interventionists, intervention strategies, and medication resistance behavior patterns with intervention outcomes in historical intervention records. This allows for a mapping between intervention success rate and context. This, combined with real-time access to interventionist availability and current resistance behavior patterns, identifies and recommends appropriate interventionists and intervention strategies. This addresses the shortcomings of existing systems, which can only identify problems but not provide specific solutions, thereby improving intervention efficiency and medication compliance.

[0024] Specifically, this method collects historical intervention records, including interventionist identities, intervention strategies, elderly individuals' medication resistance patterns, and intervention outcomes, laying the data foundation for subsequent analysis. Based on these historical records, the method analyzes the correlations between interventionists, intervention strategies, medication resistance patterns, and intervention success rates. This allows the method to learn which interventionists are most likely to succeed with which intervention strategies under specific medication resistance patterns, providing a knowledge base for recommendations. When medication resistance is detected in an elderly individual, the method obtains information about the available interventionists in the current context and the current medication resistance pattern. Subsequently, the method combines this current contextual information with the analyzed correlations to identify currently available interventionist and intervention strategy combinations with high success rates for the current resistance pattern, thereby determining recommended interventionists and intervention strategies. Finally, the method sends intervention prompts containing the recommended intervention strategies to the recommending interventionist, guiding them to take action. This approach ensures that decisions are translated into action, improving the efficiency and success rate of interventions.

[0025] Through the above scheme, this application can analyze and identify the relationship between interveners, intervention strategies, medication resistance behavior patterns and intervention success rates based on historical intervention data. Therefore, when elderly people show medication resistance behavior, it can automatically recommend suitable interveners and intervention strategies based on real-time available status. This solves the problem that existing systems cannot provide specific and executable intervention suggestions, avoids guardians from exploring solutions on their own, reduces the complexity and uncertainty of care, improves the relevance and success rate of intervention, and ultimately improves the elderly's compliance with safe medication use.

[0026] As a preferred embodiment, the solution of the present application is specifically implemented as follows: the system continuously collects and stores intervention records, which can be stored in a cloud database, including the identifier of each intervention event, the identity code of the family member or caregiver who performed the intervention, the communication method or auxiliary means used (such as verbal persuasion, mixing the medicine in juice), the behavioral characteristics displayed by the elderly at the time (such as shaking the head in refusal, keeping the mouth shut, and being irritable), and whether the intervention was ultimately successful (such as successful medication or failure to take the medicine). A data analysis module, such as a machine learning model deployed on a server, processes these historical intervention records regularly or in real time. This module performs a combined analysis of the intervener's identity code, intervention strategy, and medication resistance behavior pattern, and counts the number of successful medications and the total number of interventions under each combination, thereby calculating the intervention success rate for each scenario combination, and storing these success rates and corresponding combinations as an association relationship table. When an IoT medication monitoring system, using pill box sensors and gesture recognition cameras, detects that an elderly person is not taking medication on time and exhibits resistance, such as when the pill box remains unopened for an extended period or when the elderly person spits out medication, the system identifies the current medication resistance behavior pattern, such as resistance to a specific medication. Simultaneously, the system queries the mobile app status or schedules of all pre-defined interventionists (e.g., children or caregivers) to obtain their availability. A decision engine, such as a rule-based recommendation algorithm ranked by success rate, receives the current medication resistance behavior pattern and a list of available interventionists. It searches the association table for interventionist and intervention strategy combinations that match the current resistance behavior pattern and filters out currently unavailable interventionists. Among the remaining available combinations, the decision engine selects combinations with high intervention success rates as recommended interventionists and intervention strategies. Finally, the system sends an intervention reminder message containing a recommended intervention strategy (e.g., trying mixing medication with yogurt and playing their favorite music) to the recommended interventionist's (e.g., child) smartphone via a push message service, such as using the MQTT protocol or HTTP persistent connections, guiding them to take immediate action.

[0027] In some embodiments, step A2 comprises: A201. Combining the intervener identity, intervention strategy, and medication resistance behavior pattern in the intervention record to form multiple intervention scenario combinations; A202. For each combination of the intervention scenarios, count the total number of interventions and the number of successful interventions; A203. Calculate the intervention success rate for each intervention scenario combination based on the number of successes and the total number of interventions; A204. Associate the intervention success rate with the corresponding intervener identity, intervention strategy, and medication resistance behavior pattern to obtain the association relationship.

[0028] Among them, intervention scenario combination refers to the specific scenario classification formed by cross-combining the identity of the intervener recorded in the historical intervention data, the intervention strategy adopted, and the medication resistance behavior pattern shown by the elderly. It can be implemented using data structures such as multi-field joint indexes, hash maps or multidimensional arrays in the database to facilitate refined statistics and analysis of the intervention effects in specific situations.

[0029] In step A203, the intervention success rate can be calculated by simply dividing the number of successes by the total number of interventions. In step A204, this calculated intervention success rate, along with the corresponding interventionist identity (e.g., Child A), intervention strategy (e.g., personal persuasion), and medication resistance behavior pattern (e.g., refusal to swallow), can be stored as a new data entry in an association table or in memory as a key-value pair, for example, {"interventionist":"Child A","strategy":"personal persuasion","pattern":"refusal to swallow","success rate":0.8}. In this way, the system can establish clear success rate data for each unique combination of intervention scenarios, forming a knowledge base that can be queried and utilized.

