Credibility evaluation method and system for automobile innovation projects based on knowledge graph

Abstract

The present invention discloses a method and system for assessing the credibility of automotive innovation projects based on a knowledge graph, relating to the field of data processing technology. The method comprises: obtaining implemented automotive projects based on the knowledge graph, extracting automotive-related keywords and forming an automotive keyword library, obtaining text information, obtaining automotive-related keywords appearing in the text information and using them as benchmark words, and obtaining independent frequencies; obtaining a first correlation between every two benchmark words in implemented automotive projects, and obtaining a first benchmark credibility; obtaining the time nodes of each implemented automotive project, and obtaining a second correlation; and obtaining a second benchmark credibility corresponding to the automotive innovation project based on a benchmark model, the second correlation, and multiple independent frequencies; obtaining a target credibility based on the first benchmark credibility and the second benchmark credibility, and associating the target credibility with the automotive innovation project. The present invention has the advantages of high accuracy, high timeliness, and stability and reliability.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method and system for evaluating the credibility of an automotive innovation project based on a knowledge graph. Background Art

[0002] With the rapid development of the automotive industry, the number and technological complexity of automotive innovation projects have increased significantly. Scientifically and efficiently evaluating a project's technical feasibility and market credibility has become a key focus of the industry. Traditional evaluation methods rely primarily on expert experience or simple analysis based on statistical data, such as predictions based on keyword matching or historical project success rates. However, these methods have significant limitations. On the one hand, expert experience is significantly influenced by subjective factors, making standardized evaluation difficult. On the other hand, existing data analysis methods lack the ability to deeply explore multi-dimensional relationships within technology fields. This significantly reduces the accuracy of evaluation results, especially in complex technical scenarios (such as cross-domain technology integration and dynamic technological trends).

[0003] Currently, some research attempts to improve evaluation results by introducing knowledge graph technology, revealing the connections between technical concepts by constructing automotive knowledge bases. However, existing methods still have shortcomings when applying knowledge graphs. First, in calculating correlation, they typically only consider the co-occurrence frequency of keywords, ignoring the impact of the temporal evolution characteristics of different historical projects on the strength of technical correlations. Furthermore, traditional methods oversimplify the treatment of independent keyword frequencies and fail to consider the noise interference caused by high-frequency words, resulting in poor adaptability of evaluation models for emerging technologies or niche areas. For example, certain technology combinations frequently appear in early projects but may no longer be applicable as technology iterations occur. Existing methods do not introduce time decay mechanisms to correct such correlations. Furthermore, when assigning keyword weights, they lack nonlinear processing of frequency data, causing high-frequency but low-value words to excessively interfere with evaluation results. These issues make it difficult to meet the needs of credibility assessment for the growing number of automotive innovation projects. Summary of the Invention

[0004] In response to the deficiencies in the prior art, the present invention provides a method and system for evaluating the credibility of automotive innovation projects based on knowledge graphs.

[0005] A method for evaluating the credibility of automobile innovation projects based on a knowledge graph comprises: obtaining implemented automobile projects based on a knowledge graph, extracting multiple automobile-related keywords from the implemented automobile projects and forming an automobile keyword library, obtaining text information of the automobile innovation projects, obtaining automobile-related keywords appearing in the text information according to the text information and the automobile keyword library and using them as benchmark words, and obtaining the number of times each benchmark word appears in the text information and using them as independent frequencies; obtaining a first correlation between every two benchmark words in the implemented automobile projects, and obtaining a first benchmark credibility corresponding to the automobile innovation projects based on a benchmark model, the first correlation, and multiple independent frequencies; obtaining a time node for each implemented automobile project, and obtaining a second correlation between every two benchmark words in the implemented automobile projects based on the time node for each implemented automobile project, and obtaining a second benchmark credibility corresponding to the automobile innovation projects based on the benchmark model, the second correlation, and multiple independent frequencies; obtaining a target credibility according to the first benchmark credibility and the second benchmark credibility, and associating the target credibility with the automobile innovation project.

[0006] Optionally, obtaining the first correlation between each two benchmark words in the implemented automobile projects includes: obtaining the number of times each two benchmark words appear at the same time in each implemented automobile project and recording it as the number of correlations; and obtaining the first correlation corresponding to each two benchmark words based on multiple number of correlations corresponding to each two benchmark words.

