A method, system and computer storage medium for monitoring the effectiveness of technology transfer

Abstract

This application provides a method, system, and computer storage medium for monitoring the effectiveness of technology transfer. The method includes: acquiring raw business data from a target unit for calculating the effectiveness of technology transfer in the current statistical period; performing time consistency processing on historical business data to generate time-corrected feature data corresponding to the current statistical period; based on the time-corrected feature data and business data within the current statistical period, determining multi-dimensional effectiveness feature components reflecting revenue performance, technological attributes, and promotion and diffusion, and constructing an initial multi-dimensional effectiveness feature vector; performing scale adjustment processing on each feature component of the initial multi-dimensional effectiveness feature vector to obtain a target multi-dimensional effectiveness feature vector for the current statistical period; and performing correlation correction and fusion operations on the target multi-dimensional effectiveness feature vector corresponding to the current statistical period to generate the effectiveness value of technology transfer in the current statistical period. This application achieves automated, stable, and reproducible monitoring and calculation of the effectiveness of technology transfer.

Description

Technical Field

[0001] This application belongs to the field of data processing and intelligent monitoring technology, and in particular relates to a method, system and computer storage medium for monitoring the effectiveness of achievement transformation. Background Technology

[0002] In the process of enterprise technology transfer management, it is necessary to conduct quantitative statistics and comparative analysis of the performance of different target units in technology transfer within a statistical period to support resource investment, technology promotion, and business decision-making. Since technology transfer-related data are distributed across multiple business systems or data sources, such as finance, project management, scientific research management, intellectual property, and marketing, and the data structures and statistical definitions differ, performance statistics and analysis are often completed manually, using spreadsheets, periodic reports, or offline calculations after exporting data from various systems. In engineering implementation, this type of approach represents a typical periodic processing task involving multi-source heterogeneous data, encompassing cross-system data collection, field matching, definition alignment, and result output.

[0003] In the aforementioned statistical and analytical process, influenced by the objective laws governing technology transfer activities, data on input, application, and returns exhibit time lags or cross-cycle continuity, and the business scale and data volume vary significantly among different organizational units. Current practices typically rely on manually setting statistical definitions and calculation rules. When cross-cycle data is involved in calculations, the data volume spans significantly, and the data distribution is unstable, problems such as sensitivity to data point-in-time deviations and insufficient comparability across units can easily arise. Furthermore, multidimensional indicators may have coupling or correlation relationships; using only simple combination calculations can easily lead to duplicate measurements or amplified biases, making the final results highly susceptible to fluctuations in local indicators. In addition, offline aggregation and manual processing can easily introduce data quality issues such as omissions, errors, and duplicate statistics, further affecting the stability and reliability of the results. Summary of the Invention

[0004] Based on this, it is necessary to address the aforementioned technical problems. This application provides a method for monitoring the effectiveness of technology transfer, comprising the following steps:

[0005] Obtain the original business data of the target unit for calculating the achievement transformation efficiency of the current statistical period. The original business data includes business data within the current statistical period and historical business data earlier than the current statistical period.

[0006] The historical business data is processed for time consistency to generate time-corrected feature data corresponding to the current statistical period;

[0007] Based on the time-corrected feature data and the business data within the current statistical period, multi-dimensional performance feature components reflecting revenue performance, technical attributes, and promotion and diffusion are determined, and an initial multi-dimensional performance feature vector is constructed.

[0008] The scale adjustment process is performed on each feature component of the initial multidimensional performance feature vector to obtain the target multidimensional performance feature vector for the current statistical period.

[0009] A correlation correction and fusion operation is performed on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period.

[0010] Optionally, the step of performing correlation correction and fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value for the current statistical period includes the following steps:

[0011] Obtain the set of target multidimensional performance feature vectors corresponding to historical statistical periods;

[0012] The correlation correction parameters are determined based on the target multidimensional performance feature vector set;

[0013] The correlation correction parameter is used to perform a fusion operation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period.

[0014] Optionally, determining the correlation correction parameters based on the target multidimensional performance feature vector set includes the following steps:

[0015] Construct the covariance matrix between each feature component based on the target multidimensional performance feature vector set corresponding to the historical statistical period;

[0016] The correlation correction parameters are obtained by performing an inverse matrix operation on the covariance matrix.

[0017] Optionally, the step of performing an inverse matrix operation on the covariance matrix to obtain the correlation correction parameter further includes the following steps:

[0018] Obtain the dimension information of the covariance matrix;

[0019] Construct an identity matrix I with the same dimensions as the covariance matrix;

[0020] The covariance matrix is ​​diagonally regularized according to the following formula to obtain the modified covariance matrix ∑'.

[0021] ∑'=∑+βI,

[0022] Where ∑ is the covariance matrix and β is a regularization coefficient greater than zero;

[0023] The correlation correction parameters are obtained by performing an inverse matrix operation on the modified covariance matrix ∑'.

[0024] Optionally, the step of using the correlation correction parameter to perform a fusion operation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value for the current statistical period includes the following steps:

[0025] The target multidimensional efficiency feature vector corresponding to the current statistical period is multiplied with the correlation correction parameter to obtain the correction feature vector;

[0026] The result transformation efficiency value for the current statistical period is obtained by performing a vector inner product operation between the corrected feature vector and the target multidimensional efficiency feature vector corresponding to the current statistical period.

[0027] Optionally, the step of performing time consistency processing on the historical business data to generate time-corrected feature data corresponding to the current statistical period includes the following steps:

[0028] Obtain the statistical period to which the historical business data belongs;

[0029] Determine the time difference Δt between the statistical period to which the historical business data belongs and the current statistical period;

[0030] The historical business data is subjected to time decay weighting processing according to the following formula to obtain the time-weighted value. ,

[0031] ,

[0032] Where λ is the time decay coefficient;

[0033] The time-corrected feature data is obtained by multiplying each historical business data with its corresponding time-weighted value and then summing the results.

[0034] Optionally, the step of scaling each feature component of the initial multidimensional performance feature vector to obtain the target multidimensional performance feature vector for the current statistical period includes the following steps:

[0035] Obtain each feature component x in the initial multidimensional performance feature vector. i ;

[0036] For each characteristic component value x i The adjusted feature component x is obtained by performing nonlinear compression according to the following formula. i ',

[0037] x i '=xi / (1+αx i ), where α is a scaling factor greater than zero;

[0038] The adjusted feature component x i The target multidimensional performance feature vector is formed by combining the features in the same dimensional order as the initial multidimensional performance feature vector.

[0039] Optionally, the method for monitoring the effectiveness of technology transfer further includes the following steps:

[0040] The change in effectiveness is calculated based on the effectiveness values ​​of achievement transformation over multiple consecutive statistical periods;

[0041] The change in performance is compared with a preset threshold, and an abnormal warning signal is output when the change in performance exceeds the preset threshold.

[0042] In particular, the present invention also provides a technology transfer effectiveness monitoring system, comprising:

[0043] The acquisition module is used to acquire the original business data of the target unit for calculating the achievement transformation efficiency of the current statistical period. The original business data includes business data within the current statistical period and historical business data earlier than the current statistical period.

[0044] The processing module is used to perform time consistency processing on the historical business data and generate time-corrected feature data corresponding to the current statistical period;

[0045] The construction module is used to determine multi-dimensional performance feature components reflecting revenue performance, technical attributes, and promotion and diffusion based on the time-corrected feature data and the business data within the current statistical period, and to construct an initial multi-dimensional performance feature vector.

[0046] The scaling module is used to scale each feature component of the initial multidimensional performance feature vector to obtain the target multidimensional performance feature vector for the current statistical period.

[0047] The fusion module is used to perform correlation correction and fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period.

[0048] In particular, the present invention also provides a computer storage medium that can store program instructions, which, when executed by a processor, can realize the aforementioned method for monitoring the effectiveness of achievement transformation.

