Method, device and equipment for identifying abnormal data of working hours, medium and program product

Abstract

The invention provides a man-hour abnormal data identification method and device, equipment, a medium and a program product. The working hour abnormal data identification method comprises the steps of performing statistical characteristic analysis of data distribution on a working hour sample of a worker, and determining skewness statistics of a long tail deviation side represented by the data distribution; according to the skewness statistics, respectively adjusting a lower limit threshold and an upper limit threshold of a normal fluctuation range of the man-hour data to obtain an adjusted range; the adjusting directions of the lower limit threshold value and the upper limit threshold value are consistent with the deviation side of the long tail; and removing abnormal man-hour data exceeding the lower limit threshold and the upper limit threshold of the adjusted range in the man-hour sample, and retaining normal man-hour data in the adjusted range to obtain effective man-hour data of the worker. Therefore, more accurate effective man-hour data of workers can be obtained, and abnormal value identification is more accurate.

Description

Technical Field

[0001] This invention relates to the field of data analysis technology, and in particular to a method, apparatus, equipment, medium, and program product for identifying abnormal working hours data. Background Technology

[0002] In garment hanging production lines, to improve production scheduling accuracy and rationally determine standard working hours, it is necessary to continuously and accurately collect and statistically analyze the actual processing hours of employees. During the statistical analysis, outliers in the working hour data must be identified and removed.

[0003] Outlier identification methods in related technologies set threshold ranges. Processing times exceeding these ranges are considered outliers. However, because the threshold range is fixed once determined, it can lead to over-rejection of minor outliers or omission of serious outliers. Therefore, these outlier identification methods are not accurate enough in identifying outliers. Summary of the Invention

[0004] This application provides an improved method, apparatus, device, medium, and program product for identifying abnormal working hours data.

[0005] This application provides a method for identifying abnormal working hours data, including: Statistical characteristic analysis of the data distribution of the worker's working hours sample was performed to determine the skewness statistic of the long-tail bias of the data distribution; According to the skewness statistics, the lower and upper thresholds of the normal fluctuation range of the working hour data are adjusted respectively to obtain the adjusted range; the adjustment direction of the lower and upper thresholds is consistent with the long-tail skew side. Abnormal working hour data that exceeds the lower and upper thresholds of the adjusted range in the working hour sample are removed, and normal working hour data within the adjusted range are retained to obtain the worker's effective working hour data.

[0006] Furthermore, the step of adjusting the lower and upper thresholds of the normal fluctuation range of the working hour data according to the skewness statistics to obtain the adjusted range includes: Based on the skewness statistics, an upper threshold coefficient for adjusting the upper threshold and a lower threshold coefficient for adjusting the lower threshold are determined; the upper threshold coefficient and the lower threshold coefficient are different, and the upper threshold coefficient and the lower threshold coefficient change in the same direction, consistent with the long-tail skew side; Using the upper limit threshold coefficient and the lower limit threshold coefficient, the lower limit threshold and the upper limit threshold are adjusted accordingly to obtain the adjusted range.

[0007] Furthermore, when the skewness statistic indicates that the data distribution of the time sample is skewed to the right, the upper limit threshold coefficient increases to a first value, and the lower limit threshold coefficient increases to a second value. When the skewness statistic indicates that the data distribution of the work hour sample is skewed to the left, the upper threshold coefficient decreases to the second value, and the lower threshold coefficient decreases to the first value; the first value is greater than the second value.

[0008] Furthermore, the effective working hour data includes a sequence of effective working hours collected periodically; The statistical analysis of the effective working hour data to obtain the worker's effective actual working hour data includes: According to the time window, the effective working hours data are statistically analyzed to obtain the effective actual working hours data of workers within each time window; the period is greater than or equal to the time window.

[0009] Furthermore, the time window includes a time window of fixed duration.

[0010] Furthermore, the time window is at least longer than the production time for the worker to process one product.

[0011] Furthermore, after obtaining the workers' effective working hours data, the method further includes: Statistical analysis is performed on the effective working hour data to obtain the workers' effective actual working hour data.

[0012] Furthermore, the method also includes: If the effective actual working hours data of workers are continuously stable within multiple consecutive time windows, the time window is increased. The statistical analysis of the effective working hour data to obtain the worker's effective actual working hour data includes: Based on the increased time window, statistical analysis is performed on the effective working hour data to obtain the effective actual working hour data of workers within each time window.

[0013] Furthermore, the effective actual working hours data of the workers includes the effective actual working hours data of the workers under the combined work process; The method further includes: Based on the proportion of the standard working hours of each process under the merged process in the total standard working hours, the effective actual working hour data of the worker is broken down to obtain the actual working hour data of each process of the worker.

[0014] Furthermore, the effective actual working hours data of the workers includes the effective actual working hours data of the workers under the combined work process; The method further includes: Receive custom splitting rules; Obtain the percentage of standard working hours of the process corresponding to the custom splitting rule in the total standard working hours, split the worker's effective actual working hour data, and obtain the worker's actual working hour data for each process.

[0015] Furthermore, the method also includes: The effective actual working hours data of the workers are used to generate structured and visual reports.

[0016] Furthermore, after generating a structured visual report from the workers' effective actual working hour data, the method further includes: Push structured and visual reports to the production system.

[0017] Furthermore, after obtaining the workers' valid actual working hour data, the method further includes: The deviation between the worker's actual working hours and the standard working hours is compared to generate a deviation analysis report; the deviation analysis report includes accuracy.

