Computer performance dynamic adjusting system and method based on user operation behaviors

Abstract

The invention provides a computer performance dynamic adjusting system and method based on user operation behaviors, and relates to the technical field of computers. Comprising an event capture module, an operation semantic association module, a system resource optimization module and an adaptive correction module. The event capturing module obtains user operation data and generates an operation event sequence with a timestamp. The operation semantic association module is used for matching an operation event and an operation mode through a nonlinear time axis, and generating a resource demand vector in combination with a window focus and a process stack calling relationship. And the system resource optimization module selects a resource allocation scheme with the lowest power consumption on the basis of an ant colony algorithm under the condition that response time constraints are met, and reallocates resources. And the adaptive correction module adjusts the matching weight of the operation mode by comparing the deviation between the actual response time and the expected response time, and feeds back a corrected matching threshold. The system can dynamically adjust computer performance, improve response speed and reduce power consumption.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a computer performance dynamic adjustment system and method based on user operation behavior. Background Art

[0002] In the field of computers, traditional performance adjustment methods mainly rely on static resource allocation strategies and fixed performance parameter settings. However, with the diversification of computer application scenarios and the complexity of user operation behaviors, the existing performance adjustment methods gradually expose their limitations. For example, when users quickly switch applications or perform high-frequency input operations, the system often fails to adjust resource allocation in a timely manner, resulting in response delays or excessive power consumption. Summary of the Invention

[0003] To achieve the above objectives, the present invention is realized through the following technical solutions: A computer performance dynamic adjustment system based on user operation behavior, including:

[0004] An event capture module, which obtains the instantaneous acceleration value of the cursor displacement trajectory, the pressure gradient distribution data of the screen contact surface, the trigger frequency and duration of the input device through an interrupt trigger method, and generates an operation event sequence including timestamps;

[0005] An operation semantic association module, which generates a resource demand vector including an operation urgency coefficient and an expected response time by performing a non-linear time axis matching of the operation event sequence with an operation mode, and detecting the current window focus position and the process stack call relationship;

[0006] A system resource optimization module, which receives the operation event sequence, the operation urgency coefficient and the resource demand vector and establishes constraint conditions for computer performance, traverses the feasible solution space based on the ant colony algorithm, selects the resource configuration plan with the lowest power consumption under the constraint of meeting the expected response time, and performs real-time resource reallocation;

[0007] An adaptive correction module, which compares the deviation between the actual operation response time and the expected response time, adjusts the matching weight coefficient of the operation mode according to historical deviation data, and feeds back the corrected operation mode matching threshold to the operation semantic association module.

[0008] The event capture module triggers the real-time capture of user operation behaviors through hardware interrupts and converts them into structured event sequences. When the user performs an operation, hardware devices such as mice, keyboards, or touchscreens trigger interrupt signals, and data is obtained by listening to these interrupt signals. For the cursor displacement trajectory, the instantaneous acceleration value is obtained by calculating the displacement difference between adjacent time points and combining the time interval, thereby capturing the changes in the intensity and speed of the user's operation. For example, when the user quickly moves the mouse, the event capture module determines the urgency of the operation through the acceleration value. The larger the value, the faster the user's operation speed and the higher the operation urgency. For the pressure gradient distribution of the screen contact surface, the data of the pressure sensor is read from the touchscreen driver and converted into a standardized pressure value through normalization processing. For the trigger frequency and duration of the input device, the time stamps of the interrupt triggers are recorded, and the trigger interval and duration are calculated to analyze the user's operation rhythm. All captured data is timestamped and an operation event sequence is generated in chronological order.