[0007] Optionally, the first correlation degree corresponding to each two benchmark words is obtained according to a plurality of correlation times corresponding to each two benchmark words, and is expressed as: , ;in, is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project.

[0008] Optionally, the baseline model in obtaining the first baseline credibility corresponding to the automobile innovation project based on the baseline model, the first correlation, and multiple independent frequencies is expressed as: ;in, is the first benchmark credibility, is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of first degree of association, is the independent frequency corresponding to the P-th benchmark word, is the number of benchmark words.

[0009] Optionally, obtaining the second correlation between each two benchmark words in the implemented automobile projects based on the time nodes of each implemented automobile project includes: obtaining the number of simultaneous appearances of each two benchmark words in each implemented automobile project and recording it as the number of correlations; obtaining the time increase or decrease weight corresponding to each implemented automobile project according to the time nodes of each implemented automobile project; and obtaining the second correlation corresponding to each two benchmark words according to the time increase or decrease weight corresponding to each implemented automobile project and the multiple number of correlations corresponding to each two benchmark words.

[0010] Optionally, the time increase or decrease weight corresponding to each implemented automobile project is obtained according to the time node of each implemented automobile project and is expressed as: ;in, is the time increase or decrease weight corresponding to the k-th implemented automobile project, The time node for automotive innovation projects, is the time node of the kth implemented automobile project, is the technology decay cycle coefficient.

[0011] Optionally, the second correlation degree corresponding to each two benchmark words is obtained based on the time increase / decrease weight corresponding to each implemented automobile project and the multiple correlation times corresponding to each two benchmark words, and is expressed as: , ;in, is the second correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project, The time increase or decrease weight corresponding to the k-th implemented automobile project.

[0012] Optionally, the target credibility is obtained according to the first reference credibility and the second reference credibility as follows: , ;in, is the target credibility, is the first benchmark credibility, is the second benchmark credibility, is the basic weight, is the attenuation weight.

[0013] A knowledge graph-based automobile innovation project credibility assessment system is also provided. The system is used to implement a knowledge graph-based automobile innovation project credibility assessment method. The system includes: an acquisition module, which is used to acquire implemented automobile projects based on the knowledge graph, extract multiple automobile field keywords from the implemented automobile projects and form an automobile keyword library, and acquire text information of the automobile innovation projects, acquire automobile field keywords appearing in the text information according to the text information and the automobile keyword library and use them as benchmark words, and acquire the number of times each benchmark word appears in the text information as an independent frequency; a first evaluation module, which is used to acquire a first correlation between every two benchmark words in the implemented automobile projects, and acquire a first benchmark credibility corresponding to the automobile innovation project based on a benchmark model, the first correlation and multiple independent frequencies; a second evaluation module, which is used to acquire a time node for each implemented automobile project, and acquire a second correlation between every two benchmark words in the implemented automobile project based on the time node for each implemented automobile project, and acquire a second benchmark credibility corresponding to the automobile innovation project based on the benchmark model, the second correlation and multiple independent frequencies; an evaluation decision module, which is used to acquire a target credibility based on the first benchmark credibility and the second benchmark credibility, and associate the target credibility with the automobile innovation project.

[0014] Optionally, the first evaluation module is further used to: obtain the number of times each two benchmark words appear at the same time in each implemented automotive project and record it as the number of associations; and obtain the first degree of association corresponding to each two benchmark words based on multiple number of associations corresponding to each two benchmark words.

[0015] The beneficial effects of the present invention are embodied in:

[0016] The entire knowledge graph-based credibility assessment method for automotive innovation projects effectively addresses core issues in traditional automotive innovation project credibility assessments, such as static technology relevance, missing time evolution features, and interference from high-frequency words, through the deep integration of knowledge graphs and multi-dimensional dynamic assessment models. This significantly improves the accuracy, timeliness, and adaptability of the assessment results. A dynamically updated automotive keyword library is first constructed, based on which a dual-path correlation calculation model is adopted: the first baseline credibility focuses on the stability analysis of historical technology combinations, while the second baseline credibility introduces a time-based weighting mechanism, assigning higher weights to recent technology hotspots while suppressing the attenuation of outdated technology combinations. The two generate target credibility through a configurable fusion strategy, which not only retains the stability assessment of mature technology solutions but also strengthens the forward-looking prediction of hot innovations. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following briefly describes the drawings required for the specific embodiments or the description of the prior art. Similar elements or parts are generally identified by similar reference numerals throughout the drawings. Elements or parts in the drawings are not necessarily drawn to scale.