[0018] This application provides a device for identifying abnormal working hours data, including: The skewness statistic determination module is used to perform statistical characteristic analysis on the data distribution of the worker's working hour sample and determine the skewness statistic representing the long-tail skewness of the data distribution. The adjusted range determination module is used to adjust the lower limit threshold and the upper limit threshold of the normal fluctuation range of the working hour data according to the skewness statistics to obtain the adjusted range; the adjustment direction of the lower limit threshold and the upper limit threshold is consistent with the long tail skew side. The working hour data processing module is used to remove abnormal working hour data from the working hour sample that exceeds the lower and upper thresholds of the adjusted range, and retain normal working hour data within the adjusted range to obtain the worker's effective working hour data. The actual working hour data determination module is used to perform statistical analysis on the effective working hour data to obtain the worker's effective actual working hour data.

[0019] This application provides an electronic device, including one or more processors, for implementing the time anomaly data identification method described above.

[0020] This application provides a computer-readable storage medium having a program stored thereon that, when executed by a processor, implements the method described in any of the preceding claims.

[0021] This application provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the method described in any of the preceding claims.

[0022] In some embodiments, the method for identifying abnormal work hour data in this application statistically analyzes the data distribution of worker work hour samples and dynamically adjusts the lower and upper thresholds of the normal fluctuation range of work hour data. This ensures that the adjusted range aligns with the actual work hour data distribution, better reflecting normal fluctuations. The dynamic scaling of the data skew allows the lower and upper thresholds to flexibly adapt to the degree of distribution skewness. Furthermore, by removing abnormal work hour data according to the adjusted range and retaining normal work hour data, more accurate effective work hour data for workers can be obtained, making outlier identification more accurate. Moreover, by representing the skewness of the long-tail bias in the data distribution, extreme outliers do not unduly affect the statistical calculation of work hours, while normal long-tail data is fully preserved. This adaptive capability to data skewness effectively reduces false positives for normal samples in the long-tail region, significantly providing highly reliable actual work hour data and optimizing the accuracy of actual work hour calculations. Attached Figure Description

[0023] Figure 1 The figure shown is a histogram of the processing data distribution of the first worker in the time anomaly data identification method provided in this application embodiment; Figure 2 The figure shown is a histogram of the processing data distribution of the second worker in the time anomaly data identification method provided in this application embodiment; Figure 3 The figure shown is a quantile-quantile plot of the processing time distribution of the first worker in the time anomaly data identification method provided in this application embodiment; Figure 4 The figure shown is a quantile-quantile plot of the processing time distribution of the second worker in the time anomaly data identification method provided in this application embodiment; Figure 5 The diagram shown is a flowchart of the method for identifying abnormal working hours data provided in an embodiment of this application; Figure 6 As shown Figure 5 The diagram shows a comparison of the accuracy of different outlier identification methods for worker-process combinations, along with other outlier identification methods and related technologies. Figure 7 As shown Figure 5 The chart shows a comparison of recall rates between methods for identifying abnormal working hours data and other outlier identification methods, along with related technologies. Figure 8 As shown Figure 5 The chart shows a comparison of retention rates (LRA) of long-tailed normal samples in skewed data with other outlier identification methods and related technologies for identifying abnormal working hours data. Figure 9 The diagram shown is a structural schematic of an electronic device provided in an embodiment of this application. Detailed Implementation

[0024] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with one or more embodiments of this specification. Rather, they are merely examples of apparatuses and methods consistent with some aspects of one or more embodiments of this specification as detailed in the appended claims.

[0025] It should be noted that the steps of the corresponding methods are not necessarily performed in the order shown and described in this specification in other embodiments. In some other embodiments, the methods may include more or fewer steps than described in this specification. Furthermore, a single step described in this specification may be broken down into multiple steps in other embodiments; and multiple steps described in this specification may be combined into a single step in other embodiments.

[0026] Through exploratory analysis of a large amount of production data, the inventors discovered that the distribution of single-piece processing time for hanging workers during the production process is not a conventional normal distribution.

[0027] like Figure 1 and Figure 2 The histogram showing the work hours of the first and second workers on a given workday is shown, with the first worker's employee number being 286 and the second worker's employee number being 336. If the data follows a normal distribution, the histogram should be approximately symmetrical, with the mean and median close together and the decay rates of the left and right tails being roughly equal. However, in the actual data, the work hour distributions of both the first and second workers exhibit a main peak skewed to the left, with a significant long tail on the right. A large number of samples are concentrated in shorter work hour intervals, and the mean differs considerably from the median, presenting a typical right-skewed long-tailed distribution.

[0028] like Figure 3 and Figure 4 The diagram shows the quantile-quantile (Q-Q) plots for the first and second workers. If the data follows a normal distribution, the sample quantiles should be distributed along a reference line. However, the actual results show that the upper quartiles are systematically above the reference line, and the deviation increases gradually with increasing quantiles. The lower quartiles are mostly slightly below or close to the reference line, indicating that the data as a whole exhibits significant positive skewness and heavy-tailed characteristics, clearly deviating from the normality assumption.

[0029] Outlier identification methods, which combine relevant technologies, generally include the symmetric thresholding method and the weakly modified thresholding method. The implementation process of the symmetric thresholding method is as follows: A symmetric threshold interval is set with the mean as the center; data points are iterated through, and those exceeding the set symmetric threshold interval are identified as outliers.

[0030] The implementation process of the aforementioned weakly corrected thresholding method is as follows: Sort the dataset from smallest to largest, and calculate the lower quartile and upper quartile. Then, calculate the interquartile range and the correction threshold, and set the weakly corrected threshold range. Data points exceeding the aforementioned weakly corrected threshold range are identified as outliers.