[0009] The operation semantic association module receives the operation event sequence, synchronizes all operation events on the same time axis by interpolating and aligning the time axis, and normalizes the feature values of each operation event, mapping the data to the range [0,1]. For example, when the user performs mouse movement and keyboard input simultaneously, the two types of operation events are synchronized through time axis alignment to keep the operation complete and consistent. The operation semantic association module performs a non-linear time axis matching between the current operation event sequence and the operation patterns of user behaviors, calculates the similarity between the operation event sequence and the operation patterns, and determines the most matching operation pattern. First, the key features in the operation event sequence and the operation patterns are extracted, such as the instantaneous acceleration of cursor displacement, the pressure gradient value of the screen contact surface, the trigger frequency of the input device, etc., and these feature values are normalized and mapped to the range [0,1]. The time axes of the operation event sequence and the operation patterns are interpolated and aligned to make them synchronous in the time dimension. The Euclidean distance between the operation event sequence and the operation patterns is calculated. The feature values at each time point are regarded as points in a multi-dimensional space, and by calculating the Euclidean distance between the two, where and Denote the normalized value of the operation event sequence and the operation mode on the i-th feature; n represents the number of features; D(P, Q) represents the Euclidean distance between the operation event sequence P and the operation mode Q, and the smaller the distance, the higher the similarity; during the process, in combination with the current window focus position and the process stack call relationship, context weighting is performed on the similarity to improve the accuracy of matching; finally, according to the preset similarity threshold, the most matching operation mode with the smallest Euclidean distance and exceeding the threshold is selected; assume that the Euclidean distances between the operation event sequence and three operation modes, mode A, mode B, and mode C, are calculated, and the results are mode A: DA = 0.15, mode B: DB = 0.25, mode C: DC = 0.18 respectively. According to the preset threshold T = 0.2, it is filtered out that DA and DC satisfy D ≤ 0.2, while DB does not, and mode A is the smallest. Therefore, it is determined that the operation event sequence most matches mode A.

[0010] The system resource optimization module parses the expected response time in the resource demand vector as a hard time threshold to ensure that the system completes the response to the current operation within this time limit; for example, when the user performs a video editing operation, the system sets the system to complete resource allocation within 100 milliseconds according to the expected response time; generates resource priority weights according to the dynamic value of the operation urgency coefficient, and the higher the urgency coefficient, the greater the resource allocation weight and the higher the priority; at the same time, real-time collects the resource occupancy status of the current system, such as CPU utilization rate and memory occupancy rate, and in combination with the process stack call relationship, determines the feasible range of resource allocation, and constructs a set of constraint conditions including time constraint, resource weight constraint, and dynamic resource capacity constraint.

[0011] The generation process of the urgency coefficient dynamically determines the urgency of user operations by combining the characteristics of user operation behaviors and predefined rules; extracts key features from the operation event sequence, including the instantaneous acceleration value of cursor displacement, the pressure gradient value of the screen contact surface, and the trigger frequency of the input device; normalizes the feature values and maps them to the range of [0, 1], sets fixed thresholds for each feature to judge the urgency of user operations. For example, the instantaneous acceleration threshold of cursor displacement is set to 0.7, the pressure gradient threshold of the screen contact surface is set to 0.6, and the trigger frequency threshold of the input device is set to 0.5; based on feature extraction and threshold judgment, combines the current window focus position and the process stack call relationship to judge the context environment of user operations. For example, if the user is using video editing software, the context environment is determined to be of high urgency; if the user is using text editing software, the context environment is determined to be of low urgency; if any one of the feature values of the instantaneous acceleration of cursor displacement, the pressure gradient of the screen contact surface, or the trigger frequency of the input device exceeds its corresponding threshold and the context environment is of high urgency, the urgency coefficient is set to 1.0, indicating the highest urgency; if the feature value exceeds the threshold but the context environment is of low urgency, the urgency coefficient is set to 0.7, indicating medium urgency; if all feature values do not exceed the threshold but the context environment is of high urgency, the urgency coefficient is set to 0.5, indicating relatively low urgency; if all feature values do not exceed the threshold and the context environment is of low urgency, the urgency coefficient is set to 0.3, indicating the lowest urgency.

[0012] The system uses the ant colony algorithm to model the resource allocation problem as a constrained multi-objective optimization problem; maps the path of each ant to a resource allocation combination and introduces constraint conditions as mandatory rules for path selection; during the search process, ants select the initial path according to the pheromone concentration and heuristic function, but only retain the feasible paths that meet all constraint conditions, including that the response time of the path does not exceed the expected threshold and the resource allocation does not exceed the system's real-time capacity; in the pheromone update mechanism, incorporates the operation urgency coefficient into the calculation of pheromone increment, and the specific formula is: , where represents the pheromone increment of path ij; E is a constant representing the pheromone intensity, for example, set to 0.5 for low urgency and 0.1 for medium urgency; F is the operation urgency coefficient; represents the response time of path ij; for operations with high urgency, the weight of the pheromone increment of its corresponding path is increased to guide subsequent ants to preferentially explore the resource allocation scheme of high-priority resources; at the same time, dynamically adjusts the heuristic function according to the system's real-time load, and through multiple rounds of iteration, converges to a resource allocation scheme that meets all constraint conditions and has the lowest power consumption, and triggers real-time reallocation.

