Media data fluctuation analysis method and system based on artificial intelligence

Abstract

The invention belongs to the technical field of data analysis, and particularly discloses a media data fluctuation analysis method and system based on artificial intelligence. The method comprises the following steps: acquiring target media data, identifying abnormal change points in the target media data by applying a trend model, and judging whether abnormal fluctuation exists or not through a fluctuation identification model based on the calculated number of continuous abnormal change points; by determining an abnormal fluctuation interval, executing optimization processing on data in the interval by adopting a self-adaptive correction strategy; and respectively calculating a forward correction parameter and a backward correction parameter based on the change amplitude of the data point set before the central point and the change amplitude of the data point set after the central point. According to the method, the abnormal fluctuation interval is defined by positioning the abnormal fluctuation, non-key noise interference is filtered, smooth transition between the optimized data and the data on the two sides of the abnormal fluctuation interval is achieved through correction verification, and the accuracy and credibility of abnormal fluctuation recognition in the media data are improved.

Description

Technical Field

[0001] This invention belongs to the field of data analysis technology, specifically relating to a method and system for media data fluctuation analysis based on artificial intelligence. Background Technology

[0002] With the development of digital information, the public's channels for obtaining information and communicating have gradually shifted from offline to online media such as social networking platforms. Through big data analysis of online media, especially the analysis of its dynamic fluctuations, it is possible to predict market trends and manage brand reputation.

[0003] In existing technologies, there are many methods for fluctuation analysis of media data, which tend to make global adjustments or reconstructions to the entire dataset in order to obtain smooth or regular analysis results. Among them, the overall recalculation of historical data and new data consumes huge computing resources, resulting in low processing efficiency, which in turn leads to a lengthy analysis process and affects the real-time performance of fluctuation monitoring. Furthermore, existing technologies lack the ability to identify data fluctuations in their global adjustment strategies. They struggle to distinguish between abnormal fluctuations caused by key events and normal random fluctuations in data, and are prone to incorrectly smoothing or removing normal and analytically valuable fluctuation characteristics, leading to distorted analysis results. This, in turn, undermines the originality and authenticity of the data and causes changes in the processed data, resulting in insufficient reliability of the analytical conclusions.

[0004] In view of this, this application proposes a media data fluctuation analysis method and system based on artificial intelligence. Summary of the Invention

[0005] The present invention aims to address the shortcomings of existing technologies by providing a method and system for media data fluctuation analysis based on artificial intelligence.

[0006] To achieve the above-mentioned objectives, the technical solution adopted by the present invention is as follows: An artificial intelligence-based method for analyzing media data fluctuations includes the following steps: Identify abnormal fluctuations in target media data; When abnormal fluctuations are identified in the target media data, the following steps are performed: determine the abnormal fluctuation range corresponding to the abnormal fluctuation; and perform optimization processing on the data within the abnormal fluctuation range. After performing optimization processing on the data within the abnormal fluctuation range, a correction verification is performed. The correction verification includes: calculating the change range between the optimized data and the data before optimization processing; when the calculated change range is not within the preset allowable range, the step of determining the abnormal fluctuation range is triggered to be executed again. The process of identifying abnormal fluctuations in target media data includes: applying a trend model to analyze the changes between adjacent data points in the target media data to identify abnormal change points; calculating the number of consecutive abnormal change points based on the abnormal change points; and applying a fluctuation identification model to determine that the fluctuations in the target media data are abnormal fluctuations when the number of consecutive abnormal change points exceeds a preset consecutive number threshold.

[0007] Preferably, determining the abnormal fluctuation range corresponding to the abnormal fluctuation includes: Mark the abnormal change points that constitute abnormal fluctuations as reference points; use the reference points as the processing objects, reposition the reference points to determine the center point; Based on the center point, a preset range is extended along the time axis in both directions before and after the center point to define the abnormal fluctuation range; If multiple reference points exist within the abnormal fluctuation range, the multiple reference points are merged to update the center point, and the abnormal fluctuation range is updated based on the updated center point.