[0030] This solution systematically refines intervention records to precisely establish the correlation between interventionists, intervention strategies, medication resistance behavior patterns, and intervention success rates. Specifically, by comprehensively combining the interventionist identity, intervention strategy, and medication resistance behavior patterns contained in the intervention records, multiple unique intervention scenario combinations are generated. This step elevates historical intervention events from a single dimension to a multi-dimensional contextual consideration, enabling subsequent analysis to target specific, reproducible intervention scenarios. Next, for each identified intervention scenario combination, the system counts the total number of interventions and the number of successful interventions in the historical data. This statistical process provides a direct data basis for quantifying the intervention effect in each specific scenario and ensures the objectivity of subsequent success rate calculations. Based on this statistical count, the system calculates the intervention success rate for each intervention scenario combination based on the number of successful interventions and the total number of interventions. This quantifies the abstract intervention effect into a concrete numerical value, providing a unified metric for comparing the effects of different scenario combinations. Ultimately, the calculated intervention success rate is explicitly bound to the corresponding intervener identity, intervention strategy, and medication resistance behavior pattern, thereby obtaining the desired association relationship. The establishment of this association relationship enables the system to transform from a rough "association obtained by analysis" into a refined, operational knowledge base, which clearly records the historical intervention success rate when a specific intervener and a specific intervention strategy face a specific medication resistance behavior pattern. This not only solves the problem of how to accurately calculate and obtain these associations, but also provides a solid data foundation and decision-making basis for subsequent systems to intelligently recommend the most likely successful interveners and intervention strategies from historical data based on the current medication resistance behavior pattern and the available status of the intervener when monitoring the medication resistance behavior of the elderly, thereby improving the accuracy and effectiveness of intervention recommendations.

[0031] Preferably, step A202 may include: B1. For each intervention scenario combination, count the total number of direct interventions and the number of direct successes; B2. Determine whether the total number of direct interventions is lower than the preset sample threshold; B3. When the total number of direct interventions is not less than the preset sample threshold, the total number of direct interventions and the number of direct successes are used as the total number of interventions and the number of successes for the intervention scenario combination; B4. When the total number of direct interventions is lower than the preset sample threshold, determine the corresponding macro-intervention scenario category based on the interventionist identity and medication resistance behavior pattern of the intervention scenario combination; B5. Count the total number of macro interventions and macro successes under the macro intervention scenario category; B6. Calculate the total number of interventions and the number of successes for the intervention scenario combination based on the total number of macro interventions, the number of macro successes, the total number of direct interventions, and the number of direct successes.

[0032] The total number of direct interventions and direct successes are the total number of interventions and successes for the corresponding intervention scenario combinations directly recorded in the intervention records.

[0033] The preset sample threshold is a key parameter used to determine whether the direct historical data for a specific combination of intervention scenarios is statistically significant. This threshold can be configured based on experience, statistical principles, or the system's requirements for data reliability. For example, it can be set to a minimum number of interventions, below which data is considered insufficient to independently support reliable success rate calculations.

[0034] Macro-intervention context categories are defined by generalizing the interventionist identity and medication resistance behavior patterns for specific intervention context combinations into broader categories. This generalization allows the system to aggregate more historical data, providing a more robust statistical basis for specific combinations where data is sparse. For example, a specific interventionist (e.g., "Child A") can be generalized to a more general category (e.g., "Family Member"), or a specific medication resistance behavior (e.g., "Refusal to Swallow") can be generalized to a more abstract category (e.g., "Oral Resistance").

[0035] In step B6, the calculation of the total number of interventions and successes for the intervention scenario combination involves fusing the direct statistical data for the specific intervention scenario combination with the macro-statistical data for the corresponding macro-intervention scenario category. This fusion can employ a variety of statistical methods, such as weighted averaging, Bayesian estimation, or other data smoothing techniques, aiming to balance specificity with statistical robustness, thereby obtaining an estimate that reflects the unique characteristics of the combination while benefiting from broader statistical patterns.

[0036] By introducing these features, the overall working principle of this solution is optimized, effectively addressing the data sparsity issue. First, the system performs preliminary direct statistics for each intervention scenario combination, obtaining its most granular historical data. Subsequently, the system intelligently assesses the reliability of this direct data by comparing it against a preset sample threshold. When sufficient direct data is available, the system directly uses this data as the total number of interventions and successes for that combination, ensuring accuracy in the absence of sufficient data. However, when direct data is insufficient to support reliable calculations, the system no longer relies on unreliable small sample data. Instead, it generalizes the combination to a broader macro-intervention scenario category based on the interventionist's identity and medication resistance behavior patterns. This generalization allows the system to leverage a larger amount of historical data within this macro-category, resulting in more statistically reliable macro-total number of interventions and macro-successes. Finally, the system cleverly fuses the direct data for a specific intervention scenario combination with the macro-category data, ensuring that the resulting total number of interventions and successes reflects the specificity of that combination while benefiting from broader statistical patterns. This provides a relatively accurate and robust success rate estimate in the presence of sparse data. This optimized statistical process provides more reliable input for calculating intervention success rates in the subsequent step A203, which in turn enables step A204 to establish more accurate correlations. Ultimately, this enables step A4 to make decisions based on more precise and reliable information when determining recommended interventionists and intervention strategies, thereby improving the effectiveness of the recommended strategies and the performance of the entire elderly care safety monitoring system.

