A Discrete Observation Method and System Based on Time-Point Series Data

Abstract

This invention provides a discrete observation method and system based on time-series data, belonging to the field of data processing technology. Specifically, it includes: when the discrete analysis requirement type does not belong to the target requirement type, determining the discrete coefficients in each stable period based on the observation and analysis results of time-series data from stable periods and time-series data from other monitoring devices; determining the relevant data types of time-series data from other monitoring devices based on the stability of the discrete coefficients; determining the deviation of the discrete coefficients of the relevant data types at different stable periods; performing observation and analysis processing of the time-series data of the relevant data types based on the deviation; and determining the discrete observation and analysis method of the time-series data of the target data type based on the observation and analysis results of the time-series data of each relevant data type, combined with the discrete analysis requirement type, thereby improving the reliability of the monitoring data.

Description

Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a discrete observation method and system based on time-point series data. Background Technology

[0002] Point-in-time series are absolute numerical sequences reflecting the state of a phenomenon at a specific point in time, and they have wide applications in finance, economics, monitoring equipment, and other fields. Common analytical methods include correlation analysis, cointegration analysis, and regression analysis; however, these methods are computationally complex and struggle to quickly identify opportunities or risks when new data is generated. Therefore, a method is needed that can simplify the processing of two sets of point-in-time series and determine their degree of dispersion in real time.

[0003] Furthermore, for time-series data with correlations, such as multiple monitoring nodes monitoring the same pipeline, or servers and communication equipment in the same data link, their monitoring data often exhibit a certain degree of correlation. Therefore, how to determine the discrete observation method between monitoring nodes based on the degree of dispersion of time-series data between monitoring nodes and other monitoring nodes, and in combination with the anomalies in the monitoring data of correlated monitoring nodes, in order to ensure the reliability of the anomaly identification results of the monitoring node's monitoring data, has become an urgent technical problem to be solved.

[0004] Therefore, there is an urgent need for a discrete observation method and system based on point-in-time series data. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted:

[0006] Specifically, this application provides a discrete observation method based on point-in-time series data, which includes:

[0007] S1 uses the parsing results of the time-point sequence data of the target data type of the monitoring equipment to determine the stable period of the target data type, and determines the discrete analysis requirement type of the target data type based on the historical operation data of the stable period.

[0008] S2 When the discrete analysis requirement type does not belong to the target requirement type, based on the observation and analysis results of the time-point series data in the stable period and the time-point series data of other monitoring devices, the discrete coefficient in each stable period is determined, and based on the stability of the discrete coefficient, the relevant data type of the time-point series data in other monitoring devices is determined.

[0009] S3 determines the deviation of the discrete coefficients of the relevant data types at monitoring time points in different stable periods, and performs observation and analysis processing methods for the time point series data of the relevant data types based on the deviation. Based on the observation and analysis processing results of the time point series data of each relevant data type, and in combination with the discrete analysis requirement type, determines the discrete observation and analysis method for the time point series data of the target data type.

[0010] The beneficial effects of this invention are as follows:

[0011] Based on the observation and analysis results of time-series data during stable periods and time-series data from other monitoring devices, the coefficient of dispersion in each stable period is determined. This enables the analysis of the correlation and dispersion of two time-series data based on the time-series data between the monitoring devices and other monitoring devices during stable periods. It also lays the foundation for determining the relevant data types based on the analysis results of the dispersion, and thus realizing the discrete observation and analysis processing between the target data type and the data type.

[0012] Based on the observation and analysis results of point-in-time series data of various related data types and the discrete analysis requirements, the discrete observation and analysis method for point-in-time series data of the target data type is determined. This method considers both the observation and analysis requirements of the target data type's point-in-time series data itself and the reliability of the observation and analysis results of point-in-time series data of related data types. In this way, a targeted discrete observation and analysis method is determined, ensuring the reliability of the monitoring, analysis, and processing of the target data type's point-in-time series data.

[0013] Furthermore, the monitoring equipment includes monitoring equipment that monitors physical signals, including electrical signals.

[0014] Furthermore, the other monitoring devices are those that have a correlation with the monitoring device in terms of monitoring location or other monitoring devices that have a correlation with the time-point sequence data of the target data type that conforms to the constraints of natural laws.

[0015] Furthermore, the target data type is determined based on the data type of the monitoring data from the monitoring device.

[0016] Furthermore, the historical operating data of the stable period includes the historical running length of the stable period and the distribution data of the stable period.

[0017] Furthermore, the method for determining the discrete observation analysis method for the time-point sequence data of the target data type is as follows:

[0018] Based on the observation and analysis results of time point series data of related data types, the identification results of abnormal fluctuation time points of the time point series data of the related data types are determined;

[0019] Based on the identification results of the abnormal fluctuation time points, the relevant data types of the abnormal fluctuation time points are determined;

[0020] Based on the relevant data types, the relevant data types at points of abnormal fluctuation, and the discrete analysis requirements, a discrete observation and analysis method for the point-in-time series data of the target data type is determined.

[0021] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned discrete observation method based on point-in-time series data when running the computer program.

[0022] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0024] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0025] Figure 1 This is a flowchart of a discrete observation method based on point-in-time series data;

[0026] Figure 2 This is a flowchart illustrating the method for determining the discrete analysis requirement type of the target data type;

[0027] Figure 3 This is a flowchart illustrating the method for determining the relevant data types;

[0028] Figure 4 This is a flowchart illustrating the method for determining the observation, analysis, and processing of point-in-time series data of related data types. Detailed Implementation

[0029] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0030] Example 1

[0031] like Figure 1 As shown, this application provides a discrete observation method based on point-in-time series data, specifically including:

[0032] S1 uses the parsing results of the time-point sequence data of the target data type of the monitoring equipment to determine the stable period of the target data type, and determines the discrete analysis requirement type of the target data type based on the historical operation data of the stable period.

[0033] Furthermore, the monitoring equipment includes monitoring equipment that monitors physical signals, including electrical signals.

[0034] Furthermore, the other monitoring devices are those that have a correlation with the monitoring device in terms of monitoring location or other monitoring devices that have a correlation with the time-point sequence data of the target data type that conforms to the constraints of natural laws.