[0013] The adaptive correction module assigns a higher weight to recent deviations and a lower weight to historical deviations, enabling the system to quickly adapt to changes in user operation behavior. For example, assuming the sliding window size is 5, the response time deviations of the last 5 operations are D1, D2, D3, D4, D5, and the corresponding weights are W1, W2, W3, W4, W5, where W5 > W4 > W3 > W2 > W1. The calculation formula for the weighted average deviation is , when the deviation exceeds the preset threshold, for example, the threshold is set to 0.2, analyze the matching success rate of the corresponding operation mode in the historical matching records. If the matching results of a certain operation mode frequently lead to large deviations, reduce the weight coefficient of this operation mode proportionally. If the matching deviation of a certain mode is small in a similar context environment, increase its weight coefficient.

[0014] The correction of the weight coefficient combines the current window focus position and the process stack call relationship to ensure that the correction process has context relevance. The corrected weight coefficient is converted into an operation mode matching threshold through a dynamic threshold mapping mechanism and fed back to the operation semantics association module. During the mapping process, the upper and lower limits of the matching threshold are adjusted non-linearly according to the change amplitude of the weight coefficient. For example, when the weight coefficient decreases from 0.8 to 0.6, the matching threshold of the corresponding operation mode increases from 0.7 to 0.8, reducing the probability of its being selected. When the weight coefficient increases from 0.5 to 0.7, the matching threshold decreases from 0.6 to 0.5, increasing its matching priority. When updating the threshold, retain some weight parameters of the previous cycle to prevent the matching strategy from oscillating due to instantaneous deviation fluctuations. For example, assuming the weight coefficient of the previous cycle is 0.7 and the weight coefficient calculated in the current cycle is 0.6, the module updates the threshold to 0.75 instead of directly adjusting it to 0.8 for a smooth transition.

[0015] The present invention provides a computer performance dynamic adjustment system based on user operation behavior, having the following beneficial effects:

[0016] 1. By capturing user operation behavior in real time and dynamically adjusting system resource allocation, the present invention solves the problems of untimely resource allocation and response delay in a computer system when dealing with complex user operations. The system intelligently allocates resources according to the urgency and context environment of user operations, ensuring smooth response even during high-frequency operations or rapid application switching, and significantly improving the user experience.

[0017] 2. By introducing non-linear time-axis matching and adaptive correction, and optimizing resource allocation in combination with the ant colony algorithm, the present invention enables the system to select the solution with the lowest power consumption on the premise of meeting the response time constraint, achieving a balance between performance and energy efficiency and effectively reducing system energy consumption.

[0018] 3. By dynamically adjusting the matching weights and thresholds, the present invention solves the problems of insufficient system adaptability and easy policy oscillation caused by instantaneous deviation. The system can adaptively optimize the matching strategy according to historical operation data, improving the stability and intelligence level of the system and reducing misjudgment of user operations. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic flow diagram of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Apparently, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0021] The event capture module obtains the operation behavior of the user in real time through the hardware interruption mechanism. When the user operates the computer, such as moving the mouse, clicking the keyboard or using the touch screen, the corresponding hardware device will trigger an interruption signal. By listening to these interruption signals, the original operation data is obtained. Taking the capture of the cursor displacement trajectory as an example, the displacement difference between adjacent time points is calculated, and the instantaneous acceleration value is obtained by combining the time interval. The specific calculation formula is as follows: , where a(t) represents the instantaneous acceleration at time t, is the displacement difference between adjacent time points, is the time interval; through this acceleration value, the changes in the intensity and speed of the user's operation are captured. For example, when the user quickly moves the mouse to browse the web page, a larger acceleration value indicates a high degree of operation urgency. For the pressure gradient distribution of the screen contact surface, the original data of the pressure sensor is read from the touch screen driver and normalized to convert it into a standardized pressure value. The normalization formula is: , where, is the normalized pressure value, P is the original pressure value, and are the maximum and minimum values of the pressure respectively; for the trigger frequency and duration of the input device, the system records the time stamps of the interruption triggers, calculates the trigger interval and the duration, so as to analyze the operation rhythm of the user. All the captured data is timestamped and an operation event sequence is generated in chronological order.