[0031] As can be seen, both the symmetric thresholding method and the weakly modified thresholding method mentioned above are offline static methods. Once the threshold is determined, it remains fixed, which can lead to over-removal of minor anomalies or the omission of serious anomalies. Therefore, the outlier identification methods of these technologies are not accurate enough in outlier identification.

[0032] To address the inaccuracy of outlier identification methods in the aforementioned related technologies, the present application provides a method for identifying outlier work time data. This method statistically analyzes the distribution of worker work time samples and dynamically adjusts the lower and upper thresholds of the normal fluctuation range of work time data. This ensures that the adjusted range aligns with the actual work time data distribution, better reflecting normal fluctuations. The dynamic scaling of the data skew allows the lower and upper thresholds to flexibly adapt to the degree of distribution skewness. Furthermore, by removing outlier work time data according to the adjusted range and retaining normal work time data, more accurate effective work time data for workers can be obtained, making outlier identification more accurate. By representing the skewness of the long-tail bias in the data distribution, extreme outliers do not unduly affect the statistical calculation of work time, while normal long-tail data is fully preserved. This adaptive capability to data skewness effectively reduces false positives for normal samples in the long-tail region, significantly providing highly reliable actual work time data and optimizing the accuracy of actual work time calculation.

[0033] Figure 5 The diagram shows a flowchart of the method for identifying abnormal working hours data provided in an embodiment of this application.

[0034] like Figure 5 As shown, the method for identifying abnormal working hours may include, but is not limited to, the following steps 110 to 130: Step 110: Perform statistical characteristic analysis on the data distribution of the workers' working hour sample to determine the skewness statistic representing the long-tail bias of the data distribution.

[0035] Long-tail skewness refers to the asymmetry in data distribution, reflecting asymmetry in the dataset. This long-tail skewness can be caused by extreme values ​​or sparse sample points, resulting in a distribution where one side of the data is longer than the other. In work hour data, this long-tail characteristic often manifests as some workers having significantly higher or lower-than-average work hours. Analyzing the characteristics of long-tail skewness allows for more accurate identification of potential outliers and avoids misjudging normal but sparsely distributed data. Thus, by fully utilizing statistical skewness indices to quantitatively describe the shape of the data distribution, we can effectively distinguish and appropriately handle long-tail regions.

[0036] In step 110, the worker's time sample may include, but is not limited to, the processing time sample data of the worker in each of the multiple processes. Correspondingly, the valid actual time data of the worker in step 140 below may include, but is not limited to, the valid actual time data of the worker under the combined processes. See below for details.

[0037] In step 110, the worker's time sample may include, but is not limited to, at least one of the worker's processing time samples for a single operation. Correspondingly, the worker's valid actual time data in step 140 below may include, but is not limited to, the worker's valid actual time data for a single operation. See below for a detailed explanation.

[0038] The sample data on workers' processing times at each stage of the process reflects their time investment in different processes. By recording and analyzing the processing times at each stage in detail, it is possible to further identify which processes may have efficiency issues or anomalies. This effectively covers the characteristics of different processes, thereby improving the comprehensiveness and accuracy of the data analysis.

[0039] The selection method for the processing time samples of workers in a single process step involves separately statistically analyzing the processing time of each individual process step, which allows for a clearer identification of outlier distribution characteristics within that specific process. Furthermore, comprehensive analysis of processing time sample data from multiple processes helps to uncover potential correlations between processes, thus providing a basis for subsequent customized time breakdown. This not only enhances sensitivity to local anomalies but also provides a more reliable basis for global time management.

[0040] In step 110, based on the results of the skewness statistics and the actual distribution of the work hour data, the threshold can be adjusted to identify potential abnormal work hour data. This reflects the actual distribution of the work hour data, effectively capturing outliers and providing a more reliable data foundation for subsequent analysis.

[0041] Step 120: According to the skewness statistics, adjust the lower and upper thresholds of the normal fluctuation range of the working hour data to obtain the adjusted range; the adjustment direction of the lower and upper thresholds is consistent with the long-tail skewed side.

[0042] The aforementioned skewness statistic is an indicator used to measure the degree of asymmetry in the distribution of work-hour data. By calculating the skewness statistic, it can be determined whether the data distribution exhibits a long tail and its skewness. If the skewness is positive, it indicates that the work-hour data distribution is right-skewed, and abnormal work-hour data may be concentrated in the higher work-hour range. If the skewness is negative, it indicates that the work-hour data distribution is left-skewed, and abnormal work-hour data may be concentrated in the lower work-hour range. Thus, the skewness statistic becomes an important basis for adjusting the threshold of the normal fluctuation range, thereby more accurately reflecting the actual distribution of work-hour data.

[0043] Step 130: Remove abnormal working hour data from the working hour sample that exceeds the lower and upper limits of the adjusted range, and retain normal working hour data within the adjusted range to obtain the worker's effective working hour data.

[0044] This article defines work hour data that is below the lower limit of the adjusted range or above the upper limit of the adjusted range as abnormal work hour data. Work hour data that falls within the adjusted range is defined as normal work hour data.

[0045] Based on this, step 130 above can remove abnormal working hour data that is below the lower limit threshold or above the upper limit threshold of the adjusted range, retaining normal working hour data within the adjusted range. This allows for a more accurate distinction between normal working hour fluctuations and outliers, ensuring the accuracy and reliability of valid working hour data. This process not only reduces the interference of outliers on the overall data but also better reflects the actual working time distribution characteristics of workers. Furthermore, the retained normal working hour data provides a high-quality foundation for further analysis, helping to improve the scientific rigor and rationality of subsequent decision-making.

