Project innovation level evaluation method and device, terminal equipment and storage medium
Through word frequency statistics and network structure analysis combined with convolutional neural network and recurrent neural network, the accuracy and comprehensiveness of project innovation level evaluation in large-scale data is solved, and a more explanatory project innovation evaluation is achieved.
Patent Information
- Application Number
- CN202510688726.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-05-27
- Publication Date
- 2025-08-15
AI Technical Summary
The existing technology is difficult to accurately and comprehensively evaluate the level of innovation of projects in large-scale data, resulting in inaccurate evaluation results and lack of general applicability.
The text data is processed through word frequency statistics to obtain text features, build a network structure of numerical data, and use convolutional neural networks and recurrent neural networks to extract spatial local features and timing features respectively, and fuse the two to generate evaluation results.
It improves the accuracy and comprehensiveness of project innovation level assessment, reduces the dependence on expert knowledge in specific industries, and enhances the general use of evaluation methods.
Smart Images
Figure CN120494580A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of big data analysis technology, and in particular to a project innovation level assessment method, apparatus, terminal device and storage medium. Background Art
[0002] With the development of information technology, the amount of data has exploded. Especially in the problem of power grid project management, it is necessary to screen out projects with a certain level of innovation from the massive data composed of a large number of projects.
[0003] The traditional method is usually for experts to evaluate and read each project book, give corresponding evaluation results, and finally obtain the innovation level of each project based on the evaluation results of multiple experts. However, this method is difficult to process large-scale data, and different project books contain different professional terms, colloquial expressions and language styles, which require experts in the corresponding fields to accurately evaluate. Therefore, some existing technologies use big data to analyze different projects, thereby improving the efficiency and versatility of project creativity level evaluation. However, since each project's data sources are extensive, including text data such as project documents, scientific research results, market feedback, team member information, etc., as well as numerical data such as project budgets, time schedules, market share, technical indicators, etc., the data structure of each project is complex, and it is difficult to directly extract effective features that can accurately reflect the project's innovation level from these data, resulting in inaccurate evaluation results and an inability to comprehensively and objectively evaluate the project's innovation capabilities, affecting the final evaluation and optimization effects.
[0004] Therefore, how to ensure the versatility of the project innovation level evaluation method and at the same time improve the accuracy and comprehensiveness of the evaluation results is a technical problem that needs to be solved at present. Summary of the Invention
[0005] The present application provides a project innovation level assessment method, apparatus, terminal device and storage medium, which can solve the problems of poor versatility of project innovation level assessment methods in the prior art and low accuracy and comprehensiveness of evaluation results.
[0006] An embodiment of the present application provides a method for evaluating a project's innovation level, including:
[0007] Acquire data of the project to be evaluated; wherein the data of the project to be evaluated includes text data and numerical data;
[0008] Processing the text data to obtain text features by using a word frequency statistics method;
[0009] constructing a network structure according to the numerical data, and calculating numerical features according to the network structure;
[0010] Calculating spatial local features through a convolutional neural network based on the text features and the numerical features;
[0011] Calculating time series features through a recurrent neural network based on the text features and the numerical features;
[0012] The spatial local features and the temporal features are integrated to obtain an evaluation result.
[0013] Compared with the existing technology, the above embodiment has the following beneficial effects: it captures implicit core innovations from text data through word frequency statistics, and at the same time, through the numerically constructed network structure, it captures explicit results that can verify innovation from numerical data, avoiding the traditional method of relying on a single feature extraction method to process different types of characters; at the same time, it focuses on extracting the spatial local features of the data through the convolutional neural network, and captures the dynamic changes of the data in the time series through the recurrent neural network, thereby effectively evaluating the degree of innovation of the project throughout the planning process and reducing the interference caused by the fluctuations of the spatial local features extracted by the convolutional neural network. Therefore, by processing text data and numerical data separately, combining the cooperation of convolutional neural networks and recurrent neural networks, the comprehensiveness and accuracy of the final evaluation results are effectively improved. Finally, through a unified feature extraction method and the cooperation of convolutional neural networks and recurrent neural networks, the dependence of project innovation evaluation in different fields on the knowledge of specific industry experts is reduced, ensuring the versatility of the evaluation method.
