Multi-dimensional big data analysis method for achievement transformation

Abstract

The invention discloses a multi-dimensional big data analysis method for achievement transformation, particularly relates to the technical field of achievement transformation, and adopts multi-dimensional analysis, a double-layer matching strategy and clustering analysis to realize recommendation of a transformation scheme. Obtaining a to-be-converted science and technology text, and extracting technical dimension features of the to-be-converted science and technology text; screening out a technical feasible conversion scheme, and calculating the science and technology dimension feature vector similarity of the to-be-converted science and technology text and the historical science and technology text based on the technical dimension feature of the to-be-converted science and technology text and the technical dimension feature of the technical feasible science and technology text in the historical record; screening out a batch of conversion schemes with high similarity; carrying out clustering analysis on a batch of conversion schemes with high similarity according to market dimension features, and counting a group proportion; and determining a conversion period index, a social influence entropy and a science and technology text distribution proportion, calculating recommendation priority coefficients by using weighted summation, and selecting a conversion scheme with the maximum recommendation priority coefficient.

Description

Technical Field

[0001] The present invention relates to the technical field of achievement transformation, and more specifically, to a multi-dimensional big data analysis method for achievement transformation. Background Art

[0002] With the rapid development of global science and technology and the deepening advancement of the knowledge economy, the transformation of scientific and technological achievements has become a key engine driving social and economic development. In recent years, while scientific research output has continued to increase, the conversion rate remains generally low. Many scientific research results fail to be promptly translated into practical productivity, creating a phenomenon of "research and market separation." Therefore, how to effectively improve the efficiency and success rate of scientific and technological achievement transformation has become a focus of both academia and industry.

[0003] Traditional methods for analyzing the transformation of scientific and technological achievements rely primarily on expert experience or simple statistical analysis, making it difficult to comprehensively and accurately assess the transformation potential and feasibility of scientific and technological achievements. These methods often overlook the combined influence of multiple factors, including technology, market, policy, and economics, resulting in a lack of scientific basis for decision-making on the transformation of scientific and technological achievements.

[0004] In order to solve the above-mentioned defects, a technical solution is now provided. Summary of the Invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a multi-dimensional big data analysis method for achievement transformation to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A multi-dimensional big data analysis method for achievement transformation, characterized by comprising the following steps:

[0008] Obtain the scientific and technological text to be transformed and extract the technical dimension features of the scientific and technological text to be transformed;

[0009] Screen out technically feasible scientific and technological texts, and calculate the similarity of the technical dimension feature vectors of the scientific and technological texts to be converted and the historical scientific and technological texts based on the technical dimension features of the scientific and technological texts to be converted and the technical dimension features of technically feasible scientific and technological texts in historical records; screen out a batch of scientific and technological texts with high similarity;

[0010] Conduct cluster analysis on a group of scientific and technological texts with high similarity according to market dimension characteristics, and calculate the distribution ratio of scientific and technological texts;

[0011] Determine the conversion cycle index, social influence entropy and scientific text distribution ratio, use weighted summation to calculate the recommendation priority coefficient, and select the conversion plan with the largest recommendation priority coefficient.

[0012] In a preferred embodiment, after obtaining the scientific text to be converted, the scientific text to be converted needs to be preprocessed, first removing noise, and then splitting the text into words; for Chinese text, Chinese word segmentation tools jieba and HanLP can be used, and for English text, NLTK and spaCy tools can be used for word segmentation, while removing stop words that have no practical significance for technical analysis; and assigning a part-of-speech tag to each word; using deep learning methods to extract features from the scientific text, and vectorizing the extracted features.

[0013] In a preferred embodiment, the historical scientific and technological texts are screened for those with successful results transformation, and the historical scientific and technological texts with technology maturity greater than a preset threshold are regarded as the scientific and technological texts with successful results transformation.