[0037] Through the implementation of the above technical solutions, this application can effectively solve the problem that when the amount of historical data for a specific intervention scenario combination is insufficient, the calculated success rate is unstable and unrepresentative. This solution ensures that a more accurate and reliable estimate of the intervention success rate can be obtained even in small sample situations by introducing hierarchical statistics and data fusion mechanisms. This directly improves the accuracy of the system's analysis of the relationship between interveners, intervention strategies, and medication resistance behavior patterns, thereby enabling the system to determine more effective and accurate recommended interveners and recommended intervention strategies based on more reliable correlations when monitoring medication resistance behavior in the elderly, thereby improving the decision-making quality and overall performance of the elderly care safety monitoring system in actual medication compliance management.

[0038] Furthermore, step B4 may include: B401. Obtain a preset generalization rule set based on the intervention scenario combination of the interventionist identity and the medication resistance behavior pattern; the generalization rule set defines a hierarchical relationship generalized from the interventionist identity and the medication resistance behavior pattern to macro categories of different granularities; B402. Based on the generalization rule set, starting from the finest-grained category corresponding to the intervention scenario combination, generalize upward step by step to generate a series of candidate macro-intervention scenario categories, and record the generalization level of each candidate macro-intervention scenario category; B403. For each candidate macro-intervention scenario category, calculate the total number of historical interventions; B404. Traverse the candidate macro-intervention scenario categories, identify the candidate macro-intervention scenario categories in which the total historical number of interventions is not less than the preset macro-sample threshold, and from these identified categories, select the candidate macro-intervention scenario category with the lowest generalization level and determine it as the macro-intervention scenario category.

[0039] The generalization rule set refers to a set of predefined rules that guide the gradual abstraction or expansion of specific intervention scenario combinations (composed of interventionist identity and medication resistance behavior patterns) into broader macro-categories. This can be implemented in the form of a tree structure, hierarchical classification table, or semantic network, where each node or level represents a macro-category of varying granularity and defines the path from one category to its parent or higher-level categories.

[0040] The generalization level refers to the degree of abstraction or distance of a macro-category relative to the finest-grained category in a generalization rule set. It can be represented by an integer value. For example, the finest-grained category can be defined as level 0, its immediate parent category as level 1, and so on. A larger level value indicates a more macro- and generalized category.

[0041] The preset macro-sample threshold is a pre-set value used to determine whether a candidate macro-intervention scenario category has sufficient historical intervention data to support reliable success rate calculation. It can be expressed as an integer representing the minimum number of interventions, for example, 20 or 30, to ensure that the data volume of the selected macro-category meets statistical reliability requirements.

[0042] This solution provides a specific method for determining macro-intervention scenario categories. It aims to address the problem of selecting macro-categories that both provide sufficient samples and maintain the finest possible granularity when direct samples are insufficient, thereby improving the accuracy of intervention success rate calculations. Specifically, when the system determines that the total number of direct interventions for a particular intervention scenario combination falls below a preset sample threshold and requires supplemental macro-category data, it first retrieves a pre-defined set of generalization rules based on the interventionist identity and medication resistance behavior patterns for the current intervention scenario combination. This set, like a hierarchical knowledge graph, defines the generalization path and hierarchical relationships from the most specific intervention scenario to macro-categories of varying granularity. This provides structured guidance for the subsequent generalization process, ensuring that generalization is based on predefined logic and hierarchy rather than arbitrary generalization, thereby ensuring the rationality and controllability of the generalization results. Next, based on this set of generalization rules, the system generalizes upwards, starting from the finest granularity category corresponding to the current intervention scenario combination. This process generates a series of potential candidate macro-intervention scenario categories and precisely records the generalization hierarchy of each candidate category. This systematic, step-by-step generalization process generates multiple potential macro-categories and documents their degree of generalization, providing a comprehensive foundation for subsequent selection, enabling the system to select from macro-categories of varying granularity. Subsequently, for each candidate macro-intervention scenario category generated, the system calculates its historical total number of interventions. This statistic assesses whether each candidate macro-category provides sufficient sample size to meet the pre-set macro-sample threshold. By counting the historical total number of interventions, the sample richness of each candidate category can be quantified, providing data support for subsequent screening. Finally, the system iterates through all candidate macro-intervention scenario categories and identifies those whose historical total number of interventions is at least the pre-set macro-sample threshold. Among these categories that meet the sample size requirement, the system selects the category with the lowest level of generalization—that is, the category that most closely resembles the fine-grained nature of the original intervention scenario combination, while ensuring sample adequacy—as the final macro-intervention scenario category. This selection mechanism ensures that while replenishing the sample size, it also maximizes the preservation of the specific information of the original intervention scenario combination, avoiding information distortion caused by overgeneralization, thereby significantly improving the accuracy and reliability of subsequent intervention success rate calculations.

[0043] Through the coordinated operation of the above steps, this solution can intelligently and strategically select the most appropriate macro-category when sample data is insufficient, effectively solving the problem that simple generalization may lead to data distortion. In the absence of direct data, it is still possible to obtain a high-precision intervention success rate, thereby improving the accuracy and effectiveness of the recommended interveners and intervention strategies in the entire elderly care safety monitoring method.