[0035] Furthermore, the target data type is determined based on the data type of the monitoring data from the monitoring device.

[0036] Furthermore, the historical operating data of the stable period includes the historical running length of the stable period and the distribution data of the stable period.

[0037] It should be noted that the stable period is the period during which the stability of the operating data of the monitoring target corresponding to the target data type meets the requirements. For example, the period during which the deviation rate between different monitoring times is within a preset deviation rate range is considered as the stable period.

[0038] In the application scenarios of this invention (a discrete observation method based on fixed-base transformation for two sets of time-point data series), particularly in fields such as industrial equipment condition monitoring, power load analysis, and environmental physical field (e.g., electromagnetic field, vibration) monitoring, the two sets of time-point data series processed often originate from specific physical monitoring equipment (e.g., current transformers, voltage sensors, temperature probes, vibration accelerometers, etc.). The operating state of the monitored target is not always stable, and usually goes through different stages such as startup, stable operation, operating condition switching, and shutdown.

[0039] Directly performing discrete analysis on monitoring data from all time periods is not only computationally inefficient, but more seriously, including data from unstable and transient processes in the analysis will severely pollute the historical dataset used to calculate the baseline μ and standard deviation σ, leading to Z-score distortion and ultimately producing misleading "false signals" or drowning out genuine "opportunity signals".

[0040] Therefore, before implementing this invention, the reliability of the input data needs to be pre-assessed. The purpose of this embodiment is to propose a method for dynamically determining whether subsequent data requires discrete analysis based on the stability characteristics of the historical operational data of the monitored target, and for determining the analysis frequency requirement (discrete analysis requirement type). This method ensures that discrete analysis is only applied to periods of high data quality and stable status, thereby guaranteeing the accuracy of the analysis results and the effectiveness of the decision-making.

[0041] Monitoring equipment and target data types: Monitoring equipment refers to sensors or instruments used to collect electrical signals (current, voltage, power), physical signals (temperature, pressure, vibration, flow rate), etc. Target data types are the specific physical quantities collected by the monitoring equipment (such as "phase A current" or "vibration of the pump body front bearing").

[0042] Stable period: refers to a continuous time period in which the monitored target is in a stable operating state and the fluctuation of its monitoring data meets the preset quality requirements. In this embodiment, a specific judgment criterion is defined: if, within a continuous period, the deviation rate (|current value - previous value| / previous value) between all adjacent monitoring times (e.g., one point per minute) falls within a preset deviation rate range (e.g., [-1%, +1%]), then that period is determined to be a "stable period". The deviation rate range can be configured according to different data types and process requirements.

[0043] Specifically, such as Figure 2 As shown, the method for determining the discrete analysis requirement type of the target data type is as follows:

[0044] S11 uses historical operating data of the target data type during stable periods to determine the distribution data of the target data type during stable periods;

[0045] Historical operational data: This refers to the collection of all identified stable periods for the target data type within a past historical period (e.g., the past 30 days). Each stable period includes its start time, end time, and duration. Based on this, two key metadata items can be extracted:

[0046] Historical runtime: The sum of the total duration of all stable periods within the historical period. Distribution data of stable periods: Primarily describes the pattern of stable periods occurring on the timeline, especially the distribution aggregated by date (day). For example, it can be used to calculate "how many days in the past 30 days had stable periods" and "what was the total duration of stable periods each day".

[0047] The system performs aggregated analysis on a daily basis based on historical operational data (e.g., the past 30 days) of the target data type. For each day, the system performs the following operations:

[0048] Extract all time periods marked as "stable periods" within the day, and sum the durations of these stable periods to obtain the "total stable duration" (unit: hours or minutes) for the day.

[0049] Record whether there is at least one stable period on the day. If there is, mark the day as "day with a stable period"; otherwise, mark it as "day without a stable period".

[0050] S12 determines the total duration of stable periods on different dates based on the distribution data, and determines the discrete analysis requirement type of the target data type based on the total duration of stable periods on different dates.

[0051] Specifically, the above steps include the following situations:

[0052] Case 1: If there are dates that do not have stable time periods, the stability of the target data type monitoring is not high. Therefore, the discrete analysis requirement type of the target data type is determined as the target requirement type.

[0053] Specifically, there is a "day without a stable time period" -> Demand type: target demand type, judgment logic: within the historical observation period, there is at least one date where the target data type does not reach the stable standard for any time period throughout the day. This indicates that the operation of the monitoring target is extremely unstable, or the operating conditions change frequently and drastically, and cannot provide a reliable and continuous high-quality data window.

[0054] Decision result: Define its demand type as "Target Demand Type". This is the most urgent and comprehensive monitoring mode. The system will trigger discrete analysis processing for all stable periods generated by this data type.

[0055] Case 2: If there are no dates without stable time periods, the average total duration of stable time periods on different dates is determined based on the total duration of stable time periods on different dates. If the average total duration of stable time periods on different dates is less than a preset duration threshold, then the discrete analysis requirement type of the target data type is determined to be a type of requirement.

[0056] No days without stable periods, but short average total stable duration per day -> Demand type: Category 1, Precondition: Every day within the historical period has at least one stable period (i.e., case 1 is negated), Judgment logic: Calculate the average of the "total stable duration" for all dates. If this average is less than a preset duration threshold (e.g., 4 hours / day), this condition is triggered. Technical meaning: Although there are stable periods every day, these stable periods are very short. High-quality data windows are scarce, and data is unreliable most of the time.

[0057] Decision result: The demand type is defined as "Type 1 demand", indicating a high degree of demand analysis.

[0058] Case 3: If the average total duration of stable periods in different dates is not less than the preset duration threshold, then the duration of different stable periods is obtained. If there are dates where the duration of stable periods does not meet the requirements, then because the duration of stable periods is short, the discrete analysis requirement type of the target data type is determined to be a type of requirement.

[0059] The average daily stable total duration meets the target, but there are "full-day short stable days" -> Demand type: Type 1 demand, prerequisite: the average daily stable total duration is not less than the preset duration threshold (e.g., ≥4 hours), thus rejecting case 2.

[0060] Judgment logic: Check if there exists a date where, although there are stable periods, the individual duration of each stable period is less than a shorter "stable period duration threshold" (e.g., each stable period is less than 30 minutes). We call such dates "full-day short stable days".