[0046] Combination Figure 5 As shown, as an optional embodiment of this application, step 130 above may include, but is not limited to, the following steps 131 to 132: Step 131: According to the skewness statistics, determine the upper threshold coefficient for adjusting the upper threshold and the lower threshold coefficient for adjusting the lower threshold; the upper threshold coefficient and the lower threshold coefficient are different, and the upper threshold coefficient and the lower threshold coefficient change in the same direction, consistent with the long-tail skew side.

[0047] Step 132: Using the upper limit threshold coefficient and the lower limit threshold coefficient, adjust the lower limit threshold and the upper limit threshold accordingly to obtain the adjusted range.

[0048] In the embodiments of this application, the upper threshold coefficient and the lower threshold coefficient change in the same direction toward the long tail bias, so as to better fit the distribution of working hours data, making it a robust statistical indicator that can characterize the skewness of working hours data and is not sensitive to abnormal working hours data, and thereby enabling more accurate statistics on skewed abnormal working hours data.

[0049] Furthermore, when the skewness statistic indicates that the data distribution of the work hour sample is skewed to the right (also known as right-skewed data), the upper limit threshold coefficient increases to a first value, and the lower limit threshold coefficient increases to a second value.

[0050] When the skewness statistic indicates that the data distribution of the time sample is skewed to the left (also known as left-skewed data), the upper threshold coefficient decreases to the second value, and the lower threshold coefficient decreases to the first value; the first value is greater than the second value.

[0051] For example, the MC (Medcouple Skewness Statistic) is determined for a sample of workers' working hours. This statistic characterizes the degree and direction of skewness in the distribution and is also called the skewness coefficient. The definition of the skewness statistic MC is: in The median of the sample of worker hours. and Less than and greater than the worker's working hours, respectively. The two halves of the work time sample. Kernel function The definition is as follows: The robustness skewness statistic The range of values ​​is Among them, skewness statistics This indicates that the time data is right-skewed, meaning the long tail is on the right side; it is a skewness statistic. This indicates that the time data is left-skewed, meaning the long tail is on the left. This indicates that the distribution is approximately symmetrical.

[0052] Based on different skewness statistics, this invention employs different threshold coefficients for the boundaries of the normal fluctuation range of working hour data, such as upper and lower thresholds: If skewness statistic The lower limit threshold coefficient is then set to a value of The upper limit threshold coefficient is taken as: If the skewness statistic The lower limit threshold coefficient is then set to a value of The upper limit threshold coefficient is taken as: .

[0053] Of course, the values ​​of 1.5, 3, and 4 for the above coefficients are merely examples to more accurately reflect the long-tail bias. However, these values ​​are not limited; any coefficient value that can better reflect the long-tail bias falls within the protection scope of this invention, and will not be listed here.

[0054] Therefore, the normal fluctuation range for determining abnormal working hour data is determined as follows: when hour: when hour: Where L and U are the lower and upper limits of the normal fluctuation range, respectively, used for determining abnormal working hours data. Compared to traditional symmetrical thresholds, the upper and lower limits of this adjusted threshold are no longer symmetrical. For a right-skewed distribution, the upper limit coefficient This will be greater than 1.5, thus extending the upper threshold further. This allows for more work-hour data to be retained in the long tail on the right, and the lower limit coefficient... If the value is less than 1.5, the lower limit threshold is tightened appropriately, thus eliminating work hour data that will not have much impact on the statistics of work hours.

[0055] Conversely, for left-skewed distributions, the above rules are reversed (the lower threshold is extended, and the upper threshold is tightened). Compared to the classic box plot in related technologies that uses a fixed coefficient of 1.5 times IQR (Interquartile Range Box Plot Method), in this embodiment, the upper and lower thresholds (i.e., the "whiskers" length coefficient) of the box plot are dynamically adjusted based on the calculated MC value.

[0056] Subsequently, work hour data below the lower threshold L or above the upper threshold U in the work hour sample are identified as abnormal work hour data and removed from the statistical analysis; the remaining data falling within the interval [L, U] are considered work hour data within the normal fluctuation range. In this way, extreme outliers will not have an undue impact on the statistical calculation of work hours, while normal long-tail data are completely preserved.

[0057] In this way, the upper and lower "whiskers" can flexibly adapt to the degree of distribution skewness based on the dynamic scaling of skewness, making the time data more effectively reflect the distribution of real time data, and thus accurately obtaining the effective actual time data of workers. Furthermore, MC skewness refers to the robust skewness statistic of Medcouple, used for outlier removal in the actual time data of garment hanging production lines. Thus, by dynamically and adaptively adjusting the outlier threshold of the IQR box plot based on the time data skewness statistic, asymmetrical scaling of the upper and lower "whiskers" lengths is achieved, thereby more accurately distinguishing between normal time fluctuations and abnormal outliers. In this way, through automated data acquisition and excellent outlier handling algorithms, highly reliable actual time data is provided, optimizing the accuracy of actual time calculation.

[0058] In this embodiment, the aforementioned adjustment method allows the threshold range to better align with the characteristics of actual working hour data distribution. When the working hour data exhibits a right skewness, an increased upper threshold coefficient better covers the long tail portion on the right, while an increased lower threshold coefficient avoids excessive restriction on the short tail portion on the left. Conversely, when the working hour data exhibits a left skewness, decreased upper and lower threshold coefficients more accurately reflect the distribution characteristics of the long tail on the left, while preventing the short tail region on the right from being misjudged as abnormal. This dynamic adjustment mechanism not only improves the accuracy of identifying abnormal working hour data but also enhances its adaptability to different skewed distributions, thus demonstrating higher robustness in complex and ever-changing real-world application scenarios.