[0044] In a specific embodiment, suppose the system detects an intervention scenario combination as "child A resists the bitter taste of the medicine for the elderly", and the total number of direct interventions is 5, which is lower than the preset sample threshold of 20 times. At this time, the system needs to determine a macro-intervention scenario category to supplement the sample. First, the system will obtain a preset generalization rule set based on the two features of "child A" and "bitter taste of the medicine". This set can define the following hierarchical relationship: Level 0: (interventionist: child A, resistance mode: bitter taste of the medicine); Level 1: (interventionist: child A, resistance mode: all medicines resistance) or (interventionist: all children, resistance mode: bitter taste of the medicine); Level 2: (interventionist: all children, resistance mode: all medicines resistance) or (interventionist: all family members, resistance mode: bitter taste of the medicine); Level 3: (interventionist: all family members, resistance mode: all reasons for resistance). Next, based on this set of generalization rules, the system begins with the finest-grained "Child A objects to the elderly person's medication due to its bitter taste" (Level 0) and generalizes upwards, generating a series of candidate macro-intervention scenario categories and recording their generalization levels. For example, candidate categories might include: "Child A objects to the elderly person's medication due to its bitter taste" (Level 0), "Child A objects to all medications taken by the elderly person" (Level 1), "All children objects to the elderly person's medication due to its bitter taste" (Level 1), "All children objects to all medications taken by the elderly person" (Level 2), "All family members objects to the elderly person's medication due to its bitter taste" (Level 2), and "All family members object to the elderly person's medication for all reasons" (Level 3). The system then calculates the total number of historical interventions for each candidate macro-intervention scenario category. Suppose the statistical results are as follows: "Child A objects to the elderly person's medication due to its bitter taste": 5 times; "Child A objects to all elderly person's medications": 15 times; "All children object to the elderly person's medication due to its bitter taste": 25 times; "All children object to all elderly person's medications": 40 times; "All family members object to the elderly person's medication due to its bitter taste": 35 times; and "All family members object to the elderly person for all reasons": 100 times. Finally, the system iterates through these candidate categories and identifies those whose total number of historical interventions is at least a preset macro-sample threshold (for example, 30 times). In this example, the categories that meet the criteria are: "All children object to all elderly person's medications" (40 times), "All family members object to the elderly person's medication due to its bitter taste" (35 times), and "All family members object to the elderly person for all reasons" (100 times). From these categories that meet the criteria, the system selects the one with the lowest level of generalization. In this example, "All children object to all medications taken by the elderly" and "All family members object to the elderly's medications due to their bitter taste" both belong to Level 2, the lowest level of generalization among the conditions. Assume that the system selects "All children object to all medications taken by the elderly" as the macro-intervention scenario category.In this way, the system can retain the specific information of the original intervention scenario combination to the maximum extent while ensuring the sample size, thereby providing a more accurate data basis for the subsequent calculation of the intervention success rate.

[0045] In some preferred embodiments, the intervention record further includes the intervention time; Step B6 includes: B601. Obtain the most recent intervention time of the intervention scenario combination, and the time distribution characteristics of the intervention records in the macro-intervention scenario category; B602. Determine the relative impact factors of the total number of direct interventions and the total number of macro interventions based on the most recent intervention time and the time distribution characteristics; B603. Using the relative impact factor, combine the total number of direct interventions with the total number of macro interventions to obtain the total number of interventions for the intervention scenario combination; B604. Use the relative impact factor to merge the direct success number and the macro success number to obtain the success number of the intervention scenario combination.

[0046] The intervention time refers to the time point when each intervention event occurs, which can be implemented using a timestamp, a date and time string, or any information format that can record time.

[0047] The most recent intervention time refers to the time point at which the most recent intervention occurred among all intervention records under an intervention scenario combination. It can be obtained by querying the intervention records associated with the intervention scenario combination in the database and selecting the record with the latest time value.

[0048] Among them, the time distribution characteristics refer to the distribution pattern of all intervention records in the macro-intervention scenario category on the time axis, which can be represented by statistical indicators such as average time, median, time span, data density curve or time series model.

[0049] Among them, the relative impact factor refers to the numerical value used to adjust the contribution ratio of direct data and macro data when integrating the two. It can be determined through preset functions, models or rules based on factors such as the timeliness, relevance or reliability of the data.

[0050] Fusion refers to combining the total number of direct interventions with the total number of macro interventions, as well as the number of direct successes with the number of macro successes, according to certain weights or proportions, to obtain representative statistical results, which can be achieved by weighted average, linear combination or other data fusion algorithms.

[0051] This application introduces intervention time information into intervention records and optimizes the steps for calculating the total number of interventions and successes for an intervention scenario combination. Specifically, when the sample size of direct intervention data is insufficient and data from the macro-intervention scenario category is needed to supplement it, the system first obtains the most recent intervention time for that intervention scenario combination, which reflects the dynamics of the combination. Simultaneously, the system obtains the temporal distribution characteristics of intervention records within the macro-intervention scenario category, enabling the system to understand the timeliness and data concentration of the macro-data set. Based on this temporal information, the system further determines the relative impact factor of the total number of direct interventions and the total number of macro-interventions. This relative impact factor is a quantified weight based on the timeliness of the direct and macro-data. It addresses the problem that simple fusion may cause old data to have an unreasonable impact on the current situation. For example, if the direct data occurred recently and the macro-data contains older records, the relative impact factor will tend to give the direct data a higher weight. The system then uses this relative impact factor to perform a weighted fusion of the total number of direct interventions and the total number of macro-interventions to obtain the total number of interventions for the intervention scenario combination. This means that the final total number of interventions is no longer a simple accumulation, but rather adjusted based on the timeliness of the data, giving more weight to recently occurring, relevant interventions. Similarly, the system uses the same relative impact factor to weight the direct success count and the macro success count to derive the success count for each intervention scenario. This ensures that the calculation of the success count fully accounts for the timeliness of the data, ensuring that the final calculated success count accurately reflects the actual success rate of the intervention in the current scenario.

