Scientific and technological achievement dynamic valuation method based on multi-source heterogeneous data fusion
By using the method of multi-source heterogeneous data fusion, utilizing brain-computer interfaces and edge computing nodes to collect data, and combining DNA computing and quantum technology for feature extraction and model updating, the problem of inaccurate valuation in existing technologies is solved, and the real-time and accuracy of dynamic valuation of scientific and technological achievements is achieved.
Patent Information
- Application Number
- CN202510794867.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-14
- Publication Date
- 2025-09-16
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies mainly rely on subjective expert ratings or historical financial data, lack the ability to integrate multimodal heterogeneous information, and are unable to respond to sudden changes in the market environment and dynamic changes in the technology life cycle in real time, resulting in inaccurate valuation of scientific and technological achievements.
Experts' subjective cognitive data is obtained through brain-computer interface devices, IoT data is collected in combination with edge computing nodes, data preprocessing is performed using federated edge learning, feature extraction is performed using DNA computing and quantum dot labeling technology, data fusion is performed using digital twin technology and reinforcement learning algorithms, a dynamic valuation model is constructed, and real-time model updates are achieved through quantum key distribution and federated learning.
It realizes the real-time collection and secure transmission of multi-source heterogeneous data, can simulate the evolution path of technology in different scenarios, update the valuation model in real time, improve the accuracy and robustness of valuation, and adapt to market changes.
Smart Images
Figure CN120655335A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of scientific research evaluation technology, and specifically to a dynamic valuation method for scientific and technological achievements by fusing multi-source heterogeneous data. Background Art
[0002] The transformation of scientific and technological achievements refers to the whole process of converting the technological achievements generated in scientific research activities (such as invention patents, technical solutions, software copyrights, etc.) into actual productive forces through market-oriented channels, covering key links such as technology research and development, pilot production, commercial application and industrial promotion.
[0003] After searching, it was found that existing technologies mainly rely on expert subjective scores or financial historical data, and lack the ability to integrate multi-source technology application scenario data, market dynamics data, and subjective cognitive data. It is difficult to integrate multimodal heterogeneous information such as subjective cognitive data collected by brain-computer interfaces, technology application scenario data generated by IoT devices, and market terminal feedback data in real time. In addition, the use of static valuation models makes it difficult to simulate the impact of dynamic scenarios such as sudden changes in the market environment and subversion of technology paths during the technology life cycle on value. Summary of the Invention
[0004] (1) Technical problems solved
[0005] In response to the shortcomings of the existing technology, the present invention provides a dynamic valuation method for scientific and technological achievements by integrating multi-source heterogeneous data, which solves the problems of the above-mentioned background technology such as "mainly relying on expert subjective scores or financial historical data".
[0006] (2) Technical solution
[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: a dynamic valuation method for scientific and technological achievements based on multi-source heterogeneous data fusion, comprising the following steps:
[0008] S1: Data Collection
[0009] Through brain-computer interface devices, experts' subjective cognitive data on scientific and technological achievements are obtained. In combination with edge computing nodes, the technology application scenario data and market terminal feedback data generated by IoT devices are collected in real time. Federated edge learning is used to achieve distributed pre-processing and encrypted transmission of data.
[0010] S2: Data cleaning and feature extraction
[0011] Encode data into molecular sequences, use the parallel processing characteristics of DNA computing to perform data deduplication and outlier detection, mine hidden features of data through molecular self-assembly characteristics, and achieve efficient feature screening based on quantum dot labeling technology;
[0012] S3: Data Fusion and Simulation
[0013] Digital twin technology is used to model the entire life cycle of scientific and technological achievements, mapping multi-source heterogeneous data to different dimensions of digital twins. Spatiotemporal dynamics models are used to simulate the interaction between data, and reinforcement learning algorithms are used to optimize data fusion weights.
[0014] S4: Dynamic Valuation Model Construction
[0015] The causal emergence algorithm is used to extract causal structures at different levels from the data. The probability distribution of the data is learned by the variational autoencoder, and adversarial training is used to generate value predictions of scientific and technological achievements in multiple scenarios.