[0049] According to the present invention, by acquiring the target unit's business data within the current statistical period and introducing historical business data earlier than the current statistical period, and combining this with time consistency processing of historical business data to generate time-corrected feature data, historical information across statistical periods can participate in performance calculation in a feature form consistent with the time benchmark of the current statistical period. This reduces statistical bias caused by time lags in input, formation, promotion, and benefit realization, and improves the stability and comparability of the results of achievement transformation performance calculation across different statistical periods. Furthermore, based on constructing multi-dimensional performance feature components reflecting benefit performance, technical attributes, and promotion diffusion, the present invention organizes each feature component into an initial multi-dimensional performance feature vector and performs scaling processing on each component of this vector. This ensures that features of different dimensions, magnitudes, and distribution patterns can still be mapped to a numerical range suitable for fusion calculation even when they differ significantly, thereby suppressing the undesirable dominant role of high-magnitude feature components in the calculation results and improving the numerical stability when multiple features participate in the calculation. Furthermore, this invention performs correlation correction and fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period. This allows the fusion process to consider the correlation structure between multidimensional feature components, suppressing duplicate measurement and bias amplification caused by feature correlation, thereby obtaining a more robust output of achievement transformation efficiency value. Therefore, this invention can achieve automated, stable, and reproducible monitoring and calculation of achievement transformation efficiency in scenarios involving multi-source heterogeneous data, cross-period data, and multidimensional feature correlation.

[0050] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, the preferred embodiments of the present invention are described in detail below. Attached Figure Description

[0051] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0052] Figure 1 A schematic flowchart of a method for monitoring the effectiveness of technology transfer according to an embodiment of the present invention is shown;

[0053] Figure 2 It shows Figure 1 A schematic flowchart of the time consistency processing method in step S200 is shown.

[0054] Figure 3 It shows Figure 1A schematic flowchart of the scaling process in step S400 is shown.

[0055] Figure 4 It shows Figure 1 A schematic flowchart of the correlation correction fusion operation method in step S500 is shown.

[0056] Figure 5 It shows Figure 4 A schematic flowchart of the method for determining the correlation correction parameter in step S502 is shown.

[0057] Figure 6 A schematic structural diagram of a computer device according to an embodiment of the present invention is shown. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0060] Figure 1 A schematic flowchart of a method for monitoring the effectiveness of technology transfer according to an embodiment of the present invention is shown. Figure 1 As shown, the method for monitoring the effectiveness of technology transfer includes steps S100 to S500.

[0061] Step S100: Obtain the original business data of the target unit for calculating the performance of results transformation in the current statistical period. The original business data includes business data within the current statistical period and historical business data earlier than the current statistical period.

[0062] The target unit refers to an organizational or business unit included in the set of monitoring objects for the effectiveness of technology transfer. It can be uniquely identified by a preset organizational identifier, business identifier, or management dimension, and can independently collect data related to the transfer of scientific and technological achievements. This target unit can be a subsidiary, branch, business unit, project management center, regional operating unit, project group management unit, or other management unit with independent data collection capabilities. Those skilled in the art should understand that the specific level and organizational form of the target unit can be set according to the actual management structure, and this embodiment of the invention does not limit this.

[0063] The current statistical period refers to the time interval corresponding to the output performance value, which can be annual, quarterly, monthly, or other preset time periods. The current statistical period can be preset by the system, configured by the user, or issued by the upper-level business system. For example, the current statistical period can be a calendar year from January 1, 2024 to December 31, 2024, or a monthly period for the fourth quarter of 2024 or December 2024. Setting the current statistical period is used to define the time base and result attribution period for this performance calculation.

[0064] Raw business data refers to the basic data set obtained from one or more data sources that can characterize the transformation process and results of the target unit's achievements. This raw business data can be obtained through preset data interfaces, database access, data synchronization tasks, or data warehouse reading. For data from semi-structured carriers such as report files and ledger files, it can also be parsed and processed to form structured records before being included in the raw business data set. The specific implementation method of data acquisition can be determined according to the deployment environment and the form of the data source, and this embodiment of the invention does not limit it in this regard.

[0065] In this embodiment, the original business data is at least divided into business data within the current statistical period and historical business data earlier than the current statistical period.

[0066] The business data within the current statistical period is used to characterize the target unit's achievement transformation activities, output performance, and promotion and application within the current statistical period. This business data within the current statistical period can include data related to achievement transformation output, data related to achievement promotion and application, and data related to achievement attributes. Specifically, data related to achievement transformation output may include, for example, revenue data, operational contribution data, achievement transformation project execution data, and productization / engineering delivery data. Data related to achievement promotion and application may include, for example, application data of achievements in engineering projects, number of projects covered, frequency of use, and application scenario information. Data records related to achievement attributes may include, for example, achievement category, achievement status, the technical direction to which the achievement belongs, the standard level or advancement level associated with the achievement, etc.

[0067] Historical business data prior to the current statistical period is used to reflect historical accumulation, cross-period continuation, or lagging impact information related to the calculation of the effectiveness of results transformation in the current statistical period. This historical business data may include historical input data or process records related to the formation of the results, historical node information related to the life cycle of the results, and historical benchmark data used to form a comparative benchmark. Historical input or process data may include, for example, historical R&D input data, historical experimental verification data, and historical pilot / demonstration implementation data. Historical node information may include, for example, the time of project initiation, the time of results formation, historical application trajectory records, and historical benefit realization trajectory records. Historical benchmark data may include, for example, historical business scale indicators and historical project load indicators, to provide a benchmark reference for cross-period comparisons.

[0068] In one example, the current statistical period is set to 2024, and the target unit is a subsidiary. In step S100, the system retrieves the subsidiary's technology transfer-related business data for 2024 from data sources such as the financial system, research management system, project management system, and intellectual property management system. It also retrieves historical business data earlier than 2024, such as R&D investment data, project initiation and formation information, and historical promotion and application data from 2020 to 2023. In another example, the current statistical period is set to the fourth quarter of 2024. The system retrieves relevant business data for that quarter and historical business data from previous quarters.

[0069] Step S200: Perform time consistency processing on historical business data to generate time-corrected feature data corresponding to the current statistical period.

[0070] Figure 2 It shows Figure 1 The flowchart shown illustrates the time consistency processing method for step S200. Figure 2 As shown, step S200 includes steps S201 to S204.

[0071] Step S201: Obtain the statistical period to which the historical business data belongs.

[0072] Historical business data is typically stored as records in the business database, with each record corresponding to a specific business occurrence period or time period. Specifically, the statistical period information is retrieved from the metadata fields, timestamp fields, or period identifier fields of the historical business data. The statistical period information can be represented by an annual identifier, quarterly identifier, monthly identifier, or other preset statistical period identifier. For example, when the statistical period is annual, the statistical period can be represented as "2023". When the statistical period is quarterly, the statistical period can be represented as "Q2 of 2024". When the statistical period is monthly, the statistical period can be represented as "March 2024".

[0073] In one implementation, historical business data contains explicit period fields, such as "statistical year," "statistical quarter," and "statistical month," which can be directly read to obtain the corresponding statistical period. In another implementation, if historical business data only contains the business occurrence time or entry time, such as timestamps or date fields, the timestamps can be mapped to the corresponding statistical period according to preset period division rules. For example, the statistical period can be determined based on which year, quarter, or month the business occurrence date belongs to.

[0074] It should be noted that the naming, format, or encoding of fields belonging to different statistical periods may vary for historical business data from different data sources. Therefore, the system can pre-configure field mapping relationships or period parsing rules to achieve unified parsing of data from different sources. For example, fields representing the year in different systems can be uniformly mapped to internal year numbers or standardized period identifiers.

[0075] Step S202: Determine the time difference Δt between the statistical period to which the historical business data belongs and the current statistical period.