[0014] Furthermore, the processing of the text data by a word frequency statistics method to obtain text features includes:
[0015] Processing the text data by word segmentation to obtain a vocabulary list including a plurality of words;
[0016] The occurrence frequency of each word in the vocabulary list is counted, and the text feature is generated according to the occurrence frequency of each word.
[0017] Compared with the existing technology, the above embodiment has the following beneficial effects: since the core innovation points of each project are usually related to the high-frequency words in the entire project document, the core innovative terms in the project can be quickly located by processing text data through the word frequency statistical method, so that the subsequent network model can effectively capture implicit innovation elements and enhance the semantic rationality of the evaluation results; at the same time, by using the statistically obtained word frequency as a text feature, it not only simplifies the complex natural language processing process in traditional text analysis, but also avoids the subjective bias of manual keyword annotation, thereby ensuring the versatility of the evaluation method.
[0018] Furthermore, constructing a network structure according to the numerical data and calculating numerical features according to the network structure includes:
[0019] Taking each value of the numerical data as a node and the association relationship between the values as an edge, and constructing the network structure according to the nodes and the edges;
[0020] The degree of each node in the network structure is counted, and the numerical feature is generated according to the degree of each node.
[0021] Compared with the existing technology, the above embodiment has the following beneficial effects: by constructing an association network of numerical data and calculating the node degree as a feature, discrete financial indicators, technical parameters, etc. are converted into a visual network structure, thereby quantifying the influence of a single numerical indicator on the overall project innovation level. For example, a high node degree may indicate that a certain technical indicator is the core driving force of project innovation. It not only intuitively displays the intrinsic connection between numerical data, but also enhances the robustness to noisy data by aggregating local information, providing more explanatory numerical features for subsequent models.
[0022] Furthermore, the calculating of spatial local features by a convolutional neural network based on the text features and the numerical features includes:
[0023] Splicing the text feature and the numerical feature to obtain a spliced feature;
[0024] Performing a convolution operation on the splicing features to obtain a convolution result;
[0025] After processing the convolution result through a preset activation function, a maximum pooling operation is performed to obtain the spatial local features.
[0026] Compared with the existing technology, the above embodiment has the following beneficial effects: through convolution operation, sliding in the spliced features, local sensitive pattern information is captured, such as the convolution operation can identify the co-occurrence pattern of keywords in the text, or the coordinated growth trend of technical indicators in the numerical data; then, through the maximum pooling operation, the data dimension is reduced while maintaining the main features to highlight the main features, thereby improving the accuracy of the final evaluation results.
[0027] Furthermore, the calculating of the time series features by a recurrent neural network based on the text features and the numerical features includes:
[0028] Wherein, the recurrent neural network includes a recurrent layer and a linear layer;
[0029] Input the splicing features into the recurrent layer one by one in chronological order, and update the hidden state of each time step;
[0030] The hidden state obtained by the last time step update is mapped to the output space through a linear layer to obtain the temporal features.
[0031] Compared with the existing technology, the above embodiment has the following beneficial effects: since the project plan in the project book is usually described in sequence according to the execution order of the project, the hidden state is updated step by step through the recurrent layer, and the evolution process of the project's innovation capability is dynamically tracked. For example, in different stages of the project book (early R&D stage, mid-term, and acceptance period), the recurrent neural network can memorize and transmit historical information to prevent short-term fluctuations from covering up long-term trends; finally, the linear layer maps the hidden state of the last time step to the output space, generating a score that reflects the dynamic characteristics of the time series, which complements the local spatial characteristics and improves the comprehensiveness of the evaluation results.
[0032] Furthermore, the step of inputting the splicing features into the recurrent layer in chronological order and updating the hidden state of each time step includes:
[0033] The equation for updating the hidden state at each time step through the recurrent layer is:
[0034] H=f(w H1 x Tk +w H2 h Tk +b H )
[0035] Among them, H is the updated hidden state; f(*) is the activation function; w H1 represents the first weight matrix input to the recurrent layer; w H2 represents the second weight matrix of the recurrent layer; b H Represents the bias vector of the recurrent layer; h Tk The hidden state at the previous time step; x Tk Represents the input data for the current time step.
