Alumni public number activity evaluation method and system
By constructing a comprehensive evaluation index system and a time-series data processing model, the problems of single indicators, inconsistent scales, and noise influence in the evaluation of alumni WeChat official account activity have been solved. This has enabled a comprehensive, accurate, and dynamic quantitative evaluation of alumni WeChat official account activity, and improved the robustness and comparability of the evaluation results.
Patent Information
- Application Number
- CN202611545119.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-08
- Publication Date
- 2026-11-03
AI Technical Summary
Existing methods for evaluating the activity of alumni WeChat official accounts suffer from problems such as a single dimension of the indicator system, mismatch in the standardization of heterogeneous indicators, inconsistent evaluation scales for different statistical periods, and insufficient utilization of historical period data, which leads to distorted evaluation results and noise.
A comprehensive evaluation index system is constructed, employing a heterogeneous data standardization model and a two-level filtering attenuation fusion model for time-series data, combined with a dynamic benchmark and a two-layer weight verification mechanism, to achieve a comprehensive, accurate, and dynamic quantitative evaluation of the activity level of alumni WeChat official accounts.
It enables a comprehensive, accurate, and dynamic quantitative assessment of the activity level of alumni WeChat official accounts, improving the robustness and universality of the assessment results. It can effectively suppress the influence of single-cycle impulse noise and capture long-term trends.
Smart Images
Figure CN123032244A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of internet data evaluation technology, specifically to a method and system for evaluating the activity level of alumni WeChat official accounts. Background Technology
[0002] WeChat official accounts are a mainstream content distribution and user operation platform in the mobile internet era. Leveraging the social network ecosystem, they achieve efficient information reach and dissemination, possessing core application value in areas such as institutional brand building, user relationship maintenance, and online service implementation. Alumni official accounts, as the core online window for universities to conduct alumni work, undertake multiple functions including information dissemination, event notification, emotional connection, and resource matching. Their operational quality directly determines the coverage and service depth of alumni work and serves as an important benchmark for measuring the effectiveness of university alumni work. Conducting scientific and quantitative evaluation of the activity level of alumni official accounts can intuitively present the current operational status, identify operational shortcomings, provide data support for iterative optimization of operational strategies, and has significant practical implications for improving the operational efficiency of alumni organizations.
[0003] However, existing methods for evaluating the activity level of WeChat official accounts have the following technical flaws:
[0004] (i) Limited Indicator Dimensions: Existing methods are mostly limited to 2 to 3 basic traffic indicators such as readership and number of posts, and use a simple linear weighting method for comprehensive scoring. This evaluation model has not established a comprehensive indicator system covering the entire chain of content production, dissemination, and interaction, and cannot fully reflect the true operational level of the public account.
[0005] (ii) Mismatch in the standardization of heterogeneous indicators: WeChat official account operation data exhibits a typical long-tail distribution characteristic—the readership and sharing of most articles are concentrated at a low level, while a few viral articles generate extremely high values far exceeding the mean. Existing technologies generally use Min-Max linear normalization or Z-score standardization to process heterogeneous indicators. Linear normalization is sensitive to extreme values, and a single viral article can dominate the overall evaluation results, leading to distorted evaluation scores that cannot truly reflect the normalized operation level of the official account.
[0006] (iii) Inconsistent evaluation scales for different statistical periods: The evaluation of alumni WeChat official account operations may be conducted at three different time granularities: daily, monthly, and yearly. The different number of natural days included in each period leads to huge differences in the magnitude of cumulative indicators (such as total number of reads). Existing technology lacks a unified scale conversion mechanism for periodic data of different granularities, making it impossible to achieve horizontal comparability of evaluation results.
[0007] (iv) Insufficient utilization of historical periodic data and lack of engineered anti-interference mechanisms: Existing methods conduct independent calculations based on fixed statistical periods, resulting in a lack of organic correlation between evaluation results from different periods. This fails to reflect the continuous impact of historical operational performance on current activity levels and makes it difficult to capture dynamic trends in activity. Furthermore, when introducing historical data for time-series fusion, existing methods fail to effectively address the technical challenge of single-period abnormal fluctuations (such as temporary hiatus or a single viral article) continuously contaminating subsequent multi-period evaluation results through time-series weighting mechanisms. Summary of the Invention
[0008] Purpose of the Invention: To address the problems of existing alumni WeChat official account activity evaluation index systems, such as single-dimensionality, mismatched data distribution, inconsistent evaluation scales, and noise influence, this invention provides a method for evaluating the activity of alumni WeChat official accounts. Considering the long-tail distribution characteristics and multi-statistical-period time-series data of alumni WeChat official account operation data, the evaluation problem is reconstructed as the problem of extracting and fusing anti-interference features of time-series data. By constructing a full-dimensional evaluation index system, combined with unified scale processing based on a dynamic benchmark surface and a two-level filtering time-series fusion mechanism, a comprehensive, accurate, and dynamic quantitative evaluation of the activity of alumni WeChat official accounts can be achieved.