[0018] Figure 1 This is a schematic diagram of the steps of the knowledge graph-based automobile innovation project credibility assessment method of the present invention;

[0019] Figure 2 This is a schematic diagram of some steps in S2 of the knowledge graph-based automobile innovation project credibility assessment method of the present invention;

[0020] Figure 3 This is a schematic diagram of some steps in S3 of the knowledge graph-based automobile innovation project credibility assessment method of the present invention. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0022] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention as claimed, but rather merely represents selected embodiments of the present invention. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without creative effort are intended to fall within the scope of protection of the present invention.

[0023] It should be noted that similar reference numerals and letters represent similar items in the following drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined or explained in subsequent drawings. In addition, the terms "first," "second," etc. are used only to distinguish the descriptions and are not to be understood as indicating or implying relative importance.

[0024] like Figure 1 As shown in the figure, a credibility assessment method for automotive innovation projects based on knowledge graph is provided, including:

[0025] S1. Acquire implemented automotive projects based on the knowledge graph, extract multiple automotive keywords from the implemented automotive projects and form an automotive keyword library, and acquire text information of automotive innovation projects. Based on the text information and the automotive keyword library, obtain automotive keywords that appear in the text information and use them as benchmark words. The number of times each benchmark word appears in the text information is obtained and used as an independent frequency.

[0026] S2. Obtain a first correlation between each two benchmark words in implemented automotive projects, and obtain a first benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the first correlation, and multiple independent frequencies;

[0027] S3. Obtaining a time point for each implemented automotive project, and based on the time point for each implemented automotive project, obtaining a second correlation between every two benchmark words in the implemented automotive project, and obtaining a second benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the second correlation, and the plurality of independent frequencies;

[0028] S4. Obtain target credibility based on the first benchmark credibility and the second benchmark credibility, and associate the target credibility with the automobile innovation project.

[0029] In this embodiment, it should be noted that in S1, based on keyword extraction and frequency statistics from the knowledge graph, a dynamic automotive keyword library is constructed and the technical characteristics of innovative projects are quantified. Specifically, all implemented automotive project data (such as technical documents, patents, or R&D reports) are first retrieved from the knowledge graph, and core terms in the project descriptions, such as "on-board sensors," "hydrogen fuel cells," and "drive-by-wire chassis," are extracted using natural language processing technology to form an automotive keyword library covering technical nodes, components, and processes. This library uses semantic disambiguation and domain filtering to ensure that keywords are strongly related to automotive innovation scenarios. For example, for new energy projects, the library may include detailed technical terms such as "solid-state battery thermal management" and "charging pile compatibility," while traditional structural projects may cover terms such as "body stiffness topology optimization" and "transmission system NVH," thereby constructing a multi-dimensional technical semantic network.

[0030] Furthermore, for the automotive innovation projects to be evaluated, their textual information (such as technical solutions and requirement specifications) is parsed, and all associated benchmark words are matched based on the keyword library. The original frequency of occurrence of each benchmark word in the text is then counted. For example, in the text of a certain intelligent driving project, "multimodal fusion algorithm" appears frequently as a benchmark word, indicating that this technology is the core of the project; while "lidar calibration" appears less frequently and may belong to an auxiliary module. This process not only captures the distribution density of technical elements but also provides basic parameters for subsequent correlation calculations. It should be noted that the dynamic update mechanism of the keyword library can incorporate emerging technical terms (such as "vehicle-road-cloud integration") to avoid the lag of the semantic library due to technological iteration and ensure that the benchmark words can accurately reflect the technical hotspots of the innovation project.

[0031] In S2, a credibility calculation based on static correlation and frequency suppression is performed, quantifying the technical plausibility of automotive innovation projects through historical technology correlation analysis and frequency optimization. Specifically, for all pairwise combinations of benchmark terms, their co-occurrence characteristics in implemented automotive projects are extracted. For example, the co-occurrence patterns of "autonomous driving algorithm" and "high-precision map" in different project documents are analyzed. By aggregating co-occurrence statistics from historical data, a first correlation is generated to represent the stability of the technology combination. This correlation reflects the strength of the technical synergy between the benchmark terms. For example, if a technology combination co-occurs frequently in many historical projects, its correlation is universal; conversely, if it only appears sporadically, it may be a marginal combination. Furthermore, the benchmark model uses a suppression function (such as logarithmic compression or threshold truncation) to reduce the noise of independent frequencies, preventing interference from high-frequency but low-correlation terms (such as the general term "vehicle system") in the credibility assessment, thereby extracting frequency contributions with actual technical value.