[0008] Preferably, the process of repositioning the reference point to determine the center point includes: Determine the maximum and minimum points within the data range adjacent to the benchmark point; based on the risk threshold, select one of the maximum and minimum points as the center point.

[0009] Preferably, the optimization processing of data within the abnormal fluctuation range includes: Based on the set of data points before and after the center point within the abnormal fluctuation range, forward correction parameters and backward correction parameters are calculated respectively; the forward correction parameters and backward correction parameters are then applied to correct the data within the abnormal fluctuation range.

[0010] Preferably, the method further includes: acquiring target media data before identifying abnormal fluctuations; The process of obtaining target media data includes: applying a time-based filtering function to determine whether the data in the original media data is invalid based on the deviation between the update time and the publication time of each media content; and removing the data that is determined to be invalid to obtain the target media data.

[0011] An artificial intelligence-based media data fluctuation analysis system includes the following modules: The fluctuation event monitoring module is used to acquire target media data and analyze the target media data using trend models and fluctuation recognition models, so as to output a fluctuation trigger signal when abnormal fluctuations are identified. The interval decision module is used to respond to fluctuation trigger signals and determine the abnormal fluctuation interval corresponding to the abnormal fluctuation. The data optimization module is used to perform optimization processing on the data within the abnormal fluctuation range determined by the interval decision module. The optimization verification module is used to calculate the change range between the optimized data and the data before optimization after the data optimization module performs optimization processing. When the change range is not within the preset allowable range, the interval decision module is triggered to redetermine the abnormal fluctuation range.

[0012] Preferably, determining the abnormal fluctuation range corresponding to the abnormal fluctuation includes: Mark the abnormal change points that constitute abnormal fluctuations as reference points; use the reference points as the processing objects, reposition the reference points to determine the center point; Based on the center point, a preset range is extended along the time axis in both directions before and after the center point to define the abnormal fluctuation range; If multiple reference points exist within the abnormal fluctuation range, the multiple reference points are merged to update the center point, and the abnormal fluctuation range is updated based on the updated center point.

[0013] Preferably, the process of repositioning the reference point to determine the center point includes: Determine the maximum and minimum points within the data range adjacent to the benchmark point; based on the risk threshold, select one of the maximum and minimum points as the center point.

[0014] Preferably, the optimization processing of data within the abnormal fluctuation range includes: Based on the set of data points before and after the center point within the abnormal fluctuation range, forward correction parameters and backward correction parameters are calculated respectively; the forward correction parameters and backward correction parameters are then applied to correct the data within the abnormal fluctuation range.

[0015] Preferably, acquiring the target media data includes: The application uses a time-based filtering function to determine whether the data in the original media data is invalid based on the deviation between the update time and the publication time of each media content; the data that is determined to be invalid is removed to obtain the target media data.

[0016] Beneficial effects The media data fluctuation analysis method provided by this invention uses a trend model to identify abnormal change points in the target media data. Based on the calculated number of consecutive abnormal change points, it uses a fluctuation identification model to determine whether there are abnormal fluctuations. This distinguishes between isolated abnormal change points that can be considered normal disturbances and the set of abnormal change points that constitute abnormal fluctuations, thereby avoiding misjudgment of occasional data jumps and filtering out non-critical noise interference, thus improving the accuracy and reliability of identifying abnormal fluctuations in media data.

[0017] When abnormal fluctuations are identified, this invention marks the abnormal change points that constitute the abnormal fluctuations as reference points. By combining the maximum and minimum points and risk thresholds in the adjacent data range, the reference points are repositioned and the center point is determined, thereby defining and updating the abnormal fluctuation range. This ensures that the optimization process can only process the data area that needs to be corrected, avoiding deviation from the processing range.