[0009] Technical solution: To achieve the above objectives, the technical solution adopted by this invention is as follows:
[0010] A method for evaluating the activity level of an alumni WeChat official account includes the following steps:
[0011] Step S1: Construct activity evaluation metrics and statistical periods covering multiple operational dimensions.
[0012] Step S2: Collect the operational data of the target alumni WeChat official account to be evaluated, and preprocess it according to the statistical period to obtain the preprocessed operational data and statistical period coefficient.
[0013] Step S3: Determine the theoretical daily baseline values for each dimension of the activity evaluation index based on the constructed activity evaluation index. Extract the actual values of each dimension of the index based on the constructed activity evaluation index and preprocessed operational data.
[0014] Step S4: Obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the real values of each dimension indicator through a heterogeneous data standardization model.
[0015] Step S5: Obtain the final activity index by using a two-stage filtering and attenuation fusion model of time series data to obtain the standard score.
[0016] Step S6: Match the activity level corresponding to the final activity index according to the level classification rules, and output the evaluation result.
[0017] Preferably, the formula for the filter attenuation fusion model is as follows:
[0018] ;
[0019] ; ;
[0020] ;
[0021] ;
[0022] in, Indicates the current period The final activity index, This is the historical influence weighting coefficient, used to control the overall contribution of all historical period data to the final result. This represents the scaling factor of the activity index benchmark. Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Weight coefficients for each dimension Indicates the first Standard scores for each dimension Indicates the distance from the current period The interval number, , This represents the total number over the historical statistical period. This is the time decay factor, used to control the decay rate of historical data as the time interval increases. Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. This is the fluctuation smoothing coefficient. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index for historical cycles at intervals.
[0023] Preferably, the first The weight coefficients for each dimension are determined using a two-layer weight verification model, which is as follows:
[0024] ;
[0025] in, For the first Weight coefficients for each dimension For expert weight fusion coefficient, Indicates the first The coefficient of variation of each dimension indicator Indicates the first The coefficient of variation of each dimension indicator Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Expert weights for each dimension indicator, Indicates the first Expert weights for each dimension indicator.
[0026] Preferably, the formula for the heterogeneous data standardization model is as follows:
[0027] ;
[0028] in, For the first Standard scores for each dimension Indicates the index dimension number. , Indicates the number of indicator dimensions. For the first The true values of indicators in each dimension, For statistical periodicity coefficients, This is the theoretical baseline value for a single day.
[0029] Preferably, the grading rules are as follows:
[0030] ;
[0031] in, Indicates activity level. Indicates the highest level. Indicates the second highest level. Indicates intermediate level. Indicates the second lowest level. This indicates the lowest level.
[0032] Preferably, the activity evaluation indicators include total number of reads, total number of posts, average number of reads, highest number of reads, total number of reads on Toutiao, total number of likes, and total number of views and shares.
[0033] Preferably, the daily theoretical benchmark value is generated by extracting the median of each dimension indicator from the statistical distribution results of the operation data of alumni WeChat official accounts across the entire industry.
[0034] Preferably, the preprocessing method includes the following steps:
[0035] The system retrieves the original operational data of the target WeChat Official Account within a specified time range through the WeChat Official Account data interface, performs deduplication verification on the original data, and removes abnormal data that exceeds the reasonable value range.
[0036] For missing indicator data, the mean interpolation method of adjacent periods is used to complete the data, and the derived indicator values are calculated based on the completed basic data.
[0037] Identify the type of the current statistical period, match the pre-stored mapping relationship between the statistical period type and the statistical period coefficient, and output the corresponding statistical period coefficient.
[0038] Preferably, the daily theoretical baseline value, weighting coefficient, and coefficient of variation are updated periodically.
[0039] Another objective of this invention is to provide an alumni WeChat official account activity evaluation system for implementing an alumni WeChat official account activity evaluation method. This system includes a data collection and preprocessing module, an indicator and statistical period coefficient module, a daily theoretical baseline and actual value module, a heterogeneous data standardization model module, a filtering attenuation and fusion module, a level classification module, and an output module, wherein:
[0040] The data collection and preprocessing module is used to collect the operational data of the target alumni WeChat official account to be evaluated, and to preprocess the data according to the statistical period to obtain the preprocessed operational data and the statistical period coefficient.