[0036] Compared with the existing technology, the above embodiment has the following beneficial effects: through the activation function and the first weight matrix and the second weight matrix, the nonlinearity of the state is realized, the model's ability to characterize complex innovation processes is enhanced, the temporal dependency of innovation capabilities is effectively captured, and the comprehensiveness of the evaluation results is improved.
[0037] Furthermore, the fusing of the spatial local features and the temporal features to obtain an evaluation result includes:
[0038] Randomly initializing the third weight matrix corresponding to each of the spatial local feature and the temporal feature;
[0039] According to the third weight matrix, a weighted sum of the spatial local features and the temporal features is calculated, and the weighted sum is used as the evaluation result.
[0040] Compared with the existing technology, the above embodiment has the following beneficial effects: through the dynamic fusion strategy, it not only retains the local sensitivity of spatial features, but also fully utilizes the dynamic evolution law of temporal features, thereby ensuring that the evaluation results maximize the reflection of the real innovation level of the project, while improving the versatility and interpretability of the evaluation method.
[0041] Another embodiment of the present application further provides a project innovation level assessment device, comprising: a data acquisition module, a text feature acquisition module, a numerical feature calculation module, a spatial local feature calculation module, a temporal feature calculation module, and a feature fusion module;
[0042] Wherein, the data acquisition module is used to acquire the project data to be evaluated; wherein, the project data to be evaluated includes text data and numerical data;
[0043] A text feature acquisition module, configured to process the text data to acquire text features by using a word frequency statistics method;
[0044] a numerical feature calculation module, configured to construct a network structure based on the numerical data and calculate numerical features based on the network structure;
[0045] A spatial local feature calculation module, configured to calculate spatial local features using a convolutional neural network based on the text features and the numerical features;
[0046] A time series feature calculation module, configured to calculate time series features through a recurrent neural network based on the text features and the numerical features;
[0047] The feature fusion module is used to fuse the spatial local features and the temporal features to obtain an evaluation result.
[0048] Another embodiment of the present application also provides a terminal device, including: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the project innovation level assessment method of the present application are implemented.
[0049] Another embodiment of the present application further provides a computer-readable storage medium item, comprising: a stored computer program, which controls the device where the computer-readable storage medium is located to execute the steps of the project innovation level assessment method of the present application when the computer program is running. BRIEF DESCRIPTION OF THE DRAWINGS
[0050] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the implementation. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0051] Figure 1 A schematic diagram of a process for evaluating a project innovation level provided in some embodiments of the present application;
[0052] Figure 2 This is a structural diagram of a project innovation level assessment device provided in some embodiments of the present application. DETAILED DESCRIPTION
[0053] To make the objectives, technical solutions, and advantages of this application more clear, the technical solutions in this application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making any creative efforts are within the scope of protection of this application.
[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.
[0055] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.
[0056] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.
[0057] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.
[0058] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).
[0059] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.
[0060] When faced with the problem of evaluating the innovation level of projects, the traditional method is usually for experts to read each project book and give corresponding evaluation results. Finally, the innovation level of each project is obtained based on the evaluation results of multiple experts. However, this method is difficult to process large-scale data, and different project books contain different professional terms, colloquial expressions and language styles, which require experts in the corresponding fields to accurately evaluate. Therefore, some existing technologies use big data to analyze different projects, thereby improving the efficiency and versatility of project innovation level evaluation. However, since each project's data comes from a wide range of sources, including text data such as project documents, scientific research results, market feedback, team member information, etc., as well as numerical data such as project budgets, time schedules, market share, technical indicators, etc., the data structure of each project is complex, and it is difficult to directly extract effective features that can accurately reflect the project's innovation level from these data, resulting in inaccurate evaluation results and an inability to comprehensively and objectively evaluate the project's innovation capabilities, affecting the final evaluation and optimization effects.
[0061] Please refer to Figure 1 To address the problems of poor versatility and low accuracy and comprehensiveness of evaluation results in existing project innovation level evaluation methods, the present invention provides a project innovation level evaluation method, including S101 to S106, specifically:
[0062] S101: Acquire project data to be evaluated; wherein the project data to be evaluated includes text data and numerical data.