[0052] In this way, when data is sparse, this solution not only supplements it with macro data but also intelligently evaluates and adjusts the contributions of direct and macro data during the fusion process, resulting in a representative total number of interventions and successes. This allows the system to provide a reliable intervention success rate even when the sample size is insufficient, thereby improving the accuracy and effectiveness of the intervention strategies recommended by the entire elderly care safety monitoring system.

[0053] Furthermore, step B602 may include: Calculating the time interval between the most recent intervention time and the current time point; Determining, based on the time distribution characteristics, an average time interval between intervention records and the current time point in the macro-intervention scenario category; Based on the time interval and the average time interval, respectively calculating the time weight of the total number of direct interventions and the time weight of the total number of macro interventions; The relative impact factors of the total direct intervention times and the total macro intervention times are determined according to the time weight of the total direct intervention times and the time weight of the total macro intervention times.

[0054] Among them, time weight refers to a value used to measure the importance of specific data (such as the total number of direct interventions or the total number of macro interventions) in the fusion process. This value is related to the recency or timeliness of the data. It can be calculated using a decay function based on time intervals (such as exponential decay, linear decay, or step decay). The smaller the time interval, the higher the weight.

[0055] Among them, the relative impact factor refers to a coefficient used to adjust the contribution ratio of the two when integrating the total number of direct interventions and the total number of macro interventions. It can be determined by the ratio or functional relationship between the time weight of the total number of direct interventions and the time weight of the total number of macro interventions to reflect the relative importance of the two in the time dimension.

[0056] To more accurately integrate direct and macro data, this solution incorporates a time-based weighting mechanism when calculating the total number of interventions and successes for an intervention scenario combination. First, the system obtains the most recent intervention time for a specific intervention scenario combination and calculates the time interval between that time and the current time point. This time interval directly quantifies the timeliness of the direct intervention data. A shorter time interval indicates more recent direct data and a higher reference value. Furthermore, based on the temporal distribution of all intervention records within the macro intervention scenario category, the system calculates the average time interval between these macro intervention records and the current time point. This average time interval provides a quantitative indicator of the overall timeliness of the macro data, reflecting the average recency of historical data within this generalized category. Based on these two time intervals, the system calculates the time weights for the total number of direct interventions and the total number of macro interventions, respectively. This process, the core of this solution, compares the specific recency of the direct data with the average recency of the macro data, assigning appropriate weights to each. For example, if the direct data is very recent, its time weight can be set higher; if the macro data is generally older, its time weight can be set relatively lower. This time interval-based weight distribution mechanism enables the system to dynamically adjust the influence of direct data and macro data during fusion, ensuring that more timely data receives higher consideration. Ultimately, the system determines the relative influence factors of the total number of direct interventions and the total number of macro interventions based on the calculated time weight of the total number of direct interventions and the time weight of the total number of macro interventions. This relative influence factor comprehensively reflects the relative importance of direct data and macro data in the time dimension, and provides an accurate proportional basis for the fusion of the two in subsequent steps. In this way, this solution ensures that when the sample size is insufficient, the reference to macro data not only takes into account its statistical value, but also its timeliness, so that the final calculated intervention success rate is closer to the actual situation and can better reflect the current trend.

[0057] This solution precisely calculates time intervals and average time intervals, and generates time weights accordingly, providing a solid data foundation and logical basis for determining relative impact factors. This allows for dynamic and intelligent adjustment of the contribution ratio of the two when integrating the total number of direct interventions with the total number of macro interventions, avoiding fusion bias caused by differences in data timeliness. Therefore, this solution not only fills the gap in the utilization of time information, but also, through refined weight calculation, makes the final intervention success rate calculation more accurate and real-time, thereby improving the practicality and reliability of the entire elderly care safety monitoring method.

[0058] In one specific embodiment, when determining the relative impact factor between the total number of direct interventions and the total number of macro interventions for a particular intervention scenario combination, the system first obtains the most recent intervention time for that intervention scenario combination. For example, if the most recent intervention occurred three days ago and the current time point is today, the calculated time interval is 3 days. The system then analyzes all historical intervention records for the corresponding macro intervention scenario category. For example, if there are 100 intervention records for that macro category, the system calculates the time interval between each of these 100 records and the current time point, then averages these time intervals. Assume the calculated average time interval is 15 days. The system then calculates a time weight based on these two time intervals. An exponential decay function can be used to calculate the time weight, for example, weight = e^(-k * time interval), where k is a preset decay coefficient. For the total number of direct interventions, the time weight can be calculated as W_d = e^(-k * t_d), where W_d is the time weight for the total number of direct interventions and t_d is the time interval between the most recent intervention time and the current time point (e.g., 3). For the total number of macro-level interventions, its time weight can be calculated as W_m = e^(-k*t_m), where W_m is the time weight of the total number of macro-level interventions, and t_m is the average time interval between the intervention record and the current time point within the macro-level intervention context category (e.g., 15). In this way, the smaller the time interval, the larger the value of the exponential decay function, thus assigning a higher weight to the data. Finally, based on the calculated time weights W_d and W_m for the total number of direct interventions and the total number of macro-level interventions, the relative impact factors of the total number of direct interventions and the total number of macro-level interventions are determined. For example, the relative impact factor can be calculated as F = W_d / (W_d+W_m). If the direct data is very recent (higher W_d) and the macro-level data is relatively older (lower W_m), the value of F will tend toward 1, indicating a greater influence from the direct data. Conversely, if the direct data is older or the macro-level data is very recent, the value of F will tend toward 0, indicating a greater influence from the macro-level data. This calculated relative impact factor F will be used in the subsequent weighted fusion of the total number of direct interventions and the total number of macro-level interventions to obtain the final total number of interventions.