[0041] The indicator and statistical period coefficient module is used to construct activity evaluation indicators and statistical periods that cover multiple operational dimensions.
[0042] The daily theoretical baseline and actual value module is used to determine the daily theoretical baseline values of each dimension indicator based on the constructed activity evaluation index. It is also used to extract the actual values of each dimension indicator based on the constructed activity evaluation index and preprocessed operational data.
[0043] The heterogeneous data standardization model module is used to obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the real values of each dimension indicator through the heterogeneous data standardization model.
[0044] The filtering attenuation fusion module is used to obtain the final activity index by filtering and attenuating the standard score through a two-stage filtering of time-series data.
[0045] The rating module is used to match the activity level corresponding to the final activity index according to the rating rules, as the evaluation result.
[0046] The output module is used to output the evaluation results.
[0047] Compared with the prior art, the present invention has the following advantages:
[0048] 1. Full-chain operation coverage: By constructing a seven-dimensional evaluation index system covering the entire chain of content production, dissemination, and interaction, we can achieve a comprehensive portrayal of the active performance of alumni WeChat official accounts and avoid evaluation bias caused by a single index.
[0049] 2. Standardization of heterogeneous indicators with strong anti-interference capabilities: By adopting a heterogeneous data standardization model to standardize heterogeneous indicators, the extreme value interference of long-tailed distributed indicators is effectively compressed, and the numerical distribution characteristics of different indicators are adapted. Combined with the dynamic benchmark mechanism constructed by the periodic coefficient, the evaluation results of daily, monthly and yearly time granularities are horizontally comparable, and the robustness and universality of the evaluation results are improved.
[0050] 3. Filtering Attenuation Fusion Model: By constructing a filtering attenuation fusion model based on two-level filtering of time series data, while establishing an organic correlation between different statistical periods, it effectively suppresses the technical defect of single-period impulse noise being continuously propagated through the attenuation mechanism. This not only smooths out the impact of short-term abnormal fluctuations but also robustly captures the long-term trend of activity.
[0051] 4. Dual-layer verification weight configuration: By combining objective weighting with expert calibration through the coefficient of variation, a dual-layer verification mechanism is established to maximize data differentiation and ensure consistency with industry understanding. Multiple sets of weight coefficients are stored synchronously to adapt to different evaluation scenarios, avoiding the one-sidedness of a single weighting method.
[0052] In summary, this invention addresses the long-tail, multi-granularity, and high-variance time-series characteristics of alumni WeChat official account operation data by establishing a tightly coupled data processing architecture of "dynamic benchmark surface - unified scale - dual-layer weighting - filtering attenuation fusion". Ultimately, it achieves dynamic, objective, and accurate quantitative evaluation of the activity level of alumni WeChat official accounts, providing reliable data support for optimizing alumni WeChat official account operations. Attached Figure Description
[0053] Figure 1 A flowchart for a method to evaluate the activity level of an alumni WeChat official account;
[0054] Figure 2 This is a schematic diagram of the structure of an alumni WeChat official account activity evaluation system. Detailed Implementation
[0055] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these examples are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.
[0056] Example 1
[0057] To address the problems of existing alumni WeChat official account activity evaluation index systems, such as limited dimensions, mismatched data distribution, inconsistent evaluation scales, and noise interference, this embodiment provides a method for evaluating alumni WeChat official account activity, such as... Figure 1 As shown, it includes the following steps:
[0058] Step S1: Construct activity evaluation metrics and statistical periods covering multiple operational dimensions.
[0059] In some examples of this invention, the activity evaluation metrics include total reads, total posts, average reads, highest read count, total read count for top articles, total likes, and total views / shares. A complete evaluation system is constructed using seven core metrics, each corresponding to different dimensions of WeChat Official Account operation, comprehensively covering the three core areas of content production, content dissemination, and user interaction.
[0060] The first metric is the total number of reads, which corresponds to the breadth of content dissemination. It refers to the total number of reads of all articles published by the official account within the statistical period, and directly reflects the overall coverage and reach of the official account's content.
[0061] The second metric is the total number of articles published, which corresponds to the content production activity dimension. It refers to the total number of articles published by the official account within the statistical period, reflecting the content production frequency and update rhythm of the operations team.
