Method and system for judging technology development stage based on adaptive similarity measurement

Abstract

This application proposes a method and system for determining technology development stages based on adaptive similarity metrics. The method includes: segmenting the Gartner curve to define multiple technology development stages; searching for technology keywords in a preset data set, generating time series data representing the time-varying quantity of scientific and technological data corresponding to the technology keywords, and preprocessing the data; dividing the preprocessed time series data into subsequences based on a preset detail resolution, connecting the subsequences to generate a main sequence, and converting the main sequence into a triplet; performing a similarity measurement between the main sequence and the curves representing each technology development stage through dynamic time warping; and determining the technology development stage with the highest degree of similarity as the development stage of the technology. This method uses the Gartner technology maturity curve and similarity metrics to make multi-scale judgments on technology development stages, improving the accuracy, applicability, and scientific nature of the judgments.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to a method and system for determining a technology development stage based on adaptive similarity measurement. Background Art

[0002] With the development of science and technology, science and technology strategy research has received increasing attention, and evaluating science and technology in different fields has important practical significance. Currently, science and technology assessment methods have shifted from traditional qualitative analysis to quantitative analysis and a combination of qualitative and quantitative analysis. Previous studies typically used traditional methods such as the Delphi method, the Analytic Hierarchy Process, and the Scenario Analysis Method to analyze the technological development status of specific fields. These methods rely heavily on expert knowledge and require judgment based on experience. While the domain authority and professionalism of the analysis results can be guaranteed, the objectivity, consistency, and comprehensiveness of the task are relatively poor. In the current era of development, emerging technologies are constantly emerging, and the demand for scientific and technological intelligence is becoming more dynamic and diverse. With the rapid development of data resources and advances in artificial intelligence technology, quantitative methods are gradually being applied to technology analysis. Technology analysis methods that combine scientific and technological data (e.g., papers) with statistical learning methods have created more possibilities for science and technology assessment.

[0003] The Hype Cycle, a tool created by Gartner Consulting to assess the visibility of new technologies and determine whether to adopt them, is a key method for assessing technology maturity, combining qualitative and quantitative methods. Emerging technology maturity methods based on the Hype Cycle are constantly emerging, becoming a key foundational principle and tool in the field of technology assessment. Based on practical needs such as understanding the development trends of a field's technology, objectively assessing technology maturity, and rationally selecting the timing of technology intervention, the Hype Cycle can be used to research and develop implementation plans to assess the maturity and development stage of a technology.

[0004] However, the generation of a Hype Cycle requires extensive expertise and domain knowledge. In practice, it's difficult to apply this method to different scenarios to grasp the progress of technological development and achieve personalized and quantitative assessments of technology maturity, technological development stages, and even the entire field. Furthermore, due to the varying timescales of technological development cycles in different fields, the scientificity, objectivity, and accuracy of Hype Cycle-based technology development stage assessment methods in related technologies are limited. Summary of the Invention

[0005] The present application aims to solve one of the technical problems in the related art at least to a certain extent.

[0006] To this end, the first purpose of this application is to propose a method for judging the technology development stage based on adaptive similarity measurement. The method performs multi-scale judgment on the technology development stage based on the Gartner technology maturity curve, an adaptive time series data feature extraction algorithm and similarity measurement. By manually adjusting the threshold, the algorithm detail feature extraction granularity is controlled to achieve adaptive processing of scientific and technological achievement time series data, and the processed time series data is compared with the Gartner curve to realize the technology development stage judgment, thereby improving the accuracy, applicability and scientificity of the judgment of the technology development stage.

[0007] The second purpose of this application is to propose a technology development stage judgment system based on adaptive similarity measurement.

[0008] A third object of the present application is to provide a non-transitory computer-readable storage medium.

[0009] To achieve the above objectives, the first embodiment of the present application proposes a method for determining the technology development stage based on an adaptive similarity metric, comprising the following steps:

[0010] Segmenting the Gartner curve to define multiple technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve;

[0011] Searching for technical keywords in a preset data set, generating time series data indicating how the amount of scientific and technological data corresponding to the technical keywords changes over time based on the search results, and preprocessing the time series data;

[0012] Dividing the preprocessed time series data into subsequences based on a preset detail resolution, connecting the subsequences to generate a main sequence to extract features from the preprocessed time series data, and converting the main sequence into triples for storage;

[0013] performing similarity measurement between the converted main sequence and the curves representing each of the technology development stages by dynamic time warping (DTW), and ranking the degree of similarity between the converted main sequence and each of the technology development stages;

[0014] The technology development stage with the highest degree of similarity is determined as the target technology development stage of the technology represented by the time series data.

[0015] Optionally, in one embodiment of the present application, dividing the pre-processed time series data into subsequences based on a preset detail resolution includes:

[0016] S1: setting the normalized time series data as the target sequence;

[0017] S2: Mark the first and last endpoints of the target sequence as first key points, and connect the first and last endpoints to generate a baseline;

[0018] S3: Calculate the vertical distance from each data point in the target sequence, excluding the first and last endpoints, to the baseline, and select the maximum distance from all vertical distances to determine the data point corresponding to the maximum distance;

[0019] S4: determining a corresponding resolution distance based on the detail resolution, comparing the maximum distance with the resolution distance, and if the maximum distance is greater than the resolution distance, marking a data point corresponding to the maximum distance as a second key point, and segmenting the target sequence into a first subsequence and a second subsequence based on the second key point;

[0020] S5: respectively updating the target sequence to each re-divided subsequence, and repeating steps S2 to S4 for each updated target sequence until no new subsequence is generated.