[0016] S5: Real-time model update system
[0017] When new data is detected, a secure data transmission channel based on quantum key distribution is used, combined with federated learning to achieve cross-node model parameter updates and instant updates of model parameters, and combined with meta-reinforcement learning algorithms to automatically adjust the model structure to adapt to data changes.
[0018] Preferably, in said S1, data collection is specifically performed through an EEG signal feature extraction algorithm to analyze the expert's evaluation intentions including but not limited to the innovation of scientific and technological achievements and market potential dimensions, and the edge computing node uses a lightweight graph neural network to perform real-time feature extraction on the sensor data.
[0019] Preferably, in the data cleaning and feature extraction of S2, after the data is converted into DNA sequences, polymerase chain reaction technology is used to perform data amplification and duplicate detection, and restriction endonucleases are used to cut and remove abnormal data sequences.
[0020] Preferably, in the data fusion and simulation deduction of S3, the evolution path of scientific and technological achievements in different market environments and technology iteration scenarios is simulated by digital twins, and the probability of value change under different evolution paths is evaluated by using the Monte Carlo tree search algorithm. In addition, a graph convolutional network is introduced for optimization, as follows:
[0021] Constructing multi-source heterogeneous data into a graph structure , where the node set Contains different types of data, edge sets Represents the relationship between data, each node There are corresponding eigenvectors , the features of all nodes form the feature matrix , is the number of nodes, The dimension of each node feature;
[0022] The graph convolutional network updates the node representation by aggregating the node’s own features and the features of its adjacent nodes. The calculation process of the layer graph convolution layer can be expressed as
[0023] in, It is The node feature matrix of the layer, initially is the adjacency matrix with self-loops added, is the original adjacency matrix, is the identity matrix; yes Diagonal node degree matrix, whose diagonal elements ; It is The trainable weight matrix of the layer; Is an activation function, such as the ReLU function .
[0024] Preferably, in the construction of the dynamic valuation model of S4, the causal emergence algorithm is used to identify the macroscopic causal structure in the data, and the causal chain is modeled time-varyingly in combination with the real-time topological update mechanism of the dynamic causal graph neural network. Specifically, the dynamic causal graph update equation is defined as:
[0025]
[0026] in, represents the causal strength matrix from node i to j at time t, is the spatiotemporal convolution kernel, extracting historical causal interaction features, For causal structure and data distribution Divergence constraint term, the demonstration process is as follows:
[0027] Step 1: Initialize the causal graph (0) Macrostructure extracted by causal emergence algorithm;
[0028] Step 2: Every time T time window data is received, the node embedding is updated through the graph convolutional network GCN:
[0029]
[0030] Step 3: Calculate the causal strength gradient based on the node embedding and use the Runge-Kutta method to solve the differential equation to update C(t);
[0031] Step 4: Remove by differentiable graph pruning technique weak causal edge of ;
[0032] The variational autoencoder learns the potential representation of the data by minimizing the reconstruction error, and combines it with the generative adversarial network to generate multimodal valuation prediction results.
[0033] Preferably, in S5, quantum entanglement is established between computing nodes at different geographical locations. When new data is generated, instantaneous synchronization of data is achieved through quantum state collapse, and the learning rate and number of network layers are automatically adjusted according to the data change frequency and model error in combination with the meta-reinforcement learning algorithm.
[0034] Preferably, during the data transmission process of S1, an encryption algorithm based on chaos theory is adopted to generate a dynamic encryption key through the sensitive dependency of the chaotic system, and perform chaotic scrambling and diffusion processing on the data.
[0035] Preferably, when the valuation results are output, a three-dimensional visual report is generated using holographic projection technology, and a brain-computer interface is combined to enable the user to interactively operate and conduct in-depth analysis of the report through thought instructions.
[0036] (3) Beneficial effects
[0037] The present invention provides a dynamic valuation method for scientific and technological achievements based on the fusion of multi-source heterogeneous data. It has the following beneficial effects:
[0038] (1) When using this dynamic valuation method for scientific and technological achievements based on the fusion of multi-source heterogeneous data, a real-time collection and preprocessing system for multi-modal heterogeneous data is constructed through the combination of brain-computer interface devices, edge computing nodes and federated edge learning. It can effectively integrate expert subjective cognitive data, technical application scenario data of IoT devices and market terminal feedback data, use chaos theory encryption algorithms to ensure data security transmission, and realize feature extraction and cleaning of multi-source data through lightweight graph neural networks and DNA computing parallel processing technology.