[0063] Preferably, in some embodiments of the present application, the text data includes project document data (such as project plans, technical reports, and research papers), market feedback data (such as user reviews and market research reports), and team member information data (such as member resumes and work results descriptions); the numerical data includes project budget data (such as the budget for each project R&D stage), time schedule data (expressed in days, months, etc.), market share data (expressed in percentages, including market share change trends since historical events), and technical indicator data (such as product performance parameters). Although the amount of numerical data is relatively small compared to text data, it contains many numerical records in chronological order and has effective information for evaluating the level of innovation.
[0064] Preferably, in some embodiments of the present application, after obtaining the data of the project to be evaluated, the process further includes preprocessing the data of the project to be evaluated to remove noise influences such as irrelevant data.
[0065] S102: Processing the text data by using a word frequency statistics method to obtain text features.
[0066] Furthermore, in some embodiments of the present application, processing the text data by a word frequency statistics method to obtain text features includes:
[0067] Processing the text data by word segmentation to obtain a vocabulary list including a plurality of words;
[0068] The occurrence frequency of each word in the vocabulary list is counted, and the text feature is generated according to the occurrence frequency of each word.
[0069] Preferably, in some embodiments of the present application, the word segmentation processing of the text data can be carried out by a word segmentation method such as a maximum matching method, a hidden Markov model, or a deep learning model-based word segmentation method. The present application does not limit the method used for word segmentation processing.
[0070] Preferably, in some embodiments of the present application, the counting of the occurrence frequency of each word in the vocabulary list includes:
[0071] A counter is used to count the number of times each word in the vocabulary list appears in the document, and the number of times a single word appears is recorded as C. Suppose the total number of words in the text data is d. Based on the formula T=C / d, the statistical feature data of the text data is obtained, where T is the frequency of occurrence of a single word. The frequency of occurrence of each word in the vocabulary list is used as the text feature, where the text feature includes the embedding vector of each word and the corresponding frequency of occurrence of each word.
[0072] In summary, it can be seen that since the core innovation points of each project are usually related to the high-frequency words in the entire project document, by processing text data through word frequency statistics, the core innovative terms in the project can be quickly located, so that the subsequent network model can effectively capture implicit innovation elements and enhance the semantic rationality of the evaluation results; at the same time, by using the statistically obtained word frequency as text features, it not only simplifies the complex natural language processing process in traditional text analysis, but also avoids the subjective bias of manual keyword annotation, ensuring the versatility of the evaluation method.
[0073] S103: Constructing a network structure according to the numerical data, and calculating numerical features according to the network structure.
[0074] Furthermore, in some embodiments of the present application, constructing a network structure based on the numerical data and calculating numerical features based on the network structure include:
[0075] Taking each value of the numerical data as a node and the association relationship between the values as an edge, and constructing the network structure according to the nodes and the edges;
[0076] The degree of each node in the network structure is counted, and the numerical feature is generated according to the degree of each node.
[0077] Preferably, in some embodiments of the present application, the numerical values of the numerical data are taken as nodes, and the associations between the numerical values are taken as edges, and the network structure is constructed according to the nodes and the edges, including: taking the numerical values as nodes, and determining the edges connecting the nodes based on the associations between the numerical values, and setting the weight values of the edges according to the associations, such as in project data, the budget numerical data of different project R&D stages are correlated with each other.
[0078] Preferably, in some embodiments of the present application, the counting of the degree of each node in the network structure and generating the numerical feature according to the degree of each node includes:
[0079] Assume that the node set of the generated network structure is M = {m1, m2, m3...m n}, where m i represents the i-th node, i∈{1,2,…,n}, where n is the total number of nodes. The number of edges connected to each node is counted to obtain the degree of each node to generate numerical features. Numerical features include the value of each node and the degree of each node.
[0080] Preferably, in some embodiments of the present application, when counting the number of edges connected to each node and obtaining the degree of each node, the weight of the connecting edges can also be introduced to determine the degree corresponding to the node based on the weight of the connecting edges.
[0081] In summary, it can be seen that this application converts discrete financial indicators, technical parameters, etc. into a visual network structure by constructing an association network of numerical data and calculating the node degree as a feature, thereby quantifying the influence of a single numerical indicator on the overall project innovation level. For example, a high node degree may indicate that a certain technical indicator is the core driving force of project innovation. It not only intuitively displays the intrinsic connection between numerical data, but also enhances the robustness to noisy data by aggregating local information, providing more explanatory numerical features for subsequent models.