[0021] Optionally, in one embodiment of the present application, determining the corresponding resolution distance based on the detail resolution includes:

[0022] Filtering out the maximum distance and the minimum distance from all the vertical distances;

[0023] The resolution distance is calculated by the following formula:

[0024]

[0025] Among them, Reso is the preset detail resolution, distance max is the maximum distance, distance min is the minimum distance, distance Reso is the resolution distance.

[0026] Optionally, in one embodiment of the present application, converting the main sequence into a triplet for storage includes: calculating the mean of the data points included in each subsequence; calculating the slope of the line connecting the first and last endpoints of each subsequence; determining the index of the first endpoint of each subsequence in the target sequence, and representing the corresponding subsequence as a triplet based on the index, mean, and slope of each subsequence.

[0027] Optionally, in one embodiment of the present application, the similarity measurement between the main sequence and each of the technology development stages is performed through dynamic time warping (DTW), including: converting the sequence segments corresponding to each of the technology development stages into triple form; constructing a distance matrix based on the converted main sequence and the sequence segments corresponding to any converted technology development stage, and determining the optimal path of the distance matrix through dynamic time warping (DTW); aligning the triplets in the sequence segments corresponding to the converted main sequence and any converted technology development stage according to the optimal path, and calculating the distance between each two corresponding triplets after alignment; calculating the sum of all distances, and taking the sum as the similarity between the main sequence and any converted technology development stage.

[0028] Optionally, in one embodiment of the present application, the distance between every two corresponding triples after alignment is calculated using the following formula:

[0029]

[0030] Among them, M 1 and M 2 Represents the mean of each two corresponding triples, q 1 and q 2 They represent the slopes of each two corresponding triplets.

[0031] Optionally, in one embodiment of the present application, preprocessing the time series data includes: performing Z-Score normalization processing on the time series data.

[0032] To achieve the above objectives, the second embodiment of the present application proposes a technology development stage judgment system based on adaptive similarity measurement, including the following modules:

[0033] a definition module, configured to segment the Gartner curve and define a plurality of technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve;

[0034] A generation module is used to search for technical keywords in a preset data set, generate time series data representing the change of the amount of scientific and technological data corresponding to the technical keywords over time based on the search results, and preprocess the time series data;

[0035] A feature extraction module is used to divide the preprocessed time series data into subsequences based on a preset detail resolution, connect the subsequences to generate a main sequence to extract features from the preprocessed time series data, and convert the main sequence into triples for storage;

[0036] a similarity measurement module, configured to measure the similarity between the converted main sequence and the curves representing each of the technology development stages by using dynamic time warping (DTW), and to rank the degree of similarity between the converted main sequence and each of the technology development stages;

[0037] The judgment module is used to determine the technology development stage with the highest similarity as the target technology development stage of the technology represented by the time series data.

[0038] Optionally, in one embodiment of the present application, the feature extraction module further includes:

[0039] A setting unit, configured to set the normalized time series data as a target sequence;

[0040] a marking unit, configured to mark the first and last endpoints of the target sequence as first key points, and connect the first and last endpoints to generate a baseline;

[0041] a calculation unit, configured to calculate a vertical distance from each data point in the target sequence, excluding the first and last endpoints, to the baseline, and to select a maximum distance from all vertical distances, and determine a data point corresponding to the maximum distance;

[0042] a segmentation unit, configured to determine a corresponding resolution distance based on the detail resolution, compare the maximum distance with the resolution distance, mark a data point corresponding to the maximum distance as a second key point if the maximum distance is greater than the resolution distance, and segment the target sequence into a first subsequence and a second subsequence based on the second key point;

[0043] The updating unit is used to update the target sequence into each re-divided subsequence, and control the marking unit, the calculation unit and the segmentation unit to repeatedly perform their own functions for each updated target sequence until no new subsequence is generated.

[0044] The technical solution provided by the embodiments of this application brings at least the following beneficial effects: This application first divides and defines the technology development stage based on the Gartner curve, then searches for papers based on technical keywords through a preset database to generate a time series curve of the change in the number of scientific papers. After completing the necessary data preprocessing operations, feature extraction and representation are performed on the target sample sequence, and the resolution of the sequence feature extraction is controlled by setting an adjustable threshold. Finally, the technology development stage is determined by similarity measurement with the Gartner curve. As a result, this application makes multi-scale judgments on the technology development stage based on the Gartner technology maturity curve, an adaptive time series data feature extraction algorithm, and a similarity measurement. When responding to the specific technology evaluation needs of different fields, the algorithm's detailed feature extraction granularity is controlled by manually adjusting the threshold, achieving adaptive processing of scientific and technological achievement time series data, which can be applied to the evaluation of technologies in various fields. Thus, a technology evaluation based on the combination of quantitative analysis of scientific and technological data and domain knowledge is achieved, promoting the combination of quantitative analysis methods of scientific and technological big data and artificial intelligence with the laws of scientific and technological development. The obtained evaluation results can provide a reference for scientific and technological intelligence analysis, and are conducive to realizing the corresponding system to increase the operability and visualization of the algorithm, improving the accuracy, applicability, and scientificity of the judgment of the technology development stage.

[0045] In order to implement the above embodiments, the third aspect of the present application also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the method for determining the technology development stage based on adaptive similarity measurement in the above embodiments is implemented.