[0014] In a preferred embodiment, the technology maturity is obtained by calculating the technology performance index and the engineering feasibility coefficient through logistic regression; the technology performance index is the ratio of the core parameter to the industry benchmark; and the engineering feasibility coefficient is jointly determined by the manufacturing yield and the supply chain maturity.

[0015] In a preferred embodiment, the scientific and technological text to be converted is obtained, its scientific and technological dimension feature vector is extracted, and the similarity of the scientific and technological dimension feature vector between the scientific and technological text to be converted and the historical scientific and technological text is calculated using cosine similarity; the scientific and technological text to be converted and the historical scientific and technological text whose similarity of the scientific and technological dimension feature vector between the scientific and technological text to be converted and the historical scientific and technological text is greater than a preset threshold are screened.

[0016] In a preferred embodiment, market dimension feature vectors are obtained, groups are divided according to the similarity of the market dimension feature vectors of scientific and technological texts, and cluster groups of different market dimension feature vectors are identified by using a K-means clustering method.

[0017] In a preferred embodiment, the number of scientific texts in each cluster center after clustering is determined, the number of scientific texts in all cluster centers is counted, and the number of scientific texts in each cluster center is divided by the number of scientific texts in all cluster centers to obtain the scientific text distribution ratio.

[0018] In a preferred embodiment, the first application time of the results, the publication time of the scientific and technological texts, and the industry benchmark period of each clustered scientific and technological text are obtained from historical records and the Internet, and the conversion cycle index is calculated using the following formula: ZS = (SC-FB) / JZ*100%; where ZS represents the conversion cycle index, SC represents the first application time of the results, FB represents the scientific and technological publication time; and JZ represents the industry benchmark period.

[0019] The calculation formula for social influence entropy is as follows: SIE=-∑(p i lnp i ); where pi is the contribution distribution probability of technology in different social fields; SIE stands for social influence entropy; the contribution distribution probability is obtained from historical records.

[0020] In a preferred embodiment, the conversion cycle index, social influence entropy and scientific text distribution ratio are determined, the conversion cycle index, social influence entropy and scientific text distribution ratio are normalized, and the recommendation priority coefficient is calculated using a weighted sum formula.

[0021] In a preferred embodiment, the recommendation priority coefficients are sorted according to numerical values, and the scientific text conversion solution with the largest numerical value is output as the recommended conversion solution.

[0022] Technical effects and advantages of the present invention:

[0023] The present invention adopts multi-dimensional analysis and a two-level matching strategy and cluster analysis to achieve the recommendation of conversion solutions. The scientific and technological texts to be converted are obtained, and the technical dimension features of the scientific and technological texts to be converted are extracted; the scientific and technological texts that are technically feasible are screened out, and the similarity of the technical dimension feature vectors of the scientific and technological texts to be converted and the historical scientific and technological texts is calculated based on the technical dimension features of the scientific and technological texts to be converted and the technical dimension features of the technically feasible scientific and technological texts in the historical records; a group of scientific and technological texts with high similarity are screened out; a group of scientific and technological texts with high similarity are clustered according to the market dimension features, and the group proportion is calculated; the conversion cycle index, social influence entropy and scientific and technological text distribution ratio are determined, and the recommendation priority coefficient is calculated using the weighted summation formula, and the conversion solution with the largest recommendation priority coefficient is selected; the multi-dimensional analysis makes the recommended conversion solution more practical. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings;

[0025] Figure 1 This is a flow chart of a multi-dimensional big data analysis method for transforming the results of the present invention. DETAILED DESCRIPTION

[0026] 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.