[0018] Within a defined abnormal fluctuation range, this invention calculates forward correction parameters and backward correction parameters based on the variation amplitudes of the data point sets before and after the center point. An adaptive correction strategy is then used to optimize the data within the range, followed by correction verification. This ensures a smooth transition between the optimized data and the data on both sides of the abnormal fluctuation range, and allows for secondary adjustments to unreasonable corrections, guaranteeing the stability and reliability of the output data. Attached Figure Description

[0019] Figure 1 This is a flowchart of the method of the present invention; Figure 2 This is a system module diagram of the present invention. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Example 1 Please refer to Figure 1 This embodiment provides a media data fluctuation analysis method based on artificial intelligence, including the following steps: By receiving raw media data containing media content, release time, and update time, target media data containing media content and corresponding time nodes is obtained. Time filtering rules are applied to the raw media data to evaluate the timeliness of the content, so as to identify and remove invalid data and ensure the accuracy of subsequent analysis. The time-based filtering rules include: obtaining the publication time and update time for each piece of data in the original media data; setting a review date based on a preset observation period used to define and evaluate the timeliness of the data; counting the number of new or updated content associated with the data after the review date, and recording the counted number as the number of review content. If the number of reviewed items is zero, it indicates that the data has not been updated effectively within the observation period. The data is then deemed invalid, and all invalid data is removed from the original media data to obtain the target media data required for subsequent steps.

[0022] Furthermore, after obtaining the target media data, the dynamic analysis stage begins: based on the target media data, statistical parameters for quantitative analysis are extracted, and a baseline trend value is established for each data point in the sequence through a preset trend model to identify data jumps in the numerical sequence. Among them, the statistical parameters are numerical sequences arranged according to time nodes, specifically the number of reads, comments, or shares counted by day or hour; The baseline trend value is used to characterize the expected behavior of data under the current trend. It can be calculated by calculating the arithmetic mean or weighted average of all data points within a preset time window before the current data point. The formula for calculating the baseline trend value is:

[0023] In the formula, This represents the baseline trend value, which is the expected value predicted based on historical data at time point t. This represents the sequence of values ​​for N data points within a preset time window preceding the current data point. The value represents the historical data point, indicating the actual observed value at time point ti; express The corresponding weight sequence; This represents the weighting coefficient, indicating the importance of different historical data points; the closer to the current time point, the higher the weight. Indicates the size of the time window, representing the number of historical data points used to calculate the trend; Indicates the current time point, representing the current index or timestamp in the data sequence; Then, the difference between the actual observed value and the baseline trend value for each data point is calculated, and this difference is used as the change value. By calculating the change value, the data jumps are distinguished from the normal growth or decline trend of the data itself. At this point, the change value is compared with the preset normal change threshold. If the absolute value of the change value exceeds the normal change threshold, the data point is identified as an abnormal change point. Then, parameters associated with the abnormal change point, including its value, time node, and corresponding media content, are extracted.

[0024] Furthermore, all identified abnormal change points are traversed along the time axis, the number of consecutively occurring abnormal change points is counted and recorded as the number of consecutive abnormal change points, and the fluctuations of the target media data are classified according to the fluctuation type judgment criteria to distinguish between isolated abnormal change points that can be regarded as noise interference and abnormal fluctuations composed of consecutive abnormal change points that indicate real events. Specifically, based on the fluctuation type judgment criteria, the fluctuation of the target media data is classified by comparing the calculated number of consecutive abnormal change points with the preset consecutive number threshold. Among them, if there are a single or a few consecutive abnormal change points, they are mostly negligible random noise, which may be caused by instantaneous errors in data acquisition; while a long series of consecutive abnormal change points may indicate the occurrence of a real event that needs attention. If the number of consecutive abnormal change points is less than or equal to the consecutive number threshold, the fluctuation is judged as a normal fluctuation. In this case, to avoid over-correction of the data, the abnormal change points that constitute this fluctuation are remarked as normal points. Conversely, if the number of consecutive abnormal change points is greater than the consecutive number threshold, the fluctuation is judged as an abnormal fluctuation. In this case, all abnormal change points that constitute the abnormal fluctuation are uniformly marked as the baseline point and used as the object to enter the next processing stage.