[0062] The third indicator is the average number of reads, which corresponds to the dimensions of content quality and audience engagement. It refers to the average number of reads for a single article within the statistical period, calculated by dividing the total number of reads by the total number of articles published. It reflects the continued appeal of the public account's content to users and the stability of the audience.
[0063] The fourth indicator is the highest number of reads, which corresponds to the dimension of viral content production capability. It refers to the highest number of reads achieved by a single article within the statistical period, reflecting the potential and peak dissemination effect of the public account in producing highly disseminated content.
[0064] The fifth indicator is the total number of reads on Toutiao, which corresponds to the core position dissemination effect dimension. It refers to the total number of reads of articles published in all Toutiao positions within the statistical period, reflecting the utilization efficiency and dissemination value of the core exposure positions of the official account.
[0065] The sixth metric is the total number of likes, which corresponds to the user recognition and interaction dimensions. It refers to the total number of likes received by all articles within the statistical period, reflecting the user's degree of recognition of the content and willingness to engage in superficial interaction.
[0066] The seventh metric is the total number of views and shares, which corresponds to the user's deep interaction and dissemination dimension. It refers to the sum of the number of views and shares of all articles within the statistical period, reflecting users' deep recognition of the content and their willingness to spread it again. It is a core indicator for measuring the content's ability to spread and spread virally.
[0067] Once the activity evaluation indicators are established, they will provide a unified standard for subsequent data collection and indicator calculation.
[0068] Step S2: Collect the operational data of the target alumni WeChat official account to be evaluated, and preprocess it according to the statistical period to obtain the preprocessed operational data and statistical period coefficient.
[0069] In some examples of the present invention, the preprocessing method includes the following steps:
[0070] Step S201: First, connect to the official data interface of the WeChat Official Accounts Platform. Retrieve the original operational data of the target Official Account within a specified time range through the Official Accounts data interface. The specified time range includes the current statistical period and a preset number of consecutive historical statistical periods. In some examples of this invention, the current year and three consecutive historical years are selected as the statistical period, covering operational data for four complete years, providing a sufficient data foundation for subsequent historical activity fusion. The retrieved original data fields include the publication time, number of reads, number of likes, number of views, number of shares, and publication location information for a single article. All data is sorted and stored according to timestamps. Deduplication verification is performed on the original data to remove abnormal data exceeding a reasonable value range. Deduplication verification refers to deduplicating read, like, and other data that are repeatedly counted for the same article, ensuring that each operational data point corresponds to only one valid count. Then, outlier identification and removal are performed. The interquartile range method is used to determine outlier data; values exceeding three times the interquartile range are identified as outliers and directly removed from the dataset to avoid extreme outlier data interfering with subsequent evaluation results.
[0071] Step S202: For missing indicator data, the mean interpolation method of adjacent periods is used to complete the missing data, and the derived indicator values are calculated based on the completed basic data.
[0072] The method of interpolating the mean of adjacent periods is adopted: for missing data at the monthly granularity, the arithmetic mean of the corresponding indicators of the two months before and after the missing month is taken as the imputed value; for missing data at the daily granularity, the arithmetic mean of the corresponding indicators of the two days before and after the missing date is taken as the imputed value.
[0073] The derived indicator values are calculated based on the basic data. The average number of reads is obtained by dividing the total number of reads by the total number of articles published within the statistical period. The total number of reads and shares is obtained by adding the number of views and shares, forming a complete original dataset of seven indicators.
[0074] Step S203: Identify the type of the current statistical period, match the pre-stored mapping relationship between the statistical period type and the statistical period coefficient, and output the corresponding statistical period coefficient.
[0075] In some examples of this invention, the type of the current statistical period is identified, a pre-stored mapping relationship between the statistical period type and the statistical period coefficient is matched, and the corresponding statistical period coefficient is output. The statistical period is annual, and the corresponding statistical period coefficient is 365. A unified conversion standard for the statistical period coefficient can eliminate evaluation scale bias caused by different time granularities, enabling horizontal comparability of daily, monthly, and annual periodic data.
[0076] Step S3: Determine the theoretical daily baseline values for each dimension of the activity evaluation index based on the constructed activity evaluation index. Extract the actual values of each dimension of the index based on the constructed activity evaluation index and preprocessed operational data.