[0061] Technical implications: While the average daily stability duration may appear to meet the target, it is composed of numerous fragmented short periods. Too short a stability period may not be sufficient to form a statistically significant local historical sample (for calculating μ and σ), thus compromising the analytical effectiveness.

[0062] Decision result: The demand type is also defined as "Type 1 Demand". The analysis strategy is the same as above, only analyzing during stable periods, and assigning lower confidence weights to the analysis results of these short stable periods.

[0063] Case 4: If there are no stable periods of time where the duration of any date does not meet the requirements, then the discrete analysis requirement type of the target data type is determined to be a type of requirement.

[0064] Ideal stable situation -> Demand type: Type II demand, prerequisite: meet the daily average stable total duration target (No case 2), and there is no "full-day short stable day" (No case 3).

[0065] Judgment logic: At this point, historical data shows that the monitored target is operating stably, with a sufficiently long and continuous stable period every day. This is the ideal data source state, which can continuously provide high-quality, long-term data windows.

[0066] Decision result: The demand type is classified as "Type II demand", which indicates that the degree of demand for relevant analysis is not high in different stable periods.

[0067] It should be noted that if the discrete analysis requirement type of the target data type is the target requirement type, then the target data type is determined to undergo discrete analysis processing during the stable period.

[0068] S2 When the discrete analysis requirement type does not belong to the target requirement type, based on the observation and analysis results of the time-point series data in the stable period and the time-point series data of other monitoring devices, the discrete coefficient in each stable period is determined, and based on the stability of the discrete coefficient, the relevant data type of the time-point series data in other monitoring devices is determined.

[0069] Furthermore, the method for determining the discrete coefficients in the stable period is as follows:

[0070] Step S1: Obtain and preprocess two sets of time point data sequences. Obtain two sets of time point data sequences, one reflecting the target data type during a stable period and the other reflecting the time point data sequences from other monitoring devices. These are denoted as sequence A = [a1, a2, ..., a...]. n The sequence B = [b1, b2, ..., b] is given by the given sequence B. n ];

[0071] Step S2: Perform fixed-base exponentiation on both sets of sequences;

[0072] Select a reference time point j, determine the values ​​of sequences A and B at the reference time point, and use them as reference values ​​A_j and B_j, respectively.

[0073] For each value a in sequence A i Divide by the base value A_j to obtain the fixed-base exponential sequence iA. Similarly, obtain the fixed-base exponential sequence iB of sequence B.

[0074] Step S3: Generate a single array by subtracting the two fixed-base exponentialized sequences obtained in Step S2 point by point to obtain the difference sequence D = [d1, d2, ..., d n ], where d_i = ia_i - ib_i;

[0075] Step S4: Calculate the coefficients of variation at each time point. Based on the difference sequence D, calculate the arithmetic mean μ and standard deviation σ, obtain the time point values ​​of the difference sequence D, denoted as x, and calculate the Z-score of the latest difference x: Z = (x - μ) / σ. The absolute value of the Z-score |Z| is the coefficient of variation at each time point.

[0076] The Z-score or the cumulative distribution probability p derived from it is used as the basis for decision-making: when |Z| exceeds a preset threshold (such as 2 or 3), it indicates that the relative relationship between the two objects has deviated significantly in a statistically significant manner.

[0077] Specifically, such as Figure 3 As shown, the method for determining the relevant data types is as follows:

[0078] In a typical three-layer processing chain of "data acquisition → preprocessing → deep analysis," the power consumption of a server is strongly correlated with its processing load. Fluctuations in the computational load of the upstream data preprocessing server (Pre-Svr) directly lead to changes in its output data volume, which in turn affects the computational load and power consumption of the downstream deep analysis server (Ana-Svr). Therefore, from a business logic perspective, the time-series power consumption data of servers in the same data chain should exhibit a conductive correlation. This embodiment aims to use statistical methods to quantitatively verify whether the power consumption data of Pre-Svr can serve as an effective "correlated data type" for analyzing and predicting the power consumption of Ana-Svr, providing data support for energy efficiency monitoring and collaborative optimization of the chain.

[0079] S31 uses the dispersion coefficients of the monitoring time points in different stable periods based on the time point data of other monitoring devices, and determines the deviation monitoring time points in the stable period based on the dispersion coefficients of the monitoring time points in different monitoring periods;

[0080] Monitoring equipment: In this step, it refers to the candidate server, that is, the upstream server Pre-Svr whose data correlation is to be evaluated.

[0081] Point-in-time data: refers to the power consumption time series data of Pre-Svr, which is collected once per minute.

[0082] Stable period: refers to the continuous time period during which the power consumption of the target server Ana-Svr is in a stable operating state. By analyzing Ana-Svr's historical data, the daily early morning off-peak business period (such as 02:00-05:00) is identified as the stable period.

[0083] Monitoring time point: The specific time point at which power consumption data is collected once per minute within each stable period.

[0084] Coefficient of Dispersion: In this invention, it specifically refers to a standardized dispersion metric based on the Z-score concept. For the power consumption value P_t of the candidate server Pre-Svr at a certain monitoring time point t, its coefficient of dispersion C_t is calculated as follows: taking all power consumption data of Pre-Svr within the entire stable time period S in which the target server is currently located as the population, calculate its mean μ_S and standard deviation σ_S, then C_t = |P_t - μ_S| / σ_S. This coefficient measures the degree of deviation of P_t from the overall distribution position of Pre-Svr under this stable link state.

[0085] Preset Discrete Coefficient Threshold: The critical value for determining whether the discrete coefficient is too large, denoted as T_cv. Based on statistical principles (similar to the 2σ principle of Z-scores), this embodiment sets T_cv = 2.0. Deviation from Monitoring Time Point: Within a certain stable period S of the target server, if the discrete coefficient C_t of the candidate server Pre-Svr at a certain monitoring time point t is greater than T_cv (2.0), then this time point t is marked as a deviation from the monitoring time point.

[0086] Example: The target server Ana-Svr was identified as having a stable period S1 from 02:00 to 04:00 on October 25th, containing 120 monitoring points (one per minute). Within this period S1, the 120 power consumption values ​​of the candidate server Pre-Svr constitute a sequence. The global mean of this sequence is calculated as μ_S1 = 850W, and the standard deviation is σ_S1 = 40W.