[0082] S104: Calculate spatial local features through a convolutional neural network based on the text features and the numerical features.
[0083] Furthermore, in some embodiments of the present application, calculating the spatial local features by a convolutional neural network based on the text features and the numerical features includes:
[0084] Splicing the text feature and the numerical feature to obtain a spliced feature;
[0085] Performing a convolution operation on the splicing features to obtain a convolution result;
[0086] After processing the convolution result through a preset activation function, a maximum pooling operation is performed to obtain the spatial local features.
[0087] Preferably, in some embodiments of the present application, after processing the convolution result by a preset activation function, a maximum pooling operation is performed to obtain the spatial local feature, specifically: processing the convolution result by a Sigmoid function, and then performing a maximum pooling operation, and outputting the final spatial local feature X by maximum pooling. cnn .
[0088] After the convolution operation, the convolution layer typically introduces nonlinear characteristics through an activation function, followed by a pooling operation. The activation function can increase the model's nonlinear expressiveness, while the pooling operation can reduce the data dimension while maintaining key features. For example, max pooling selects the maximum value within a small area as the output, thereby reducing the amount of data and highlighting key features. For project document data, pooling can filter out more representative local semantic features, reducing data complexity and making subsequent processing more efficient. For numerical data, pooling can simplify the representation of budget data, highlighting key numerical features, and helping the model better capture the main trends in the data.
[0089] In summary, it can be seen that this application uses convolution operations to slide in the spliced features to capture locally sensitive pattern information. For example, the convolution operation can identify the co-occurrence pattern of keywords in the text, or the coordinated growth trend of technical indicators in numerical data; then, the maximum pooling operation is used to reduce the data dimension while maintaining the main features to highlight the main features, thereby improving the accuracy of the final evaluation results.
[0090] S105: Calculate temporal features through a recurrent neural network based on the text features and the numerical features.
[0091] Furthermore, in some embodiments of the present application, calculating the temporal features through a recurrent neural network based on the text features and the numerical features includes:
[0092] Wherein, the recurrent neural network includes a recurrent layer and a linear layer;
[0093] Input the splicing features into the recurrent layer one by one in chronological order, and update the hidden state of each time step;
[0094] The hidden state obtained by the last time step update is mapped to the output space through a linear layer to obtain the temporal features.
[0095] Since the project plans in the project book are usually described in the order of project execution, the hidden state is updated step by time through the recurrent layer to dynamically track the evolution of the project's innovation capabilities. For example, in different stages of the project book (early R&D stage, mid-term, and acceptance period), the recurrent neural network can memorize and transmit historical information to prevent short-term fluctuations from masking long-term trends. Finally, the linear layer maps the hidden state of the last time step to the output space to generate a score that reflects the dynamic characteristics of the time series, which complements the local spatial characteristics and improves the comprehensiveness of the evaluation results.
[0096] Furthermore, in some embodiments of the present application, inputting the splicing features into the recurrent layer one by one in chronological order and updating the hidden state of each time step includes:
[0097] The equation for updating the hidden state at each time step through the recurrent layer is:
[0098] H=f(w H1 x Tk +w H2 h Tk +b H )
[0099] Among them, H is the updated hidden state; f(*) is the activation function; w H1 represents the first weight matrix input to the recurrent layer; w H2represents the second weight matrix of the recurrent layer; b H Represents the bias vector of the recurrent layer; h Tk The hidden state at the previous time step; x Tk Represents the input data for the current time step.
[0100] Through the activation function and the first weight matrix and the second weight matrix, the nonlinearity of the state is realized, the model's ability to characterize complex innovation processes is enhanced, the temporal dependence of innovation capabilities is effectively captured, and the comprehensiveness of the evaluation results is improved.
[0101] Furthermore, in some embodiments of the present application, mapping the hidden state obtained by the last time step update to the output space through a linear layer to obtain the temporal features includes:
[0102] The hidden state updated in the last time step is set to h t , the calculation formula of the time series characteristics is as follows:
[0103] X Rnn =w o h t +b o
[0104] Among them, X Rnn is the time series feature; w o is the linear layer weight matrix; b o is the linear layer bias vector.
[0105] S106: Fusing the spatial local features and the temporal features to obtain an evaluation result.