[0059] Furthermore, step B603 may include: The relative impact factor is used as the weight of the total direct intervention times, the difference between 1 and the relative impact factor is used as the weight of the total macro intervention times, and a weighted average operation is performed on the total direct intervention times and the total macro intervention times to obtain the total intervention times of the intervention scenario combination.

[0060] Specifically, weighted averaging is a statistical or mathematical method that assigns different weights to different data points, multiplies these data points by their respective weights, and then divides the sum by the total of all weights to obtain a comprehensive average. This operation can be implemented using a linear combination, for example, using the formula: Total number of interventions = (Total number of direct interventions × Relative impact factor) + (Total number of macro interventions × (1-Relative impact factor)).

[0061] This solution addresses the lack of a clear and reasonable fusion calculation method when integrating the total number of direct interventions and the total number of macro interventions through a clear weighted average calculation method. By using the relative impact factor as the weight of the total number of direct interventions and the difference between 1 and the relative impact factor as the weight of the total number of macro interventions, the system can ensure that the relative impact factor effectively guides the fusion of data, thereby improving the accuracy and reliability of the fusion results. This dynamic balance of the contributions of direct and macro data allows the calculated total number of interventions to more accurately reflect the actual frequency and sample size of specific intervention scenario combinations, providing a solid foundation for the calculation of subsequent intervention success rates, thereby improving the accuracy and effectiveness of recommended interventionists and intervention strategies.

[0062] Furthermore, step B604 may include: The relative impact factor is used as the weight of the direct success number, the difference between 1 and the relative impact factor is used as the weight of the macro success number, and a weighted average operation is performed on the direct success number and the macro success number to obtain the success number of the intervention scenario combination.

[0063] As above, the weighted average operation is performed by multiplying different data items by their corresponding weights, then summing the weighted data items and dividing them by the sum of all weights to obtain a composite result. This operation can be implemented using a linear combination, for example, using the formula: number of successes = (total number of direct interventions × relative impact factor) + (number of macro successes × (1-relative impact factor)).

[0064] This approach effectively integrates direct and macro data by using the relative impact factor as the weight for direct successes and the difference between 1 and the relative impact factor as the weight for macro successes. This approach overcomes the inaccurate calculations caused by the sparsity of direct data and fully leverages the timeliness and relevance of the data. The resulting success counts for each intervention scenario combination accurately reflect the true likelihood of success, avoiding bias caused by improper data fusion. This provides a reliable basis for subsequent calculations of intervention success rates, thereby improving the accuracy and effectiveness of recommending interventionists and intervention strategies.

[0065] In some embodiments, step A4 comprises: A401. According to the current medication resistance behavior pattern, obtain from the association relationship a plurality of interventionist and intervention strategy combinations corresponding to the current medication resistance behavior pattern, and the intervention success rate corresponding to each combination; A402. According to the available status information of each preset intervener, from the plurality of interveners and intervention strategy combinations, remove the intervener unavailable combination; A403. Among the remaining combinations of interveners and intervention strategies, select the combination with the highest intervention success rate and determine it as the recommended intervener and the recommended intervention strategy.

[0066] This application utilizes a multi-stage screening and optimization mechanism to determine recommended interventionists and intervention strategies. First, when the system detects medication resistance in an elderly individual, it identifies the current medication resistance behavior pattern. Based on this pattern, the system initially screens all interventionist and intervention strategy combinations associated with the behavior pattern from pre-established associations and obtains the historical intervention success rate for each combination. This step ensures that subsequent recommended solutions are targeted at the specific issue at hand. Second, considering practical feasibility, the system further obtains real-time availability information for each pre-set interventionist. This availability information enables the system to eliminate currently unavailable interventionist combinations from the initially screened combinations, thereby avoiding prompts to unresponsive interventionists and improving the timely response rate and implementation efficiency of the intervention. Finally, after behavioral pattern screening and availability filtering, the system ranks the remaining relevant and available interventionist and intervention strategy combinations based on each combination's historical intervention success rate and selects the combination with the highest success rate as the final recommended interventionist and intervention strategy. It is precisely because of this progressive screening and optimization process that the system is able to provide specific, actionable intervention recommendations with the highest probability of success, effectively addressing the inaccurate and suboptimal recommendations that arise from simply determining recommendations based solely on input information. This refined determination process enables the system to fully leverage previously analyzed intervention success rate information and, combined with the real-time availability of the interventionist, ensures the practicality and effectiveness of the recommendations, thereby improving the success rate of medication adherence among the elderly.

[0067] refer to Figure 2 The present application provides an Internet of Things-based elderly care safety monitoring system for monitoring elderly people's medication to improve elderly people's compliance with safe medication. The system includes an Internet of Things-based medication monitoring system 1, a server 2, and multiple interventionist terminals 3. The medication monitoring system 1 and the multiple interventionist terminals 3 are both communicatively connected to the server 2. The medication monitoring system 1 is used to monitor the medication status of the elderly, identify the elderly's medication resistance behavior and the corresponding medication resistance behavior pattern, and upload the identification results to the server 2; The server 2 is used to store intervention records and to perform: Extracting intervention records stored locally; the intervention records include the identity of the intervener of the historical medication intervention event, the intervention strategy, the elderly person's medication resistance behavior pattern, and the intervention results; Based on the intervention records, the correlation between the interventionist, intervention strategy, medication resistance behavior pattern, and intervention success rate is analyzed (the specific process can be referred to step A2 above); When the medication monitoring system 1 detects the elderly person's medication resistance behavior, it obtains the available status information and current medication resistance behavior pattern of each preset intervener (the specific process can be referred to step A3 above); Determine the recommended intervener and recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship (the specific process can be referred to step A4 above); The intervention prompt information including the recommended intervention strategy is sent to the intervener terminal 3 of the recommending intervener (the specific process can be referred to the above step A5).