[0025] Furthermore, after confirming the existence of abnormal fluctuations, the set of benchmark points constituting the abnormal fluctuations is taken as the processing object; within the time range covered by the set and the range of adjacent data, the maximum and minimum values ​​are determined; then, according to the preset risk threshold, one of the maximum and minimum values ​​is selected to reposition the benchmark point. Among them, since the initially identified benchmark point may only be the starting point or the end point of the fluctuation, while the peak or trough value of the numerical value can better represent the extreme impact of the event, the core impact location of the abnormal event can be accurately located by the maximum and minimum value points of the numerical value. Specifically, the setting of risk thresholds is related to the specific application scenario. In scenarios that focus on monitoring positive hot events, the maximum value point can be selected first; while in scenarios that focus on discovering negative events or service interruptions, the minimum value point can be selected first. This repositioned benchmark point is established as the center point. Using the center point as a reference, a preset time range is extended in both directions before and after the time axis to initially define the abnormal fluctuation range. To ensure the integrity of the range, an inspection is carried out. Specifically, the check includes: if multiple reference points exist within the abnormal fluctuation range, these reference points are merged to update the center point; the merging method includes: calculating the weighted average of the timestamps corresponding to each reference point to obtain a new center time point; Then, based on the updated center point, the abnormal fluctuation range is redefined so that the defined abnormal fluctuation range can completely cover the entire abnormal event; according to the defined abnormal fluctuation range, the data within the abnormal fluctuation range is optimized using a segmented correction strategy. Specifically, the segmented correction strategy includes: applying forward correction parameters and backward correction parameters to correct the data within the abnormal fluctuation range; wherein, the forward correction parameters are applied to correct the data before the center point; the backward correction parameters are applied to correct the data after the center point. The segmented correction process enables the corrected data segments to smoothly transition between the two ends of the range and the normal data outside. Specifically, a segmented correction strategy is adopted for the data within the abnormal fluctuation range, including: obtaining the data point set before the center point and the data point set after the center point within the abnormal fluctuation range; calculating the average value of the change amplitude between adjacent data points within the data point set before the center point as the forward correction parameter; calculating the average value of the change amplitude between adjacent data points within the data point set after the center point as the backward correction parameter; and then using the forward correction parameter and the backward correction parameter to adjust the data within the range point by point.

[0026] Furthermore, after performing optimization processing on the data within the abnormal fluctuation range, a correction check is performed to ensure the correction quality, thereby assessing whether the correction operation has introduced new and unreasonable jumps at the boundary of the range. The correction and verification process includes: calculating the variation range between the optimized data and the original data outside the range at the start and end points of the abnormal fluctuation range, and defining the allowable range of the maximum degree of change of the data sequence during normal transition. Then, it is determined whether the calculated change range is within the allowable range. If it is within the range, the optimization process is confirmed to be effective and the correction process ends. If the change range is not within the allowable range, it indicates that the correction may be too drastic or the initial abnormal fluctuation range is not accurately defined. If the change is outside the allowable range, a feedback adjustment mechanism is triggered: the abnormal fluctuation range is automatically returned and re-executed. The preset time range used to define the range is adjusted, and optimization and verification are performed again. This allows the system to learn from failed corrections and gradually improve the accuracy and success rate of subsequent correction operations.

[0027] Example 2 Please refer to Figure 2 This embodiment provides a media data fluctuation analysis system based on artificial intelligence, including the following modules: The fluctuation event monitoring module acquires target media data and processes the raw media data using a time filtering function. For each piece of media content, it compares the deviation between its update time and publication time. If the deviation does not meet the preset validity rules, the corresponding data is judged as invalid data and removed, thereby obtaining target media data for subsequent analysis. The preset validity rules include update time being much earlier than publication time or abnormal update frequency. The fluctuation event monitoring module also performs abnormal fluctuation identification on the target media data. It applies a trend model to analyze the change values ​​between adjacent data points arranged in chronological order in the target media data. When the absolute value of a certain change value exceeds the normal change threshold, the corresponding data point is identified as an abnormal change point. Based on the identified abnormal change points, the number of consecutively occurring abnormal change points is calculated. The fluctuation identification model is applied. When the number of consecutive abnormal change points is greater than the preset consecutive number threshold, the current fluctuation of the target media data is determined to be an abnormal fluctuation, and a fluctuation trigger signal is output.