[0032] For example, when evaluating an electrification platform project, the benchmark terms "battery module thermal runaway warning" and "BMS redundancy design" may show high co-occurrence correlation in historical projects, indicating strong technical coupling between the two. Meanwhile, while "fast charging protocol compatibility" has a high independent frequency, its weight is appropriately reduced after suppression, avoiding inflated credibility due to a single high-frequency term. By multiplying the mean correlation by the weighted sum of suppressed frequencies, the model strengthens the synergistic effects of the technology combination while weakening the noise impact of redundant terms. The final output, the first benchmark credibility, effectively reflects the project's advantages in technological maturity and logical coherence.

[0033] In S3, dynamic time-weighted correlation correction and credibility optimization are implemented, introducing a time evolution mechanism to dynamically calibrate technology correlations, thereby improving the adaptability of assessment results to technological trends. Specifically, by analyzing the temporal distribution characteristics of implemented automotive projects, differentiated weight coefficients are assigned to projects at different historical stages. For example, projects within the past three years receive a higher temporal weight, while projects from five years ago receive a significantly lower weight. Furthermore, a time-weighted calculation is performed on the co-occurrence relationships between benchmark terms. For example, the frequent co-occurrence of "Internet of Vehicles Security Protocol" and "OTA Upgrade" in recent projects will receive a higher weight, while combinations such as "mechanical power steering" and "traditional transmission" in earlier projects will see their weights decay due to technological obsolescence, thereby dynamically correcting for time-sensitive biases in correlation. This mechanism ensures that the correlation strength of technology combinations reflects the true state of current industry technology iterations, avoiding evaluation distortions caused by lagging historical data.

[0034] Furthermore, when calculating the second benchmark credibility, the time-weighted mean correlation value is combined with the independent frequency of the suppression process to form an evaluation indicator that takes into account both technical synergy and timeliness. For example, in a certain smart cockpit project, although the correlation between "multi-screen interactive design" and "AR-HUD" was low in early projects, recent technological breakthroughs have caused them to frequently co-occur in the latest projects. The introduction of time weighting significantly increased the contribution value of the correlation between the two. On the other hand, even if the independent frequency of outdated technologies such as "CD player module" is high, its weight will be greatly weakened by the suppression function. This dynamic adjustment enables the evaluation model to automatically capture the migration patterns of technology hotspots, such as the transition trend from "ternary lithium batteries" to "solid-state batteries" in the new energy field, thereby more accurately judging the technical foresight and feasibility of innovative projects.

[0035] In S4, multi-dimensional credibility is integrated and evaluated. Its core lies in integrating static technical correlations with dynamic time evolution characteristics to form a target credibility that comprehensively reflects the technical feasibility and trend adaptability of innovative projects. Specifically, by weighted fusion of the first benchmark credibility (based on historical co-occurrence correlation and frequency suppression) and the second benchmark credibility (based on time-weighted correlation and frequency optimization), for example, assigning a higher weight to the second benchmark credibility to highlight the impact of technological trends, thereby balancing the assessment bias between technological maturity and foresight. For example, in a hybrid project, if the "energy recovery system" and "electronic clutch module" in its technical solution perform moderately in the first benchmark credibility, but their combination has significantly improved correlation in recent projects, the second benchmark credibility is higher. After fusion, the target credibility will be closer to the current mainstream direction of technology, avoiding the risk of underestimation due to outdated historical data.

[0036] Furthermore, the target credibility generation mechanism can dynamically adjust the integration strategy in combination with field needs. For example, for disruptive innovation projects (such as all-solid-state battery research and development), if the correlation of its technical keywords in recent projects increases rapidly, even if the historical co-occurrence frequency is low, by increasing the weight coefficient of the second benchmark credibility, the target credibility can still accurately reflect its potential for technological breakthroughs; while for mature technology improvement projects (such as traditional engine energy efficiency optimization), the focus is on the stability assessment of the first benchmark credibility. Ultimately, the target credibility is bound to the project in the form of a numerical value or level. For example, a high-confidence result can be mapped to "clear technical path and low implementation risk", providing a quantitative basis for investment decisions or R&D priority division. At the same time, after quantification, it can support horizontal comparison across projects, such as identifying the difference in technical credibility between "vehicle-road collaboration" and "autonomous driving algorithm" projects, driving precise resource allocation.