[0039] (2) When using this dynamic valuation method for scientific and technological achievements based on multi-source heterogeneous data fusion, by mapping multi-source data to different dimensions of the digital twin, combined with reinforcement learning to optimize data fusion weights, Monte Carlo tree search to evaluate the probability of value changes, and graph convolutional network to update node representation, it can simulate the evolution path of technology in different scenarios in real time.
[0040] (2) When using this dynamic valuation method for scientific and technological achievements based on multi-source heterogeneous data fusion, it can be updated instantly when new data is monitored through the security parameter update mechanism that combines quantum key distribution with federated learning, and the automatic adjustment of the model structure by meta-reinforcement learning. It effectively responds to the real-time changes in market data and technical feedback during the technology life cycle, and avoids valuation deviations caused by data delays or model fixation. At the same time, the generator adopts a quantum circuit coding causal intervention strategy, and the discriminator realizes multi-time scale feature matching through dynamic mode decomposition, which further improves the real-time and robustness of multi-scenario value prediction and meets the actual needs of dynamic valuation in the transformation of scientific and technological achievements. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 This is a complete flow chart of the present invention;
[0042] Figure 2 This is a detailed flow chart of S1 of the present invention;
[0043] Figure 3 Detailed flow diagram of S2 of the present invention;
[0044] Figure 4 Detailed flow chart of S3 of the present invention;
[0045] Figure 5 This is a detailed flow chart of S5 of the present invention. DETAILED DESCRIPTION
[0046] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0047] See also Figure 1-Figure 5 The present invention provides a dynamic valuation method for scientific and technological achievements based on multi-source heterogeneous data fusion, comprising the following steps:
[0048] S1: Data Collection
[0049] Through brain-computer interface devices, we obtain experts' subjective cognitive data on scientific and technological achievements. In combination with edge computing nodes, we collect technical application scenario data and market terminal feedback data generated by IoT devices in real time. We use federated edge learning to achieve distributed preprocessing and encrypted transmission of data. We use an encryption algorithm based on chaos theory to generate dynamic encryption keys through the sensitive dependencies of chaotic systems, and perform chaotic scrambling and diffusion processing on the data. Specifically, we use the feature extraction algorithm of EEG signals to analyze the experts' evaluation intentions on dimensions such as the innovation and market potential of scientific and technological achievements. Edge computing nodes use lightweight graph neural networks to extract real-time features from sensor data.
[0050] S2: Data cleaning and feature extraction
[0051] Encode data into molecular sequences, use the parallel processing characteristics of DNA computing to remove duplicates and detect outliers, mine hidden features in the data through molecular self-assembly, and achieve efficient feature screening based on quantum dot labeling technology. Specifically, after converting the data into DNA sequences, use polymerase chain reaction (PCR) technology to amplify the data and detect duplicates, and use restriction endonucleases to cut and remove abnormal data sequences.