[0087] For the 45th monitoring point (corresponding to 02:44), the power consumption of Pre-Svr is P_45 = 950W. Calculate its coefficient of variation: C_45 = |950 - 850| / 40 = 100 / 40 = 2.5.

[0088] Since C_45 = 2.5 > T_cv = 2.0, it is determined that during the stable period S1, the monitoring point t=45 is a deviation from the monitoring point of Pre-Svr. This indicates that at this moment, the power consumption of Pre-Svr is significantly higher than its normal level under the current stable link condition.

[0089] This step forms the microscopic foundation for correlation analysis. Instead of directly comparing the shapes of the two curves, it focuses on a core question: when the downstream server (Ana-Svr) is in a stable state, is the power consumption of the upstream server (Pre-Svr) also within its normal range for this stable context? This method cleverly uses the downstream's stable state as a time benchmark and statistical background for evaluating the consistency of upstream data.

[0090] Standardization eliminates baseline differences: By using the coefficient of variation based on the statistics (μ, σ) over the entire time period, the difference in absolute power consumption between Pre-Svr and Ana-Svr is effectively eliminated, allowing the analysis to focus on relative fluctuation patterns.

[0091] Aligning with Link Transmission Logic: In a link, downstream stability often implies stable service traffic. If a significant power consumption deviation occurs upstream, it may indicate an anomaly (such as resource leakage) or the processing of abnormal data blocks. This step accurately captures these transient events that could disrupt the link's stable balance.

[0092] It provides high-quality input for subsequent aggregation analysis: the output is a clear set of "offset points" rather than fuzzy similarity scores, which allows for clear proportion calculations and logical judgments in subsequent steps.

[0093] In the above steps, S311 determines whether the time-point data series deviates from the monitoring time point in different stable periods. If so, proceed to the next step; otherwise, determine that the time-point data series does not belong to the relevant data type.

[0094] Definition: This step is an existence check. It determines whether at least one deviation from the monitoring time point has been found after analyzing all stable periods in step S31.

[0095] Example: The system traversed and analyzed Ana-Svr data over the past week (a total of 21 stable time periods). The calculation results of step S31 showed that Pre-Svr detected a total of 205 deviations from the monitoring time points in 18 of these time periods. Therefore, it was determined that "deviations from the monitoring time points" "existed," and the process proceeded to step S312.

[0096] This is a fast logical load divider. If a candidate server shows absolutely no deviation during all stable periods, it indicates that its power consumption remains perfectly within its normal fluctuation range at any stable downstream moment, exhibiting an "absolutely stable following" state. This in itself is a very strong piece of evidence of correlation and can be directly determined as a qualified data type of correlation.

[0097] Beneficial effects: Improved judgment efficiency. In this ideal situation, there is no need to perform subsequent complex percentage calculations and threshold judgments, simplifying the process. At the same time, it clarifies that "deviation" is the only regular situation requiring further refined evaluation.

[0098] S312 Based on the deviation monitoring time point data in different stable periods, determine the proportion of the number of deviation monitoring time points in different stable periods, and determine whether the average proportion of the number of deviation monitoring time points in different stable periods is greater than the preset proportion threshold. If so, determine that the time point data series does not belong to the relevant data type. If not, proceed to step S32.

[0099] The percentage of deviations from monitoring points: For a single stable time period i, its value is R_di = (number of deviations from monitoring points in this time period) / (total number of monitoring points in this time period). It reflects the density of significant fluctuations in candidate servers within this time period.

[0100] Average percentage of the number of cases: The arithmetic mean of R_di calculated over all stable periods, denoted as Avg(R_d). It represents the overall frequency of deviations in the power consumption behavior of candidate servers.

[0101] Preset quantity percentage threshold: The boundary value for determining whether the deviation is "too common", denoted as T_rd. In this embodiment, T_rd is set to 0.15 (i.e., 15%).

[0102] Example:

[0103] Calculate R_di for each stable period. For example, if there are 10 deviation points in a period with 100 monitoring points, then R_di = 0.10. Calculate R_di for the 21 periods over the past week, obtaining Avg(R_d) = 0.128. Determine: Avg(R_d) = 0.128 < T_rd = 0.15, the power consumption deviation behavior of Pre-Svr does not reach the level of "common" or "frequent". The process continues to step S32.

[0104] This step is a stability filter. It aims to eliminate candidate servers whose power consumption is consistently unstable and whose noise is excessive. If a candidate server fluctuates most of the time (high Avg(R_d) value), then its data itself is an unreliable source of noise and is unsuitable as a benchmark for stable correlation analysis.

[0105] Ensuring data quality: This step ensures that the candidate data passing through this step meets the basic stability requirements, laying the foundation for building reliable relationships. Automated screening: By setting objective statistical thresholds, automatic initial screening of candidate data sources is achieved, reducing manual intervention.

[0106] S32 determines the relevant stable periods within the stable periods based on the deviation monitoring time point data in different stable periods;

[0107] Relevant stable period: refers to a period within a certain stable timeframe of the target server where the candidate server does not deviate from the monitoring time point. That is, during this entire period, the power consumption of the candidate server remains within its normal fluctuation range relative to the current stable state of the link (all C_t ≤ T_cv).

[0108] Example: Based on all deviation monitoring time point data identified in step S31, the system examines each stable time period. It is found that among the 21 stable time periods, 14 time periods do not contain any Pre-Svr deviation points. These 14 time periods are thus identified as relevant stable time periods.

[0109] This step identifies specific time periods as evidence of the occurrence of the "ideal coordinated state." These periods directly demonstrate the correlation that "when the downstream is stable, the upstream can also remain stable."

[0110] Beneficial effects: It provides a crucial set of positive evidence for the final determination. The number and proportion of these relevant stable periods are core positive indicators for measuring the strength of the correlation.

[0111] In the above steps, S321 determines whether there is a relevant stable period in the stable period. If yes, proceed to the next step; otherwise, determine that the time point series data does not belong to the relevant data type.

[0112] This step is an existence check for positive evidence. It determines whether at least one relevant stable period was found in the analysis of step S32.