[0022] After receiving the operation event sequence generated by the event capture module, the operation semantic association module first aligns them through timeline interpolation to synchronize all operation events on the same timeline; then normalizes the eigenvalue of each operation event to map the data to the range of [0, 1]; matches the operation event sequence with the predefined operation patterns through non-linear timeline matching; during the matching process, combines the current window focus position and the process stack call relationship to calculate the similarity; the similarity calculation uses the Euclidean distance formula: , where D(P, Q) represents the Euclidean distance between the operation event sequence P and the operation pattern Q, and are the normalized values of the operation event sequence and the operation pattern on the i-th feature respectively, and n represents the number of features; the smaller the distance, the higher the similarity; for example, when the user types quickly and switches windows frequently in a text editing software, the system determines that the current operation pattern is closer to urgent document editing based on the window focus position and the process stack call relationship, and assigns a high resource priority to it; finally, according to the preset similarity threshold, selects the most matching operation pattern; assume that the Euclidean distances between the operation event sequence and three operation patterns, pattern A, pattern B, and pattern C, are calculated, and the preset threshold is 0.2, and the results are pattern A = 0.15, pattern B = 0.25, pattern C = 0.18 respectively. According to the preset threshold T = 0.2, it is filtered out that A and C satisfy less than or equal to 0.2, while B does not, and pattern A is the smallest, so it is determined that the operation event sequence most matches pattern A.

[0023] The system resource optimization module establishes constraint conditions based on the operation event sequence, the operation urgency coefficient, and the resource demand vector; parses the expected response time in the resource demand vector as a hard time threshold to ensure that the system completes the response to the current operation within this time limit; generates resource priority weights according to the operation urgency coefficient, the higher the urgency, the greater the weight; real-time collects the resource occupancy status of the current system, such as CPU utilization rate, memory occupancy rate, and combines the process stack call relationship to determine the feasible range of resource allocation; the constructed set of constraint conditions includes time constraints, resource weight constraints, and dynamic resource capacity constraints.

[0024] Based on the constraint conditions, the system uses the ant colony algorithm to model the resource allocation problem as a constrained multi-objective optimization problem; by mapping the path of each ant to a resource allocation combination and introducing the constraint conditions as the mandatory rules for path selection; during the search process, the ant selects the initial path according to the pheromone concentration and the heuristic function, but only retains the feasible paths that satisfy all the constraint conditions, including that the response time of the path does not exceed the expected threshold and the resource allocation does not exceed the real-time capacity of the system; where the heuristic function η ijHeuristic information reflecting the path, which is related to the cost or quality of the path. For example, in the resource allocation problem, the heuristic function is expressed as the reciprocal of the power consumption or response time of the path, that is , where represents the cost of path ij; when ants choose a path, they calculate the selection probability according to the product of the pheromone concentration and the heuristic function; the specific formula is: , where P ij represents the probability that an ant chooses a path from node i to node j; T ij represents the pheromone concentration of path ij; η ij represents the heuristic function value of path ij; α and β are the weight coefficients of the pheromone concentration and the heuristic function respectively, which are used to adjust the relative importance of the two in path selection; the denominator part represents the sum of the products of the pheromone concentration and the heuristic function of all feasible paths, which is used to normalize the probability.

[0025] Incorporate the operation urgency coefficient into the calculation of pheromone increment, and the specific formula is: , where represents the pheromone increment of path ij; E is a constant representing the pheromone intensity; F is the operation urgency coefficient; represents the response time of path ij; for operations with high urgency, the weight of the pheromone increment of the corresponding path is increased, guiding subsequent ants to preferentially explore the allocation scheme of high-priority resources; at the same time, the heuristic function is dynamically adjusted according to the real-time load of the system, and through multiple rounds of iteration, it converges to a resource allocation scheme that satisfies all constraint conditions and has the lowest power consumption, and triggers real-time reallocation.