[0028] The interval decision module, upon receiving a fluctuation trigger signal, marks multiple abnormal change points constituting this abnormal fluctuation as reference points. Using the reference points as the processing objects, it repositions them to determine the center point. First, it determines the maximum and minimum points of each reference point within its neighboring data range. Then, based on a preset risk threshold representing the tendency to focus on upward peaks or downward troughs in the data, it selects one of the maximum and minimum points as the center point corresponding to that reference point. Then, the interval decision module uses the center point as a basis to extend a preset time range along the time axis in both directions before and after the center point, thereby initially defining the abnormal fluctuation interval. It checks whether there are multiple reference points within this initially defined abnormal fluctuation interval to ensure the integrity of the interval. If there are multiple reference points, it means that these reference points belong to the same fluctuation event. At this time, the weighted time center or numerical center of multiple reference points is calculated to merge multiple reference points and update the center point. Based on this updated center point, the range is extended again to update and finally determine the abnormal fluctuation interval.

[0029] The data optimization module is used to optimize the data within the identified abnormal fluctuation range. It divides the data points within the abnormal fluctuation range into a set of data points before the center point and a set of data points after the center point based on whether the data points are located before or after the center point in time. The data optimization module performs calculations based on these two sets of data points to obtain forward correction parameters and backward correction parameters. The forward correction parameters can be calculated based on the trend or mean of the data points before the center point to reflect the normal state before the fluctuation occurs. The backward correction parameters can be calculated based on the trend or mean of the data points after the center point to reflect the stable state after the fluctuation ends. After calculating the forward correction parameters and the backward correction parameters, the data optimization module performs corrections on all data within the abnormal fluctuation range. It uses interpolation, fitting, or other smoothing algorithms, and uses the forward correction parameters and the backward correction parameters as boundary or constraint conditions to perform smooth transition processing on the data within the range, thereby eliminating the distortion caused by abnormal fluctuations.

[0030] The optimization verification module is used to verify the correction effect after the data optimization module performs optimization processing to ensure the logical rationality of the optimization. After the data optimization module completes the correction, the correction verification is started to calculate the change range such as the mean absolute difference, root mean square error or maximum difference value between the data after optimization and the data before optimization. The calculated change range is compared with the preset allowable change range. If the change is within the allowable range, the optimization process is considered successful. If the change is outside the preset allowable range, it indicates that the correction is too excessive, causing data distortion, or that the correction is insufficient and fails to effectively smooth the fluctuations, thus triggering a feedback signal.

[0031] The instruction range decision module receives feedback signals and redetermines the abnormal fluctuation range until the correction effect meets the requirements.

[0032] The above are merely preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A media data fluctuation analysis method based on artificial intelligence, characterized in that, Includes the following steps: Identify abnormal fluctuations in target media data; When identifying abnormal fluctuations in the target media data, perform the following: determine the abnormal fluctuation range corresponding to the abnormal fluctuation; Perform optimization processing on data within the abnormal fluctuation range; After performing optimization processing on the data within the abnormal fluctuation range, a correction verification is performed. The correction verification includes: calculating the change range between the data after optimization and the data before optimization; when the calculated change range is not within the preset allowable range, the step of determining the abnormal fluctuation range is triggered to be executed again. The process of identifying abnormal fluctuations in target media data includes: applying a trend model to analyze the changes between adjacent data points in the target media data to identify abnormal change points; calculating the number of consecutive abnormal change points based on the abnormal change points; and applying a fluctuation identification model to determine that the fluctuations in the target media data are abnormal fluctuations when the number of consecutive abnormal change points exceeds a preset consecutive number threshold.