[0113] Example: Step S32 confirms the existence of 14 relevant stable time periods. Therefore, it is determined that "existence" is achieved, and the process proceeds to S322.

[0114] Significance: If a candidate server deviates at least once during all downstream stable periods, it means that it has never achieved complete "synchronous stability" with the downstream, and there is no direct evidence of synergy between the two.

[0115] Beneficial effect: Quickly eliminates candidate data that cannot be synchronized with the downstream steady state, avoiding invalid subsequent calculations.

[0116] S322 Based on the number of relevant stable time periods, determine whether the proportion of the number of relevant stable time periods in the number of stable time periods is greater than a preset stable time period proportion threshold. If yes, determine that the time point series data belongs to the relevant data type. If no, proceed to step S33.

[0117] Definition: The percentage of relevant stable periods: R_s = (number of relevant stable periods) / (total number of stable periods).

[0118] Preset threshold for the percentage of stable periods: The threshold for determining whether cooperative stability is "dominant", denoted as T_rs. In this embodiment, T_rs is set to 0.65 (i.e., 65%).

[0119] Example: Calculate R_s = 14 / 21 ≈ 0.667. Determine: R_s ≈ 0.667 > T_rs = 0.65, the consistency rate between the power consumption data of Pre-Svr and the steady state of Ana-Svr exceeds the threshold. Based on the logic of this step, it can be directly determined that the power consumption data of Pre-Svr belongs to the relevant data type of Ana-Svr.

[0120] Significance: This is the core positive standard for determining relevance. It directly and quantitatively answers the key business question: "When the downstream server is running normally, how likely is it that the upstream server will also be running normally synchronously?"

[0121] Beneficial effect: When R_s is sufficiently high, it can provide clear and highly confident conclusions. It indicates that cooperative stability is the main pattern of the relationship between the two, and the correlation is significant.

[0122] S33 determines whether the time-point data series is a relevant data type based on the deviation monitoring time point data in different stable periods and the relevant stable period data.

[0123] It is understood that the relevant stable period is a stable period in which there is no deviation from the monitoring time point.

[0124] Specifically, the deviation monitoring time point is a monitoring time point where the coefficient of variation is greater than a preset coefficient of variation threshold.

[0125] Specifically, in the above steps, the average percentage of the number of deviation monitoring points in different stable periods is determined using deviation monitoring point data in different stable periods. Based on the average percentage of the number of deviation monitoring points in different stable periods and the percentage of the number of related stable periods in the stable periods, a correlation stability coefficient is determined. When the correlation stability coefficient is greater than a preset stability coefficient threshold, the time point data series is determined to be a related data type.

[0126] Correlation stability coefficient: A composite index that combines deviation universality Avg(R_d) and co-stability R_s. In this embodiment, its calculation formula is defined as: C = (1 - Avg(R_d)) * R_s. This design simultaneously penalizes high-frequency deviation behavior (a larger Avg(R_d) results in a smaller (1 - Avg(R_d))) and rewards high-frequency co-stability (a larger R_s). The value range of coefficient C is [0, 1], with values ​​closer to 1 indicating better overall correlation stability.

[0127] Preset stability coefficient threshold: A composite threshold used for final refined judgment, denoted as T_c. In this embodiment, T_c is set to 0.60.

[0128] Example (hypothetical scenario, used to illustrate S33):

[0129] Assume an edge case: After calculation, Avg(R_d) = 0.25 and R_s = 0.60. At this time, R_s = 0.60 < T_rs = 0.65, so the process enters S33 from S322, and calculates the relevant stability coefficient: C = (1 - 0.25) * 0.60 = 0.75 * 0.60 = 0.45.

[0130] Judgment: C = 0.45 < T_c = 0.60, the comprehensive stability does not meet the standard, and it is determined that the power consumption data of Pre-Svr does not belong to the relevant data type of Ana-Svr.

[0131] Significance: In the "fuzzy" or "competitive" situation where the collaborative stability ratio R_s does not reach the absolute dominant threshold (T_rs), a balanced and quantitative final adjudication mechanism is provided.

[0132] Refined decision-making: Avoids the arbitrariness of judgment based solely on the single index of R_s near the threshold. Comprehensive evaluation: By considering both positive evidence (collaborative stability) and negative evidence (frequency of deviation) simultaneously, a more scientific and robust decision is made. For example, even if R_s is slightly lower than 0.65, but if the deviation is extremely rare (Avg(R_d) is very small), C may still meet the standard; conversely, even if R_s is slightly higher than 0.65, but the deviation is very frequent (Avg(R_d) is very large), C may not meet the standard. This enhances the adaptability of the method in complex scenarios and the reliability of the results.

[0133] S3 determines the deviation of the coefficient of dispersion of the monitoring time points of the relevant data type in different stable periods, and based on the deviation, an observation and analysis processing method for the time series data of the relevant data type is provided. Based on the observation and analysis processing results of the time series data of each relevant data type, and in combination with the discrete analysis requirement type, a discrete observation and analysis method for the time series data of the target data type is determined.

[0134] Specifically, as Figure 4 shown, the method for determining the observation and analysis processing method of the time series data of the relevant data type is:

[0135] In complex server clusters or data processing chains, the power consumption of a target server may be statistically correlated with multiple upstream or peer servers in the same chain. After identifying multiple related data types using the aforementioned determination method, allocating appropriate monitoring resources and analysis strategies to each related data type becomes a critical issue. This embodiment aims to address how, when multiple identified related servers / data types exist, intelligently determine the most suitable point-in-time series data observation and analysis method based on the stability characteristics of each related data type and its position in a "stable cluster," thereby laying the foundation for determining targeted discrete observation and analysis methods.

[0136] In a three-layer chain of "Data Access (Access-Svr) → Real-time Processing (Stream-Svr) → Batch Processing (Batch-Svr)," the aforementioned method has been used to determine:

[0137] Target data type: Total power consumption of Batch-Svr (P_Batch).

[0138] Multiple related data types:

[0139] Related data type R1: Total power consumption of Stream-Svr (P_Stream).

[0140] Related data type R2: Total power consumption of Access-Svr (P_Access).

[0141] Related data type R3: Power consumption of storage nodes in the same rack as Batch-Svr (P_Storage).