[0076] The time difference Δt is used to quantify the periodic distance between historical business data and the current statistical period under a unified time benchmark, so as to determine the contribution weight of historical business data to the performance calculation of the current statistical period based on this periodic distance. The time difference Δt can be understood as the span of the statistical period to which the historical business data belongs relative to the current statistical period, and its value is generally a non-negative integer.

[0077] Specifically, step S202 may include: obtaining the period identifier of the current statistical period, converting the statistical period identifier of the historical business data and the current statistical period identifier into a unified period number or a unified time scale, and then determining the time difference Δt based on the difference between the two. To facilitate subsequent weighted calculations, the unit of the time difference Δt can be consistent with the type of statistical period; that is, when the statistical period is annual, the time difference Δt is in years; when the statistical period is quarterly, the time difference Δt is in quarters; and when the statistical period is monthly, the time difference Δt is in months.

[0078] As an example, when the statistical period is annual and the current statistical period is "2024", and the historical business data belongs to the statistical period of "2023", the time difference Δt can be 1. If the historical business data belongs to the statistical period of "2021", the time difference Δt can be 3. Similarly, when the statistical period is quarterly and the current statistical period is "Q4 2024", and the historical business data belongs to the statistical period of "Q2 2024", the time difference Δt can be 2. If the historical business data belongs to the statistical period of "Q4 2023", the time difference Δt can be 4. Those skilled in the art should understand that the above calculation method for the time difference Δt is based on the difference in a unified period number and is applicable to different statistical period types and different period coding rules.

[0079] In one implementation, the system pre-establishes a mapping table from period identifiers to period numbers. For example, "2024" is mapped to the number 2024, "2024 Q1 / Q2 / Q3 / Q4" are mapped to consecutive numbers, such as 2024×4+0 to 2024×4+3, and "March 2024" is mapped to 2024×12+2, etc. The time difference Δt is then calculated based on the difference between the mapped numbers. In another implementation, the system can directly calculate the period difference based on a date range. For example, the start and end dates of the current statistical period and the corresponding period are converted into a unified time coordinate, and the time difference Δt is calculated based on a preset period length.

[0080] It should be noted that when the statistical period of historical business data is later than the current statistical period, such as when the time stamp is abnormal due to data backfilling or adjustment of statistical standards, the system can mark the data as an abnormal record and perform removal, zeroing or merging processing to avoid interfering with subsequent time weighting. This abnormality handling strategy can be configured by the system.

[0081] Step S203: Perform time-attenuation weighting on the historical business data according to the following formula to obtain the time-weighted value. ,

[0082] ,

[0083] Where λ is the time decay coefficient.

[0084] This weighting value reflects the objective law that the contribution of historical business data to the calculation of the effectiveness of the transformation of results in the current statistical period decays over time. By introducing a time weighting value, historical information earlier than the current statistical period can be mapped to the current statistical period in the form of an equivalent contribution without changing the attribution of the results in the current statistical period. This reduces the calculation bias caused by cross-period lag and improves the stability and comparability of cross-period monitoring results.

[0085] In this embodiment, the weighting value ω t It satisfies the following property: when Δt=0, ω t =1 indicates that data from the same period as the current statistical period maintains full contribution; as Δt increases, ω t Monotonically decreasing, meaning that earlier historical data contributes less to the current cycle. ω t The value range is (0, 1], which facilitates numerical stabilization processing in subsequent weighted calculations. These characteristics enable the system to perform differentiated processing on historical data across different time spans.

[0086] In this embodiment, the time decay coefficient λ is used to control the decay rate, and its value can be preset by the system or configured by the user. Generally, the larger λ is, the faster the decay, indicating that the impact of historical data on the current statistical period weakens more quickly. The smaller λ is, the slower the decay, indicating that historical data still has a longer period of influence on the current statistical period. To adapt to different business scenarios, λ can be set according to the type of result, the life cycle of the result, or industry experience. For example, for results with rapid technological iteration and short update cycles, a larger λ can be set; for results in the construction industry that require a longer verification and promotion cycle, a smaller λ can be set. This invention does not limit the specific value of λ, as long as it can achieve a weighted effect of decay as Δt increases.

[0087] In one example, the current statistical period is 2024, and the statistical period for a certain historical business data is 2023. Therefore, Δt = 1, and when λ = 0.2, ω t =e -0.2 ≈0.82. If the historical business data belongs to the statistical period of 2021, then Δt=3, corresponding to ω t =e -0.6 ≈0.55. This shows that historical data closer to 2024 has a higher weight in subsequent calculations.

[0088] In the results transformation efficiency monitoring scenario of this embodiment, there is an objective misalignment between historical business data and business data within the current statistical period on the time axis. In particular, process data such as R&D investment, experimental verification, and pilot demonstrations often precede the statistical period for revenue realization and promotion and dissemination. If historical data is processed using equal-weighted accumulation or simple amortization, it is prone to two types of technical problems: First, the contribution of early historical data to the current period is over-amplified, causing efficiency calculation to be overly sensitive to historical long-tail data, thereby leading to cross-period result drift; Second, the correlation between recent historical data and current data is diluted, making the monitoring results slow to respond to real business changes, reducing the timeliness and stability of efficiency monitoring. To solve the above problems, the time decay weighting formula used in this application is used to weight historical business data, so that the contribution of historical data decreases monotonically with the time difference Δt between it and the current statistical period, thereby achieving alignment of contributions of historical information in the time dimension.

[0089] Furthermore, the aforementioned time-weighted method with exponential decay has significant advantages in this embodiment. The output value of the exponential decay function, i.e., the time-weighted value, is stably limited to the (0,1) interval, which naturally suppresses the cumulative amplification effect of long-term historical data and ensures that the weight changes continuously and smoothly as the period difference increases, thereby avoiding the abrupt discontinuity problems introduced by segmented thresholds or equal-weighting. On this basis, the decay rate is adjustable through the decay coefficient λ, enabling the monitoring system to adaptively adjust the decay intensity of historical contributions according to the type of results, industry verification cycle, or business chain lag characteristics. Thus, this embodiment can map historical inputs and process data to the current statistical period in a time-consistent equivalent contribution form without changing the attribution of the results in the current statistical period, reducing the calculation bias caused by cross-period misalignment, improving the responsiveness of monitoring results to current business changes, and enhancing the stability and comparability of performance values ​​between different statistical periods.

[0090] Step S204: Multiply each historical business data with its corresponding time-weighted value and perform a weighted sum to obtain time-corrected feature data.

[0091] In this embodiment, step S204 may include: for multiple data records belonging to different historical statistical periods within the same type of historical business data, such as the same indicator type, the same result item, the same business dimension, or the same set of data fields, reading their corresponding historical values ​​and weighting them with the corresponding time-weighted value ω. t Multiplying them together gives the weighted term ω. t ·RD t Then, the weighted terms are summed to obtain the time-corrected feature data F corresponding to the current statistical period. adj The time-corrected feature data F adj The calculation formula is:

[0092] ,

[0093] Among them, RD t This represents the historical business data value corresponding to the historical statistical period t. Through the above weighted summation process, the historical business data is mapped to the current statistical period in the time dimension, enabling subsequent steps to perform feature component calculation and fusion operations under a unified time benchmark.

[0094] In one example, the current statistical period is 2024, and the historical business data is the R&D investment data from 2021 to 2023, which are respectively RD 2021 =100、RD 2022 =120、RD 2023 =150, where the unit can be ten thousand yuan or other units of measurement. If the time weighting values ​​obtained in step S203 are ω 2021 =0.55、ω 2022 =0.67、ω 2023 =0.82, then the time-corrected feature data obtained in step S204 is:

[0095] F adj =0.55×100+0.67×120+0.82×150;

[0096] The time-corrected feature data F adj It can be used as a time-adjusted R&D investment or time-adjusted cost feature corresponding to 2024, and then used to construct the feature components related to revenue performance, thereby reducing the impact of cross-cycle misalignment on performance calculation.