[0046] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which

[0048] Figure 1 A flowchart of a method for determining a technology development stage based on an adaptive similarity measurement proposed in an embodiment of the present application;

[0049] Figure 2 A schematic diagram of a specific Gartner curve proposed in an embodiment of the present application;

[0050] Figure 3 A schematic diagram of a curve corresponding to a technology development stage proposed in an embodiment of the present application;

[0051] Figure 4 A schematic diagram of a curve corresponding to another technical development stage proposed in an embodiment of the present application;

[0052] Figure 5 A schematic diagram of a curve corresponding to another technical development stage proposed in an embodiment of the present application;

[0053] Figure 6 A schematic diagram of a curve corresponding to another technical development stage proposed in an embodiment of the present application;

[0054] Figure 7 A schematic diagram of a curve showing the change in the number of search results for a specific technical keyword over time proposed in an embodiment of the present application;

[0055] Figure 8 A flowchart of a specific method for extracting features from time series data proposed in an embodiment of the present application;

[0056] Figure 9 This is a flowchart of a specific method for measuring the similarity between a master sequence and a technology development stage through dynamic time warping (DTW) proposed in an embodiment of the present application;

[0057] Figure 10 A flowchart of a specific method for determining a technology development stage based on adaptive similarity measurement proposed in an embodiment of the present application;

[0058] Figure 11 This is a structural diagram of a technology development stage judgment system based on adaptive similarity measurement proposed in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0060] It should be noted that, in relation to the relevant technologies, the applicant has found that, in combination with the Hype Cycle's identification of technical problems faced in the technology development stage according to the field's technical evaluation needs, machine learning methods can be introduced to analyze scientific and technological data to improve the applicability of technology evaluation methods.

[0061] Similarity metrics are a typical method used in machine learning to solve time series problems. They can be divided into lockstep metrics and elastic similarity metrics. Elastic similarity metrics, such as Dynamic Time Warping (DTW), can find the optimal alignment between sequences by shrinking or stretching the time axis. However, in the actual task of technology assessment, relying on scientific and technological data such as scientific papers, simply judging the temporal trends of scientific and technological achievements based on similarity metrics suffers from poor scalability and difficulty coordinating in multi-task scenarios. Therefore, it is necessary to add adjustable control parameters to control the granularity of feature extraction from time series data, thereby realizing a multi-scale technology assessment method applicable to different fields.

[0062] Based on this, the applicant proposed a method and system for judging the technology development stage based on adaptive similarity measurement. This application is based on the basic idea of ​​the Gartner curve and similarity measurement, develops a multi-scale time series processing method and implements a corresponding system. It targets the needs of technology evaluation in different fields and makes it suitable for the evaluation of technology in various fields. It can achieve the technical analysis goal based on the combination of quantitative analysis of scientific and technological data and domain knowledge, which is conducive to the realization of technology development stage assessment and technology development risk control.

[0063] A method and system for determining a technology development stage based on an adaptive similarity measurement according to an embodiment of the present invention will be described below with reference to the accompanying drawings.

[0064] Figure 1 This is a flow chart of a method for determining a technology development stage based on adaptive similarity measurement proposed in an embodiment of the present application, as shown in FIG. Figure 1 The method comprises the following steps:

[0065] Step S101 : segmenting the Gartner curve to define multiple technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve.

[0066] Among them, the Gartner Hype Cycle is a technology maturity curve (Hype Cycle) created by Gartner Consulting. The technology maturity curve can be used to combine qualitative and quantitative methods to evaluate technology maturity. The Hype Cycle can be used to explore the basic laws of technological innovation evolution. It is an effective tool for comprehensively grasping the development trend of technological innovation and objectively evaluating the maturity of technological innovation.

[0067] Among them, such as Figure 2 As shown in the figure, the technology maturity curve divides the maturity of technology development into five periods according to the changes over time: technology germination period, expectation expansion period, bubble trough period, steady improvement period and actual production peak period. Figure 2The horizontal axis represents time, and the dotted lines in the figure divide the Gartner Hype cycle into different periods.

[0068] Specifically, the Gartner curve is segmented, and multiple technology development stages are defined based on the segmentation results, where each technology development stage includes two adjacent periods in the Gartner curve. Specifically, in the embodiments of the present application, the Gartner curve is used as the research basis, and the curve is segmented according to the technology development stages included in the Hype cycle curve. Each segmented curve segment includes two development stages in the Hype cycle to reflect the changing process of technology development. That is, in this embodiment, each two adjacent periods in the above five periods are segmented into a curve segment, and the technology development stage is defined based on each curve segment.

[0069] As an example, the segmented curve is Figures 3 to 6 As shown, Figure 3 It means "the stage from technology budding to expansion" (abbreviated as "budding-expansion"), Figure 4 It means "the trough period of technology expansion turning into bubble" (abbreviated as "expansion-bubble"), Figure 5 It means "the transition from the trough of bubble to the bright period" (abbreviated as "bubble-bright"), Figure 6 This represents the transition from the bright period to the peak of actual production (abbreviated as "bright-peak"). The horizontal axis of the graph represents time, and the vertical axis represents the value of the original Gartner curve. This division into stages of technological development facilitates subsequent research and verification based on actual data.

[0070] In the embodiment of the present application, after the division is performed in this way, the curve shape of each sub-segment can include the stage of the technology and the possible transition state. Compared with the division method of a single period, the division rule of the present application can include richer curve information while avoiding the neglect of key information. Figures 3 to 6 The segmentation results show that the curves of different periods have distinct shape characteristics, which are unique and can be distinguished from each other. Therefore, they can be used as a basis for judgment. This facilitates subsequent research using time series similarity metrics to determine the similarity between the target curve and each subsegment, thereby obtaining information about the period of the technology.

[0071] Step S102 : searching for technical keywords in a preset data set, generating time series data indicating how the quantity of scientific and technological data corresponding to the technical keywords changes over time based on the search results, and preprocessing the time series data.

[0072] The preset dataset is an experimental dataset used for evaluating the technology development stage, and the technology keywords are keywords that represent the technology to be evaluated. The dataset and technology keywords can be set based on the actual situation and needs of the technology development stage evaluation.