[0142] We need to determine the observation and analysis methods for the three related data types P_Stream, P_Access, and P_Storage respectively.

[0143] Known historical data:

[0144] Over the past week, P_Batch had a total of 21 stable periods. Within these 21 periods: P_Stream had 18 related stable periods (meaning it was stable within these 18 periods), P_Access had 15 related stable periods, and P_Storage had 12 related stable periods.

[0145] S41 determines the relevant stable period of the relevant data type within the stable period based on the aforementioned deviation.

[0146] In the above steps, S411 determines the correlation coefficient of the relevant data type based on the proportion of the relevant stable time periods in the stable time periods, and determines whether the correlation coefficient of the relevant data type is greater than the preset coefficient threshold. If yes, the observation and analysis processing method of the time point series data of the relevant data type is determined to be the preset analysis and processing method. If no, proceed to step S42.

[0147] Relevant stable period: For each stable period of the target data type P_Batch, if the related data type P_Stream does not deviate from the monitoring time point within that period, then that period is a relevant stable period of P_Stream.

[0148] Correlation coefficient: This refers to the proportion of stable periods of the related data type P_Stream to the total number of stable periods of the target data type P_Batch. Corr_R1 = N_correlated_R1 / N_total = 18 / 21 ≈0.857. This coefficient directly quantifies the degree of co-stability between P_Stream and P_Batch.

[0149] Preset coefficient threshold: The threshold for determining whether the cooperative stability is "extremely high", set to T_high = 0.85.

[0150] Preset analysis and processing method: This refers to real-time parsing and processing of point-in-time series data of relevant data types at different times, instantly calculating the deviation rate between each new data point and the previous point. If the deviation rate exceeds a preset threshold, it is immediately marked as an abnormal fluctuation point. This method has strong real-time performance and can ensure data reliability, but it has high computational overhead.

[0151] Example: Calculate the correlation coefficient of P_Stream: Corr_R1 = 0.857, and determine: Corr_R1 ≈ 0.857 > T_high = 0.85.

[0152] Decision: Due to the extremely high degree of coordination stability between P_Stream and P_Batch (over 85%), their behavior is highly predictable and reliable. Therefore, a pre-defined analysis and processing method (i.e., high-frequency real-time monitoring) is directly adopted for P_Stream.

[0153] Assign the most stringent monitoring strategies to highly stable and highly correlated data types. This type of data is the most reliable "weathervane" for understanding the state of the target server, and its data quality needs to be carefully protected; any minor anomaly may be of great significance.

[0154] Implement real-time protection for key related indicators to ensure that anomalies can be detected as soon as possible, providing high-fidelity input for state awareness of core business links.

[0155] S42 will identify the relevant data types that belong to the relevant stable periods in different relevant stable periods as stable relevant data types;

[0156] In the above steps, in one possible embodiment, if there is no stable related data type in the relevant stable time period of the related data type, then the observation analysis and processing method of the time series data of the related data type is determined to be the preset analysis and processing method.

[0157] Identify "stable related data types" and construct a stable correlation network. Stable related data types refer to other related data types that are also stable during all their respective stable periods for a given related data type (such as P_Access). That is, during the periods when P_Access and P_Batch are co-stable, examine whether other related data types such as P_Stream and P_Storage are also stable simultaneously. This describes the group relationship of "co-stability" among multiple related data types.

[0158] Example:

[0159] Analyze 15 relevant stable periods of P_Access.

[0160] It was found that P_Stream (R1) was in a stable state in 14 of the 15 time periods (which are also the relevant stable time periods of P_Stream), and P_Storage (R3) was in a stable state in 8 of the 15 time periods.

[0161] Therefore, for P_Access: the set of stable related data types is {P_Stream} in 14 time periods, and the set of stable related data types is {P_Stream, P_Storage} in 8 time periods.

[0162] During the remaining time period, no other related data types remained stable at the same time.

[0163] This step goes beyond bilateral relationships and examines multivariate co-stability patterns. If a related data type is frequently stable alongside other related data types, it indicates that its stability is not isolated but embedded in a broader stability pattern. Therefore, its stability has little impact on the observation results of the target data type, resulting in lower requirements for observational analysis.

[0164] S43 determines the observation, analysis, and processing method for the time-point series data of the relevant data type based on the relevant stable time periods of the relevant data type and the stable relevant data types in different relevant stable time periods.

[0165] It should be noted that the stable related data type refers to the related data type other than the aforementioned related data type, which belongs to the relevant stable period of the related data type.

[0166] It should be noted that in the above steps, the weight value of the relevant stable period is determined based on the number of stable related data types in the relevant stable period. The comprehensive weight value is determined based on the sum of the weight values ​​of the relevant stable periods of the relevant data types. When the comprehensive weight value is greater than the preset weight threshold, the observation and analysis processing method of the time-point series data of the relevant data type is determined as the preset analysis and processing method. In other cases, the observation and analysis processing method of the time-point series data of the relevant data type is determined as the second preset analysis and processing method.

[0167] It is understood that the weight value of the relevant stable period is determined based on the number of stable relevant data types in the relevant stable period, and the more stable data types in the relevant stable period, the smaller the weight value of the relevant stable period.

[0168] It should be noted that the preset analysis and processing method involves real-time parsing of the relevant data types at different times to determine whether there are any abnormal fluctuations in the relevant data types, thereby ensuring the reliability of the time-series data of the relevant data types.

[0169] Weighting of relevant stable periods: Each relevant stable period is assigned a weight, which is inversely proportional to the number of "stable relevant data types" within that period. The formula can be set as: Weight_i = 1 / (1 + Count_Stable_i), where Count_Stable_i is the number of stable relevant data types (excluding itself) within that period. The larger the number, the smaller the weight, because its stability is confirmed by more "peers," so its stability has a lower impact on the correlation analysis results of the target data type, hence the smaller the weight coefficient.

[0170] Total weight value: The sum of the weight values ​​of all relevant stable periods, Total_Weight = Σ(Weight_i). Preset weight threshold: The critical value that determines which analysis and processing method to use, set as T_weight.

