Method and device for predicting quotation track mode, medium and product

Abstract

The invention discloses a citation trajectory mode prediction method, device, medium and product, and relates to the field of citation trajectory mode prediction.The method comprises the steps that complete citation data of papers in different fields are obtained, and a paper citation time sequence is constructed; taking a paper citation time sequence as a citation track; clustering by adopting a K-Shape clustering algorithm, an improved elbow rule and a contour coefficient method according to the quotation tracks of the papers in different fields; each clustering category is a citation trajectory mode corresponding to the citation trajectory; training an LSTM (Long Short Term Memory) neural network model adopting an adaptive sliding window mechanism by utilizing the clustered citation tracks; and utilizing the trained LSTM long short term memory neural network model to predict a citation trajectory mode and a future trend of a to-be-evaluated paper. According to the method, various existing modes of the paper quotation trajectory can be efficiently identified and deeply analyzed, and probability inference is carried out on future conditions of related paper quotation trajectories.

Description

Technical Field

[0001] The present application relates to the field of citation trajectory pattern prediction, and in particular to a citation trajectory pattern prediction method, device, medium and product. Background Art

[0002] Citation trajectory pattern prediction is an efficient tool for evaluating the future influence trend of papers. Each pattern of citation trajectory maps the evolution law of the influence of a specific type of paper. By identifying the trend of these patterns over time, we can not only evaluate the historical influence of a single paper and the intrinsic characteristics of various types of knowledge dissemination, but also predict its future academic influence and evolution trajectory. However, with the surge in the amount of papers and citation data, traditional analysis methods such as manually set parameters, simple curve fitting, and reliance on fixed standards face significant limitations. They do not have the ability to capture the diversity and complexity of citation patterns, and it is difficult to identify and predict papers that do not follow typical growth and decline patterns, especially papers that did not receive enough attention in the early stages but later showed significant influence. When dealing with large-scale, heterogeneous citation data, traditional methods often ignore the microscopic changes in the trajectory of paper citations, resulting in inaccurate evaluation results.

[0003] Therefore, in order to fill the gap in current technology in the field of complex citation dynamic analysis, it is urgent to provide a citation trajectory pattern prediction method. Summary of the invention

[0004] The purpose of this application is to provide a citation trajectory pattern prediction method, device, medium and product that can efficiently identify and deeply analyze various existing patterns of paper citation trajectories, and make probabilistic inferences about the future situation of related paper citation trajectories.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a citation trajectory pattern prediction method, the citation trajectory pattern prediction method comprising: Obtain complete citation data of papers in different fields and construct a time series of paper citations; and use the time series of paper citations as citation tracks; the complete citation data includes: the number of citations of the paper, the citation list, and the citation time; According to the citation tracks of papers in different fields, K-Shape clustering algorithm and improved elbow rule and silhouette coefficient method are used for clustering; each cluster category is the citation track pattern corresponding to the citation track; The clustered citation trajectories are used to train the LSTM long short-term memory neural network model with an adaptive sliding window mechanism; The trained LSTM long short-term memory neural network model is used to predict the citation trajectory patterns and future trends of the papers to be evaluated.

[0006] Optionally, the acquisition of complete citation data of papers in different fields and the construction of paper citation time series specifically include: Get complete citation data for papers in different fields; Preprocessing the complete citation data; the preprocessing includes: data cleaning, denoising, interpolation or resampling, missing value processing and dimensionless processing; Construct the paper citation time series based on the preprocessed complete citation data.

[0007] Optionally, clustering is performed using a K-Shape clustering algorithm and an improved elbow rule and silhouette coefficient method according to the citation tracks of papers in different fields, specifically including: Using the formula Determine the elbow point of the citation trajectory within the cluster; Using the formula Determine the silhouette coefficient of the citation trajectories within the cluster; use Determine the optimal number of clusters; in, is the elbow point, is the sum of squared errors, Elbow is the elbow rule function, is the adjustment factor obtained based on the citation trajectory, , is a volatility indicator determined by the standard deviation of citation trajectories, is the fluctuation frequency of the citation trajectory, is the long-term trend of the citation trajectory, , and are weights, , is the silhouette coefficient, To measure the morphological similarity between the citation trajectory and the cluster center, To measure the morphological difference between the citation trajectory and the nearest non-belonging cluster center, is the optimal number of clusters, The function indicates that the number of clusters with the largest average silhouette coefficient among the candidate clusters preliminarily determined according to the elbow rule is taken as the optimal number of clusters.