[0077] In some examples of this invention, the daily theoretical benchmark value is generated by extracting the median of each dimension indicator from the statistical distribution results of the operational data of alumni WeChat official accounts across the entire industry. The specific steps are as follows:
[0078] First, the pre-stored statistical distribution results of the operation data of alumni WeChat official accounts across the entire industry are retrieved, and the theoretical daily benchmark values corresponding to each dimension indicator are extracted. The theoretical daily benchmark values are generated by taking the median of the operation data of 1200 selected alumni WeChat official accounts nationwide over 12 consecutive months. This avoids the mean deviation problem caused by top-performing accounts and accounts that have been inactive for a long time, ensuring that the benchmark values align with the normal operation level of the vast majority of ordinary alumni WeChat official accounts. Using the median instead of the mean avoids the mean deviation problem caused by top-performing accounts and accounts that have been inactive for a long time, ensuring that the benchmark values align with the normal operation level of the vast majority of ordinary alumni WeChat official accounts. In some examples of this invention, the daily benchmark value for reading is 100,000, the daily benchmark value for posting is 8, the daily benchmark value for likes is 10,000, and the daily benchmark value for views and shares is 5,000.
[0079] Based on the obtained preprocessed operational data, the values of the corresponding indicators are extracted from the preprocessed operational data (the original dataset of seven indicators) according to the activity evaluation indicators and the statistical period, and used as the true values of each dimension indicator.
[0080] Step S4: Obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the real values of each dimension indicator through a heterogeneous data standardization model.
[0081] In some examples of this invention, the heterogeneous data standardization model formula is as follows:
[0082] ;
[0083] in, For the first Standard scores for each dimension Indicates the index dimension number. , Indicates the number of indicator dimensions. For the first The true values of indicators in each dimension, For statistical periodicity coefficients, This is the theoretical baseline value for a single day.
[0084] Indicators such as WeChat Official Account readership and sharing numbers follow an approximate log-normal distribution—most samples are concentrated in the low-value range, while a few samples are scattered in the extremely high-value range, forming a long tail. The heterogeneous data standardization model of this invention can compress this long-tail distribution into a narrower numerical range, preventing a single viral article from excessively inflating the overall evaluation score, and making the evaluation results more reflective of the normalized operation level of the WeChat Official Account rather than occasional viral events.
[0085] Step S5: Obtain the final activity index by using a two-stage filtering and attenuation fusion model of time series data to obtain the standard score.
[0086] The method for establishing the filtering attenuation fusion model is as follows:
[0087] Step S501, Calculation of the original activity index for the current period:
[0088] The formula for calculating the current period's original activity index is as follows:
[0089]
[0090] in, Indicates the current period Original Activity Index This is the time index for the current statistical period. This represents the scaling factor of the activity index benchmark. Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Weight coefficients for each dimension Indicates the first Standard scores for each dimension.
[0091] The benchmark scaling factor for the activity index can be specified directly, or it can be calculated by collecting a sample of alumni WeChat official account operations across the entire industry (e.g., 1200 universities for 12 consecutive months), calculating the weighted standardized score of the sample, and using the median of the weighted standardized score as the benchmark scaling factor for the activity index.
[0092] Step S502, the The weight coefficients for each dimension are determined using a two-layer weight verification model.
[0093] The construction method of the two-layer weighted verification model is as follows:
[0094] Objective weighting of coefficient of variation:
[0095]
[0096] in, This indicates that the coefficient of variation is objectively weighted. Indicates the first The coefficient of variation of each dimension indicator Indicates the first The coefficient of variation of each dimension indicator Indicates the index dimension number. , Indicates the number of indicator dimensions.
[0097] Expert calibration method:
[0098]
[0099] in, Indicates expert weighting. Indicates the first Expert weights for each dimension indicator, Indicates the first Expert weights for each dimension indicator.
[0100] In some examples of this invention, the method for determining expert weights is as follows: Ten experts, including those in charge of alumni affairs at universities and experts in new media operations, are invited to independently score the importance of the seven dimensions on a 100-point scale. The average of all expert scores is taken to obtain the expert score for each dimension. After normalizing the expert scores, the expert weights are obtained. Finally, the objective weights and expert weights are weighted and merged proportionally to obtain the final set of weight coefficients. The expert weight fusion coefficient is... Take 0.4.
[0101] In some examples of this invention, the weights for expert total read count are 0.25, total number of articles are 0.10, average read count is 0.20, highest read count is 0.10, total number of top-reads is 0.15, total number of likes is 0.10, and total number of views and shares is 0.10. The system synchronously stores multiple sets of weight coefficients adapted to different evaluation scenarios, which can be flexibly switched according to the evaluation focus.