[0097] It should be noted that the construction of time-corrected feature data can be carried out separately for different types of historical business data. For example, for quantities that can be accumulated, such as input, frequency, and coverage, a weighted summation can be used to obtain the time-corrected cumulative amount. For quantities that can form a ratio or intensity, such as input intensity per unit size or application intensity per unit project, the time-corrected cumulative amount can be further combined with the corresponding benchmark quantity in the current statistical period for ratio or normalization processing. For discrete information such as the project initiation time, formation time, stage characteristics, and status characteristics of historical nodes, since they are not suitable for direct accumulation, the system preferably maps them to computable numerical features or vector features before participating in time consistency processing, and then performs weighted aggregation based on time weighting values ​​to generate time-corrected feature data corresponding to the current statistical period.

[0098] In one embodiment, when the set of stages has a preset chronological order, ordinal mapping is performed on the stage characteristics. Specifically, a set of stages for the product transformation lifecycle is predefined, and each stage is assigned a monotonically increasing stage number. For example, "R&D," "pilot," "promotion," and "industrialization" are mapped to 1, 2, 3, and 4, respectively.

[0099] In another embodiment, one-hot encoding mapping is performed on stage features or multi-state features. Specifically, the dimensions of the state set or stage set are predefined, and the states corresponding to the historical statistical period t are mapped to a set of one-hot encoded vectors. For example, if the stage set is "R&D, pilot, promotion, industrialization", then when the historical statistical period t is in the "promotion" stage, the one-hot encoded vector can be represented as (0, 0, 1, 0). For binary state features, such as whether the achievement appraisal has been completed or whether the industrialization stage has been entered, they can be mapped to numerical features of 0 or 1. Through the above mapping and weighted convergence processing, discrete information such as stage features and state features can participate in the subsequent calculation and fusion operation of performance feature components in a numerical form consistent with the current statistical period time base.

[0100] The time-corrected feature data formed in step S204 enables historical business data to participate in the performance calculation process of the current statistical period in a unified, continuous and computable form. This allows historical information across statistical periods to participate in subsequent fusion calculations in a feature form consistent with the time base of the current statistical period, improving the numerical stability and reproducibility of cross-period data fusion calculation results, and suppressing the undesirable bias and excessive influence of the accumulation of long-term historical data on the performance value of results transformation.

[0101] Step S300: Based on the time-corrected feature data and the business data within the current statistical period, determine the multi-dimensional performance feature components that reflect revenue performance, technical attributes, and promotion and diffusion, and construct an initial multi-dimensional performance feature vector.

[0102] The purpose of constructing the initial multidimensional efficiency feature vector is to decompose the achievement transformation efficiency into several computable and fusionable feature components under a unified time benchmark, and to organize the feature components into a vectorized representation for subsequent scaling and correlation correction fusion operations. Specifically, step S300 uses the time-corrected feature data obtained in step S200 as the basic input after cross-cycle compensation, and combines it with the business data within the current statistical period to calculate the revenue performance feature component, technical attribute feature component, and promotion and diffusion feature component, thereby forming an initial multidimensional efficiency feature vector to characterize the achievement transformation efficiency status of the target unit in the current statistical period.

[0103] The revenue performance component characterizes the economic output generated by the transformation of research results within the current statistical period. Historical inputs earlier than the current statistical period are included in the calculation with equivalent contributions processed for time consistency, thus reducing statistical bias caused by earlier historical inputs and later returns. To adapt to different monitoring objectives, this embodiment provides two complementary calculation methods: a net contribution method and an efficiency method.

[0104] In the net contribution method, the calculation of this revenue performance characteristic component includes the following steps: obtaining the revenue amount R based on revenue-related business data within the current statistical period. cur The equivalent input I is obtained based on time-corrected feature data. adj According to formula x profit =R cur -I adj The return performance characteristic component x is calculated. profit In a specific example, the current statistical period is 2024, and the revenue R is obtained by summarizing the revenue-generating business data of the target unit in 2024. cur The value is 500, with units such as ten thousand yuan. Meanwhile, historical business data related to inputs earlier than 2024, after time consistency processing in steps S203 to S204, yields time-corrected characteristic data for inputs corresponding to 2024, denoted as equivalent input amount I. adj Equivalent input I adj For example, 318, in units of ten thousand yuan. Then the characteristic component of the return performance is: x profit =500-318=182. This result indicates that after adjusting historical investments for 2024, the net contribution of technology transfer in 2024 is 1.82 million yuan.

[0105] In the efficiency-oriented approach, the calculation method for this revenue performance characteristic component includes the following steps: obtaining the revenue amount R based on revenue-related business data within the current statistical period. cur The equivalent input I is obtained based on time-corrected feature data. adj According to formula x profit =R cur / (I adj +ε) calculates the return performance characteristic component x profit Here, ε is a preset minimum positive number to prevent the denominator from being zero. This efficiency-based method outputs the current period revenue generated per unit of equivalent input. It is a normalized intensity indicator, suitable for horizontal comparisons between target units of different sizes where greater comparability is needed. For example, when the revenues of two units differ significantly, the efficiency-based method better reflects input-output efficiency. In a specific example, the revenue R... cur The equivalent input is 500. adjIf ε is 318 and takes a minimum value, then the characteristic component of the return performance is: x profit ≈500 / 318=1.57. This result indicates that, based on the equivalent input adjusted for time consistency, each unit of equivalent input in 2024 corresponds to approximately 1.57 units of revenue output.

[0106] It should be noted that the revenue R cur and equivalent input I adj The specific composition can be configured based on data availability and business requirements. The calculation method for the revenue performance characteristic components is not limited to the above-mentioned difference or ratio forms, as long as the current revenue and time-adjusted input can be quantitatively correlated under the same time benchmark.

[0107] Technical attribute feature components are used to characterize the technical quality, advancement level, or standardization degree of the achievements of the target unit within the current statistical period. This ensures that the characteristicization of achievement transformation efficiency not only reflects economic output but also demonstrates the impact of the achievement's technical attributes on transformation efficiency. In this embodiment, technical attribute feature components can be generated based on the achievement list data, achievement attribute data, and standard level and advancement level data related to the achievements within the current statistical period.

[0108] In one implementation, the method for calculating the technical attribute feature components includes the following steps: obtaining the result set G within the current statistical period; determining the technical attribute level of each result g in the result set G, such as standard level, advancement level, or preset technical level; and determining the corresponding attribute mapping coefficient n based on the technical attribute level. g And count the number m of results corresponding to the mapping coefficient of this attribute. g ; for each result m g and n g Perform aggregation operations to obtain the technical attribute feature components x tech The technical attribute feature component x tech For example, the calculation formula can be:

[0109] .

[0110] In a specific example, the current statistical period is 2024. The target unit produced and included a total of 8 achievements in 2024, of which 2 were classified as "domestically leading," 5 as "domestically advanced or below," and 1 as "internationally advanced or above." Based on the preset level mapping rules, the system sets the mapping coefficient for "domestically advanced or below" to 1, the mapping coefficient for "domestically leading" to 1.5, and the mapping coefficient for "internationally advanced or above" to 2. Therefore, the technical attribute feature component can be calculated as: x tech=5×1+2×1.5+1×2=10. This result indicates the overall level of the technical attributes of the target unit's achievements in 2024. Higher-level achievements contribute more to the technical attribute feature components due to their larger mapping coefficients, thereby improving the system's sensitivity to identifying high-quality achievements.

[0111] It should be noted that the aforementioned level mapping coefficients and level classification methods can be preset by the system or configured by the user, and can be adjusted according to industry standard systems, enterprise internal standard systems, or achievement evaluation systems. This embodiment does not limit the specific categories of technical attribute levels, as long as the technical attributes of the achievement can be converted into calculable quantitative coefficients and participate in feature component aggregation. In addition, in some implementations, to avoid the technical attribute feature components being overly dominated by quantity due to differences in the scale of the achievement, the aggregation results can also be normalized or nonlinearly compressed. The relevant processing can be uniformly implemented in the subsequent scaling steps of this embodiment.