[0059] Combination Figure 5 As shown, the above method also includes: step 140, performing statistical analysis on the effective working hour data to obtain the workers' effective actual working hour data.

[0060] The average of effective working hours can be used as the actual working hours data for workers. Furthermore, statistical indicators such as the average, median, and standard deviation of effective working hours data can be used to further quantify the distribution characteristics of workers' actual working hours. These statistical results reflect the workload and efficiency of workers within a certain time frame, providing a basis for developing reasonable working hour management strategies. Simultaneously, analyzing the distribution pattern of effective working hours data can identify potential work patterns or regularities, thereby helping managers optimize resource allocation and task assignment. In addition, grouping or classifying effective working hour data can reveal the impact of different time periods and different task types on working hours, laying the foundation for the design of personalized management solutions.

[0061] The effective working hours data includes a sequence of effective working hours collected periodically.

[0062] In the first example, the aforementioned period can be greater than or equal to the time window. This allows for the collection of effective working hours in stages throughout the day according to the period, improving the real-time performance of data acquisition compared to collecting data only once a day.

[0063] In the second example, the time window includes a fixed-length time window. The aforementioned period can be greater than 20 minutes but less than 12 hours. For example, the period could be half an hour or one hour. In practical applications, selecting an appropriate time window length requires comprehensive consideration of the data collection frequency, system processing capacity, and business needs. Shorter time windows can capture the changing trends of working hour data more promptly, but may increase the computational burden on the system; while longer time windows, although reducing the pressure on data processing, may lead to a lag in outlier identification. Therefore, when setting the time window, a trade-off is made between the distribution characteristics of historical data and the requirements of real-time monitoring.

[0064] In the third example, the aforementioned cycle is specifically at least longer than the production time for the worker to process one product. This avoids analytical bias caused by missing data due to incomplete production of a single product. Simultaneously, this setting helps to more accurately capture the details of working time variations, providing a more reliable data foundation for subsequent outlier identification. Furthermore, by linking the cycle to the duration, the processing requirements of different products can be dynamically adapted, further enhancing the flexibility and applicability of the method.

[0065] Specifically, when the period equals the time window, the working hours data within this time window can be collected and denoted as the sample set: Calculate the first quartile Q1, median Q2, and third quartile Q3 of the sample, and obtain the interquartile range: , Represents the sample set ， This represents the first sample in the sample set. This represents the second sample in the sample set. This represents the nth sample in the sample set. The total number of samples.

[0066] The collected work time data is stored in a structured database. The database structure includes, but is not limited to, the following fields: (1) Work process number: a unique number that identifies different work processes; (2) Employee ID: a unique identifier for the operating employee; (3) Hanger ID (Identification): a unique ID that identifies each hanger; (4) Start time: the timestamp of the hanger entering the work process; (5) End time: the timestamp of the hanger leaving the work process.

[0067] Accordingly, step 140 above may include the following step 141: Step 141, statistically analyze the effective working hours data according to the time window to obtain the effective actual working hours data of workers in each time window.

[0068] In scenarios involving both merged and non-merged processes, the system employs a sliding window algorithm to dynamically analyze continuous effective time sequences, optimizing time complexity and improving the efficiency of acquiring merged process time. This algorithm maintains a fixed-length time window and calculates the moving average and standard deviation of time within the window in real time as the window slides. The sliding window algorithm described above achieves dynamic management of work time data by maintaining a fixed-length time window. Within a sliding window, a valid actual work time data is obtained. When new valid actual work time data is introduced, the window slides forward one step. During this process, the oldest valid actual work time data is removed, and the new valid actual work time data is added to the window. This design ensures that the window always maintains the latest valid actual work time data, making the statistical results more timely and accurate. In this way, it is possible to quickly respond to changes in the production process and reflect the actual work time situation of each process in a timely manner.

[0069] After the window slides, the system does not need to re-traverse the entire valid actual working hour data; instead, it only recalculates the mean and standard deviation for the updated data within the window. This dynamic calculation method significantly reduces the computational burden and improves real-time performance. Specifically, the mean can be updated using simple addition and subtraction, and the standard deviation can be adjusted based on previous values. This efficient calculation strategy not only saves time but also helps optimize the use of system resources, ensuring smooth operation even under high loads.

[0070] The final output statistics reflect the timeliness characteristics of the process. Through sliding window processing, short-term fluctuations in the data are effectively eliminated, providing smooth statistical information. This smoothing process makes the trends in effective actual working hours data more apparent, helping decision-makers make more accurate judgments. Simultaneously, ensuring the real-time nature of effective actual working hours data allows users to quickly obtain the latest and most accurate data and adjust production strategies accordingly.

[0071] In this embodiment, the changing trends and fluctuation characteristics of working hours data within different time windows can be captured in greater detail. For each time window, the statistical analysis results can reflect the concentrated distribution of workers' actual working hours and potential abnormal fluctuations within that time period. This method not only helps to improve the granularity of data analysis but also provides a more accurate basis for further identifying potential abnormal working hours data. Simultaneously, by reasonably setting the relationship between the period and the time window, it ensures that the statistical results reflect both short-term fluctuations and long-term trends, thereby providing a reliable basis for anomaly determination in subsequent steps.