[0106] Furthermore, in some embodiments of the present application, the fusing of the spatial local features and the temporal features to obtain an evaluation result includes:
[0107] Randomly initializing the third weight matrix corresponding to each of the spatial local feature and the temporal feature;
[0108] According to the third weight matrix, a weighted sum of the spatial local features and the temporal features is calculated, and the weighted sum is used as the evaluation result.
[0109] Furthermore, in some embodiments of the present application, the randomly initialized third weight matrices corresponding to the spatial local features and the temporal features include: randomly initializing third weight matrices W1 and W2 and a bias vector B, wherein W1 and W2 are randomly taken based on uniform distribution U(-0.1, 0.1).
[0110] Furthermore, in some embodiments of the present application, calculating a weighted sum of the spatial local features and the temporal features according to the third weight matrix, and using the weighted sum as the evaluation result, includes:
[0111] The spatial local feature X cnn And the timing feature X Rnn The evaluation result is obtained by combining the following formula with the third weight matrix W1 and W2 and the bias vector B
[0112]
[0113] The above fusion method combines the spatial local features extracted by the convolutional neural network and the temporal features processed by the recurrent neural network. It is no longer a simple form of original text data or numerical data, but a comprehensive feature vector obtained through a complex feature extraction and fusion process. For example, for a certain technical keyword in the project document data, its frequency of occurrence in the original text and its positional relationship in the document may be reflected in a certain feature form after processing by the convolutional neural network, and the semantic coherence and logical relationship of the keyword in the entire project document collection may be reflected after processing by the recurrent neural network, thereby reflecting the comprehensive performance of the project in multi-source data such as text and numerical values.
[0114] It can be seen from the above embodiments that this application uses a dynamic fusion strategy to retain the local sensitivity of spatial features and fully utilize the dynamic evolution laws of temporal features, thereby ensuring that the evaluation results maximize the reflection of the real innovation level of the project, while improving the versatility and interpretability of the evaluation method.
[0115] In summary, compared with the existing technology, the project innovation level assessment method provided by the embodiment of the present application has the following beneficial effects: it captures implicit core innovations from text data through word frequency statistics, and at the same time, through the numerically constructed network structure, it captures explicit results that can verify innovation from numerical data, avoiding the traditional method of relying on a single feature extraction method to process different types of characters; at the same time, it focuses on extracting the spatial local features of the data through the convolutional neural network, and captures the dynamic changes of the data in the time series through the recurrent neural network, thereby effectively assessing the degree of innovation of the project throughout the planning process and reducing the interference caused by the fluctuations of the spatial local features extracted by the convolutional neural network. Therefore, by processing text data and numerical data separately, and combining the convolutional neural network with the recurrent neural network, the comprehensiveness and accuracy of the final assessment results are effectively improved. Finally, through a unified feature extraction method and the cooperation of the convolutional neural network and the recurrent neural network, the dependence of project innovation assessment in different fields on the knowledge of experts in specific industries is reduced, and the versatility of the assessment method is guaranteed.
[0116] like Figure 2 As shown, based on the above-mentioned method embodiment, an embodiment of the present application provides a project innovation level assessment device, including: a data acquisition module 201, a text feature acquisition module 202, a numerical feature calculation module 203, a spatial local feature calculation module 204, a temporal feature calculation module 205 and a feature fusion module 206.
[0117] Furthermore, in some embodiments of the present application, the data acquisition module 201 is used to acquire the project data to be evaluated; wherein, the project data to be evaluated includes text data and numerical data; the text feature acquisition module 202 is used to process the text data through a word frequency statistics method to obtain text features; the numerical feature calculation module 203 is used to construct a network structure based on the numerical data, and calculate numerical features based on the network structure; the spatial local feature calculation module 204 is used to calculate the spatial local features through a convolutional neural network based on the text features and the numerical features; the temporal feature calculation module 205 is used to calculate the temporal features through a recurrent neural network based on the text features and the numerical features; the feature fusion module 206 is used to fuse the spatial local features and the temporal features to obtain an evaluation result.
[0118] Furthermore, in some embodiments of the present application, the text data is processed by a word frequency statistics method to obtain text features, including: word segmentation processing of the text data to obtain a vocabulary list including several words; counting the frequency of occurrence of each word in the vocabulary list, and generating the text features based on the frequency of occurrence of each word.