[0112] The promotion and diffusion feature component is used to characterize the application coverage and diffusion intensity of the results of a target unit in engineering projects or business scenarios within the current statistical period, reflecting the degree of penetration from the formation to the application and promotion of the results. In this embodiment, the promotion and diffusion feature component can be generated based on promotion-related business data such as the application records of the results within the current statistical period, the number of covered projects, the application frequency, and the scale of project clusters. To facilitate subsequent fusion calculations and cross-unit comparisons, the promotion and diffusion feature component can be quantitatively expressed by combining application intensity and coverage breadth.

[0113] In one implementation, the method for calculating the diffusion characteristic component includes the following steps: obtaining the application records of the results within the current statistical period, and determining the diffusion frequency F and the number of covered projects Z; in one implementation, the target unit's project scale benchmark N within the current statistical period can also be obtained, such as the total number of projects under construction or the project load index, for scaling processing; and determining the diffusion quantification result x based on the diffusion frequency F and the number of covered projects Z. spread The formula for calculating the diffusion characteristic component can be, for example, as follows:

[0114] ;

[0115] In this context, log(F·Z+1) is used to suppress the amplification effect of extremely high-frequency data on the results, and the denominator N is used to scale the results of the generalization of target units of different sizes. When there is no need to introduce a scale benchmark, the denominator can be omitted or replaced with other preset scaling terms.

[0116] In a specific example, the current statistical period is 2024. The target unit's promotion frequency (F) of a certain achievement in engineering projects within 2024 is 25, the number of projects covered (Z) is 10, and the target unit's total number of projects under construction (N) is 80. Then, the promotion and diffusion characteristic component can be calculated as: x spread =log(25×10+1) / 80. The resulting promotion and diffusion feature components can comprehensively reflect the frequency and coverage of promotion, and suppress the bias caused by high-frequency repeated use of a single project through logarithmic gain, while achieving comparability across target units through scale benchmark.

[0117] In different implementations, the promotion and diffusion feature component can be further incorporated with factors such as the number of years of continuous promotion, application scenario type, and key project weight. The introduction and quantification of these factors can be configured according to business needs.

[0118] In this embodiment, when the revenue performance characteristic component x profit Technical attribute feature component x tech And the quantitative results of the promotion and diffusion x spread After determination, step S300 combines the feature components in a preset dimensional order to form an initial multidimensional performance feature vector, for example: X=(x profit , x tech , x spread When the dimension is expanded, it is represented as X=(x1, x2,...,x...). n It should be understood that the vector dimension n can be expanded or trimmed according to business needs, data availability, and configuration rules. For example, a revenue sub-component, a technology sub-component, or a promotion sub-component can be introduced, thereby representing the initial multidimensional performance feature vector in a higher-dimensional form. This embodiment does not limit the specific value of the vector dimension. Through the above vectorization construction, the performance conversion efficiency of the target unit in the current statistical period is transformed into a structured, computable multidimensional representation, providing a unified data input format and computational basis for the scaling processing in step S400 and the correlation correction and fusion operation in step S500.

[0119] Step S400: Perform scaling processing on each feature component of the initial multidimensional efficiency feature vector to obtain the target multidimensional efficiency feature vector for the current statistical period.

[0120] In this embodiment, since the feature components in the initial multidimensional performance feature vector may have differences in dimensions, numerical magnitudes, and distribution skewness, the scaling process maps each feature component to a numerical range more suitable for fusion operation. This reduces the undesirable dominant effect of a single high-magnitude component on the result in subsequent fusion operations and improves the numerical stability and comparability of features of different dimensions in the fusion stage.

[0121] Figure 3It shows Figure 1 The flowchart shown illustrates the scaling process in step S400. Figure 3 As shown, step S400 includes steps S401 to S403.

[0122] Step S401: Obtain each feature component x in the initial multidimensional performance feature vector i .

[0123] In one implementation, the system stores the initial multidimensional performance feature vector as an array, vector object, or key-value pair structure, for example, with feature name as key and feature value as value. Step S401 reads the feature component value x of each dimension sequentially by traversing the data structure. i In another implementation, the system can parse the initial multidimensional performance feature vector according to a preset dimensional order table or feature template to ensure that the subsequent scaling processing maintains a consistent dimensional order and calculation method across different statistical periods or different target units.

[0124] It should be noted that, to ensure the numerical validity of the scaling calculation, step S401 may also include a validity check of the feature components, such as determining whether the feature components are missing, non-numerical, or abnormal extreme values. When a missing or abnormal value is detected, the system can execute a preset data cleaning strategy, such as zeroing, truncation, interpolation, or marking abnormalities, to avoid unstable scaling results caused by abnormal input.

[0125] Step S402: For each feature component value x i The adjusted feature component x is obtained by performing nonlinear compression according to the following formula. i ',

[0126] x i '=x i / (1+αx i ), where α is a scaling factor greater than zero.

[0127] This nonlinear compression process maps feature components of different magnitudes and distributions to a numerical range more suitable for fusion computation. The nonlinear compression function exhibits a monotonically increasing characteristic, meaning that the larger the original feature component value, the larger the adjusted feature component value x. i The larger x remains, the more it maintains the feature ranking relationship. Simultaneously, the function exhibits asymptotic saturation properties, as x... i When the value is large, the adjusted characteristic component value x i The growth rate of ' decreases, thereby suppressing the undesirable dominant role of high-level features in the subsequent fusion operation results.

[0128] The above nonlinear compression function satisfies the following property: when x iWhen x approaches 0 i 'Approximately equal to x i This maintains high resolution for small-scale features; when x i When the denominator term 1+αx increases, i Increase, leading to x i The relative growth slows down, thus compressing large values. Through this process, the feature components in the initial multidimensional efficiency feature vector are balanced on a numerical scale, improving the numerical stability of subsequent correlation correction fusion operations.

[0129] In this embodiment, the scaling factor α is used to control the compression intensity. Generally, the larger α is, the more pronounced the compression effect; the smaller α is, the weaker the compression effect. The scaling factor α can be preset by the system, configured by the user, or adaptively determined based on the sample distribution of historical statistical periods. For example, the system can set α based on the numerical range, mean, or quantile statistical results of each feature component within a historical statistical period, so that different feature components fall into similar numerical ranges after compression.

[0130] In one example, if a certain feature component takes the value x i =182, take α as 0.01, then x i =182 / (1+0.01×182)≈64.5. If the other characteristic component takes the value x j =10, then x i =10 / (1+0.01×10)≈9.1. This shows that the larger numerical feature components are significantly compressed, while the smaller numerical feature components maintain good discriminative power, making each feature component more suitable for subsequent fusion calculations. It should be noted that step S402 can perform the above nonlinear compression process on all feature components one by one. In some implementations, compression can be performed only on preset high-order feature components, while maintaining an identity mapping or using a smaller α value for low-order feature components to further improve the discriminative power of the feature components.

[0131] Step S403: Adjust the characteristic component x i 'The target multidimensional performance feature vector is formed by combining the features in the same dimensional order as the initial multidimensional performance feature vector.'

[0132] In this embodiment, the system maintains the number of dimensions and semantics of the initial multidimensional performance feature vector unchanged, and only changes the original feature component value x of each dimension. i Replace it with its corresponding adjusted characteristic component value x iThis process yields the target multidimensional performance feature vector. The system combines these vectors according to a preset dimension order table or feature template to ensure that the target multidimensional performance feature vectors generated for different statistical periods and target units maintain consistency in dimension arrangement, thereby avoiding input misalignment in subsequent matrix operations due to inconsistent dimension order.