[0073] For example, this application can set the Web of Science (WOS) database as an experimental data set for the research and judgment of the technology development stage. Web of Science is a comprehensive information service platform that supports literature retrieval in natural sciences, social sciences, arts and humanities. Its data comes from various resources such as journals, books, patents, conference records and network resources. It can provide a scientific and sufficient basis for qualitative research on the technology development stage. Therefore, this application will use this as a data set for the online verification and implementation stage of technology evaluation. When it is necessary to judge the current stage of technological development of machine learning, you can search for the keywords "machine learning" or "machine learning" in WOS.

[0074] In the embodiment of the present application, the search results of specific technical keywords in the preset data set are used to generate the change data of the quantity data according to the time development, so as to generate the corresponding time series data set as the basis for technical judgment. Continuing with the above search for "machine learning" as an example, the search results in WOS, that is, the scientific and technological data corresponding to the technical keywords, are obtained, and the corresponding time series data set is generated according to the time series data set corresponding to the search results. Figure 7 As shown in the figure, the time series data of the number of scientific and technological data such as relevant papers, patents and works over the years, Figure 7 The horizontal axis is time and the vertical axis is the number of search results.

[0075] Furthermore, the time series data is preprocessed.

[0076] In one embodiment of the present application, due to the different research conditions in different fields, when conducting technical evaluations based on the actual conditions of each field and the actual needs of intelligence analysts, in order to offset the result differences caused by orders of magnitude differences, the generated time series data can be standardized.

[0077] As a possible implementation method, Z-Score normalization is performed on the generated time series data, and data of different magnitudes are converted into Z-Score values ​​of uniform measurement for comparison. For the generated target sequence S = (s1, s2, s3…s n ), calculate the standardized sequence S′=(s′1,s′2,…,s′ n ), calculated as follows:

[0078]

[0079] Among them, μ is the mean of each data in the original sequence S, and σ is the standard deviation of each data in the original sequence S.

[0080] It should be noted that this application may also select other annotation processing methods according to actual conditions, such as the maximum and minimum standardization method or the decimal calibration method. Before performing the standardization process, this application may also perform other preprocessing steps such as removing error data from the generated time series data, and complementing and supplementing the time series data with data contained in other data sets, so as to further improve the accuracy and comprehensiveness of the acquired time series data, which is conducive to improving the accuracy of judgments in subsequent technical development stages.

[0081] Step S103 : divide the preprocessed time series data into subsequences based on a preset detail resolution, connect the subsequences to generate a main sequence to extract features from the preprocessed time series data, and convert the main sequence into triples for storage.

[0082] Specifically, feature extraction is performed on the preprocessed time series data. Since the characteristics of technological development in different fields are different, the time series data corresponding to different fields also have different characteristics. In order to meet the research needs of intelligence analysts in different scenarios, this application sets the control parameters of the resolution of time series data feature extraction, that is, the preset detail resolution, to achieve the extraction of curve features through parameter adjustment. The detailed fluctuation characteristics of the time series data can be ignored as needed, and only its main trends are recorded, which is suitable for the analysis of overall trends.

[0083] In specific implementation, in one embodiment of the present application, for the normalized sequence S′, the present application controls the resolution of feature extraction through a preset detail resolution parameter, namely, the parameter Reso. When a larger Reso value is set, the trend of the sequence can be roughly extracted. When a relatively small Reso value is set, the local features of the sequence can be grasped in more detail. In order to more clearly illustrate the specific implementation process of feature extraction based on the detail resolution parameter of the present application, a specific method for feature extraction of time series data proposed in this embodiment is exemplified. Figure 8 As shown, the method includes the following steps:

[0084] S1: Set the normalized time series data as the target sequence.

[0085] Specifically, the preset detail resolution, i.e., the given Reso value, is obtained. Assume that the standardized sequence S′=(s′1,s′2,…,s′ n ), with S′ as the target sequence for current subsequence division.

[0086] S2: Mark the first and last endpoints of the target sequence as the first key points, and connect the first and last endpoints to generate a baseline.

[0087] Specifically, the first and last endpoints s′1 and s′ of the sequence S′n Mark them as key points and connect them to get the baseline l 1n .

[0088] S3: Calculate the vertical distance from each data point in the target sequence except the first and last endpoints to the baseline, select the maximum distance from all vertical distances, and determine the data point corresponding to the maximum distance.

[0089] Specifically, calculate the remaining points s′2,…,s′ in S′ one by one n-1 to l 1n The vertical distance (s′ i ,l 1n ), where i = 2, 3, ..., n-1, find the point s' that maximizes the distance among all the calculated vertical distances d And record it, and record the distance corresponding to the point as distance max =(s′ d ,l 1n ).

[0090] S4: Determine the corresponding resolution distance based on the detail resolution, compare the maximum distance with the resolution distance, and if the maximum distance is greater than the resolution distance, mark the data point corresponding to the maximum distance as the second key point, and segment the target sequence into the first subsequence and the second subsequence according to the second key point.

[0091] Specifically, for distance max To make a judgment, first determine the corresponding resolution distance based on the detail resolution Reso , then compare the maximum distance with the resolution distance. If distance max >distance Reso , then s′ d Mark as the key point, and split S′ into L1(s′1,s′2,…,s′ d ) and L2(s′ d+1 ,s′ d+2 ,…,s′ n ), that is, the target sequence is divided into the first subsequence and the second subsequence according to the second key point. If the distance max ≤distance Reso , then only s′1 and s′ are retained n are important points, and the rest are ignored.

[0092] Among them, distance Reso It is a pre-set control parameter that can determine the corresponding resolution distance based on the detail resolution ResoIn specific implementation, you can first compare and filter out the maximum distance among all vertical distances. max and minimum distance min Calculate the distance based on the maximum distance, minimum distance and preset Reso value Reso In the embodiment of the present application, the parameter Reso is represented as the distance from all points in the current sequence except the endpoints to the baseline (distance Reso ) and the maximum distance between them max and minimum distance min The resolution distance can be calculated by the following formula:

[0093]

[0094] Among them, Reso is the preset detail resolution, distance max is the maximum distance, distance min is the minimum distance, distance Reso is the resolution distance, in Reso, distance max and distance min When the value is known, the distance can be calculated by substituting it into the above formula Reso .