[0171] The second preset analysis and processing method is a condition-triggered, non-real-time batch processing analysis method. For example, "when no stable period is identified for the relevant data type within the most recent preset time period (e.g., 3 days), an analysis window (e.g., 6 hours) is initiated to identify and process abnormal fluctuation points in the data within that window." It significantly reduces computational overhead while maintaining basic reliability.

[0172] Example (continued from P_Access):

[0173] Calculate the weights for each time period:

[0174] For the 14 time periods accompanied only by P_Stream: Count_Stable=1, Weight = 1 / (1+1)=0.5.

[0175] For the 8 time periods accompanied by P_Stream and P_Storage: Count_Stable=2, Weight =1 / (1+2)≈0.333.

[0176] For a single time period without a companion: Count_Stable=0, Weight = 1 / (1+0)=1.

[0177] Calculate the overall weight: Total_Weight = (14 * 0.5) + (8 * 0.333) + (1 * 1) = 7 + 2.664 + 1 = 10.664. Decision judgment: Set T_weight = 12.0. Since 10.664 < 12.0, determine to use the second preset analysis method (condition-triggered analysis) for P_Access.

[0178] It should be noted that the abnormal fluctuation point is the point in time when the deviation rate of the monitoring data from the previous point in time is greater than a preset deviation rate threshold (e.g., 5%).

[0179] It should be noted that the second preset analysis and processing method is to identify and process abnormal fluctuation points in the future unit time period when the relevant data type does not have a stable period in the most recent preset time period, so as to ensure the reliability of the time point series data.

[0180] Specifically, the method for determining the discrete observation analysis method for the point-in-time series data of the target data type is as follows:

[0181] In the above steps: 1) The discrete analysis requirement type of the target data type (such as Batch-Svr power consumption) was determined (reflecting its own operational stability level); 2) Multiple related data types (such as upstream server power consumption) were identified, and the observation and analysis processing method (real-time monitoring or triggered analysis) was dynamically determined for each related data type. These related data types will become the benchmark or control sequence for discrete observation and analysis of the target data (such as Z-score calculation, anomaly detection).

[0182] This embodiment aims to solve the final execution decision problem: how to comprehensively consider the stability requirements of the target data type (discrete analysis requirement type) and the real-time health status of all related data types (whether there are abnormal fluctuations), and dynamically determine the discrete observation and analysis execution strategy to be adopted for the target data type. This ensures that the analysis action itself is intelligent, flexible, and risk-controllable.

[0183] Target data type (Target Metric, TM): The total power consumption P_Batch of Batch-Svr. Its discrete analysis requirement type has been assessed as a second-class requirement type based on its historical stability (its operation is relatively stable).

[0184] Correlated data types (CMs): Three related data types have been identified:

[0185] CM1: Stream-Svr power consumption P_Stream, using a preset analysis and processing method (real-time monitoring).

[0186] CM2: Access-Svr power consumption P_Access, using the second preset analysis and processing method (trigger-based analysis).

[0187] CM3: Storage-Svr power consumption P_Storage, using a preset analysis and processing method (real-time monitoring).

[0188] S51 uses the observation and analysis results of the time point series data of the relevant data type to determine the identification result of the abnormal fluctuation time points of the time point series data of the relevant data type;

[0189] Abnormal fluctuation points: Problem points identified in real time according to the observation and analysis processing methods configured for each relevant data type. For CM1 and CM3 using the "preset analysis and processing method" (real-time monitoring), an abnormal fluctuation point refers to any moment when the power consumption deviation rate relative to the data of the previous minute is greater than 5%. For CM2 using the "second preset analysis and processing method" (trigger-based analysis), analysis will only be initiated and abnormal fluctuation points may be identified if it has not been identified as a stable period within the past 3 days.

[0190] Example:

[0191] Within the past 3-day assessment window: CM1 (P_Stream): The monitoring system reported 2 instances of abnormal fluctuations. CM2 (P_Access): No instances of abnormal fluctuations were reported. CM3 (P_Storage): The monitoring system reported 0 instances of abnormal fluctuations.

[0192] S52 determines the relevant data types of the abnormal fluctuation time points based on the identification results of the abnormal fluctuation time points;

[0193] Example: Based on the results of S51, the relevant data types for the time points with abnormal fluctuations are determined as follows: CM1, and the relevant data types for the time points without abnormal fluctuations are determined as follows: CM2, CM3.

[0194] S53 determines the discrete observation and analysis method for the time-point sequence data of the target data type based on the relevant data type, the relevant data type at the time point with abnormal fluctuations, and the discrete analysis requirement type.

[0195] Discrete observation analysis method: refers to the specific analysis actions performed on the target data type P_Batch. The core is "under what conditions, and with which related data types, to perform discrete analysis (such as calculating the Z-score of the difference sequence)".

[0196] Discrete observation and analysis with related data types: This refers to comparing the target data sequence P_Batch with one or more healthy (currently without abnormal fluctuations) related data sequences within a selected stable period (such as base transformation, calculation of statistical indicators) to discover deviations of the target data from the benchmark.

[0197] Output operation and maintenance commands: When discrete analysis detects a significant deviation (such as Z score > 3), commands such as "Check Batch-Svr load" and "View upstream data backlog" are automatically generated.

[0198] Specifically, Case 1: If the discrete analysis requirement type of the target data type is a single requirement type:

[0199] Sub-case 1: If the number of related data types is less than the preset threshold for the number of related data types or if all related data types have abnormal fluctuation points, then within the preset time period in the future, the discrete observation and analysis method for the time-point sequence data of the target data type is to perform discrete observation and analysis processing with related data types as long as there is a stable time period.

[0200] Sub-case 2: If the number of related data types is not less than the preset threshold for the number of related data types and all related data types do not have abnormal fluctuation points, then within the preset time period in the future, the discrete observation and analysis method for determining the time-point sequence data of the target data type is to perform discrete observation and analysis processing with related data types during stable periods with a duration greater than the preset time period threshold.

[0201] Briefly describe the decision-making logic when the target data type is a type of requirement (with poor inherent stability):

[0202] Scenario assumption: P_Batch is a type of requirement, and the relevant data types are the same as above (CM1 is abnormal, CM2 and CM3 are normal).

[0203] Decision-making process (Scenario 1):

[0204] The number of related data types is 3, which is not less than the threshold N_cm_threshold=2, and not all related data types are abnormal (CM2 and CM3 are normal).