[0008] Optionally, measure the morphological similarity between the citation trajectory and the cluster center Specifically include: ; in, is the citation track, Citation Track With cluster center The shape similarity between For all citation tracks and cluster centers The maximum value of the distance.

[0009] Optionally, measure the morphological difference between the citation trajectory and the nearest non-belonging cluster center Specifically include: ; in, Citation Track The nearest cluster center outside the cluster to which it belongs.

[0010] Optionally, the method of using the trained LSTM long short-term memory neural network model to predict the citation trajectory pattern and future trend of the paper to be evaluated specifically includes: Using the formula Identify papers to be evaluated Citation track sequence ; is the output function of the trained LSTM long short-term memory neural network model, are the parameters of the trained LSTM long short-term memory neural network model; Using the formula Determine the citation track sequence The citation trajectory pattern of candidate attributions ;in, Citation track sequence Belong to Citation Track Pattern The probability of , is the probability distribution function of the classifier output, is a mathematical function; use Determine the citation track sequence Future Trends ;in, It is a function that predicts future trends based on current citation trajectory patterns and feature representations.

[0011] Optionally, the method uses the trained LSTM long short-term memory neural network model to predict the citation trajectory pattern and future trend of the paper to be evaluated, and then further includes: Visualize the citation trajectory patterns and future trends of the papers being evaluated.

[0012] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the citation trajectory pattern prediction method.

[0013] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which implements the citation trajectory pattern prediction method when executed by a processor.

[0014] In a fourth aspect, the present application provides a computer program product, including a computer program, which implements the citation trajectory pattern prediction method when executed by a processor.

[0015] According to the specific embodiments provided in this application, this application discloses the following technical effects: The present application provides a citation trajectory pattern prediction method, device, medium and product, which clusters the citation trajectories of papers in different fields using the K-Shape clustering algorithm and the improved elbow rule and silhouette coefficient method; each clustering category (cluster) is a citation trajectory pattern corresponding to the citation trajectory, and the optimized clustering method is combined with advanced time series analysis technology to achieve accurate prediction and in-depth understanding of the paper citation trajectory pattern; while improving clustering accuracy, citation trajectories with similar morphological characteristics are effectively summarized; an adaptive sliding window mechanism is used in the LSTM long short-term memory neural network model to capture long-term dependencies in time series data, and the ability to capture dynamic information of citation trajectories is enhanced through dynamic sliding window technology; adaptive dynamic analysis of periodic fluctuations, trend changes and the immediate impact of emergencies on citation rates of paper citation behaviors. It reduces the reliance on manually annotated data and simplifies the data preprocessing process, thereby reducing costs while improving the flexibility and adaptability of the system. This application can effectively process large-scale heterogeneous data and capture the diversity of citation trajectories. It can not only identify typical patterns of "evergreen" (i.e., the paper continues to receive stable citations after publication) and "short-lived" (high initial citation rate but then rapidly decays), but also discover "delayed blooming" (few citations in the early stage, gradually receiving more attention over time) and other more complex or rare citation patterns, comprehensively reflecting the rich patterns of knowledge dissemination and providing powerful tools for scientific research evaluation, knowledge diffusion research and policy making. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0017] Figure 1 Schematic diagram of a flow chart of a citation trajectory pattern prediction method in one embodiment of the present application.

[0018] Figure 2The figure is a schematic diagram of the clustering process of the K-shape clustering algorithm in one embodiment of the present application.

[0019] Figure 3 This is a schematic diagram of the LSTM long short-term memory neural network model structure. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0022] In an exemplary embodiment, Figure 1 As shown, a citation trajectory pattern prediction method is provided, which includes the following steps 201 to 208. Among them: S101, obtain the complete citation data of papers in different fields, and construct the paper citation time series; and use the paper citation time series as the citation trajectory ; The complete citation data includes: the number of citations of the paper, the citation list and the citation time.

[0023] Citation Track It is defined as: a paper from the date of publication to Years, citation count per year ,This sequence represents the dynamic evolution of the paper’s influence over time and is a quantitative representation of the paper’s citation trajectory.