2. The media data fluctuation analysis method based on artificial intelligence according to claim 1, characterized in that, The determination of the abnormal fluctuation range corresponding to the abnormal fluctuation includes: Mark the abnormal change points that constitute abnormal fluctuations as reference points; use the reference points as the processing objects, reposition the reference points to determine the center point; Based on the center point, a preset range is extended along the time axis in both directions before and after the center point to define the abnormal fluctuation range; If multiple reference points exist within the abnormal fluctuation range, the multiple reference points are merged to update the center point, and the abnormal fluctuation range is updated based on the updated center point.

3. The media data fluctuation analysis method based on artificial intelligence according to claim 2, characterized in that, The process of repositioning the reference point to determine the center point, using the reference point as the processing object, includes: Determine the maximum and minimum points within the data range adjacent to the benchmark point; based on the risk threshold, select one of the maximum and minimum points as the center point.

4. The media data fluctuation analysis method based on artificial intelligence according to claim 1, characterized in that, The optimization processing of data within the abnormal fluctuation range includes: Based on the set of data points before and after the center point within the abnormal fluctuation range, forward correction parameters and backward correction parameters are calculated respectively; the forward correction parameters and backward correction parameters are then applied to correct the data within the abnormal fluctuation range.

5. The media data fluctuation analysis method based on artificial intelligence according to claim 1, characterized in that, The method further includes: acquiring target media data before identifying abnormal fluctuations; The process of obtaining target media data includes: applying a time-based filtering function to determine whether the data in the original media data is invalid based on the deviation between the update time and the publication time of each media content; and removing the data that is determined to be invalid to obtain the target media data.

6. A media data fluctuation analysis system based on artificial intelligence, characterized in that, Includes the following modules: The fluctuation event monitoring module is used to acquire target media data and analyze the target media data using trend models and fluctuation recognition models, so as to output a fluctuation trigger signal when abnormal fluctuations are identified. The interval decision module is used to respond to fluctuation trigger signals and determine the abnormal fluctuation interval corresponding to the abnormal fluctuation. The data optimization module is used to perform optimization processing on the data within the abnormal fluctuation range determined by the interval decision module. The optimization verification module is used to calculate the change range between the optimized data and the data before optimization after the data optimization module performs optimization processing. When the change range is not within the preset allowable range, the interval decision module is triggered to redetermine the abnormal fluctuation range.

7. The media data fluctuation analysis system based on artificial intelligence according to claim 6, characterized in that, The determination of the abnormal fluctuation range corresponding to the abnormal fluctuation includes: Mark the abnormal change points that constitute abnormal fluctuations as reference points; use the reference points as the processing objects, reposition the reference points to determine the center point; Based on the center point, a preset range is extended along the time axis in both directions before and after the center point to define the abnormal fluctuation range; If multiple reference points exist within the abnormal fluctuation range, the multiple reference points are merged to update the center point, and the abnormal fluctuation range is updated based on the updated center point.

8. The media data fluctuation analysis system based on artificial intelligence according to claim 7, characterized in that, The process of repositioning the reference point to determine the center point, using the reference point as the processing object, includes: Determine the maximum and minimum points within the data range adjacent to the benchmark point; based on the risk threshold, select one of the maximum and minimum points as the center point.

9. A media data fluctuation analysis system based on artificial intelligence according to claim 6, characterized in that, The optimization processing of data within the abnormal fluctuation range includes: Based on the set of data points before and after the center point within the abnormal fluctuation range, forward correction parameters and backward correction parameters are calculated respectively; the forward correction parameters and backward correction parameters are then applied to correct the data within the abnormal fluctuation range.

10. A media data fluctuation analysis system based on artificial intelligence according to claim 6, characterized in that, The acquisition of target media data includes: The application uses a time-based filtering function to determine whether the data in the original media data is invalid based on the deviation between the update time and the publication time of each media content; the data that is determined to be invalid is removed to obtain the target media data.

Images