[0102] The two-layer weighted verification model is then:
[0103]
[0104] in, For the first Weight coefficients for each dimension For expert weight fusion coefficient, Indicates the first The coefficient of variation of each dimension indicator Indicates the first The coefficient of variation of each dimension indicator Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Expert weights for each dimension indicator, Indicates the first Expert weights for each dimension indicator.
[0105] Method for determining the coefficient of variation: Based on the sample data of alumni WeChat official account operations across the industry, calculate the coefficient of variation for each dimension indicator. The coefficient of variation is the ratio of the sample standard deviation to the sample mean of the corresponding dimension indicator. The larger the coefficient of variation, the stronger the distinguishing ability of the indicator, and the higher the weight should be assigned.
[0106] Step S503: Perform a smoothing filter on the original activity index:
[0107] ; ;
[0108] ;
[0109] ;
[0110] in, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. This is the fluctuation smoothing coefficient. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, This represents the total number over the historical statistical period. Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index for historical cycles at intervals.
[0111] Step S504: Perform time-decay fusion on the smoothed activity index:
[0112] ;
[0113] In summary, the filtering attenuation fusion model is obtained as follows:
[0114] ;
[0115] ; ;
[0116] ;
[0117] ;
[0118] In the case of a one-sided smoothing scenario for the first and last historical cycles, the neighborhood term corresponding to the missing side in the formula is set to 0, and the denominator is adjusted to 1 to ensure the rationality of the weight calculation. Indicates the current period The final activity index, This is the historical influence weighting coefficient, used to control the overall contribution of all historical period data to the final result. This represents the scaling factor of the activity index benchmark. Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Weight coefficients for each dimension Indicates the first Standard scores for each dimension Indicates the distance from the current period The interval number, , This represents the total number over the historical statistical period. This is the time decay factor, used to control the decay rate of historical data as the time interval increases. Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. This is the fluctuation smoothing coefficient. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index for historical cycles at intervals.
[0119] The initial attenuation weights for each historical period are calculated based on the attenuation coefficient. All initial attenuation weights are then normalized to obtain standardized attenuation weights that sum to 1. This ensures that changes in the number of historical periods do not alter the overall contribution magnitude of the historical items, improving the parameter's universality and adaptability. Historical periods with longer time intervals correspond to lower attenuation weights, reflecting the pattern of greater impact in the near term and less impact in the long term.
[0120] In some examples of this invention, for scenarios involving annual long-term trend assessment, the attenuation coefficient... A value of 0.2 is relatively small, indicating a slower decay rate of historical data in the long term, thus preserving more long-term operational characteristics; historical influence weighting coefficient. A value of 0.3 is used to control the overall contribution of all historical period data to the final result, reflecting both the continuity of historical operations and using the current period's performance as the core of the evaluation; fluctuation smoothing coefficient. A value of 0.7 is used to moderately suppress the transmission effect of extreme fluctuations in a single period while preserving the authenticity of the original periodic data. For monthly short-term operational assessment scenarios, the attenuation coefficient can be adjusted to 0.4 and the fluctuation smoothing coefficient to 0.8 to accelerate the attenuation of historical data, reduce the smoothing intensity, and highlight the weight of recent operational performance.
[0121] During the calculation, the activity index of each historical period is first sorted according to its distance from the current period, and the ordinal number of the time interval corresponding to each historical period is determined. The ordinal number of the historical period closest to the current period is 1, and the ordinal number of the farthest historical period is 1. In this embodiment The value is 3.
[0122] For the most recent historical period with ordinal number 1, only the data from its more distant adjacent historical periods are used for one-sided smoothing calculation; for ordinal numbers... For the furthest historical period, only the data from the adjacent historical period on its closest side is used for one-sided smoothing calculation; for historical periods in the middle, data from the adjacent historical periods on both sides are used for two-sided smoothing calculation. The smoothing process balances the authenticity of the original data with the strength of anomaly suppression through the fluctuation smoothing coefficient. This can effectively avoid extreme situations such as a sudden increase in popular content or temporary suspension of updates within a single period from continuously interfering with the evaluation results of subsequent multiple periods through the decay mechanism, thereby improving the robustness of the evaluation.
[0123] The filtering attenuation fusion model of this invention, through the organic correlation between different statistical periods and the combination of heterogeneous data standardization model, can not only smooth the impact of short-term abnormal fluctuations on the evaluation results, but also accurately capture the long-term trend of activity, making the evaluation results more in line with the real state of public account operation, and significantly improving the robustness and horizontal comparability of the evaluation results.