[0024] S101 specifically includes: Get complete citation data for papers in different fields; The complete citation data is preprocessed; the preprocessing includes: data cleaning, denoising, interpolation or resampling, missing value processing and dimensionless processing; the missing value processing includes: forward filling, backward filling or average filling; the dimensionless processing is to perform maximum and minimum normalization (also known as Min-Max normalization) on the data to ensure that all data dimensions are scaled to the interval [0, 1] to eliminate the impact caused by different citation bases between different papers, thereby ensuring that the citation trajectory sequence only reflects the relative change trend of the paper's influence. The specific formula is as follows: ; In the formula, Indicates The paper is published in The number of citations per year, Indicates The lowest number of citations for the papers during this period, Indicates The highest citation frequency of the papers during this period, Indicates The paper is published in The normalized value of the citation frequency in the year, and the processed citation trajectory sequence is expressed as .

[0025] Construct the paper citation time series based on the preprocessed complete citation data.

[0026] S102, based on the citation tracks of papers in different fields, the K-Shape clustering algorithm and the improved elbow rule and silhouette coefficient method are used for clustering; each cluster category is the citation track pattern corresponding to the citation track; S102 specifically includes: S21, clustering based on K-shape clustering algorithm; The specific process of S21 is as follows: S211, calculate the similarity between citation trajectories: use the K-shape clustering algorithm based on the similarity of the fluctuation shape of the citation trajectory for unsupervised analysis. Considering the displacement or scaling differences on the time axis, the shape-based distance SBD algorithm is used to compare the similarity of two citation trajectories, denoted as To evaluate any two citation tracks and citation tracks The shape similarity between them ensures accurate matching in complex patterns. The specific formula is as follows: ; In the formula, is the length of the time series; is a potential shape transformation of the time series points, such as normalization or smoothing.

[0027] S212, calculate the shape distance matrix: use the K-Shape algorithm to calculate the shape distances between all citation track pairs and obtain the shape distance matrix ,in, Indicates the paper and the shape distance between paper j; S213, Citation Trajectory Clustering: Based on Shape Distance Matrix , using K-Shape clustering algorithm to divide the papers into groups, denoted as . Each cluster Represents a collection of papers with similar citation trajectories, with each category reflecting a unique citation trajectory pattern.

[0028] ; In the formula, Representative A cluster contains a collection of papers with similar citation patterns, i.e., citation trajectory patterns.

[0029] S22, optimize the clustering effect of K-shape algorithm by improving elbow rule and silhouette coefficient method, and enhance the number of citation pattern categories K reliability; S22 specifically includes: S221, in clustering algorithms, the function of the elbow rule is to help determine the optimal number of clusters. Based on this, this application introduces this method into the K-shape algorithm to improve the clustering performance of citation trajectory data with significant dynamic characteristics. Specifically, this application defines a dynamic coefficient related to the amplitude and frequency of time series fluctuations , and apply it to the calculation process of the elbow rule. The purpose of this is to make the elbow rule more sensitive to the turning point of the impact of the change in the number of clusters on the overall error, thereby enhancing the sensitivity of the algorithm to the characteristics of the time series, making the clustering results more accurate and better reflecting the true dynamic characteristics of the time series, where: Using the formula Determine the elbow point of the citation trajectory within the cluster; in, is the elbow point, is the sum of squared errors, Elbow is the elbow rule function, is the adjustment factor obtained based on the citation trajectory, , is a volatility indicator determined by the standard deviation of citation trajectories, is the fluctuation frequency of the citation trajectory, which can be estimated by the main frequency components obtained by Fourier transform. The long-term trend of the citation trajectory can be obtained by fitting a trend line and calculating its slope. , and are weights, ; S222, in clustering algorithms, the function or advantage of the silhouette coefficient method is to help evaluate the quality of clustering, that is, to measure the closeness of a sample to its cluster and the distinction from other clusters. Based on this, this application introduces this method into the K-shape algorithm to improve the clustering performance of citation trajectory data with significant dynamic characteristics. Specifically, the silhouette coefficient method is optimized by calculating the similarity and difference of the time series morphology, and the morphological similarity index is introduced. and morphological difference index , and integrate it into the silhouette coefficient The purpose of this is to allow the silhouette coefficient method to more accurately identify the number of categories of citation trajectories, thereby enhancing the algorithm's sensitivity to time series characteristics, making the clustering results more accurate and better reflecting the true dynamic characteristics of time series; among them: Using the formula Determine the silhouette coefficient of the citation trajectories within the cluster; is the silhouette coefficient, To measure the morphological similarity between the citation trajectory and the cluster center, To measure the morphological difference between the citation trajectory and the nearest non-belonging cluster center, ; is the citation track, Citation Track With cluster center The shape similarity between For all citation tracks and cluster centers The maximum value of the distance, used for normalization . ; Citation Track The nearest cluster center outside the cluster to which it belongs.