[0027] The present invention adopts multi-dimensional analysis, a two-level matching strategy, and clustering analysis to achieve the recommendation of transformation solutions. Obtain the scientific and technological text to be transformed, and extract the technical dimension features of the scientific and technological text to be transformed; screen out the scientifically and technologically feasible scientific and technological texts, and calculate the similarity of the scientific and technological dimension feature vectors between the scientific and technological text to be transformed and the scientific and technological texts with scientifically and technologically feasible technical dimension features in the historical records based on the technical dimension features of the scientific and technological text to be transformed; screen out a batch of scientific and technological texts with high similarity; perform clustering analysis on the batch of scientific and technological texts with high similarity according to the market dimension features, and count the group ratio; determine the transformation cycle index, social influence entropy, and scientific and technological text distribution ratio, use the weighted summation formula to calculate the recommendation priority coefficient, and select the transformation solution with the largest recommendation priority coefficient.

[0028] Example 1

[0029] A multi-dimensional big data analysis method for achievement transformation in the present invention, as Figure 1 shown, includes the following steps:

[0030] Obtain the scientific and technological text to be transformed, and extract the technical dimension features of the scientific and technological text to be transformed;

[0031] Screen out the scientifically and technologically feasible scientific and technological texts, and calculate the similarity of the scientific and technological dimension feature vectors between the scientific and technological text to be transformed and the scientific and technological texts with scientifically and technologically feasible technical dimension features in the historical records based on the technical dimension features of the scientific and technological text to be transformed; screen out a batch of scientific and technological texts with high similarity;

[0032] Perform clustering analysis on the batch of scientific and technological texts with high similarity according to the market dimension features, and count the group ratio;

[0033] Determine the transformation cycle index, social influence entropy, and scientific and technological text distribution ratio, use weighted summation to calculate the recommendation priority coefficient, and select the transformation solution with the largest recommendation priority coefficient.

[0034] Specifically:

[0035] Preprocess the scientific and technological text to be transformed, aiming to clean the original text and prepare for subsequent analysis; specifically, first remove noise, such as removing non-technical content such as irrelevant symbols, punctuation marks, numbers, HTML tags, advertisements, etc.; then split the text into smaller units. For Chinese texts, Chinese word segmentation tools (such as jieba, HanLP, etc.) can be used, and for English texts, tools such as NLTK and spaCy can be used for word segmentation, and at the same time remove stop words such as "de", "shi", "zai" that are common but have no practical significance for technical analysis; and assign词性 labels (such as nouns, verbs, adjectives, etc.) to each word, which helps to identify important technical entities.

[0036] Use named entity recognition technology to automatically identify technical entities in text, such as equipment, technical terms, product names, company names, technical methods, etc.

[0037] Using a syntactic analysis tool (such as Stanford CoreNLP) to perform dependency parsing can help identify the relationships between technical entities in a sentence. Dependency trees can clarify the grammatical relationships between words and help understand the connections between technical entities. For example:

[0038] Technical entity relationship: such as "battery management system (technical entity) is used for (relationship) electric vehicles (technical entity)".

[0039] Attribute relationship: such as "5G technology (technology entity) supports (relationship) high-speed data transmission (attribute)".

[0040] Based on these parsing results, the technical dimension features in the text can be extracted, and the relationship between various technical entities can be clearly understood.

[0041] We use deep learning methods like BERT and BiLSTM-CRF for feature extraction. These models can capture more complex contextual information and help identify more accurate technical dimension features.

[0042] BERT: Through the deep semantic understanding capabilities of the pre-trained language model, it can identify more fine-grained technical dimension features.

[0043] BiLSTM-CRF: Bidirectional LSTM combined with the conditional random field model can efficiently perform sequence labeling tasks such as named entity recognition and technical triple extraction (entity-relationship-attribute).

[0044] To use the extracted technical dimension features in subsequent analysis, these features can be vectorized. Using TF-IDF vectorization, words are represented as vectors, which effectively expresses the weight of the words and extracts the most critical technical dimension features from the technical literature.

[0045] Quantize the feature vectors of historical scientific and technological texts: feature vector of scientific and technological dimension: ai = (ai1, ai2, ..., aio); feature vector of market dimension: di = (di1, di2, ..., din).

[0046] It should be noted that aio represents the value of the i-th scientific text on the o-th feature. The subsequent expressions are similar and will not be repeated here.