[0095] As can be seen from the above, when Reso is set to 1, only the two endpoints of the target sequence are retained as key points, and the rest of the points are ignored; when Reso is set to 0, all points in the current sequence are considered key points. Therefore, by adjusting the Reso value, the detail resolution of the target sequence can be adjusted.

[0096] S5: Update the target sequence to each re-divided subsequence respectively, and repeat steps S2 to S4 for each updated target sequence until no new subsequence is generated.

[0097] Specifically, L1(s′1,s′2,…,s′ d ) and L2(s′ d+1 ,s′ d+2 ,…,s′ n ) is the target sequence for the next round of subsequence division. Repeat the above steps S2 to S4 for L1 and L2 respectively to divide L1 and L2 into subsequences respectively, and through iteration, divide the subsequences generated in each round through the method of steps S2 to S4 until no new subsequences are generated, that is, there are no key points in the subsequences generated in the previous round, and then the iteration ends.

[0098] Furthermore, after the iteration is completed, each subsequence is connected to generate a main sequence, that is, all subsequences are connected end to end to obtain a main sequence L generated after feature extraction of the preprocessed time series data.

[0099] Therefore, through the method of extracting features from time series data in the embodiment of the present application, feature extraction can be performed according to the set detail resolution parameters when the sequence of time series data is too long or the local fluctuations are too much, effectively avoiding the influence of noise or unimportant detail features on the effect and efficiency of the final similarity measurement, and improving the applicability of the method for judging the technical development stage of the present application in practical applications.

[0100] Furthermore, the main sequence is converted into triples for storage. In this application, each subsequence is represented as a triple, thereby converting the connected main sequence into triples for storage as a basis for subsequent similarity measurement.

[0101] As a possible implementation method, converting the main sequence into a triplet includes the following steps: first, calculating the mean of the data points contained in each subsequence, then calculating the slope of the line connecting the first and last endpoints of each subsequence, and determining the index of the first endpoint of each subsequence in the target sequence. Then, based on the index, mean, and slope of each subsequence, the corresponding subsequence is represented as a triplet.

[0102] For example, the main sequence L is represented by calculating the mean and slope of each subsequence in the main sequence L. The main sequence L containing k subsequences can be written as: L = (<p1,M1,q1> ,<p2,M2,q2> ,…, <p k ,M k ,q k >), where p represents the index of the starting point of the current subsequence in the original sequence, M represents the mean of the data points contained in the subsequence, and q represents the slope of the line connecting the starting point and the end point of the subsequence, which can be calculated according to the actual situation of the time series data.

[0103] Step S104 : perform similarity measurement between the converted main sequence and the curves representing each technological development stage through dynamic time warping (DTW), and rank the similarity between the converted main sequence and each technological development stage.

[0104] Specifically, based on the technological development stages derived from the Hype cycle classification in step S101, the similarity between the feature-extracted sample sequence and each technological development stage is measured. It should be noted that because the length of the feature-extracted main sequence may differ from the length of the curves representing each technological development stage, dynamic time warping (DTW) is used as a metric to measure the similarity between the main sequence and the curves representing each development stage.

[0105] Among them, DTW can calculate the similarity between two sequences of different lengths and rhythms. DTW can automatically distort the sequence (for example, extending and shortening the time series) to make the shapes of the two sequences as consistent as possible, making it easier to calculate the similarity between the two sequences.

[0106] In order to more clearly illustrate the specific implementation process of measuring the similarity between the main sequence and each technology development stage by DTW in this application, a specific method of measuring the similarity between the main sequence and the technology development stage by dynamic time warping DTW proposed in the embodiment of this application is exemplified. Figure 9 As shown, the method includes the following steps:

[0107] S901: Convert the sequence segments corresponding to each technology development stage into triples.

[0108] Specifically, since the morphological characteristics of the hype cycle are obvious and there are no excessive local details, there is no need to extract its features. Therefore, the sequence segments corresponding to each technological development stage can be directly converted into triplets for similarity comparison.

[0109] For example, the time series data of generated technical keywords is used as the time series S 1 , taking any current technology development stage for comparison as the time series S 2 , S 1 The result after feature extraction and triple conversion is S 2 The result after triple conversion is

[0110] S902: Construct a distance matrix based on the converted main sequence and the sequence segments corresponding to any converted technology development stage, and determine the optimal path of the distance matrix through dynamic time warping (DTW).

[0111] Specifically, according to L 1 and L 2The number of triplets in determines the size of the distance matrix, i.e., an m×n distance matrix is ​​constructed. DTW is then used to find the optimal path through the matrix grid and record the path. The optimal path can be the path with the highest adjustment efficiency and alignment.

[0112] S903: Align the converted main sequence and the triplets in the sequence segment corresponding to any converted technology development stage according to the optimal path, and calculate the distance between every two corresponding triplets after the alignment.

[0113] Specifically, according to the optimal path obtained in S902, L 1 and L 2 Align the triplets in and calculate the aligned L 1 and L 2 As a possible implementation, the distance between each two corresponding triplets after alignment can be calculated using the following formula:

[0114]

[0115] Among them, M 1 and M 2 Represents the mean of each two corresponding triples, q 1 and q 2 They represent the slopes of each two corresponding triplets.

[0116] For example, suppose the two triplets currently being calculated are and The distance calculation formula is:

[0117]

[0118] S904: Calculate the sum of all distances, and use the sum as the similarity between the main sequence and any converted technology development stage.