[0072] As an optional embodiment of this application, the above-mentioned method for identifying abnormal working hours may also include, but is not limited to: increasing the time window when the effective actual working hours data of workers within multiple consecutive time windows are detected to be continuously stable.

[0073] Accordingly, step 140 above may include, but is not limited to, statistically analyzing the effective working hour data according to the increased time window to obtain the effective actual working hour data of workers within each time window.

[0074] In this embodiment, by increasing the time window, data fluctuations can be further smoothed, thus allowing for a clearer observation of long-term trends in work hour data. This method is particularly suitable for scenarios where short-term fluctuations are small, but significant changes may occur over longer time spans. Furthermore, after adjusting the time window, the statistical analysis results will focus more on the overall work hour distribution characteristics, rather than changes in local details. This not only improves the efficiency of work hour data analysis but also reduces interference from short-term fluctuations. Simultaneously, the operation of increasing the time window can be flexibly configured according to actual needs to meet the analysis accuracy requirements of different scenarios, providing more robust work hour data support for subsequent anomaly detection.

[0075] As an optional embodiment of this application, the above method may also include, but is not limited to: First, according to the proportion of the standard working hours of each process under the merged process in the total standard working hours, the effective actual working hour data of the worker is split to obtain the actual working hour data of each process of the worker.

[0076] In the first step above, the process is broken down according to the proportion of the standard working hours of each process under the merged process in the total standard working hours, so as to ensure that the working hours of each process of the workers are allocated in accordance with the actual operation logic.

[0077] In this embodiment, this breakdown method allows for a more precise assessment of the time consumption of each process, thus providing data support for process optimization. Simultaneously, this method helps identify which processes may have abnormal time consumption, providing clues for further analysis. Furthermore, based on the actual time data of each process, comparative analysis can be performed with historical data to discover potential efficiency improvement opportunities or the root causes of abnormal fluctuations. This refined processing not only enhances the targeting of data analysis but also provides a more reliable basis for subsequent decision-making.

[0078] As an optional embodiment of this application, the above-described method for identifying abnormal working hours data may also include, but is not limited to, the following first and second steps: The first step is to receive the custom splitting rules.

[0079] In this context, splitting rules refer to customizable division rules for work processes based on specific conditions. These rules can be flexibly set by users according to actual production needs, such as dividing processes based on process type, processing time range, or equipment usage. By receiving custom splitting rules, the system can more accurately identify the differences between different processes, thus providing more detailed data support for subsequent time analysis. This method not only improves the accuracy of anomaly detection but also enhances the system's adaptability and scalability, enabling it to meet the needs of diverse production scenarios.

[0080] The above-mentioned modified splitting rules are simple adjustments made to the existing splitting rules to adapt to the needs of flexible production, enabling the system to adapt to the needs of flexible production and providing enterprises with greater flexibility and adaptability.

[0081] The second step is to obtain the proportion of standard working hours of the process corresponding to the custom splitting rule in the total standard working hours, and to split the worker's effective actual working hour data to obtain the actual working hour data of each process of the worker.

[0082] In this embodiment, the custom-defined splitting rules allow for flexible adaptation to the time calculation needs of different production environments. By proportionally allocating the total standard working hours to each process, the time consumption of each step can be more accurately reflected. Furthermore, refining the actual working hour data not only helps to clarify the true time consumption of each process but also provides data support for further optimizing the production process. This rule-based automated processing significantly improves the accuracy and efficiency of time data analysis.

[0083] As an optional embodiment of this application, the above-described method for identifying abnormal working hours may also include, but is not limited to, generating structured and visualized reports from the worker's valid actual working hour data.

[0084] This structured, visual report, containing fields such as process ID, average working hours, and sample size, allows for a clear view of each worker's working hours distribution across different time periods, facilitating production management and decision-making. The report can also include multi-dimensional data analysis, such as summaries by process, product type, or work group.

[0085] As an optional embodiment of this application, the above-described method for identifying abnormal working hours may also include, but is not limited to, pushing the structured visual report to the production system after generating a structured visual report from the worker's valid actual working hour data. The visual report contains the broken down valid actual working hour data for each worker or the valid actual working hour data for each worker's individual process, to reflect all the actual production hours for each process.

[0086] The aforementioned push of structured and visualized reports to the production system is achieved by pushing real-time work hour data to the production system through a standardized API interface, so as to facilitate real-time adjustment of production line takt time and production scheduling.

[0087] As an optional embodiment of this application, the above-described method for identifying abnormal working hours may also include, but is not limited to, after obtaining the worker's valid actual working hours data, comparing the deviation between the worker's actual working hours data and the standard working hours, and generating a deviation analysis report; the deviation analysis report includes accuracy.

[0088] In this embodiment, a deviation analysis report is automatically generated, comparing actual working hours with the theoretical working hour database to drive iterative updates of standard working hours. Furthermore, by analyzing the accuracy of the deviation report, production processes can be further optimized and work efficiency improved. Based on the deviation analysis results, the system automatically adjusts worker task allocation to reduce future working hour deviations.

[0089] This invention combines an outlier identification algorithm based on MC skewness with a sliding window dynamic calculation scheme, further enhancing the practical application value of the method. To verify its effectiveness, embodiments of this application compare it with traditional methods such as the Mean Method, Median Absolute Deviation Method (MAD), and IQR box plot method using various metrics.

[0090] Figure 6 As shown Figure 5 The diagram shows a comparison of the accuracy of the time anomaly data identification method and related technologies, as well as other anomaly identification methods, on worker-process combinations.