[0047] Screen historical scientific and technological texts for those whose achievements have been successfully transformed into scientific and technological texts. Specifically, historical scientific and technological texts with a technology maturity greater than Q are considered to have successfully transformed into scientific and technological texts. Each historical scientific and technological text has a corresponding transformation plan.

[0048] Further technical maturity is obtained by the following formula: D = 1-e -(aTPI+bEFC) ; Wherein D represents technology maturity; TPI represents technology performance index. Furthermore, technology performance index is the ratio of core parameters to industry benchmarks (such as chip computing power / international top level × 100%). The larger the technology performance index, the greater the technology maturity, and vice versa. EFC represents engineering feasibility coefficient. Furthermore, engineering feasibility coefficient is jointly determined by manufacturing yield (mass production line yield ≥ 95% gets 1 point, and 0.1 point is deducted for every 5% decrease) and supply chain maturity (localization rate of key materials × 0.3 + equipment autonomy × 0.7). Manufacturing yield and supply chain maturity are further normalized, and then weighted summation is performed to obtain engineering feasibility coefficient, and the weights are set by the staff themselves. The larger the engineering feasibility coefficient, the greater the technology maturity, and vice versa. a and b are the logistic regression coefficients of technology performance index and engineering feasibility coefficient, respectively, and both are greater than zero.

[0049] The specific Q is set by the staff themselves, which can control the completeness of the results transformation.

[0050] The eigenvalues are normalized or standardized to reduce the impact of scale differences. This process is a conventional method in the prior art and will not be described in detail here.

[0051] Obtain the technology text to be converted and extract its technology dimension feature vector. The extraction process is the same as above, and obtain the technology dimension feature vector y = (y1, y2, ..., yo) of the technology text to be converted.

[0052] Use cosine similarity to calculate the similarity between the feature vectors of the technology dimension of the technology text to be converted and the historical technology text. The formula is as follows: Where k is the feature index, which ranges from 1 to m and traverses all features, so that the cosine similarity formula can calculate the similarity between multi-dimensional feature vectors.

[0053] The scientific texts to be converted whose similarity between the scientific dimension feature vectors and the historical scientific texts is greater than a preset threshold are screened; a group of scientific texts with a high similarity to the scientific texts to be converted are determined to facilitate subsequent screening.

[0054] Obtain the market dimension feature vector, divide the groups according to the similarity of the market dimension features of the successfully converted scientific and technological texts, identify cluster groups with different market dimension features, and analyze the most representative main features within the clustered groups.

[0055] The scientific and technological texts to be transformed are matched with successfully transformed scientific and technological texts to determine similar successfully transformed scientific and technological texts, and then the conversion plan recommendation strategy is comprehensively determined based on the group to which the similar successfully transformed scientific and technological texts belong and the main characteristics of the group.

[0056] Specifically, cluster groups of feature vectors of different market dimensions are identified and cluster analysis is performed on similar successfully converted scientific and technological texts using the K-means clustering method;

[0057] K-means is an unsupervised learning algorithm that aims to partition data into K non-overlapping clusters. Its core idea is to iteratively optimize the squared distance from each data point to the centroid (mean point) of its cluster, thereby partitioning the data into K non-overlapping clusters.

[0058] Obtain market dimension feature vectors, divide scientific and technological texts into groups according to their similarity, and identify cluster groups of different market dimension feature vectors using the K-means clustering method; specifically, the following steps are included:

[0059] Obtain the market dimension feature vector: di = (di1, di2, ..., din); market dimension features include conversion fields, applicable regions, etc.

[0060] Initialize the cluster centers and use the elbow rule to calculate the SSE (within-cluster sum of squares) for different K values (2-10). Select the inflection point where the SSE decreases and determine the optimal number of clusters K. Each center ck is an n-dimensional vector:

[0061] ck=(ck1,ck2,...,ckn); where k represents the kth cluster center, and k=1,2,...,K. ckj represents the value of the kth cluster center on the jth feature, which is initially randomly selected.