[0052] S3: Data Fusion and Simulation
[0053] Digital twin technology is used to model the entire life cycle of scientific and technological achievements. Multi-source heterogeneous data is mapped to different dimensions of the digital twin. The interaction between data is simulated using a spatiotemporal dynamics model. The data fusion weight is optimized by combining a reinforcement learning algorithm. Specifically, the evolution path of scientific and technological achievements in different market environments and technology iteration scenarios is simulated using digital twins. The Monte Carlo tree search algorithm is used to evaluate the probability of value change under different evolutionary paths. In addition, a graph convolutional network is introduced for optimization, as follows:
[0054] Constructing multi-source heterogeneous data into a graph structure , where the node set Contains different types of data, edge sets Represents the relationship between data, each node There are corresponding eigenvectors , the features of all nodes form the feature matrix , is the number of nodes, The dimension of each node feature;
[0055] The graph convolutional network updates the node representation by aggregating the node’s own features and the features of its adjacent nodes. The calculation process of the layer graph convolution layer can be expressed as:
[0056] in, It is The node feature matrix of the layer, initially is the adjacency matrix with self-loops added, is the original adjacency matrix, is the identity matrix; yes Diagonal node degree matrix, whose diagonal elements ; It is The trainable weight matrix of the layer; Is an activation function, such as the ReLU function ;
[0057] S4: Dynamic Valuation Model Construction
[0058] The causal emergence algorithm is used to extract causal structures at different levels from the data, and the probability distribution of the data is learned by combining the variational autoencoder. In addition, a quantum annealing-inspired hierarchical attention mechanism is embedded in the variational autoencoder. The algorithm formula is as follows:
[0059] First, construct the Hamiltonian in the VAE encoder:
[0060]
[0061] in, , is the quantum fluctuation intensity that decays over time. The demonstration process is as follows:
[0062] Step 1: Map the eigenvectors to the Ising model spin space Step 2: Perform the quantum annealing process:
[0063]
[0064] Step 3: Measure the final state to get the attention weight distribution:
[0065] Step 4: Use weighted features As input to the VAE encoder.
[0066] Through adversarial training, the value prediction of scientific and technological achievements in multiple scenarios is generated. Specifically, the causal emergence algorithm is used to identify the macro causal structure in the data, and the real-time topological update mechanism of the dynamic causal graph neural network is combined to model the time-varying causal chain. Specifically, the dynamic causal graph update equation is defined as:
[0067]
[0068] in, represents the causal strength matrix from node i to j at time t, is the spatiotemporal convolution kernel, extracting historical causal interaction features, For causal structure and data distribution Divergence constraint term, the demonstration process is as follows:
[0069] Step 1: Initialize the causal graph (0) Macrostructure extracted by causal emergence algorithm;
[0070] Step 2: Every time T time window data is received, the node embedding is updated through the graph convolutional network GCN:
[0071]
[0072] Step 3: Calculate the causal strength gradient based on the node embedding and use the Runge-Kutta method to solve the differential equation to update C(t);
[0073] Step 4: Remove by differentiable graph pruning technique weak causal edge of ;
[0074] The variational autoencoder learns the potential representation of the data by minimizing the reconstruction error, and combines it with a generative adversarial network to generate multimodal valuation prediction results. When the valuation results are output, holographic projection technology is used to generate a three-dimensional visual report. Combined with the brain-computer interface, users can interact with the report and conduct in-depth analysis through mental instructions.
[0075] A hybrid quantum-classical generative adversarial network is constructed, in which the generator adopts a quantum circuit coding causal intervention strategy and the discriminator realizes multi-time scale feature matching through dynamic mode decomposition.
[0076] S5: Real-time model update system
[0077] When new data is detected, a secure data transmission channel based on quantum key distribution is used, combined with federated learning to achieve cross-node model parameter updates and instant updates of model parameters. The model structure is automatically adjusted to adapt to data changes in combination with the meta-reinforcement learning algorithm. Specifically, quantum entanglement is established between computing nodes in different geographical locations. When new data is generated, instantaneous synchronization of data is achieved through quantum state collapse. The meta-reinforcement learning algorithm is combined to automatically adjust the learning rate and number of network layers according to the frequency of data changes and model errors.
[0078] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.