[0119] Specifically, after calculating the distance between each two corresponding triples in sequence using the calculation formula in S903, all distances are summed up, and the result represents S 1 and S 2 similarity.

[0120] Furthermore, the next stage of technological development is selected as the time series S 2 ,Through this similarity measurement method, the similarity between the main sequence and each technology development stage can be measured in turn, that is, the main sequence can be measured with the curves of the "germination-expansion", "expansion-bubble", "bubble-bright" and "bright-peak" stages respectively.

[0121] Then, the similarity measurement results are sorted, that is, after comparing the various similarity measurement results, the similarity between the main sequence and each technology development stage is sorted in descending order.

[0122] Step S105 : determining the technology development stage with the highest similarity as the target technology development stage of the technology represented by the time series data.

[0123] Specifically, the technological development with the highest degree of similarity to the main sequence is obtained based on the sorting results, and the technological development stage with the highest degree of similarity is used as the development stage of the technology represented by the searched time series data, that is, as the development stage of the technology to be judged, thereby realizing the judgment of the technological development stage.

[0124] In one embodiment of the present application, the generated judgment results can be displayed on the human-computer interaction interface. For example, on the interactive interface of the application for judging the technology development stage, a technology keyword search box and a development stage judgment result display area are set. After the user searches for technology keywords according to actual needs, the development stage of the corresponding technology is displayed in the display area after calculation, which improves the convenience of users in operating and obtaining judgment results.

[0125] Therefore, this method judges the technology development stage by measuring the similarity between the time series data of WOS technology retrieval results and the GartnerHype Cycle that divides technology development stages. Specifically, by using adjustable parameters to control the resolution of time series data feature extraction, multi-scale time series data feature extraction is performed on technology-related time series data, thereby improving the accuracy of judgment.

[0126] In summary, the method for determining the technology development stage based on adaptive similarity measurement in the embodiment of the present application first divides and defines the technology development stage based on the Gartner curve, then searches for papers based on technical keywords through a preset database to generate a time series curve of the number of scientific papers. After completing the necessary data preprocessing operations, feature extraction and representation are performed on the target sample sequence, and the resolution of the sequence feature extraction is controlled by setting an adjustable threshold. Finally, the technology development stage is determined by performing similarity measurement with the Gartner curve. Thus, the method makes multi-scale judgments on the technology development stage based on the Gartner technology maturity curve and similarity measurement. When responding to the specific technology evaluation needs of different fields, the algorithm details feature extraction granularity is controlled by manually adjusting the threshold, realizing adaptive processing of scientific and technological achievement time series data, which can be applied to the evaluation of technologies in various fields. Thus, a technology evaluation based on the combination of quantitative analysis of scientific and technological data and domain knowledge is realized, which promotes the combination of quantitative analysis methods of scientific and technological big data and artificial intelligence with the laws of scientific and technological development, and is conducive to realizing the corresponding system to increase the operability and visualization of the algorithm, thereby improving the accuracy, applicability and convenience of determining the technology development stage.

[0127] In order to more clearly illustrate the method for determining the technology development stage based on the adaptive similarity metric in the embodiment of the present application, the following is combined with Figure 10 A specific embodiment is used for illustration. Figure 10 As shown, the method includes the following steps:

[0128] S10: Define the technology development stage based on the Gartner Hype cycle.

[0129] In this step, the Gartner curve is used as the research basis, and the curve is divided according to the technology development stages included in the Hype cycle curve. Each curve contains two development stages in the Hype cycle, reflecting the changing process of technology development. The main stages include "technology germination to expansion stage" (abbreviated as "germination-expansion"), "technology expansion to bubble trough period" (abbreviated as "expansion-bubble"), "bubble trough period to bright period" (abbreviated as "bubble-bright"), and "bright period to actual production peak period" (abbreviated as "bright-peak"). The division of technology development stages lays the foundation for further research and verification based on actual data.

[0130] S20: Generate time series data based on scientific and technological big data according to the searched technical keywords, and preprocess the data.

[0131] In this step, we generate quantitative data from WOS search results for specific technical keywords and then analyze their changes over time to create a corresponding time series dataset, which serves as the basis for technology judgment. Due to the varying research conditions in different fields, we perform Z-score normalization on the generated time series data to offset differences in results caused by orders of magnitude.

[0132] S30: Extract and represent features of time series data based on detail resolution.

[0133] In this step, to meet the research needs of different technical fields, a multi-scale similarity measurement method is proposed. This method allows for adjusting thresholds based on actual conditions and extracting features at different scales of the curve. The time series data is divided into subsequences based on the detail resolution (Reso). All subsequences that meet the requirements are iteratively extracted and concatenated end-to-end to form a master sequence. The extracted master sequence is then stored as a triplet, which serves as the basis for the similarity measurement.

[0134] S40: Measure the similarity between the main sequence and the Gartner curve.

[0135] In this step, the research samples are compared with the Hype cycle and ranked according to the degree of similarity with different stages.

[0136] Specifically, based on the segmented sequence after Hypecycle division, the similarity between the sample sequence after feature extraction and different technical segments is measured respectively. In order to meet the similarity measurement between time series data of different lengths, dynamic time warping (DTW) is used as the measurement method, and the sequences are sorted according to the degree of similarity with different stages.

[0137] S50: The R&D judgment system realizes the research and judgment of the technology development stage.

[0138] In this step, a corresponding system is implemented, using the development stage with the highest degree of similarity as the development stage of the technology represented by the time series sample to determine the development stage of the technology. To make the operation and implementation of this method more convenient, a corresponding system is developed to display the overall results of the analysis.