[0037] In summary, the entire knowledge graph-based automotive innovation project credibility assessment method, through the deep integration of knowledge graphs and multi-dimensional dynamic assessment models, effectively solves core issues in traditional automotive innovation project credibility assessments, such as the static nature of technology relevance, the lack of time evolution characteristics, and interference from high-frequency words. This significantly improves the accuracy, timeliness, and adaptability of the assessment results. Specifically, a dynamically updated automotive keyword library is first constructed. Based on this, a dual-path relevance calculation model is adopted: the first baseline credibility focuses on the stability analysis of historical technology combinations, while the second baseline credibility introduces a time-based weighting mechanism, assigning higher weights to recent technology hotspots while suppressing the attenuation of outdated technology combinations. The two generate a target credibility through a configurable fusion strategy, which not only retains the stability assessment of mature technology solutions but also strengthens the forward-looking prediction of hot innovations.

[0038] like Figure 2 As shown, in one embodiment, obtaining the first correlation between each two reference words in the implemented automobile project in S2 includes:

[0039] S21, obtaining the number of times each two benchmark words appear together in each implemented automotive project and recording the number of associations;

[0040] S22 . Obtain a first correlation degree corresponding to each two benchmark words according to a plurality of correlation times corresponding to each two benchmark words.

[0041] In this embodiment, it should be noted that, specifically, in S21, for each pair of benchmark words, all implemented automotive project documents (e.g., patents and technical reports) are traversed to count the number of times the two words co-occur within the same project. For example, the co-occurrence of "autonomous driving algorithm" and "multi-sensor fusion" in a smart driving project, or the co-occurrence of "battery thermal management" and "BMS control" in a new energy project, is counted. This process is automated using text mining techniques, such as sliding window or semantic dependency analysis, to identify contextual co-occurrence relationships between technical terms, avoiding misjudgments caused by simple proximity. For example, when analyzing an in-vehicle communications project, "V2X protocol" and "low-latency transmission" may co-occur across multiple paragraphs due to functional coupling within the technical solution description. However, "in-vehicle entertainment system" and "battery module," while appearing in the same document, have zero co-occurrence due to their unrelated technologies. This accurately captures the true relevance of the technology combination.

[0042] S22 is the correlation integration and normalization stage, whose goal is to transform scattered co-occurrence features into standardized technical correlation indicators. Specifically, the co-occurrence counts of each benchmark word pair across all historical projects are aggregated and analyzed. For example, the average co-occurrence strength across different projects is calculated. If a technology combination (such as "steer-by-wire" and "redundant control system") consistently co-occurs across most projects, the average value is high; conversely, if it only appears sporadically in a single project, the average value is low. Further normalization is performed to map the values ​​to a range from 0 to 1. For example, if a benchmark word pair has the highest total co-occurrence count in the database, the normalized result approaches 1, indicating a strong technical correlation; while a value close to 0 indicates a marginalized combination. This process makes correlations across different technical fields comparable. For example, the correlation between "hydrogen fuel cell stacks" and "proton exchange membranes" in the new energy sector can be compared horizontally with the correlation between "lightweight vehicle body" and "carbon fiber materials" in the structural design sector, providing a unified benchmark for evaluating cross-disciplinary innovation projects.

[0043] In one embodiment, in S22, the first correlation degree corresponding to each two benchmark words is obtained based on the multiple correlation times corresponding to each two benchmark words, and is expressed as:

[0044] , ;in,

[0045] is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project.

[0046] In this embodiment, it should be noted that the expression , represents the average number of times two benchmark words co-occur in all implemented automotive projects. For example, if the benchmark word pair ("autonomous driving algorithm", "high-precision map") appears 100 times in 100 historical projects, the average value is 1 time / project. This can eliminate the impact of the number of projects. Specifically, by dividing by the total number of projects , to avoid statistical bias caused by the number of projects. For example, if a technology combination appears 10 times in 10 projects (on average 1 time / project), its correlation should be equivalent to that of appearing 100 times in 100 projects (on average 1 time / project), rather than the latter being mistakenly judged as a stronger correlation due to its higher total number. The universality of technology combinations can also be quantified. Specifically, the average value reflects the prevalence of technology combinations in historical projects. High-frequency co-occurrence (such as an average of 5 times / project) indicates stable technology synergy, while low-frequency co-occurrence (such as an average of 0.1 times / project) may be a random combination.