[0026] The adaptive correction module obtains the response time series of the last N operations through a sliding window, and performs weighted average processing on the deviation amount of each operation, where the weight of the recent deviation is higher than that of the historical deviation; the calculation formula of the weighted average deviation amount is: , where, represents the weighted average deviation amount, is the deviation amount of the i-th operation, is the corresponding weight; when the deviation exceeds the preset threshold, analyze the matching success rate of the corresponding operation mode in the historical matching records; if a certain operation mode frequently causes a large deviation, reduce the weight coefficient of this operation mode proportionally; if the matching deviation is small in a similar context environment, increase its weight coefficient; the correction of the weight coefficient combines the current window focus position and the process stack call relationship to ensure that the correction process has context relevance; the corrected weight coefficient is converted into an operation mode matching threshold through a dynamic threshold mapping mechanism and fed back to the operation semantics association module; during the mapping process, the upper and lower limits of the matching threshold are non-linearly adjusted according to the change range of the weight coefficient; for example, when the weight coefficient decreases from 0.8 to 0.6, the matching threshold of the corresponding operation mode increases from 0.7 to 0.8, reducing the probability of its being selected; when the weight coefficient increases from 0.5 to 0.7, the matching threshold decreases from 0.6 to 0.5, increasing its matching priority; when updating the threshold, retain some weight parameters of the previous cycle to prevent the matching strategy from oscillating due to instantaneous deviation fluctuations.

[0027] Through the collaborative work of modules such as event capture, operation semantics association, system resource optimization, and adaptive correction, the present invention realizes the dynamic adjustment of computer performance based on user operation behavior; in different application scenarios, the system can capture user operation behavior in real time, dynamically adjust resource allocation, and effectively improve the response speed and energy efficiency ratio of the computer.

[0028] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A computer performance dynamic adjustment system based on user operation behavior, characterized in that, Including: An event capture module, which obtains the instantaneous acceleration value of the cursor displacement trajectory, the pressure gradient distribution data of the screen contact surface, the trigger frequency and duration of the input device through an interrupt trigger method, and generates an operation event sequence including timestamps; An operation semantics association module, which generates a resource requirement vector including an operation urgency coefficient and an expected response time by performing a non-linear time axis matching of the operation event sequence with the operation mode, and detecting the current window focus position and the process stack call relationship; A system resource optimization module, which receives the operation event sequence, the operation urgency coefficient and the resource requirement vector, establishes the constraint conditions of the computer performance, traverses the feasible solution space based on the ant colony algorithm, selects the resource configuration scheme with the lowest power consumption under the constraint of meeting the expected response time, and performs real-time resource reallocation; An adaptive correction module, which compares the deviation between the actual operation response time and the expected response time, adjusts the matching weight coefficient of the operation mode according to the historical deviation data, and feeds back the corrected operation mode matching threshold to the operation semantics association module.

2. The computer performance dynamic adjustment system based on user operation behavior according to claim 1, characterized in that: The event capture module captures the user's operation behavior in real time and converts it into a structured event sequence; when the user operates, the hardware device triggers an interrupt signal, and the original data is obtained by listening to the interrupt signal; for the cursor displacement trajectory, the module obtains the instantaneous acceleration value by calculating the displacement difference between adjacent time points; for the pressure gradient distribution of the screen contact surface, the event capture module reads the original data of the pressure sensor from the touch screen driver and normalizes it; for the trigger frequency and duration of the input device, the trigger interval and duration are calculated by recording the timestamps of the interrupt triggers; all data are timestamped and an operation event sequence is generated in chronological order.

3. A computer performance dynamic adjustment system based on user operation behavior according to claim 1, characterized in that: The operation semantics association module receives the operation event sequence, aligns the time axis of the operation event sequence through interpolation to synchronize all operation events on the same time axis, normalizes the eigenvalue of each operation event, and maps the data to the range of [0, 1]; performs a non-linear time axis matching of the current operation event sequence with the operation mode of the user behavior, and determines the most matching operation mode by calculating the similarity between the operation event sequence and the operation mode; during the process, the operation semantics is further refined by combining the current window focus position and the process stack call relationship.

4. A computer performance dynamic adjustment system based on user operation behavior according to claim 3, characterized in that: The operation semantics association module judges the application program or interface element that the user is currently operating by detecting the current window focus position, and analyzes the context environment of the user's operation in combination with the process stack call relationship; Based on the analysis, a resource requirement vector is generated, which includes two key parameters: an operation urgency coefficient and an expected response time; among them, the operation urgency coefficient reflects the urgency of the user's operation, and this coefficient is dynamically calculated according to the characteristics of the operation mode and the context environment; the expected response time represents the expected response time required for the system to complete the current operation, and this parameter is dynamically adjusted according to the historical response time of the operation mode and the current system load situation.