[0091] like Figure 6As shown, for the calculation of working hours for each worker-process combination, the accuracy of this method (MC-Adjusted) is significantly improved compared to the traditional algorithm. This is mainly due to its adaptive ability to data skewness, which effectively reduces false positives to normal samples in the long-tail region, i.e., the number of false positives (FP) is greatly reduced.

[0092] Figure 7 As shown Figure 5 The chart shows a comparison of recall rates between methods for identifying abnormal working hours data and other outlier identification methods.

[0093] like Figure 7 As shown, this method also performs well and stably in terms of recall. Even when traditional methods have excessively wide thresholds due to extreme values, leading to missed detections, this method can still accurately detect real anomalies due to its robustness.

[0094] Figure 8 As shown Figure 5 The chart shows a comparison of the retention rate (LRA, Long-tailed Regular-sample Acceptance Rate) of other outlier identification methods for identifying abnormal working hours and related technologies in skewed data.

[0095] This application defines the Long Tail Retention Rate (LRA) as an indicator to measure the retention of non-outliers after cleaning. Figure 8 The results show that for skewed data, this method can effectively accommodate "slow but normal" work cycles by dynamically expanding the upper fence, while its robustness avoids excessive interference of outliers on the threshold. In contrast, the symmetrical fence method, due to its overly tight upper fence, causes a large number of long-tailed normal samples to be misjudged, and its LRA value is significantly lower.

[0096] In summary, this invention improves the outlier identification method by using MC skewness, thereby enhancing the accuracy and practicality of time data analysis and ultimately improving the production efficiency and management level of garment production lines.

[0097] Based on the same inventive concept as the above method, this application also provides a device for identifying abnormal working hours data, which may include the following modules: The skewness statistic determination module is used to perform statistical characteristic analysis on the data distribution of the worker's working hour sample and determine the skewness statistic representing the long-tail skewness of the data distribution. The adjusted range determination module is used to adjust the lower limit threshold and the upper limit threshold of the normal fluctuation range of the working hour data according to the skewness statistics to obtain the adjusted range; the adjustment direction of the lower limit threshold and the upper limit threshold is consistent with the long tail skew side. The working hour data processing module is used to remove abnormal working hour data from the working hour sample that exceeds the lower and upper thresholds of the adjusted range, and retain normal working hour data within the adjusted range to obtain the worker's effective working hour data. As one embodiment, the device further includes: an actual working hour data determination module, used to perform statistical analysis on the effective working hour data of the worker after obtaining the effective working hour data of the worker, to obtain the worker's effective actual working hour data.

[0098] As one embodiment, the device further includes: The time window adjustment module is used to increase the time window when the effective actual working hours data of workers within multiple consecutive time windows are detected to be continuously stable. The actual working hours data determination module is specifically used to perform statistical analysis on the effective working hours data according to the increased time window to obtain the effective actual working hours data of workers within each time window.

[0099] As one embodiment, the worker's effective actual working hours data includes the worker's effective actual working hours data under the merged process; the device further includes: a first splitting module, used to split the worker's effective actual working hours data according to the proportion of the standard working hours of each process under the merged process in the total standard working hours, to obtain the worker's actual working hours data for each process.

[0100] As one embodiment, the worker's effective actual working hours data includes the worker's effective actual working hours data under the merged process; the device further includes: a custom rule module for receiving custom splitting rules; The second splitting module is used to obtain the proportion of standard working hours of the process corresponding to the custom splitting rule in the total standard working hours, split the worker's effective actual working hour data, and obtain the actual working hour data of each process of the worker.

[0101] As one embodiment, the device further includes a report generation module, used to generate a structured and visual report from the worker's valid actual working hour data.

[0102] As one embodiment, the device further includes a data push module, used to push the structured visualization report to the production system after generating the structured visualization report from the worker's effective actual working hour data.

[0103] As one embodiment, the apparatus further includes: an analysis and comparison module, configured to, after obtaining the worker's valid actual working hour data, compare the deviation between the worker's actual working hour data and the standard working hour data, and generate a deviation analysis report; the deviation analysis report includes accuracy.

[0104] Each module of the above-mentioned device corresponds to the steps of the above-mentioned method. The specific implementation process of the function and role of each module in the above-mentioned device can be found in the implementation process of the corresponding steps in the above-mentioned method, which can achieve the same technical effect, and will not be repeated here.

[0105] This application provides an electronic device, including the above-described abnormal working time data identification device.

[0106] The method for identifying abnormal working hours data in this application is applied to electronic devices. These electronic devices can be PC (Personal Computer) devices. The PC devices can include, but are not limited to, desktop computers, tablet computers, or laptop computers.

[0107] Set up an interface for inputting information on the webpage of the PC device. In this way, the user can fill in the required information through the displayed input interface to obtain the user's desired information and achieve the user's purpose.

[0108] Figure 9 The diagram shown is a structural schematic of the electronic device 50 provided in an embodiment of this application.

[0109] like Figure 9 As shown, the electronic device 50 includes one or more processors 51 for implementing the time anomaly data identification method as described above.

[0110] In some embodiments, electronic device 50 may include storage medium 59. For example, computer-readable storage medium may store a program that can be invoked by processor 51, and may include non-volatile storage medium. In some embodiments, electronic device 50 may include memory 58 and interface 57. In some embodiments, electronic device 50 may also include other hardware depending on the specific application.

[0111] The computer-readable storage medium of this application embodiment stores a program thereon, which, when executed by the processor 51, is used to implement the working time anomaly data identification method described above.