[0062] Calculate the distance and assign the sample, calculate its Euclidean distance with all K cluster centers, and assign the user to the cluster with the nearest center. The formula for Euclidean distance is: Where D ik represents the distance between the i-th scientific text and the k-th cluster center;

[0063] Assign scientific and technological texts to the cluster Ck with the smallest distance, and calculate the assignment as: cluster(d i )=argmin k D ik ; where cluster(d i ) is the cluster label of scientific text, that is, assigned to the nearest cluster center Ck, argmin k D ik Indicates finding the cluster center k with the minimum distance.

[0064] Update the cluster center. When all scientific documents are assigned to their respective cluster centers, calculate the mean vector of each cluster to update the cluster center. The calculation of the updated cluster center ck is as follows: |Ck| represents the number of scientific and technological texts currently belonging to cluster Ck.

[0065] If the change in the updated cluster center compared to the previous center is less than a set threshold ∈, which is the model convergence condition, the algorithm ends.

[0066] Determine the number of scientific texts in each cluster center after clustering, count the number of scientific texts in all cluster centers, and divide the number of scientific texts in each cluster center by the number of scientific texts in all cluster centers to obtain the scientific text distribution ratio bl; a high scientific text distribution ratio indicates that the technology is more suitable for this market field and is relatively mature in this field.

[0067] The main characteristics of the group include the conversion cycle index and social influence entropy; the specific calculation process is as follows;

[0068] Furthermore, the first application time of the results, the publication time of the scientific and technological texts, and the industry benchmark cycle of each clustered scientific and technological text were obtained from historical records and the Internet, and the conversion cycle index was calculated using the following formula: ZS = (SC-FB) / JZ*100%; where ZS represents the conversion cycle index, SC represents the first application time of the results, FB represents the scientific and technological publication time; and JZ represents the industry benchmark cycle.

[0069] Furthermore, the calculation formula for social influence entropy is as follows: SIE = -∑(p i lnp i ); where p i It is the probability distribution of technology's contribution to different social fields (environment, medical care, education, etc.); SIE stands for social influence entropy; specifically, the contribution distribution probability is obtained from historical records.

[0070] Determine the conversion cycle index, social influence entropy and scientific and technological text distribution ratio, normalize the conversion cycle index, social influence entropy and scientific and technological text distribution ratio, and use the weighted summation formula to calculate the recommendation priority coefficient, the formula is as follows: T = -αZS + βSIE + γbl; where T represents the recommendation priority coefficient; bl represents the scientific and technological text distribution ratio, the larger the scientific and technological text distribution ratio, the larger the recommendation priority coefficient, and vice versa; ZS represents the conversion cycle index, the larger the conversion cycle index, the smaller the recommendation priority coefficient, and vice versa; SIE represents the social influence entropy, the larger the social influence entropy, the larger the recommendation priority coefficient, and vice versa; α, β, and γ represent the weight coefficients of the conversion cycle index, social influence entropy, and scientific and technological text distribution ratio, respectively. The specific ones are set by the staff in this field as needed.

[0071] The recommended priority coefficients are sorted according to their numerical values, and the scientific text conversion plan with the largest numerical value is output as the recommended conversion plan.

[0072] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0073] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0074] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0075] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0076] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A multi-dimensional big data analysis method for achievement transformation, characterized by: The following steps are involved: Obtain the scientific and technological text to be transformed and extract the technical dimension features of the scientific and technological text to be transformed; Screen out technically feasible scientific and technological texts, and calculate the similarity of the technical dimension feature vectors of the scientific and technological texts to be converted and the historical scientific and technological texts based on the technical dimension features of the scientific and technological texts to be converted and the technical dimension features of technically feasible scientific and technological texts in historical records; screen out a batch of scientific and technological texts with high similarity; Conduct cluster analysis on a group of scientific and technological texts with high similarity according to market dimension characteristics, and calculate the distribution ratio of scientific and technological texts; Determine the conversion cycle index, social influence entropy and scientific text distribution ratio, use weighted summation to calculate the recommendation priority coefficient, and select the conversion plan with the largest recommendation priority coefficient.