[0068] In a specific embodiment, the system can operate as follows: The IoT-based medication monitoring system 1 can be composed of a smart pill box, a wearable device, and multiple environmental sensors. The smart pill box has built-in medication access sensors and lid-opening sensors to monitor whether medication has been removed and whether the pill box has been opened. The wearable device can be a smart bracelet that monitors the elderly person's physiological indicators and activity status. The environmental sensors can include a millimeter-wave radar for contactless monitoring of the elderly person's posture and activity area, and a smart speaker for analyzing voice emotions. When the smart pill box detects that medication has not been taken on time, or the millimeter-wave radar identifies that the elderly person has not moved for an extended period of time with an abnormal posture, or the smart speaker identifies that the elderly person is depressed, the medication monitoring system 1 integrates this information to identify possible medication resistance behaviors, such as "refusal to swallow" or "emotional resistance," and determines the corresponding medication resistance behavior pattern. These identification results are then uploaded to the server 2 via the wireless network. Server 2 can be a computing instance deployed in the cloud, such as a Linux-based virtual machine. It runs a database management system, such as PostgreSQL, to store intervention records and an application service program to handle business logic. Intervention records can be stored in a database table containing fields such as "interventionist ID," "intervention strategy description," "resistance behavior pattern code," "intervention result (success / failure)," and "timestamp." Server 2 retrieves these intervention records periodically or as needed. To analyze the correlation between interventionists, intervention strategies, medication resistance behavior patterns, and intervention success rates, Server 2's application service program can run a data analysis module. This module can employ statistical analysis methods, such as calculating the historical success rate for each "interventionist-intervention strategy-resistance behavior pattern" combination and storing these success rates in a pre-calculated lookup table. For example, if Child A's success rate for the "mixed yogurt" strategy against the "refusal to swallow" pattern is 80%, this information will be recorded. When the medication monitoring system 1 detects new medication resistance behavior, Server 2 communicates with the interventionist terminal 3 to obtain the available status information for each pre-set interventionist. For example, the intervener terminal 3 (which can be a family member's smartphone application) can report its online status, geographic location (to determine whether it is at home) or manually set "busy" status in real time. At the same time, server 2 will receive the current medication resistance behavior pattern, such as "emotional resistance". The decision module of server 2 will determine the recommended intervener and recommended intervention strategy based on this real-time information and pre-established associations. For example, if the current resistance pattern is "emotional resistance", and "child B" is currently available and its historical records show that the "play favorite music" strategy has the highest success rate when dealing with "emotional resistance", then the server will determine "child B" as the recommended intervener and "play favorite music" as the recommended intervention strategy.Finally, the server sends a recommended intervention strategy to Child B's smartphone app. For example, a push notification might read, "The elderly person is experiencing emotional resistance. We recommend playing their favorite music to assist with medication." This allows the interventionist to act according to the system's specific recommendations, improving the effectiveness of the intervention.

[0069] This system intelligently recommends interventionists and strategies, effectively addressing the lack of specific, actionable guidance for caregivers facing elderly individuals' medication resistance. Instead of simply informing caregivers about the cause of the problem, the system provides clear recommendations on "what to do" based on historical experience and real-time context. This significantly reduces the complexity and uncertainty of caregivers' independent search for solutions, making interventions more targeted and effective, thereby directly improving elderly individuals' actual medication adherence and reducing the burden of care.

[0070] The above description is merely an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, various modifications and variations of the present application are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for monitoring elderly people's safety based on the Internet of Things, which is used to monitor the elderly's medication to improve their compliance with medication safety. The steps of the method include: A1. Obtain intervention records; these records include the identity of the intervener, intervention strategy, the elderly individual's medication resistance behavior patterns, and intervention outcomes. A2. Based on the intervention records, analyze the correlation between the interventionist, intervention strategy, medication resistance behavior pattern, and intervention success rate; A3. When the elderly person's medication resistance behavior is monitored, the available status information of each preset interventionist and the current medication resistance behavior pattern are obtained; A4. Determine a recommended intervener and a recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship; A5. Send intervention prompt information including the recommended intervention strategy to the recommended interventionist.

2. The method for monitoring elderly care safety based on the Internet of Things according to claim 1, characterized in that: Step A2 includes: A201. Combining the intervener identity, intervention strategy, and medication resistance behavior pattern in the intervention record to form multiple intervention scenario combinations; A202. For each combination of the intervention scenarios, count the total number of interventions and the number of successful interventions; A203. Calculate the intervention success rate for each intervention scenario combination based on the number of successes and the total number of interventions; A204. Associate the intervention success rate with the corresponding intervener identity, intervention strategy, and medication resistance behavior pattern to obtain the association relationship.