[0133] In one implementation, the system stores the target multidimensional performance feature vector as an array or vector object, along with dimension indices. In another implementation, the system stores the target multidimensional performance feature vector as a key-value pair structure, where the key is the dimension name and the value is the adjusted feature component, and the dimension order is achieved during output or computation using a dimension ordering table. The target multidimensional performance feature vector generated in step S403 can be directly used as input data for subsequent correlation correction fusion operations, ensuring dimensional consistency and feasibility in subsequent calculations.

[0134] Step S500: Perform correlation correction and fusion calculation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value for the current statistical period.

[0135] Figure 4 It shows Figure 1 The flowchart shown is a schematic representation of the correlation correction fusion operation method in step S500. Figure 4 As shown, step S500 includes steps S501 to S503.

[0136] Step S501: Obtain the set of target multidimensional performance feature vectors corresponding to the historical statistical period.

[0137] The set of target multidimensional performance feature vectors corresponding to the historical statistical period is used to provide a historical sample basis for the estimation of subsequent correlation correction parameters. In one embodiment, the set of target multidimensional performance feature vectors corresponding to the historical statistical period consists of the target multidimensional performance feature vectors of L consecutive historical statistical periods within a preset rolling time window. Assuming the current statistical period is T, the historical statistical periods covered by the rolling time window can be {TL, T-L+1,..., T-1}, and the corresponding set of target multidimensional performance feature vectors can be represented as: {x T-L ', x T-L+1 ',..., x T-1 '}, where x k ' represents the target multidimensional efficiency feature vector corresponding to the kth statistical period.

[0138] In a preferred embodiment, the rolling time window is updated in a sliding manner as the statistical period updates. When the statistical period changes from T to T+1, the system removes the target multidimensional performance feature vector corresponding to the earliest statistical period TL from the set and adds the target multidimensional performance feature vector corresponding to the newly generated statistical period T, so as to keep the set always composed of L consecutive historical statistical periods. Through the rolling update mechanism, the correlation correction parameter can be continuously updated over time to adapt to changes in the business environment, data distribution, and organizational operating status. It should be noted that the parameter L can be preset by the system or configured by the user, and can be adjusted according to the data volume, period granularity, and computational stability requirements.

[0139] Step S502: Determine the correlation correction parameters based on the target multidimensional performance feature vector set.

[0140] Figure 5 It shows Figure 4 The flowchart illustrates the method for determining the correlation correction parameter in step S502. Figure 5 As shown, step S502 includes steps S5021 and S5022.

[0141] Step S5021: Construct the covariance matrix between each feature component based on the target multidimensional efficiency feature vector set corresponding to the historical statistical period.

[0142] In one embodiment, assuming the dimension of the target multidimensional performance feature vector is n, each target multidimensional performance feature vector can be represented as: X k '=(x k,1 ', x k,2 ', ..., x k,n '). Where, k∈{TL, T-L+1,..., T-1}, x k,i ' represents the adjusted value of the i-th feature component in the k-th statistical period. The system uses the historical vector set as a sample, calculates the sample mean of each feature component, and further calculates the covariance between any two feature components, thus forming an n×n covariance matrix ∑. The element ∑ in the i-th row and j-th column of the covariance matrix ∑ is... ij This is used to reflect the degree of coordinated change between the i-th and j-th feature components over a historical statistical period. In one example, assume the target multidimensional performance feature vector has a dimension of 3, and X... k '=(x k,profit ', x k,tech ', x k,spread '), taking a rolling time window length L of 3, and assuming the target multidimensional performance feature vectors corresponding to the three historical statistical periods are: X 2021 =(60, 9, 0.05), X 2022=(65,10, 0.06), X 2023 =(70, 10, 0.08). First, calculate the mean of each feature component in the historical sample set, for example: profit =(60+65+70) / 3=65, tech =(9+10+10) / 3=9.67, spread =(0.05+0.06+0.08) / 3=0.0633. Then, calculate the covariance between any two-dimensional feature components. Taking the covariance between the profit performance feature component and the technical attribute feature component as an example, it can be calculated as follows:

[0143] ,

[0144] Substituting the above sample data, we can obtain:

[0145] 2021: (60-65)×(9-9.67)=(-5)×(-0.67)=3.35;

[0146] 2022: (65-65)×(10-9.67)=0×0.33=0;

[0147] 2023: (70-65)×(10-9.67)=5×0.33=1.65;

[0148] Summing and then dividing by L-1=2, we get:

[0149] .

[0150] Similarly, we can calculate the covariance between the revenue performance feature component and the promotion and diffusion feature component, the covariance between the technical attribute feature component and the promotion and diffusion feature component, and the variance term of each component itself, i.e., the diagonal elements. Finally, we fill the above covariance values ​​in dimensional order to form a covariance matrix ∑, for example:

[0151] ;

[0152] The matrix is ​​symmetric about the main diagonal, i.e., ∑ i,j =∑ j,i .

[0153] Through the above construction method, the covariance matrix can characterize the fluctuation intensity of each characteristic component and its inter-correlation degree within the historical statistical period, providing a computational basis for correcting the correlation structure in the subsequent step S5022.

[0154] Step S5022: Perform inverse matrix operation on the covariance matrix to obtain the correlation correction parameters.

[0155] In this embodiment, the covariance matrix is ​​assessed for invertibility or numerical stability before performing the inverse matrix operation. If the covariance matrix satisfies a preset stability condition, the direct inversion method is executed; otherwise, a regularized inversion method is used. The preset stability condition characterizes the numerical stability requirement for the covariance matrix to be suitable for direct inversion matrix operations. In one embodiment, the preset stability condition includes any one or a combination of invertibility, full rank, condition number, and eigenvalue conditions.

[0156] The invertibility condition is used to determine whether the covariance matrix possesses the basic invertibility of directly performing inverse matrix operations. In one embodiment, the invertibility condition includes: the absolute value of the determinant of the covariance matrix is ​​greater than a preset lower threshold, to avoid amplifying the numerical error of the inverse matrix operation when the determinant is close to zero. Optionally, the invertibility condition may also include: the minimum value of the main diagonal elements of the covariance matrix is ​​greater than a preset lower threshold, to exclude numerical instability caused by excessively small main diagonal elements.

[0157] The full-rank condition is used to determine whether the covariance matrix is ​​a full-rank matrix. In one embodiment, the full-rank condition includes: the rank of the covariance matrix is ​​equal to the order of the covariance matrix, to avoid rank deficiency that would prevent the inverse matrix operation from being performed.

[0158] The condition number condition is used to determine whether the covariance matrix is ​​a numerically stable invertible matrix. In one embodiment, the condition number condition includes: the condition number of the covariance matrix is ​​less than a preset upper threshold, wherein the upper threshold is used to limit the degree of error amplification during the inverse matrix operation, thereby improving the stability of the calculation results.

[0159] Eigenvalue conditions are used to determine whether the covariance matrix is ​​close to singular. In one embodiment, the eigenvalue condition includes: the smallest eigenvalue of the covariance matrix is ​​greater than a preset lower threshold, where the lower threshold is a threshold greater than zero, to avoid instability in inverse matrix operations when the matrix is ​​close to singular.

[0160] In a preferred embodiment, the system uses condition number conditions and / or eigenvalue conditions as the primary basis for determining numerical stability. When the covariance matrix satisfies the preset stability conditions, a direct inversion method is executed. When the covariance matrix does not satisfy the preset stability conditions, a regularized inversion method is adopted.

[0161] The direct inversion method refers to directly performing inverse matrix operations on the covariance matrix to obtain the correlation correction parameters. In one embodiment, this regularized inversion method includes:

[0162] Step 1: Obtain the dimension information of the covariance matrix;

[0163] Step 2: Construct an identity matrix I with the same dimensions as the covariance matrix;

[0164] Step 3: Perform diagonal regularization on the covariance matrix according to the following formula to obtain the corrected covariance matrix ∑'.