2. The multi-dimensional big data analysis method for achievement transformation according to claim 1 is characterized by: After obtaining the scientific and technological text to be converted, it is necessary to preprocess the scientific and technological text to be converted. First, remove the noise, and then split the text into words. For Chinese text, Chinese word segmentation tools such as jieba and HanLP can be used, and for English text, NLTK and spaCy tools can be used for word segmentation. At the same time, stop words that have no practical significance for technical analysis are removed; and a part-of-speech tag is assigned to each word. Deep learning methods are used to extract features from the scientific and technological text, and the extracted features are vectorized.

3. The multi-dimensional big data analysis method for achievement transformation according to claim 1 is characterized by: The screening of historical scientific and technological texts for successful transformation of achievements takes historical scientific and technological texts whose technical maturity is greater than a preset threshold as scientific and technological texts with successful transformation of achievements.

4. The multi-dimensional big data analysis method for achievement transformation according to claim 3 is characterized by: The technology maturity is obtained by calculating the technology performance index and engineering feasibility coefficient through logistic regression; the technology performance index is the ratio of the core parameters to the industry benchmark; the engineering feasibility coefficient is jointly determined by the manufacturing yield and the supply chain maturity.

5. The multi-dimensional big data analysis method for achievement transformation according to claim 1 is characterized by: Obtain the scientific and technological text to be converted, extract its scientific and technological dimension feature vector, and use cosine similarity to calculate the similarity between the scientific and technological dimension feature vector of the scientific and technological text to be converted and the historical scientific and technological text; screen out the scientific and technological texts whose similarity between the scientific and technological dimension feature vector of the scientific and technological text to be converted and the historical scientific and technological text is greater than a preset threshold.

6. The multi-dimensional big data analysis method for achievement transformation according to claim 1 is characterized by: Obtain the market dimension feature vector, divide the scientific and technological texts into groups according to the similarity of the market dimension feature vector, and identify the clustering groups of different market dimension feature vectors through the K-means clustering method.

7. The multi-dimensional big data analysis method for achievement transformation according to claim 6 is characterized by: Determine the number of scientific texts in each cluster center after clustering, count the number of scientific texts in all cluster centers, and divide the number of scientific texts in each cluster center by the number of scientific texts in all cluster centers to obtain the scientific text distribution ratio.

8. The multi-dimensional big data analysis method for achievement transformation according to claim 1 is characterized by: The first application time of each clustered scientific and technological text, the publication time of the scientific and technological text, and the industry benchmark cycle are obtained from historical records and the Internet, and the conversion cycle index is calculated using the following formula: ZS = (SC-FB) / JZ*100%; Among them, ZS represents the conversion cycle index, SC represents the time of first application of results, FB represents the time of technology release; JZ represents the industry benchmark cycle; The calculation formula for social influence entropy is as follows: SIE=-∑(p i lnp i ); where p i is the contribution distribution probability of technology in different social fields; SIE stands for social influence entropy; the contribution distribution probability is obtained from historical records.

9. The multi-dimensional big data analysis method for achievement transformation according to claim 1, characterized in that: Determine the conversion cycle index, social influence entropy and scientific and technological text distribution ratio, normalize the conversion cycle index, social influence entropy and scientific and technological text distribution ratio, and use the weighted sum formula to calculate the recommendation priority coefficient.

10. The multi-dimensional big data analysis method for achievement transformation according to claim 9, characterized in that: The recommended priority coefficients are sorted according to their numerical values, and the scientific text conversion plan with the largest numerical value is output as the recommended conversion plan.

Images