[0139] This method divides and defines technology development stages based on the Gartner Hype Cycle. It retrieves papers based on technical keywords using the WOS database to generate a time series curve of the number of scientific papers. After completing necessary data preprocessing operations, it extracts and represents features from the target sample sequence. It controls the resolution of sequence feature extraction by setting an adjustable threshold. It determines the technology development stage by measuring similarity with the Gartner Hype Cycle, and achieves a corresponding system to increase the operability and visualization of the algorithm.

[0140] In order to implement the above embodiment, the present application also proposes a technology development stage judgment system based on adaptive similarity measurement. Figure 11 This is a structural diagram of a technology development stage judgment system based on adaptive similarity measurement proposed in an embodiment of the present application.

[0141] like Figure 11 As shown, the technology development stage judgment system based on adaptive similarity measurement includes: a definition module 100, a generation module 200, a feature extraction module 300, a similarity measurement module 400 and a judgment module 500.

[0142] The definition module 100 is used to segment the Gartner curve and define multiple technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve.

[0143] The generation module 200 is used to search for technical keywords in a preset data set, generate time series data indicating the change in the amount of scientific and technological data corresponding to the technical keywords over time based on the search results, and preprocess the time series data.

[0144] The feature extraction module 300 is used to divide the preprocessed time series data into subsequences based on a preset detail resolution, connect the subsequences to generate a main sequence to extract features from the preprocessed time series data, and convert the main sequence into triples for storage.

[0145] The similarity measurement module 400 is used to measure the similarity between the converted main sequence and the curves representing each technology development stage through dynamic time warping (DTW), and to rank the similarity between the converted main sequence and each technology development stage.

[0146] The judgment module 500 is used to determine the technology development stage with the highest similarity as the target technology development stage of the technology represented by the time series data.

[0147] Optionally, in one embodiment of the present application, the feature extraction module 300 further includes: a setting unit for setting the time series data after normalization as a target sequence; a marking unit for marking the first and last endpoints of the target sequence as first key points, and connecting the first and last endpoints to generate a baseline; a calculation unit for calculating the vertical distance from each data point in the target sequence except the first and last endpoints to the baseline, and screening out the maximum distance from all vertical distances to determine the data point corresponding to the maximum distance; a segmentation unit for determining the corresponding resolution distance based on the detail resolution, comparing the maximum distance with the resolution distance, if the maximum distance is greater than the resolution distance, marking the data point corresponding to the maximum distance as a second key point, and segmenting the target sequence into a first subsequence and a second subsequence according to the second key point; an updating unit for updating the target sequence to the first subsequence and the second subsequence, respectively, and controlling the marking unit, the calculation unit and the segmentation unit to repeatedly execute their own functions for each updated target sequence until no new subsequence is generated.

[0148] Optionally, in one embodiment of the present application, the segmentation unit is specifically configured to: screen out a maximum distance and a minimum distance from all vertical distances;

[0149] The resolution distance is calculated using the following formula:

[0150]

[0151] Among them, Reso is the preset detail resolution, distance max is the maximum distance, distance min is the minimum distance, distance Reso is the resolution distance.

[0152] Optionally, in one embodiment of the present application, the feature extraction module 300 is specifically used to: calculate the mean of the data points contained in each subsequence; calculate the slope of the line connecting the first and last endpoints of each subsequence; determine the index of the first endpoint of each subsequence in the target sequence, and represent the corresponding subsequence as a triple according to the index, mean and slope of each subsequence.

[0153] Optionally, in one embodiment of the present application, the similarity measurement module 400 is specifically used to: convert the sequence segments corresponding to each technological development stage into triple form; construct a distance matrix based on the converted main sequence and the sequence segments corresponding to any converted technological development stage, and determine the optimal path of the distance matrix through dynamic time warping DTW; align the triplets in the sequence segments corresponding to the converted main sequence and any converted technological development stage according to the optimal path, and calculate the distance between each two corresponding triplets after alignment; calculate the sum of all distances, and take the sum as the similarity between the main sequence and any converted technological development stage.

[0154] Optionally, in one embodiment of the present application, the similarity measurement module 400 is specifically configured to calculate the distance between each two corresponding triples after alignment using the following formula:

[0155]

[0156] Among them, M 1 and M 2 Represents the mean of each two corresponding triples, q 1 and q 2 They represent the slopes of each two corresponding triplets.

[0157] Optionally, in one embodiment of the present application, the generation module 200 is further configured to perform Z-Score normalization processing on the time series data.

[0158] In summary, the technology development stage judgment system based on adaptive similarity measurement in the embodiment of the present application performs multi-scale judgments on the technology development stage based on the Gartner technology maturity curve, an adaptive time series data feature extraction algorithm, and similarity measurement. When responding to specific technology evaluation needs in different fields, the system manually adjusts the threshold to control the granularity of the algorithm's detailed feature extraction, thereby achieving adaptive processing of scientific and technological achievement time series data and being applicable to the evaluation of technologies in various fields. Thus, it achieves technology evaluation based on the combination of quantitative analysis of scientific and technological data and domain knowledge, promotes the combination of quantitative analysis methods of scientific and technological big data and artificial intelligence with the laws of scientific and technological development, and is conducive to realizing the corresponding system to increase the operability and visualization of the algorithm, thereby improving the accuracy, applicability, and scientific nature of the judgment of the technology development stage.

[0159] In order to implement the above-mentioned embodiments, the present invention also proposes a non-temporary computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the method for determining the technology development stage based on adaptive similarity measurement described in the embodiment of the first aspect of the present application.