3. The method for monitoring elderly care safety based on the Internet of Things according to claim 2, characterized in that: Step A202 includes: B1. For each intervention scenario combination, count the total number of direct interventions and the number of direct successes; B2. Determine whether the total number of direct interventions is lower than the preset sample threshold; B3. When the total number of direct interventions is not less than the preset sample threshold, the total number of direct interventions and the number of direct successes are used as the total number of interventions and the number of successes for the intervention scenario combination; B4. When the total number of direct interventions is lower than the preset sample threshold, determine the corresponding macro-intervention scenario category based on the interventionist identity and medication resistance behavior pattern of the intervention scenario combination; B5. Count the total number of macro interventions and macro successes under the macro intervention scenario category; B6. Calculate the total number of interventions and the number of successes for the intervention scenario combination based on the total number of macro interventions, the number of macro successes, the total number of direct interventions, and the number of direct successes.

4. The method for monitoring elderly care safety based on the Internet of Things according to claim 3, characterized in that: Step B4 includes: B401. Obtain a preset generalization rule set based on the intervention scenario combination of the interventionist identity and the medication resistance behavior pattern; the generalization rule set defines a hierarchical relationship generalized from the interventionist identity and the medication resistance behavior pattern to macro categories of different granularities; B402. Based on the generalization rule set, starting from the finest-grained category corresponding to the intervention scenario combination, generalize upward step by step to generate a series of candidate macro-intervention scenario categories, and record the generalization level of each candidate macro-intervention scenario category; B403. For each candidate macro-intervention scenario category, calculate the total number of historical interventions; B404. Traverse the candidate macro-intervention scenario categories, identify the candidate macro-intervention scenario categories in which the total historical number of interventions is not less than the preset macro-sample threshold, and from these identified categories, select the candidate macro-intervention scenario category with the lowest generalization level and determine it as the macro-intervention scenario category.

5. The method for monitoring elderly care safety based on the Internet of Things according to claim 3, characterized in that: The intervention record also includes the intervention time; Step B6 includes: B601. Obtain the most recent intervention time of the intervention scenario combination, and the time distribution characteristics of the intervention records in the macro-intervention scenario category; B602. Determine the relative impact factors of the total number of direct interventions and the total number of macro interventions based on the most recent intervention time and the time distribution characteristics; B603. Using the relative impact factor, combine the total number of direct interventions with the total number of macro interventions to obtain the total number of interventions for the intervention scenario combination; B604. Use the relative impact factor to merge the direct success number and the macro success number to obtain the success number of the intervention scenario combination.

6. The method for monitoring elderly care safety based on the Internet of Things according to claim 5, characterized in that: Step B602 includes: Calculating the time interval between the most recent intervention time and the current time point; Determining, based on the time distribution characteristics, an average time interval between intervention records and the current time point in the macro-intervention scenario category; Based on the time interval and the average time interval, respectively calculating the time weight of the total number of direct interventions and the time weight of the total number of macro interventions; The relative impact factors of the total direct intervention times and the total macro intervention times are determined according to the time weight of the total direct intervention times and the time weight of the total macro intervention times.

7. The method for monitoring elderly care safety based on the Internet of Things according to claim 6, characterized in that: Step B603 includes: The relative impact factor is used as the weight of the total direct intervention times, the difference between 1 and the relative impact factor is used as the weight of the total macro intervention times, and a weighted average operation is performed on the total direct intervention times and the total macro intervention times to obtain the total intervention times of the intervention scenario combination.

8. The method for monitoring elderly care safety based on the Internet of Things according to claim 6, characterized in that: Step B604 includes: The relative impact factor is used as the weight of the direct success number, the difference between 1 and the relative impact factor is used as the weight of the macro success number, and a weighted average operation is performed on the direct success number and the macro success number to obtain the success number of the intervention scenario combination.

9. The method for monitoring elderly care safety based on the Internet of Things according to claim 1, characterized in that: Step A4 includes: A401. According to the current medication resistance behavior pattern, obtain from the association relationship a plurality of interventionist and intervention strategy combinations corresponding to the current medication resistance behavior pattern, and the intervention success rate corresponding to each combination; A402. According to the available status information of each preset intervener, from the plurality of interveners and intervention strategy combinations, remove the intervener unavailable combination; A403. Among the remaining combinations of interveners and intervention strategies, select the combination with the highest intervention success rate and determine it as the recommended intervener and the recommended intervention strategy.

10. An Internet of Things-based elderly care safety monitoring system for monitoring elderly people's medication to improve their compliance with medication safety, characterized by: The system includes an Internet of Things-based medication monitoring system, a server, and multiple interventionist terminals, wherein the medication monitoring system and the multiple interventionist terminals are both communicatively connected to the server; The medication monitoring system is used to monitor the medication status of the elderly, identify the elderly's medication resistance behavior and the corresponding medication resistance behavior pattern, and upload the identification results to the server; The server is used to store intervention records and to perform: Extracting intervention records stored locally; the intervention records include the identity of the intervener of the historical medication intervention event, the intervention strategy, the elderly person's medication resistance behavior pattern, and the intervention results; Based on the intervention records, the correlation between the interventionist, intervention strategy, medication resistance behavior pattern and intervention success rate was analyzed; When the medication monitoring system detects medication resistance behavior of the elderly, obtaining available status information and current medication resistance behavior pattern of each preset intervener; Determining a recommended intervener and a recommended intervention strategy based on the available status information of each preset intervener, the current medication resistance behavior pattern, and the association relationship; Sending intervention prompt information including the recommended intervention strategy to the intervener terminal of the recommended intervener.

Images