[0047] Further, Normalization is implemented, compressing the output range of the average co-occurrence count to (0, 1). This means that when the average co-occurrence count is low (e.g., x=1), the output value increases slowly (approximately 0.73); when the count is high (e.g., x=5), the output approaches 1 but the growth rate slows significantly (approximately 0.99). This prevents high-frequency but low-value terms (such as the general term "vehicle system") from being overly amplified due to their high absolute frequency. Furthermore, the correlation of all technology combinations is mapped to the same scale (0, 1), facilitating horizontal comparison. For example, "battery thermal management (BMS control)" in the new energy sector and "body stiffness (lightweight materials)" in the traditional sector can be evaluated using the same criteria.

[0048] In summary, if we directly use the total frequency or linear normalization, high-frequency but weakly technically related words (such as "vehicle system" appears frequently in multiple projects) will lead to an artificially high correlation. Through this expression, even if a pair of benchmark words co-occurs abnormally frequently in a few projects (such as the second project in a certain project), =50), its contribution will be diluted by the average value (for example, when n=100, the contribution of a single item will only increase by 0.5), and the increase in correlation after normalization and compression will be limited.

[0049] Assume that two pairs of benchmark words: Benchmark word pair A: appear 10 times in 10 items on average (total frequency 100), then =10, ≈1.0. Base word pair B: It co-occurs once in 100 items on average (total frequency 100), then =1, ≈0.73. Although the total frequencies are the same, benchmark word pair A is judged to be strongly associated due to its high frequency concentration in a few items, while benchmark word pair B is reasonably weakened due to its widespread low-frequency co-occurrence.

[0050] In one embodiment, the baseline model in S2 for obtaining the first baseline credibility corresponding to the automobile innovation project based on the baseline model, the first correlation, and the multiple independent frequencies is expressed as:

[0051] ;in,

[0052] is the first benchmark credibility, is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of first degree of association, is the independent frequency corresponding to the P-th benchmark word, is the number of benchmark words.

[0053] In this embodiment, it should be noted that Represents the average first relevance of all benchmark word pairs. For example, if a project contains three benchmark words (A, B, C), then we need to calculate 、 、 There are 3 pairs of correlations, and their average is calculated. In this way, firstly, the average value can be used to reflect the synergy balance level of all technical combinations in the project. For example, if the average correlation value of a project is 0.9 (close to 1), it indicates that most combinations in its technical solutions have strong historical synergy; if the average value is only 0.3, the technical logic may be loose or rely on marginal combinations; secondly, the deviation of the number of benchmark words can be eliminated by dividing by the logarithm. (Total number of first correlations) This method avoids bias in reliability due to the number of base words in a project. For example, a project with five base words requires calculating 10 pairs of correlations, while a project with three base words only calculates three pairs. Averaging enhances comparability.

[0054] Furthermore, the second part of the expression In the example, the nonlinear compression of the independent frequencies of the benchmark words is reflected. The numerator is the original frequency of a single benchmark word. ; Denominator ,in is the total frequency of all benchmark words; suppress high , the logarithmic function grows slowly, so the higher the total frequency, the greater the increase in the denominator, thus compressing each contribution; while retaining low The effective signal can be retained regardless of the total frequency changes A certain weight is given to avoid that the effective signals in the technical sub-sectors are completely drowned out.

[0055] Furthermore, the correlation mean is finally multiplied by the suppressed frequency sum, that is, = mean correlation × sum of suppression frequencies. First, the mean correlation reflects the synergy of the technology combination, and the sum of suppression frequencies reflects the density of technical elements. The multiplication of the two ensures that the credibility meets both the "rational technical logic" and "prominent core elements" requirements. For example, if a project has a high mean correlation but an extremely low sum of frequencies (concise technical solution description), or an extremely high sum of frequencies but a low mean correlation (stacked with high-frequency irrelevant terms), its credibility will be suppressed. The product operation strengthens the synergistic effect of high correlation and reasonable frequency. For example, the product of a mean correlation of 0.8 and a sum of suppression frequencies of 20 (16) is significantly higher than the product of a correlation of 0.5 and a sum of suppression frequencies of 30 (15), reflecting that technical synergy takes precedence over simple element stacking.