[0119] Furthermore, in some embodiments of the present application, constructing a network structure based on the numerical data and calculating numerical features based on the network structure include: taking each numerical value of the numerical data as a node, and the association relationship between each numerical value as an edge, and constructing the network structure based on the nodes and the edges; counting the degree of each node in the network structure, and generating the numerical features based on the degree of each node.
[0120] Furthermore, in some embodiments of the present application, the spatial local features are calculated through a convolutional neural network based on the text features and the numerical features, including: splicing the text features and the numerical features to obtain spliced features; performing a convolution operation on the spliced features to obtain a convolution result; and processing the convolution result through a preset activation function, and then performing a maximum pooling operation to obtain the spatial local features.
[0121] Furthermore, in some embodiments of the present application, the temporal features are calculated through a recurrent neural network based on the text features and the numerical features, including: wherein the recurrent neural network includes a recurrent layer and a linear layer; the splicing features are input into the recurrent layer one by one in chronological order, and the hidden state of each time step is updated; through the linear layer, the hidden state updated in the last time step is mapped to the output space to obtain the temporal features.
[0122] Furthermore, in some embodiments of the present application, inputting the splicing features into the recurrent layer one by one in chronological order and updating the hidden state of each time step includes:
[0123] The equation for updating the hidden state at each time step through the recurrent layer is:
[0124] H=f(w H1 x Tk +w H2 h Tk +b H )
[0125] Among them, H is the updated hidden state; f(*) is the activation function; w H1 represents the first weight matrix input to the recurrent layer; w H2 represents the second weight matrix of the recurrent layer; b H Represents the bias vector of the recurrent layer; h Tk The hidden state at the previous time step; x Tk Represents the input data for the current time step.
[0126] Preferably, in some embodiments of the present application, the activation function f(*) can adopt a Sigmoid function or a ReLu function, and the present application does not limit the activation function used.
[0127] Furthermore, in some embodiments of the present application, the fusion of the spatial local features and the temporal features to obtain an evaluation result includes: randomly initializing a third weight matrix corresponding to each of the spatial local features and the temporal features; calculating the weighted sum of the spatial local features and the temporal features according to the third weight matrix, and using the weighted sum as the evaluation result.
[0128] In summary, compared with the prior art, the project innovation level assessment device provided by the embodiment of the present application has the following beneficial effects: it captures implicit core innovations from text data through word frequency statistics, and at the same time, through the numerically constructed network structure, it captures explicit results that can verify innovation from numerical data, avoiding the traditional method of relying on only a single feature extraction method to process different types of characters; at the same time, it focuses on extracting the spatial local features of the data through the convolutional neural network, and captures the dynamic changes of the data in the time series through the recurrent neural network, thereby effectively assessing the degree of innovation of the project throughout the planning process and reducing the interference caused by the fluctuations of the spatial local features extracted by the convolutional neural network. Therefore, by processing text data and numerical data separately, and combining the convolutional neural network with the recurrent neural network, the comprehensiveness and accuracy of the final assessment results are effectively improved. Finally, through a unified feature extraction method and the cooperation of the convolutional neural network and the recurrent neural network, the dependence of project innovation assessment in different fields on the knowledge of experts in specific industries is reduced, and the versatility of the assessment method is guaranteed.
[0129] It can be understood that the above-mentioned device embodiment corresponds to the method embodiment of the present application, which can implement any of the above-mentioned method embodiments of the present application to provide a project innovation level assessment method.
[0130] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules may be selected according to actual needs to achieve the purpose of the present embodiment. In addition, in the drawings of the device embodiments provided herein, the connection relationship between the modules indicates that there is a communication connection between them, which may be implemented as one or more communication buses or signal lines. Those skilled in the art may understand and implement the present invention without inventive effort.
[0131] Based on the above-mentioned embodiment of the project innovation level assessment method, another embodiment of the present application provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the project innovation level assessment method of any embodiment of the present application is implemented.
[0132] For example, in this embodiment, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more module elements may be a series of computer program instruction segments capable of completing specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal device.
[0133] The terminal device may be a computing device such as a desktop computer, a notebook computer, a PDA, a cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.
[0134] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, connecting various parts of the entire terminal device using various interfaces and lines.