[0165] ∑'=∑+βI,

[0166] Where ∑ is the covariance matrix and β is the regularization coefficient that is greater than zero;

[0167] Step four: Perform inverse matrix operations on the corrected covariance matrix ∑' to obtain the correlation correction parameters.

[0168] In this embodiment, when the number of historical samples is limited or the feature components are strongly correlated, the covariance matrix may exhibit numerical instability such as a determinant close to zero or an excessively large condition number. This makes direct inversion operations prone to amplifying computational errors or even preventing inversion altogether. By introducing a regularization term into the main diagonal of the covariance matrix, the diagonal elements of the modified covariance matrix are increased overall, thereby improving the matrix's invertibility and reducing the risk of numerical instability in inverse matrix operations.

[0169] In a specific example, assuming the target multidimensional performance feature vector has a dimension of 2, the covariance matrix would be, for example:

[0170] ∑= ;

[0171] It can be seen that the off-diagonal elements of the matrix are close to the diagonal elements, indicating that the two-dimensional eigencomponents are highly correlated. In this case, the covariance matrix is ​​prone to becoming singular, and direct inversion may lead to numerical instability. Therefore, we choose a regularization coefficient β of 0.1 to construct the identity matrix I:

[0172] I= ;

[0173] The corrected covariance matrix is ​​then:

[0174] ∑'=∑+βI= ;

[0175] Compared to the original covariance matrix, the modified covariance matrix has increased diagonal elements, thus improving its invertibility. The inverse matrix operation is then performed on the modified covariance matrix to obtain the correlation correction parameters. The inverse matrix operation can be implemented using numerical linear algebra algorithms, such as Gaussian elimination, LU decomposition, or other equivalent inversion algorithms.

[0176] It should be noted that the regularization coefficient can be preset by the system or configured by the user, and can be adjusted according to the number of historical samples, the dimension of the feature vector, or the numerical stability index of the covariance matrix. For example, when the system detects that the condition number of the covariance matrix exceeds a preset threshold or the minimum eigenvalue is lower than a preset threshold, the regularization coefficient can be increased. When the numerical stability of the covariance matrix is ​​good, a smaller regularization coefficient can be used to reduce the disturbance to the original correlation structure.

[0177] Step S503: Use the correlation correction parameter to perform a fusion operation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value for the current statistical period.

[0178] Step S503 includes the following steps: performing matrix multiplication on the target multidimensional efficiency feature vector corresponding to the current statistical period and the correlation correction parameter to obtain the correction feature vector; performing a vector inner product operation on the correction feature vector and the target multidimensional efficiency feature vector corresponding to the current statistical period to obtain the achievement transformation efficiency value of the current statistical period.

[0179] Specifically, suppose the target multidimensional efficiency feature vector corresponding to the current statistical period is: X T '=(x T,1 ', x T,2 ',x T,3 The correlation correction parameter is matrix W, which is obtained from the inverse of the corrected covariance matrix. Then, matrix multiplication yields the corrected eigenvector Y, i.e., Y = W × X. T Then, the vector inner product operation is performed to obtain the result transformation efficiency value S. T Then S T =(X T ') T ×Y.

[0180] In a specific example, assuming the current statistical period is 2024, the target multidimensional effectiveness feature vector corresponding to the current statistical period is: X 2024 =(60, 9, 0.05), assuming the correlation correction parameter W obtained in step S502 is:

[0181] W= ;

[0182] First, matrix multiplication is performed to obtain the corrected eigenvector Y:

[0183] Y=W×X 2024 '= = ;

[0184] Then, the vector dot product operation is performed:

[0185] S2024 =X 2024 '×Y=60×1.209+9×1.14+0.05×0.25;

[0186] This yields the achievement transformation efficiency value for the current statistical period, with the numerical result directly calculated and output by the system. Through the above process, the fusion operation not only considers the numerical magnitude of each feature component but also reflects the correlation structure between feature components through correlation correction parameters. This ensures that, when there is a correlation between feature components, the fusion result can more stably reflect the comprehensive contribution of multidimensional features.

[0187] In an optional embodiment, the method for monitoring the effectiveness of technology transfer may further include the following steps: calculating the change in effectiveness based on the effectiveness values ​​of technology transfer over multiple consecutive statistical periods; comparing the change in effectiveness with a preset threshold, and outputting an abnormal warning signal when the change in effectiveness exceeds the preset threshold.

[0188] In this optional embodiment, the above steps are used to monitor the time series changes of the results transformation efficiency based on the output of the results transformation efficiency value for the current statistical period, so as to automatically output early warning results when abnormal fluctuations occur in the efficiency value. Multiple consecutive statistical periods can be a preset number of adjacent statistical periods, such as the most recent 2, 3, 4, or more statistical periods.

[0189] In one implementation, the change in efficiency can be determined by the difference between the efficiency values ​​of results transformation in adjacent statistical periods. Assume the current statistical period is T, the previous statistical period is T-1, and the efficiency value of results transformation in the current statistical period is S. T The achievement transformation efficiency value for the previous statistical period was S. T-1 Then the change in efficiency ΔS T This can be expressed as: ΔS T =|S T -S T-1 By taking the absolute value, the warning logic can be applied to both abnormal increases and abnormal decreases. In another implementation, the change in performance can also be determined using a volatility index based on a multi-period window, such as calculating the variance or standard deviation of the performance value sequence within a preset window to characterize the intensity of performance value fluctuations in the short term.

[0190] In this optional embodiment, a preset threshold is used to define the acceptable fluctuation range and the abnormal fluctuation range. The preset threshold can be preset by the system or configured by the user, and can be adjusted according to the distribution of performance values ​​in historical statistical periods, the business scale of the target unit, or the monitoring scope. For example, the system can adaptively determine the preset threshold based on the mean, standard deviation, or quantile statistical results of performance values ​​in historical statistical periods to improve the adaptability of the early warning judgment to different target units.

[0191] When the calculated change in performance exceeds a preset threshold, the system outputs an abnormal warning signal. This warning signal may include a warning flag, a warning level parameter, a warning cause identifier, or a warning notification message, and can be output to a display terminal, business system, or message push module via a preset interface. In one implementation, the warning level can be graded based on the magnitude of the performance change exceeding the threshold; for example, the magnitude of exceeding the threshold can be mapped to high, medium, and low levels to facilitate automatic triggering or manual verification of subsequent handling strategies.

[0192] According to the embodiments of the present invention, by acquiring the target unit's business data within the current statistical period and introducing historical business data earlier than the current statistical period, and combining this with time consistency processing of the historical business data to generate time-corrected feature data, historical information across statistical periods can participate in performance calculation in a feature form consistent with the time benchmark of the current statistical period. This reduces statistical bias caused by time lags in input, formation, promotion, and benefit realization, and improves the stability and comparability of the results of achievement transformation performance calculation across different statistical periods. Furthermore, based on constructing multi-dimensional performance feature components reflecting benefit performance, technical attributes, and promotion diffusion, the embodiments of the present invention organize each feature component into an initial multi-dimensional performance feature vector, and perform scale adjustment processing on each component of this vector. This ensures that features of different dimensions, magnitudes, and distribution patterns can still be mapped to a numerical range suitable for fusion calculation even when they differ significantly, thereby suppressing the undesirable dominant role of high-magnitude feature components in the calculation results and improving the numerical stability when multiple features participate in the calculation. Furthermore, this embodiment of the invention performs correlation correction fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period. This allows the fusion process to consider the correlation structure between multidimensional feature components, suppressing duplicate measurement and bias amplification caused by feature correlation, thereby obtaining a more robust output of achievement transformation efficiency value. Therefore, this embodiment of the invention can achieve automated, stable, and reproducible monitoring and calculation of achievement transformation efficiency in scenarios involving multi-source heterogeneous data, cross-period data, and the presence of multidimensional feature correlations.