[0112] This application provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the method described in any of the preceding claims.

[0113] This application also provides a computer program stored in a computer-readable storage medium, for example... Figure 9 The storage medium 59, and when the processor executes the computer program, causes the processor 51 to perform the method described above.

[0114] This application may take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented using any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer-readable storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0115] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this specification should be included within the scope of protection of this specification.

[0116] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Without further limitation, an element qualified by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A work hour abnormal data recognition method characterized by comprising: The method comprises: performing statistical characteristic analysis on data distribution of a work time sample of a worker to determine a skewness statistical quantity representing a long-tail skew side of the data distribution; adjusting a lower threshold and an upper threshold of a normal fluctuation range of the work time data according to the skewness statistical quantity, to obtain an adjusted range; the adjustment directions of the lower threshold and the upper threshold are consistent with the long-tail skew side; removing abnormal work time data beyond the lower threshold and the upper threshold of the adjusted range from the work time sample, and retaining normal work time data within the adjusted range, to obtain effective work time data of the worker.

2. The work hour abnormality data recognition method according to Claim 1, characterized by, The method further comprises: determining an upper threshold coefficient for adjusting the upper threshold and a lower threshold coefficient for adjusting the lower threshold according to the skewness statistical quantity; the upper threshold coefficient and the lower threshold coefficient are different, and the upper threshold coefficient and the lower threshold coefficient change in the same direction and are consistent with the long-tail skew side; adjusting the lower threshold and the upper threshold according to the upper threshold coefficient and the lower threshold coefficient, to obtain the adjusted range.

3. The work hour abnormality data recognition method according to claim 2, characterized by, When the skewness statistical quantity represents that the data distribution of the work time sample represents a long-tail skew to the right side, the upper threshold coefficient becomes a first value, and the lower threshold coefficient becomes a second value; When the skewness statistical quantity represents that the data distribution of the work time sample represents a long-tail skew to the left side, the upper threshold coefficient becomes the second value, and the lower threshold coefficient becomes the first value; the first value is greater than the second value.

4. The work hour abnormal data recognition method according to Claim 1, characterized by, After obtaining the effective work time data of the worker, the method further comprises performing statistical analysis on the effective work time data to obtain effective actual work time data of the worker.

5. The work hour abnormality data recognition method according to Claim 4, characterized by, The effective work time data comprises an effective work time sequence collected according to a period; The method further comprises: when it is detected that the effective actual work time data of the worker in a plurality of continuous time windows is stable, increasing the time window; The method further comprises:

6. The work hour abnormality data recognition method according to Claim 5, characterized by, performing statistical analysis on the effective work time data according to the increased time window to obtain the effective actual work time data of the worker in each time window.

7. The work hour abnormality data recognition method according to Claim 5, characterized by, The effective actual work time data of the worker comprises effective actual work time data of the worker in a combined process; 8. The work hour abnormality data recognition method according to Claim 5, characterized by, The method further comprises: splitting the effective actual work time data of the worker according to a proportion of standard work time of each process in a combined process in total standard work time, to obtain actual work time data of each process of the worker. ​ ​ 9. The work hour abnormal data recognition method according to any one of claims 1 to 8, characterized by, ​ ​ ​ 10. The work hour abnormal data recognition method according to any one of claims 1 to 8, characterized by, The effective actual working hour data of the worker includes effective actual working hour data of the worker in a combined process; The method further includes: receiving a custom split rule; obtaining a proportion of standard working hours of a process corresponding to the custom split rule in total standard working hours, splitting the effective actual working hour data of the worker to obtain actual working hour data of each process of the worker.

11. The work hour abnormal data recognition method according to any one of claims 1 to 8, characterized by, The method further includes: generating a structured visual report based on the effective actual working hour data of the worker.

12. The work hour abnormality data recognition method according to Claim 11, characterized by, After the structured visual report is generated based on the effective actual working hour data of the worker, the method further includes: pushing the structured visual report to a production system.

13. The work hour abnormal data recognition method according to any one of claims 1 to 8, characterized by, After the effective actual working hour data of the worker is obtained, the method further includes: comparing the actual working hour data of the worker with the standard working hour to generate a deviation analysis report; the deviation analysis report includes accuracy.

14. An abnormal work hours data identifying apparatus characterized by comprising: It includes: a skewness statistical quantity determination module configured to analyze statistical characteristics of data distribution of a working hour sample of a worker, and determine a skewness statistical quantity representing a long-tail skew side of the data distribution; an adjusted range determination module configured to adjust a lower threshold value and an upper threshold value of a normal fluctuation range of working hour data according to the skewness statistical quantity, respectively, to obtain an adjusted range; the adjustment directions of the lower threshold value and the upper threshold value are consistent with the long-tail skew side; a working hour data processing module configured to remove abnormal working hour data in the working hour sample that exceeds the lower threshold value and the upper threshold value of the adjusted range, and retain normal working hour data in the adjusted range, to obtain effective working hour data of the worker; an actual working hour data determination module configured to statistically analyze the effective working hour data to obtain effective actual working hour data of the worker.

15. An electronic device, comprising: One or more processors are configured to implement the working hour abnormal data identification method according to any one of claims 1 to 13.

16. A computer readable storage medium characterized by: A program is stored on the computer readable medium, and when executed by a processor, the program implements the working hour abnormal data identification method according to any one of claims 1 to 13.

17. A computer program product comprising computer programs / instructions, characterized in that, The computer program / instruction is executed by a processor to implement the working hour abnormal data identification method according to any one of claims 1 to 13.

Images