[0124] Step S6: Match the activity level corresponding to the final activity index according to the level classification rules, and output the evaluation result.
[0125] In some examples of this invention, the hierarchical classification rules are as follows:
[0126] ;
[0127] in, Indicates activity level. Indicates the highest level. Indicates the second highest level. Indicates intermediate level. Indicates the second lowest level. This indicates the lowest level.
[0128] Five activity levels are set up using preset grading rules, each corresponding to a continuous activity index range. The endpoints of these ranges are based on industry benchmark data. The highest level corresponds to an activity index of at least 1200; the second highest level corresponds to an activity index between 1000 and 1200; the middle level corresponds to an activity index between 800 and 1000; the second lowest level corresponds to an activity index between 600 and 800; and the lowest level corresponds to an activity index below 600. The boundary values for these five levels are based on the quantile of the activity index of alumni WeChat official accounts across all universities in China. The grading results can be directly used for quantitative scoring in the annual performance evaluation of alumni work at universities.
[0129] In some examples of this invention, the daily theoretical baseline value, weighting coefficient, and coefficient of variation are updated periodically. The update process is based on statistical analysis of newly added sample data from the operation of alumni WeChat official accounts across the industry, and the parameter values are adjusted in conjunction with the scoring and calibration results of domain experts. The adjusted parameters are stored, overriding the original parameters, and are applied to all subsequent evaluation calculations.
[0130] Example 2
[0131] To address the problems of existing alumni WeChat official account activity evaluation index systems, such as limited dimensions, mismatched data distribution, inconsistent evaluation scales, and noise interference, this embodiment provides an alumni WeChat official account activity evaluation system for implementing alumni WeChat official account activity evaluation methods, such as... Figure 2 As shown, it includes a data acquisition and preprocessing module, an indicator and statistical period coefficient module, a daily theoretical baseline and actual value module, a heterogeneous data standardization model module, a filtering attenuation and fusion module, a level classification module, and an output module, wherein:
[0132] The data collection and preprocessing module is used to collect the operational data of the target alumni WeChat official account to be evaluated, and to preprocess the data according to the statistical period to obtain the preprocessed operational data and the statistical period coefficient.
[0133] The indicator and statistical period coefficient module is used to construct activity evaluation indicators and statistical periods that cover multiple operational dimensions.
[0134] The daily theoretical baseline and actual value module is used to determine the daily theoretical baseline values of each dimension indicator based on the constructed activity evaluation index. It is also used to extract the actual values of each dimension indicator based on the constructed activity evaluation index and preprocessed operational data.
[0135] The heterogeneous data standardization model module is used to obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the real values of each dimension indicator through the heterogeneous data standardization model.
[0136] The filtering attenuation fusion module is used to obtain the final activity index by filtering and attenuating the standard score through a two-stage filtering of time-series data.
[0137] The rating module is used to match the activity level corresponding to the final activity index according to the rating rules, as the evaluation result.
[0138] The output module is used to output the evaluation results.
[0139] In summary, this invention, by fully executing the aforementioned evaluation process, completes a quantitative assessment of the annual activity of target university alumni WeChat official accounts. The entire calculation process is logically clear, the parameter determination method is scientific and reproducible, and the evaluation results comprehensively reflect the multi-dimensional operational performance of the official account in the current period while also demonstrating the continuous impact of historical operational status, effectively avoiding the one-sidedness and interference of extreme values in single-period evaluations. The practical application of this embodiment verifies the feasibility and applicability of the technical solution of this invention in real-world scenarios, providing reliable data support for adjusting the operational strategies of university alumni WeChat official accounts, and has practical application value for improving the operational efficiency of alumni work.
[0140] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.
Claims
1. A method for evaluating the activity level of an alumni WeChat official account, characterized in that, Includes the following steps: Step S1: Construct activity evaluation indicators and statistical periods covering multiple operational dimensions; Step S2: Collect the operational data of the target alumni WeChat official account to be evaluated, and preprocess it according to the statistical period to obtain the preprocessed operational data and statistical period coefficient; Step S3: Determine the theoretical daily baseline values of each dimension indicator based on the constructed activity evaluation index; extract the actual values of each dimension indicator based on the constructed activity evaluation index and preprocessed operational data. Step S4: Obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the actual values of each dimension indicator through a heterogeneous data standardization model. Step S5: Obtain the final activity index by applying the standard score to a two-stage filtering and attenuation fusion model of time-series data; Step S6: Match the activity level corresponding to the final activity index according to the level classification rules, and output the evaluation result.