[0193] Specifically, this invention also provides a technology transfer efficiency monitoring system, including an acquisition module, a processing module, a construction module, a scaling module, and a fusion module. The acquisition module acquires raw business data from the target unit for calculating the technology transfer efficiency of the current statistical period. The raw business data includes business data within the current statistical period and historical business data earlier than the current statistical period. The processing module performs time consistency processing on the historical business data to generate time-corrected feature data corresponding to the current statistical period. The construction module, based on the time-corrected feature data and the business data within the current statistical period, determines multi-dimensional efficiency feature components reflecting revenue performance, technical attributes, and promotion and diffusion, and constructs an initial multi-dimensional efficiency feature vector. The scaling module performs scaling processing on each feature component of the initial multi-dimensional efficiency feature vector to obtain the target multi-dimensional efficiency feature vector for the current statistical period. The fusion module performs correlation correction and fusion operations on the target multi-dimensional efficiency feature vector corresponding to the current statistical period to generate the technology transfer efficiency value for the current statistical period.

[0194] It should be understood that the functions of each module in the above system correspond one-to-one with the corresponding steps of the aforementioned method for monitoring the effectiveness of achievement transformation. Those skilled in the art can understand the specific implementation of the above system based on reading the aforementioned method embodiments, and will not be elaborated here.

[0195] Each module in the aforementioned technology transfer effectiveness monitoring system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the computer device's memory as software, so that the processor can call and execute the corresponding operations of each module.

[0196] In one embodiment, Figure 6 A schematic structural diagram of a computer device according to an embodiment of the present invention is shown, such as Figure 6 The computer device described herein can be a server, a terminal, or a cloud computing node, and its internal structure diagram can be as follows: Figure 6As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for monitoring the effectiveness of technology transfer. The display unit is used to create a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0197] Those skilled in the art will understand that Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0198] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for monitoring the effectiveness of results transformation.

[0199] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described method for monitoring the effectiveness of results transformation.

[0200] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0201] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0202] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0203] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring the effectiveness of technology transfer, characterized in that, Includes the following steps: Obtain the original business data of the target unit for calculating the achievement transformation efficiency of the current statistical period. The original business data includes business data within the current statistical period and historical business data earlier than the current statistical period. Perform time consistency processing on the historical business data to generate time-corrected feature data corresponding to the current statistical period, including the following steps: Obtain the statistical period to which the historical business data belongs; Determine the time difference Δt between the statistical period to which the historical business data belongs and the current statistical period; The historical business data is subjected to time decay weighting processing according to the following formula to obtain the time weighting value ω. t , ， Wherein, λ is the time decay coefficient, used to control the decay rate; The time-corrected feature data is obtained by multiplying each historical business data with its corresponding time-weighted value and then summing the results. Based on the time-corrected feature data and the business data within the current statistical period, multi-dimensional performance feature components reflecting revenue performance, technical attributes, and promotion and diffusion are determined, and an initial multi-dimensional performance feature vector is constructed. n represents the vector dimension; The target multidimensional efficiency feature vector for the current statistical period is obtained by scaling each feature component of the initial multidimensional efficiency feature vector, including the following steps: Obtain each feature component x in the initial multidimensional performance feature vector. i ; For each characteristic component value x i The adjusted feature component x is obtained by performing nonlinear compression according to the following formula. i ', x i '=x i / (1+αx i ), where α is a scale adjustment coefficient greater than zero, used to control compressive strength, based on each characteristic component x within the historical statistical period. i The numerical range, mean, or quantile statistical results can be set. The adjusted feature component x i The target multidimensional performance feature vector is formed by combining the features in the same dimensional order as the initial multidimensional performance feature vector. The process of performing correlation correction and fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value for the current statistical period includes the following steps: Obtain the set of target multidimensional performance feature vectors corresponding to historical statistical periods; Construct the covariance matrix between each feature component based on the target multidimensional performance feature vector set corresponding to the historical statistical period; Obtain the dimension information of the covariance matrix; Construct an identity matrix I with the same dimensions as the covariance matrix; The covariance matrix is ​​diagonally regularized according to the following formula to obtain the modified covariance matrix ∑'. ∑’=∑+βI， Where ∑ is the covariance matrix and β is a regularization coefficient greater than zero; Perform inverse matrix operations on the modified covariance matrix ∑' to obtain the correlation correction parameters; The correlation correction parameter is used to perform a fusion operation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period.

2. The method for monitoring the effectiveness of technology transfer according to claim 1, characterized in that, The step of fusing the target multidimensional efficiency feature vector corresponding to the current statistical period using the correlation correction parameter to generate the achievement transformation efficiency value for the current statistical period includes the following steps: The target multidimensional efficiency feature vector corresponding to the current statistical period is multiplied with the correlation correction parameter to obtain the correction feature vector; The result transformation efficiency value for the current statistical period is obtained by performing a vector inner product operation between the corrected feature vector and the target multidimensional efficiency feature vector corresponding to the current statistical period.

3. The method for monitoring the effectiveness of technology transfer according to claim 1, characterized in that, It also includes the following steps: The change in effectiveness is calculated based on the effectiveness values ​​of achievement transformation over multiple consecutive statistical periods; The change in performance is compared with a preset threshold, and an abnormal warning signal is output when the change in performance exceeds the preset threshold.

4. A technology transfer effectiveness monitoring system, characterized in that, include: The acquisition module is used to acquire the original business data of the target unit for calculating the achievement transformation efficiency of the current statistical period. The original business data includes business data within the current statistical period and historical business data earlier than the current statistical period. The processing module is used to perform time consistency processing on the historical business data and generate time-corrected feature data corresponding to the current statistical period, including the following steps: Obtain the statistical period to which the historical business data belongs; Determine the time difference Δt between the statistical period to which the historical business data belongs and the current statistical period; The historical business data is subjected to time decay weighting processing according to the following formula to obtain the time weighting value ω. t , ， Wherein, λ is the time decay coefficient, used to control the decay rate; The time-corrected feature data is obtained by multiplying each historical business data with its corresponding time-weighted value and then summing the results. The construction module is used to determine multi-dimensional performance feature components reflecting revenue performance, technical attributes, and promotion and diffusion based on the time-corrected feature data and the business data within the current statistical period, and to construct an initial multi-dimensional performance feature vector. n represents the vector dimension; The scaling module is used to scale each feature component of the initial multidimensional performance feature vector to obtain the target multidimensional performance feature vector for the current statistical period, including the following steps: Obtain each feature component x in the initial multidimensional performance feature vector. i ; For each characteristic component value x i The adjusted feature component x is obtained by performing nonlinear compression according to the following formula. i ', x i '=x i / (1+αx i ), where α is a scale adjustment coefficient greater than zero, used to control compressive strength, based on each characteristic component x within the historical statistical period. i The numerical range, mean, or quantile statistical results can be set. The adjusted feature component x i The target multidimensional performance feature vector is formed by combining the features in the same dimensional order as the initial multidimensional performance feature vector. The fusion module is used to perform correlation correction and fusion operations on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period, including the following steps: Obtain the set of target multidimensional performance feature vectors corresponding to historical statistical periods; Construct the covariance matrix between each feature component based on the target multidimensional performance feature vector set corresponding to the historical statistical period; Obtain the dimension information of the covariance matrix; Construct an identity matrix I with the same dimensions as the covariance matrix; The covariance matrix is ​​diagonally regularized according to the following formula to obtain the modified covariance matrix ∑'. ∑’=∑+βI， Where ∑ is the covariance matrix and β is a regularization coefficient greater than zero; Perform inverse matrix operations on the modified covariance matrix ∑' to obtain the correlation correction parameters; The correlation correction parameter is used to perform a fusion operation on the target multidimensional efficiency feature vector corresponding to the current statistical period to generate the achievement transformation efficiency value of the current statistical period.

5. A computer storage medium, characterized in that, The computer storage medium can store program instructions, which, when executed by a processor, can implement the achievement transformation efficiency monitoring method as described in any one of claims 1 to 3.

Images