[0205] Therefore, we proceed to sub-case 2.

[0206] Decision: The discrete observation analysis method is to "perform discrete observation analysis on stable periods with a duration greater than a preset duration threshold within the next day, along with relevant data types."

[0207] By performing discrete observation and analysis on point-in-time series data of the target data type, it is possible to promptly identify and handle abnormal situations in the point-in-time series data of the target data type, such as deviation monitoring points, and make decisions based on the data at the deviation monitoring points.

[0208] It should be noted that the decision processing includes outputting operation and maintenance instructions, outputting market divergence instructions, etc.

[0209] Another understandable scenario is scenario 2: if the discrete analysis requirement type of the target data type is a type II requirement:

[0210] Entering Case 2: Because the discrete analysis requirement type of the target P_Batch is a type II requirement.

[0211] Sub-case analysis: Not all relevant data types exhibit abnormal fluctuations at certain points (CM2 and CM3 are healthy). Therefore, proceed to sub-case 1.

[0212] Sub-case 1 decision: The discrete observation analysis method is determined as follows: During stable periods of more than 30 minutes in the next day, discrete observation analysis is performed on relevant data types.

[0213] Meaning: In the next execution cycle, the system will:

[0214] a. Select a stable period: Within the next day, the discrete analysis is triggered only when P_Batch itself enters a stable period lasting more than 30 minutes.

[0215] b. Selecting a benchmark: During this stable period, select relevant data types (CM2 and CM3 in this example) that currently have no abnormal fluctuations as the benchmark sequence.

[0216] c. Execution Analysis: Perform basis transformation and Z-score analysis on the sequences of P_Batch and CM2 and CM3 respectively.

[0217] d. Ignore abnormal benchmarks: CM1 is not selected as the current analysis benchmark because it has recently experienced abnormal fluctuations, and using it as a benchmark may introduce noise or lead to misjudgment.

[0218] Sub-case 1: Based on the abnormal fluctuation time point data of different related data types, when it is determined that all related data types do not have abnormal fluctuation time points, the discrete observation and analysis method for the time point series data of the target data type is to perform discrete observation and analysis processing with related data types during stable periods with a duration greater than the preset duration threshold.

[0219] Sub-case 2: If all related data types have abnormal fluctuation points, based on the abnormal fluctuation points of different related data types, determine whether there are related data types in a unit time period where the number of abnormal fluctuation points does not meet the requirements. If so, proceed to the next step. If not, within a future preset time period, determine that the discrete observation and analysis method for the time point series data of the target data type is to perform discrete observation and analysis processing with related data types in stable time periods with a duration greater than the preset time period threshold.

[0220] The data types within a unit time period whose number of abnormal fluctuation points does not meet the requirements are designated as fluctuation data types. The fluctuation quantity ratio is determined based on the proportion of the fluctuation data type among the related data types. When the fluctuation quantity ratio is greater than a preset quantity ratio threshold, the discrete observation and analysis method for the time series data of the target data type is to perform discrete observation and analysis processing with the related data types during stable periods with a duration greater than a preset duration threshold. When the fluctuation quantity ratio is not greater than the preset quantity ratio threshold, the discrete observation and analysis method for the time series data of the target data type is to perform discrete observation and analysis processing with the related data types during stable periods with a duration greater than a preset duration threshold.

[0221] Example 2

[0222] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned discrete observation method based on point-in-time series data when running the computer program.

[0223] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, so the descriptions are relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0224] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0225] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A discrete observation method based on point-in-time series data, characterized in that, Specifically, it includes: Based on the parsing results of the time-point sequence data of the target data type of the monitoring equipment, the stable period of the target data type is determined, and the discrete analysis requirement type of the target data type is determined based on the historical operation data of the stable period. The monitoring device is a server, and the target data type is the server's power consumption; When the discrete analysis requirement type does not belong to the target requirement type, the discrete coefficient in each stable period is determined based on the observation and analysis results of the time series data in the stable period and the time series data of other monitoring devices. Based on the stability of the discrete coefficient, the relevant data type of the time series data in other monitoring devices is determined. The other monitoring devices are other servers that are on the same data link as the server. The deviation of the discrete coefficients of the relevant data types at monitoring time points in different stable periods is determined. Based on the deviation, an observation and analysis processing method for the time point series data of the relevant data types is performed. Based on the observation and analysis processing results of the time point series data of each relevant data type, and combined with the discrete analysis requirement type, a discrete observation and analysis method for the time point series data of the target data type is determined. The method for determining the discrete analysis requirement type of the target data type is as follows: Based on historical operational data of the target data type during stable periods, determine the distribution data of the target data type during stable periods; Based on the distribution data, the total duration of stable periods in different dates is determined. Based on the total duration of stable periods in different dates, the discrete analysis requirement type of the target data type is determined. The discrete analysis requirement type includes the target requirement type, a first-class requirement type, and a second-class requirement type. The method for determining the discrete observation analysis method for the point-in-time series data of the target data type is as follows: Based on the observation and analysis results of time point series data of related data types, the identification results of abnormal fluctuation time points of the time point series data of the related data types are determined; Based on the identification results of the abnormal fluctuation time points, the relevant data types of the abnormal fluctuation time points are determined; Based on the relevant data types, the relevant data types at points of abnormal fluctuation, and the discrete analysis requirements, a discrete observation and analysis method for the point-in-time series data of the target data type is determined.

2. The discrete observation method based on point-in-time series data as described in claim 1, characterized in that, If the discrete analysis requirement type of the target data type is the target requirement type, then it is determined that the target data type will be subjected to discrete analysis processing during the stable period.

3. The discrete observation method based on point-in-time series data as described in claim 1, characterized in that, If the discrete analysis requirement type of the target data type is a single requirement type: If the number of related data types is less than the preset threshold for the number of related data types, or if all related data types have abnormal fluctuation points, then the discrete observation and analysis method for determining the time-point sequence data of the target data type is to perform discrete observation and analysis processing with related data types as long as there is a stable period.

4. A computer system, comprising: A memory and processor connected by communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a discrete observation method based on point-in-time series data as described in any one of claims 1-3.

Images