Claims
1. A dynamic valuation method for scientific and technological achievements based on multi-source heterogeneous data fusion, characterized by: The following steps are involved: S1: Data Collection Through brain-computer interface devices, experts' subjective cognitive data on scientific and technological achievements are obtained. In combination with edge computing nodes, the technology application scenario data and market terminal feedback data generated by IoT devices are collected in real time. Federated edge learning is used to achieve distributed pre-processing and encrypted transmission of data. S2: Data cleaning and feature extraction Encode data into molecular sequences, use the parallel processing characteristics of DNA computing to perform data deduplication and outlier detection, mine hidden features of data through molecular self-assembly characteristics, and achieve efficient feature screening based on quantum dot labeling technology; S3: Data Fusion and Simulation Digital twin technology is used to model the entire life cycle of scientific and technological achievements, mapping multi-source heterogeneous data to different dimensions of digital twins. Spatiotemporal dynamics models are used to simulate the interaction between data, and reinforcement learning algorithms are used to optimize data fusion weights. S4: Dynamic Valuation Model Construction The causal emergence algorithm is used to extract causal structures at different levels from the data. The probability distribution of the data is learned by the variational autoencoder, and adversarial training is used to generate value predictions of scientific and technological achievements in multiple scenarios. S5: Real-time model update system When new data is detected, a secure data transmission channel based on quantum key distribution is used, combined with federated learning to achieve cross-node model parameter updates and instant updates of model parameters, and combined with meta-reinforcement learning algorithms to automatically adjust the model structure to adapt to data changes.
2. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: In S1, data collection is specifically performed through an EEG signal feature extraction algorithm to analyze experts' evaluation intentions on dimensions including but not limited to the innovation of scientific and technological achievements and market potential. The edge computing node uses a lightweight graph neural network to perform real-time feature extraction on sensor data.
3. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: In the data cleaning and feature extraction of S2, after the data is converted into DNA sequences, polymerase chain reaction technology is used to perform data amplification and duplicate detection, and restriction endonucleases are used to cut and remove abnormal data sequences.
4. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: In the data fusion and simulation deduction of S3, the evolution path of scientific and technological achievements in different market environments and technology iteration scenarios is simulated through digital twins. The Monte Carlo tree search algorithm is used to evaluate the probability of value change under different evolution paths. In addition, a graph convolutional network is introduced for optimization, as follows: Constructing multi-source heterogeneous data into a graph structure , where the node set Contains different types of data, edge sets Represents the relationship between data, each node There are corresponding eigenvectors , the features of all nodes form the feature matrix , is the number of nodes, The dimension of each node feature; The graph convolutional network updates the node representation by aggregating the node’s own features and the features of its adjacent nodes. The calculation process of the layer graph convolution layer can be expressed as in, It is The node feature matrix of the layer, initially is the adjacency matrix with self-loops added, is the original adjacency matrix, is the identity matrix; yes The diagonal node degree matrix of ; It is The trainable weight matrix of the layer; Is an activation function, such as the ReLU function .
5. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: In the construction of the dynamic valuation model of S4, the causal emergence algorithm is used to identify the macro causal structure in the data, and the real-time topological update mechanism of the dynamic causal graph neural network is combined to model the time-varying causal chain. Specifically, the dynamic causal graph update equation is defined as: in, represents the causal strength matrix from node i to j at time t, is the spatiotemporal convolution kernel, extracting historical causal interaction features, For causal structure and data distribution Divergence constraint term, the demonstration process is as follows: Step 1: Initialize the causal graph (0) Macrostructure extracted by causal emergence algorithm; Step 2: Every time T time window data is received, the node embedding is updated through the graph convolutional network GCN: Step 3: Calculate the causal strength gradient based on the node embedding and use the Runge-Kutta method to solve the differential equation to update C(t); Step 4: Remove by differentiable graph pruning technique Weak causal edges <0.2; The variational autoencoder learns the potential representation of the data by minimizing the reconstruction error, and combines it with the generative adversarial network to generate multimodal valuation prediction results.
6. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: In S5, quantum entanglement is established between computing nodes in different geographical locations. When new data is generated, instantaneous synchronization of the data is achieved through quantum state collapse. Combined with the meta-reinforcement learning algorithm, the learning rate and the number of network layers are automatically adjusted according to the frequency of data changes and model errors.
7. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 1 is characterized by: During the data transmission process of S1, an encryption algorithm based on chaos theory is adopted to generate a dynamic encryption key through the sensitive dependency of the chaotic system, and perform chaotic scrambling and diffusion processing on the data.
8. The method for dynamic valuation of scientific and technological achievements based on multi-source heterogeneous data fusion according to claim 5 is characterized by: When the valuation results are output, holographic projection technology is used to generate a three-dimensional visual report, and the brain-computer interface is combined to enable the user to interact with and conduct in-depth analysis of the report through thought instructions.