[0135] Based on the above-mentioned method embodiments, another embodiment of the present application provides a computer-readable storage medium, including a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method described in any of the above-mentioned method embodiments of the present application.
[0136] Wherein, the module / unit integrated in the device / terminal equipment, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, the present application implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program, when executed by the processor, can implement the steps of the above-mentioned various method embodiments. Wherein, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device that can carry the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunications signal, and a software distribution medium, etc.
Claims
1. A method for evaluating project innovation level, characterized in that: include: Acquire data of the project to be evaluated; wherein the data of the project to be evaluated includes text data and numerical data; Processing the text data by using a word frequency statistics method to obtain text features; constructing a network structure according to the numerical data, and calculating numerical features according to the network structure; Calculating spatial local features through a convolutional neural network based on the text features and the numerical features; Calculating time series features through a recurrent neural network based on the text features and the numerical features; The spatial local features and the temporal features are integrated to obtain an evaluation result.
2. A project innovation level assessment method according to claim 1, characterized in that: The processing of the text data by a word frequency statistics method to obtain text features includes: Processing the text data by word segmentation to obtain a vocabulary list including a plurality of words; The occurrence frequency of each word in the vocabulary list is counted, and the text feature is generated according to the occurrence frequency of each word.
3. A project innovation level assessment method according to claim 1, characterized in that: The step of constructing a network structure based on the numerical data and calculating numerical features based on the network structure includes: Taking each value of the numerical data as a node and the association relationship between the values as an edge, and constructing the network structure according to the nodes and the edges; The degree of each node in the network structure is counted, and the numerical feature is generated according to the degree of each node.
4. A project innovation level assessment method according to claim 1, characterized in that: The calculating of spatial local features by a convolutional neural network based on the text features and the numerical features includes: Splicing the text feature and the numerical feature to obtain a spliced feature; Performing a convolution operation on the splicing features to obtain a convolution result; After processing the convolution result through a preset activation function, a maximum pooling operation is performed to obtain the spatial local features.
5. A project innovation level assessment method according to claim 4, characterized in that: The calculating of the time series features by a recurrent neural network based on the text features and the numerical features includes: Wherein, the recurrent neural network includes a recurrent layer and a linear layer; Input the splicing features into the recurrent layer one by one in chronological order, and update the hidden state of each time step; The hidden state obtained by the last time step update is mapped to the output space through a linear layer to obtain the temporal features.
6. A project innovation level assessment method according to claim 5, characterized in that: The step of inputting the splicing features into the recurrent layer in chronological order and updating the hidden state of each time step includes: The equation for updating the hidden state at each time step through the recurrent layer is: H=f(w H1 x Tk +w H2 h Tk +b H ) Among them, H is the updated hidden state; f(*) is the activation function; w H1 represents the first weight matrix input to the recurrent layer; w H2 represents the second weight matrix of the recurrent layer; b H Represents the bias vector of the recurrent layer; h Tk The hidden state at the previous time step; x Tk Represents the input data for the current time step.
7. A project innovation level assessment method according to claim 1, characterized in that: The fusing of the spatial local features and the temporal features to obtain an evaluation result includes: Randomly initializing the third weight matrix corresponding to each of the spatial local feature and the temporal feature; According to the third weight matrix, a weighted sum of the spatial local features and the temporal features is calculated, and the weighted sum is used as the evaluation result.
8. A project innovation level assessment device, characterized in that: include: Data acquisition module, text feature acquisition module, numerical feature calculation module, spatial local feature calculation module, temporal feature calculation module and feature fusion module; Wherein, the data acquisition module is used to acquire the project data to be evaluated; wherein, the project data to be evaluated includes text data and numerical data; A text feature acquisition module, configured to process the text data to acquire text features by using a word frequency statistics method; a numerical feature calculation module, configured to construct a network structure based on the numerical data and calculate numerical features based on the network structure; A spatial local feature calculation module, configured to calculate spatial local features using a convolutional neural network based on the text features and the numerical features; A time series feature calculation module, configured to calculate time series features through a recurrent neural network based on the text features and the numerical features; The feature fusion module is used to fuse the spatial local features and the temporal features to obtain an evaluation result.
9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a method for evaluating the innovation level of a project as described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a project innovation level assessment method as described in any one of claims 1 to 7.