[0030] S223, combined with the improved elbow rule and the optimized silhouette coefficient method, this application adaptively and dynamically adjusts the clustering category, thereby intelligently determining the optimal number of clusters This process not only considers the number of clusters, but also the quality of clusters, effectively improving the flexibility and accuracy of citation trajectory pattern recognition; among them: use Determine the optimal number of clusters; is the optimal number of clusters, The function indicates that the number of clusters with the largest average silhouette coefficient among the candidate clusters preliminarily determined according to the elbow rule is taken as the optimal number of clusters.

[0031] In an exemplary embodiment, Figure 2 As shown, the clustering method process in this application is as follows: (1) For each pair of citation tracks, the normalized cross-correlation value is calculated to assess the shape similarity between them, and the centroid of each cluster is calculated using the cross-correlation based distance measure (SBD). The centroid of each cluster is determined by comparing the shape similarity of the citation tracks.

[0032] (2) For each considered k value, calculate and record the error metric for each clustering, usually the within-cluster sum of squares (WCSS). Draw a graph of the relationship between k value and WCSS, and combine the results of the improved elbow rule and the silhouette coefficient method. For each considered k value, calculate the silhouette coefficient and elbow point of the samples in each cluster, find the k value with the largest average silhouette coefficient and the point where the rate of decline of the WCSS curve slows down significantly, and preliminarily determine several possible candidate cluster numbers based on the elbow rule. Then, select the one with the largest average silhouette coefficient as the optimal cluster number from the candidate number. A higher average silhouette coefficient indicates better clustering quality.

[0033] (3) Randomly select K time series as the initial cluster centers. Use the distance measurement method unique to K-Shape clustering, called shape-based distance (SBD), to calculate the distance between each time series and each cluster center.

[0034] (4) Based on the calculated shape distance matrix, each citation track is assigned to the cluster center that is closest to it.

[0035] (5) For each cluster, calculate the average time series of all its members as the new cluster center.

[0036] (6) Repeat steps (4) and (5) until the cluster center no longer changes or the predetermined number of iterations is reached. In each iteration, K-Shape clustering will reallocate the time series to the nearest cluster center and update the cluster center.

[0037] (7) Assign each citation trajectory to the final cluster center to obtain the final clustering result. The clustering results can be grouped according to the shape characteristics of the time series.

[0038] (8) Use a time series diagram to display the characteristics of each cluster. Analyze the common characteristics of each cluster and identify different time series patterns or trends.

[0039] S103, using the clustered citation trajectories to train an LSTM long short-term memory neural network model using an adaptive sliding window mechanism; like Figure 3As shown in the figure, the LSTM long short-term memory neural network model is defined using a deep learning framework (such as TensorFlow, PyTorch, etc.); the network architecture of the LSTM long short-term memory neural network model includes the number of layers of the LSTM neural network, the number of units in each layer, and the input and output dimensions. Add the necessary fully connected layer (Dense layer) for classification, and the output of the LSTM layer (usually the hidden state) can be passed to the fully connected layer for classification. At the output of the fully connected layer, a softmax activation function (for multi-classification problems) is used to generate a probability distribution.

[0040] In S103, an adaptive sliding window mechanism is constructed to capture the dynamic change characteristics of the citation time series, so as to enhance the adaptability and prediction effect of the LSTM long short-term memory neural network model. The specific process is as follows: S31, sliding window size adaptive adjustment: window size Adaptive adjustment based on the volatility of citation trajectories: ; In the formula, Citation Track The standard deviation of the trend, a measure of volatility; is a function that adjusts the window size according to volatility.