2. The method for evaluating the activity level of alumni WeChat official accounts according to claim 1, characterized in that: The formula for the filtering attenuation fusion model is as follows: ； ； ； ； ； in, Indicates the current period The final activity index, Historical influence weighting coefficient, This represents the scaling factor of the activity index benchmark. Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Weight coefficients for each dimension Indicates the first Standard scores for each dimension Indicates the distance from the current period The interval number, , This represents the total number over the historical statistical period. The time decay factor, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. This is the fluctuation smoothing coefficient. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates to The smoothed activity index obtained after performing neighborhood sliding smoothing. Indicates the distance from the current period The The original activity index of historical cycles at intervals, Indicates the distance from the current period The The original activity index for historical cycles at intervals.
3. The method for evaluating the activity level of alumni WeChat official accounts according to claim 2, characterized in that: The first The weight coefficients for each dimension are determined using a two-layer weight verification model, which is as follows: ； in, For the first Weight coefficients for each dimension For expert weight fusion coefficient, Indicates the first The coefficient of variation of each dimension indicator Indicates the first The coefficient of variation of each dimension indicator Indicates the index dimension number. , Indicates the number of indicator dimensions. Indicates the first Expert weights for each dimension indicator, Indicates the first Expert weights for each dimension indicator.
4. The method for evaluating the activity level of alumni WeChat official accounts according to claim 3, characterized in that: The formula for the heterogeneous data standardization model is as follows: ； in, For the first Standard scores for each dimension Indicates the index dimension number. , Indicates the number of indicator dimensions. For the first The true values of indicators in each dimension, For statistical periodicity coefficients, This is the theoretical baseline value for a single day.
5. The method for evaluating the activity level of alumni WeChat official accounts according to claim 4, characterized in that: The ranking rules are as follows: ； in, Indicates activity level. Indicates the highest level. Indicates the second highest level. Indicates intermediate level. Indicates the second lowest level. This indicates the lowest level.
6. The method for evaluating the activity level of alumni WeChat official accounts according to any one of claims 1-5, characterized in that: The activity evaluation metrics include total reads, total posts, average reads, highest read count, total read count for top articles, total likes, and total views and shares.
7. The method for evaluating the activity level of alumni WeChat official accounts according to claim 6, characterized in that: The theoretical baseline value for a single day is generated by extracting the median of each dimension indicator from the statistical distribution results of the operation data of alumni WeChat official accounts across the entire industry.
8. The method for evaluating the activity level of alumni WeChat official accounts according to claim 7, characterized in that: The preprocessing method includes the following steps: The system retrieves the original operational data of the target WeChat Official Account within a specified time range through the WeChat Official Account data interface, performs deduplication verification on the original data, and removes abnormal data that exceeds the reasonable value range. For missing indicator data, the mean interpolation method of adjacent periods is used to complete the data, and the derived indicator values are calculated based on the completed basic data. Identify the type of the current statistical period, match the pre-stored mapping relationship between the statistical period type and the statistical period coefficient, and output the corresponding statistical period coefficient.
9. The method for evaluating the activity level of alumni WeChat official accounts according to claim 8, characterized in that: The daily theoretical baseline value, weighting coefficient, and coefficient of variation are updated regularly.
10. An evaluation system based on the alumni WeChat official account activity evaluation method according to claim 1, characterized in that: It includes modules for indicators and statistical period coefficients, data acquisition and preprocessing, daily theoretical benchmark and actual values, heterogeneous data standardization model, filtering attenuation and fusion, level classification, and output. The indicator and statistical period coefficient module is used to construct activity evaluation indicators and statistical periods covering multiple operational dimensions; The data collection and preprocessing module is used to collect the operational data of the target alumni WeChat official account to be evaluated, and to preprocess the data according to the statistical period to obtain the preprocessed operational data and the statistical period coefficient. The daily theoretical baseline and actual value module is used to determine the daily theoretical baseline value of each dimension indicator based on the constructed activity evaluation index; and to extract the actual value of each dimension indicator based on the constructed activity evaluation index and preprocessed operational data. The heterogeneous data standardization model module is used to obtain the standard scores of each dimension indicator by using the daily theoretical baseline value, statistical period coefficient, and the real values of each dimension indicator through the heterogeneous data standardization model. The filtering attenuation and fusion module is used to obtain the final activity index by filtering and attenuating the standard score through a two-stage filtering and fusion model of time-series data. The rating module is used to match the activity level corresponding to the final activity index according to the rating rules, as the evaluation result. The output module is used to output the evaluation results.