[0160] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, if schematic expressions of the above terms are used in multiple embodiments or examples, it does not mean that these embodiments or examples are the same. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine the different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0161] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0162] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0163] The logic and / or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include the following: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0164] It should be understood that various parts of the present application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used to implement: a discrete logic circuit having a logic gate circuit for implementing a logic function on a data signal, an application-specific integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0165] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0166] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0167] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A method for determining the technology development stage based on adaptive similarity measurement, characterized in that: The following steps are involved: Segmenting the Gartner curve to define multiple technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve; Searching for technical keywords in a preset data set, generating time series data indicating how the amount of scientific and technological data corresponding to the technical keywords changes over time based on the search results, and preprocessing the time series data; Divide the preprocessed time series data into subsequences based on a preset detail resolution, connect the subsequences to generate a main sequence to extract features from the preprocessed time series data, and convert the main sequence into a triplet for storage, wherein the preset detail resolution is a control parameter for setting the resolution of time series data feature extraction; performing similarity measurement between the converted main sequence and the curves representing each of the technology development stages by dynamic time warping (DTW), and ranking the degree of similarity between the converted main sequence and each of the technology development stages; The technology development stage with the highest degree of similarity is determined as the target technology development stage of the technology represented by the time series data.

2. The judgment method according to claim 1, characterized in that: The sub-sequence division of the pre-processed time series data based on a preset detail resolution includes: S1: Set the time series data after normalization as the target sequence; S2: Mark the first and last endpoints of the target sequence as first key points, and connect the first and last endpoints to generate a baseline; S3: Calculate the vertical distance from each data point in the target sequence, excluding the first and last endpoints, to the baseline, and select the maximum distance from all vertical distances to determine the data point corresponding to the maximum distance; S4: determining a corresponding resolution distance based on the detail resolution, comparing the maximum distance with the resolution distance, and if the maximum distance is greater than the resolution distance, marking a data point corresponding to the maximum distance as a second key point, and segmenting the target sequence into a first subsequence and a second subsequence based on the second key point; S5: respectively updating the target sequence to each re-divided subsequence, and repeating steps S2 to S4 for each updated target sequence until no new subsequence is generated.

3. The judgment method according to claim 2, characterized in that: The determining the corresponding resolution distance based on the detail resolution includes: Filtering out the maximum distance and the minimum distance from all the vertical distances; The resolution distance is calculated by the following formula: Among them, Reso is the preset detail resolution, distance max is the maximum distance, distance min is the minimum distance, distance Reso is the resolution distance.

4. The judgment method according to claim 2, characterized in that: The step of converting the main sequence into a triple form for storage includes: Calculate the mean of the data points contained in each subsequence; Calculating the slope of the line connecting the first and last endpoints of each subsequence; The index of the first endpoint of each subsequence in the target sequence is determined, and the corresponding subsequence is represented as a triple according to the index, mean, and slope of each subsequence.

5. The judgment method according to claim 1, characterized in that: The similarity measurement between the main sequence and each of the technology development stages by using dynamic time warping (DTW) includes: Converting the sequence segments corresponding to each of the technology development stages into triples; constructing a distance matrix based on the converted main sequence and the sequence segments corresponding to any converted technology development stage, and determining an optimal path of the distance matrix by dynamic time warping (DTW); Aligning the triplets in the converted main sequence and the sequence segments corresponding to any converted technology development stage according to the optimal path, and calculating the distance between every two corresponding triplets after the alignment; The sum of all distances is calculated, and the sum is taken as the similarity between the main sequence and any of the converted technology development stages.

6. The judgment method according to claim 5, characterized in that: The distance between each two corresponding triplets after alignment is calculated using the following formula: Among them, M 1 and M 2 Represents the mean of each two corresponding triples, q 1 and q 2 They represent the slopes of each two corresponding triplets.

7. The judgment method according to claim 1, characterized in that: The preprocessing of the time series data includes: performing Z-Score normalization processing on the time series data.

8. A technology development stage judgment system based on adaptive similarity measurement, characterized in that: include: a definition module, configured to segment the Gartner curve and define a plurality of technology development stages, wherein each technology development stage includes two adjacent periods in the Gartner curve; A generation module is used to search for technical keywords in a preset data set, generate time series data representing the change of the amount of scientific and technological data corresponding to the technical keywords over time based on the search results, and preprocess the time series data; A feature extraction module is configured to divide the preprocessed time series data into subsequences based on a preset detail resolution, connect the subsequences to generate a main sequence to extract features from the preprocessed time series data, and convert the main sequence into a triplet for storage, wherein the preset detail resolution is a control parameter for the resolution of feature extraction of the time series data; a similarity measurement module, configured to measure the similarity between the converted main sequence and the curves representing each of the technology development stages by using dynamic time warping (DTW), and to rank the degree of similarity between the converted main sequence and each of the technology development stages; The judgment module is used to determine the technology development stage with the highest similarity as the target technology development stage of the technology represented by the time series data.

9. The judgment system according to claim 8, characterized in that: The feature extraction module further includes: A setting unit is used to set the time series data after normalization as the target sequence; a marking unit, configured to mark the first and last endpoints of the target sequence as first key points, and connect the first and last endpoints to generate a baseline; a calculation unit, configured to calculate a vertical distance from each data point in the target sequence, excluding the first and last endpoints, to the baseline, and to select a maximum distance from all vertical distances, and determine a data point corresponding to the maximum distance; a segmentation unit, configured to determine a corresponding resolution distance based on the detail resolution, compare the maximum distance with the resolution distance, mark a data point corresponding to the maximum distance as a second key point if the maximum distance is greater than the resolution distance, and segment the target sequence into a first subsequence and a second subsequence based on the second key point; The updating unit is used to update the target sequence into each re-divided subsequence, and control the marking unit, the calculation unit and the segmentation unit to repeatedly perform their own functions for each updated target sequence until no new subsequence is generated.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for determining the technology development stage based on the adaptive similarity measurement as described in any one of claims 1 to 7 is implemented.

Images