5. A computer performance dynamic adjustment system based on user operation behavior according to claim 1, characterized in that: The system resource optimization module establishes constraint conditions for computer performance based on the operation event sequence, operation urgency coefficient, and resource demand vector; parses the expected response time in the resource demand vector as a hard time threshold, enabling the system to complete the response to the current operation within this time limit; generates resource priority weights according to the dynamic value of the operation urgency coefficient, where a higher urgency coefficient results in a greater resource allocation weight. Realtime collects the resource occupancy status of the current system, and determines the feasible range of resource allocation in combination with the process stack call relationship; constructs a set of constraint conditions, including time constraints, resource weight constraints, and dynamic resource capacity constraints.

6. The computer performance dynamic adjustment system based on user operation behavior according to claim 5, characterized in that: Based on the constraint conditions, the system resource optimization module uses the ant colony algorithm to model the resource allocation problem as a constrained multi-objective optimization problem; the algorithm maps the path of each ant to a resource allocation combination and introduces constraint conditions as mandatory rules for path selection; during the search process, ants select the initial path according to the pheromone concentration and heuristic function, but only retain the feasible paths that meet all constraint conditions, including that the response time of the path does not exceed the expected threshold and the resource allocation does not exceed the system's realtime capacity; in the pheromone update mechanism, the operation urgency coefficient is incorporated into the calculation of the pheromone increment: for operations with high urgency, the pheromone increment weight corresponding to their paths is increased, guiding subsequent ants to preferentially explore resource allocation schemes for high-priority resources; the heuristic function is dynamically adjusted according to the system's realtime load, and through multiple iterations, converges to a resource allocation scheme that meets all constraint conditions and has the lowest power consumption, and triggers realtime reallocation.

7. A computer performance dynamic adjustment system based on user operation behavior according to claim 1, characterized in that: The adaptive correction module obtains the response time sequence of the most recent N operations through a sliding window, and performs weighted average processing on the deviation amount of each operation, where the weight of recent deviations is higher than that of historical deviations, reflecting the dynamic change trend of user operation behavior; based on the calculated deviation amount, when the deviation amount exceeds the preset threshold, analyzes the matching success rate of the corresponding operation mode in the historical matching records. If the matching result of a certain operation mode frequently leads to large deviations, then proportionally reduces the weight coefficient of this operation mode; if a certain mode has a small matching deviation in a similar context environment, then increases its weight coefficient.

8. The computer performance dynamic adjustment system based on user operation behavior according to claim 7, characterized in that: The weight coefficient combines the current window focus position and the process stack call relationship to perform context-related constraints on the correction of the weight coefficient; the corrected weight coefficient is converted into an operation mode matching threshold through a dynamic threshold mapping mechanism and fed back to the operation semantic association module; during the mapping process, according to the change amplitude of the weight coefficient, non-linearly adjusts the upper and lower limits of the matching threshold. When the weight coefficient decreases, the matching threshold of the corresponding operation mode increases, reducing the probability of its being selected; when the weight coefficient increases, the matching threshold decreases, increasing its matching priority; retain part of the weight parameters of the previous cycle during threshold update to prevent oscillation of the matching strategy caused by instantaneous deviation fluctuations.

9. A computer performance dynamic adjustment method based on user operation behavior, based on the system described in claim 1, characterized in that, Including: Obtains the instantaneous acceleration value of the cursor displacement trajectory, the pressure gradient distribution data of the screen contact surface, the trigger frequency and duration of the input device through an interrupt trigger method, and generates an operation event sequence including timestamps. By performing non-linear time-axis matching between the operation event sequence and the operation mode, and detecting the current window focus position and the process stack call relationship, a resource demand vector containing the operation urgency coefficient and the expected response time is generated; Receive the operation event sequence, the operation urgency coefficient, and the resource demand vector, and establish the constraint conditions of computer performance. Based on the ant colony algorithm, traverse the feasible solution space, select the resource allocation plan with the lowest power consumption under the constraint of meeting the expected response time, and perform real-time resource reallocation; Compare the deviation between the actual operation response time and the expected response time, adjust the matching weight coefficient of the operation mode according to the historical deviation data, and feedback the corrected operation mode matching threshold to the operation semantics association module.

Images