[0041] S32, sliding window step size adaptive adjustment: window step size It can be adaptively adjusted based on the first-order difference of the citation trajectory: ; ; In the formula, Citation Track The standard deviation, a measure of volatility, is a function that adjusts the step size based on the difference. If the difference is large, the step size may be shorter to capture rapid changes; if the difference is small, the step size may be longer to observe longer-term trends.

[0042] S33, define an objective function to combine the effects of window size and step size on prediction error and optimize them: ; In the formula, is the trade-off coefficient, and is the penalty term for window size and step length, which is used to ensure that the window size and step length are within a reasonable range and avoid excessive computational complexity or information loss caused by being too small or too large. , the optimal window size and step size can be found, so that the model can better adapt to the characteristics of citation trajectory data during prediction and improve the accuracy and robustness of prediction.

[0043] S34, using the clustering results, that is, a collection of papers with similar citation behaviors, as prior knowledge for training the LSTM long short-term memory neural network model, and fine-tuning the parameters of the LSTM long short-term memory neural network model through grid search , in order to enhance its sensitivity to time series data, improve the recognition accuracy of different citation trajectory patterns, and ensure that the model can accurately understand and predict various citation patterns. The specific formula is as follows: ; ; ; ; In the formula, Represents the parameter set of the LSTM long short-term memory neural network model; is the input gate weight; is the hidden layer weight; is the bias; the loss function Used to measure the accuracy of LSTM long short-term memory neural network model prediction, using cross entropy As a loss function; is the citation trajectory sequence predicted by the LSTM long short-term memory neural network model; is the actual citation trajectory sequence; is the learning rate; is the loss function Parameters The gradient of is the optimization function, It is a citation trajectory pattern validation dataset used to fine-tune the parameters of the LSTM long short-term memory neural network model; are the parameters of the optimized LSTM long short-term memory neural network model.

[0044] S104, using the trained LSTM long short-term memory neural network model to predict the citation trajectory pattern and future trend of the papers to be evaluated.

[0045] S104 specifically includes: Using the formula Identify papers to be evaluated Citation track sequence ; is the output function of the trained LSTM long short-term memory neural network model, are the parameters of the trained LSTM long short-term memory neural network model; Using the formula Determine the citation track sequence The citation trajectory pattern of candidate attributions ;in, Citation track sequence Belong to The probability of a citation trajectory pattern, , is the probability distribution function of the classifier output, is a mathematical function; use Determine the citation track sequence Future Trends ;in, It is a function that predicts future trends based on current citation trajectory patterns and feature representations.

[0046] S104 and later also include: Visualize the citation trajectory patterns and future trends of the papers being evaluated.

[0047] Based on the visual analysis of clustering results, understand the characteristics and development trends of different citation trajectory categories. Quantitatively describe and analyze the shape characteristics of each cluster category, such as period, amplitude, etc. Apply this application to actual citation analysis scenarios, such as scientific research evaluation, knowledge discovery, etc. Present the results of the analysis in intuitive charts, tables, etc. for easy understanding and interpretation by users. Explain and illustrate the results in detail, and explain the meaning behind them and their potential application value.

[0048] The citation trajectory sequence not only reflects the dynamic evolution of the paper's influence, but also reveals its acceptance and development trend in the academic field. The citation trajectory pattern is a series of representative and recognizable citation behavior patterns identified and defined by in-depth analysis of the citation history and frequency changes of a large number of papers. These patterns include, but are not limited to, rapid rise followed by stability, gradual growth with mutation points, continuous stability, and explosive growth followed by stability, etc., and this application can be automatically updated and adjusted as the paper is published.

[0049] This application uses the K-Shape algorithm's sensitivity to capturing the shape features of time series, and optimizes the K-Shape algorithm by integrating the elbow rule and the silhouette coefficient method, so as to more accurately identify the shape features of time series and improve the accuracy of clustering. At the same time, the LSTM long short-term memory neural network model is used to capture the long-term dependencies in time series data, and the dynamic sliding window technology is used to enhance its ability to capture the dynamic information of citation trajectories. Combined with the preliminary grouping results of the K-Shape clustering algorithm, it takes into account the popularity cycle of papers, the natural decline of academic influence, and the impact of incidental events on citation rates, and completes the probability prediction of future trends of papers.

[0050] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a citation trajectory pattern prediction method is implemented.

[0051] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0052] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0053] 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, stored data, displayed data, 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 relevant data must comply with relevant regulations.

[0054] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile 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), magnetoresistive 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 may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0055] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0056] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, 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 specification.

[0057] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A citation trajectory pattern prediction method, characterized in that: The citation trajectory pattern prediction method comprises: Obtain complete citation data of papers in different fields and construct a time series of paper citations; and use the time series of paper citations as citation tracks; the complete citation data includes: the number of citations of the paper, the citation list, and the citation time; According to the citation tracks of papers in different fields, K-Shape clustering algorithm and improved elbow rule and silhouette coefficient method are used for clustering; each cluster category is the citation track pattern corresponding to the citation track; The clustered citation trajectories are used to train the LSTM long short-term memory neural network model with an adaptive sliding window mechanism; The trained LSTM long short-term memory neural network model is used to predict the citation trajectory patterns and future trends of the papers to be evaluated.

2. The citation trajectory pattern prediction method according to claim 1, characterized in that: The method of obtaining complete citation data of papers in different fields and constructing a time series of paper citations specifically includes: Get complete citation data for papers in different fields; Preprocessing the complete citation data; the preprocessing includes: data cleaning, denoising, interpolation or resampling, missing value processing and dimensionless processing; Construct the paper citation time series based on the preprocessed complete citation data.

3. The citation trajectory pattern prediction method according to claim 1, characterized in that: According to the citation tracks of papers in different fields, the K-Shape clustering algorithm and the improved elbow rule and silhouette coefficient method are used for clustering, which specifically includes: Using the formula Determine the elbow point of the citation trajectory within the cluster; Using the formula Determine the silhouette coefficient of the citation trajectories within the cluster; use Determine the optimal number of clusters; in, is the elbow point, is the sum of squared errors, Elbow is the elbow rule function, is the adjustment factor obtained based on the citation trajectory, , is a volatility indicator determined by the standard deviation of citation trajectories, is the fluctuation frequency of the citation trajectory, is the long-term trend of the citation trajectory, , and are weights, , is the silhouette coefficient, To measure the morphological similarity between the citation trajectory and the cluster center, To measure the morphological difference between the citation trajectory and the nearest non-belonging cluster center, is the optimal number of clusters, The function indicates that the number of clusters with the largest average silhouette coefficient among the candidate clusters preliminarily determined according to the elbow rule is taken as the optimal number of clusters.

4. The citation trajectory pattern prediction method according to claim 3, characterized in that: Measuring the morphological similarity between citation tracks and cluster centers Specifically include: ； in, is the citation track, Citation Track With cluster center The shape similarity between For all citation tracks and cluster centers The maximum value of the distance.

5. The citation trajectory pattern prediction method according to claim 4, characterized in that: Measures the morphological difference between the citation trajectory and the nearest non-belonging cluster center Specifically include: ； in, Citation Track The nearest cluster center outside the cluster to which it belongs.

6. The citation trajectory pattern prediction method according to claim 1, characterized in that: The trained LSTM long short-term memory neural network model is used to predict the citation trajectory pattern and future trend of the paper to be evaluated, specifically including: Using the formula Identify papers to be evaluated Citation track sequence ; is the output function of the trained LSTM long short-term memory neural network model, are the parameters of the trained LSTM long short-term memory neural network model; Using the formula Determine the citation track sequence The citation trajectory pattern of candidate attributions ;in, Citation track sequence Belong to Citation Track Pattern The probability of , is the probability distribution function of the classifier output, is a mathematical function; use Determine the citation track sequence Future Trends ;in, It is a function that predicts future trends based on current citation trajectory patterns and feature representations.

7. The citation trajectory pattern prediction method according to claim 1, characterized in that: The method uses the trained LSTM long short-term memory neural network model to predict the citation trajectory pattern and future trend of the paper to be evaluated, and then includes: Visualize the citation trajectory patterns and future trends of the papers being evaluated.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the citation trajectory pattern prediction method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the citation trajectory pattern prediction method described in any one of claims 1 to 7 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the citation trajectory pattern prediction method described in any one of claims 1 to 7 is implemented.

Images