[0056] For example: Project A: total frequency 1000, frequency after suppression and =1000 / 7.9≈126.6. Project B: Total frequency 200, frequency after suppression and =200 / 6.3≈31.7. Although the total frequency of item A is 5 times higher, the frequency difference after suppression is only 4 times. Furthermore, further constraints are applied by combining the mean relevance to avoid a single high-frequency word dominating the evaluation.

[0057] Further explanation: To evaluate the credibility of a smart driving project, it contains three benchmark words (A: multi-sensor fusion, B: environmental perception algorithm, C: redundant control), with independent frequencies F1=50, F2=30, F3=20, and correlation =0.9, =0.7, =0.8. Then the mean correlation is 0.8, the total frequency is 100, 17.8, and finally the first benchmark credibility is 14.24.

[0058] Comparison scenario: If another project contains 3 benchmark words (A: battery thermal management, B: autonomous driving algorithm, C: in-car entertainment), the independent frequency is artificially high (F1=200, F2=150, F3=150, total frequency=500), but the average correlation is low ( =0.1, =0.1, =0.1, mean=0.1). Then 69.25, and finally the first benchmark credibility is 6.9.

[0059] Then, item 1: Although the frequency is low, it has reasonable credibility (14.24) due to its high correlation.

[0060] Project 2: Despite its artificially high frequency, its credibility (6.9) is no higher than Project 1 due to its low mean correlation. Comparison conclusion: Project 1's technical solution (high synergy + reasonable frequency) is superior to Project 2 (low synergy + high-frequency stacking).

[0061] like Figure 3 As shown, in one embodiment, obtaining the second correlation between each two benchmark words in the implemented automobile projects based on the time nodes of each implemented automobile project in S3 includes:

[0062] S31, obtaining the number of times each two benchmark words appear together in each implemented automotive project and recording the number of associations;

[0063] S32. Obtaining the time increase or decrease weight corresponding to each implemented automobile project according to the time node of each implemented automobile project;

[0064] S33 , obtaining a second correlation degree corresponding to each two benchmark words according to the time increase / decrease weight corresponding to each implemented automobile project and a plurality of correlation times corresponding to each two benchmark words.

[0065] In this embodiment, it should be noted that in S31, for each pair of benchmark words, all implemented automotive project documents are traversed to count the number of co-occurrences of the two in the same project, such as the co-occurrence frequency of "Internet of Vehicles Protocol" and "OTA Upgrade" in a certain intelligent connected project. This is the same as the static statistics in S21, except that the number of associations in S31 will be used as the second correlation degree in S33 in combination with the time-dependent weighting. For example, a pair of benchmark words may co-occur frequently in early projects, but the frequency is reduced in recent projects. At this time, the original data needs to be retained to support dynamic correction. This step ensures the integrity of historical data through time-independent co-occurrence statistics and avoids information loss due to pre-weighting.

[0066] In S32, the number of associations is weighted differently according to the time distance of the project. Specifically, a time increase or decrease weight is assigned to each implemented automotive project. For example, the weight of the project within the past three years is 1.8, and the weight of the project five years ago is 0.5. The weight value decays nonlinearly with time (such as exponential decay). The weight range is limited to between 0 and 2, which not only retains the reference value of historical data (the weight is not 0) but also highlights the priority of recent projects. For example, for the technology combination of "solid-state battery" and "thermal runaway warning", if it only appears frequently in projects in the past two years, the high weight of recent projects will significantly increase its correlation; while the combination of "mechanical supercharger" and "carburetor" in early projects has a weight close to 0 due to technological obsolescence, thereby suppressing the interference of outdated technology on the evaluation.

[0067] In S33, time weights are integrated with co-occurrence counts to generate a dynamically modified second degree of correlation. For example, in early projects, the correlation counts between "AR-HUD" and "multi-screen interaction" were low and the weights were attenuated. However, in recent projects, due to technological breakthroughs, the correlation counts have surged and the weights have increased. After integration, their second degree of correlation is significantly higher than the static correlation. This mechanism enables the correlation strength of technology combinations to dynamically reflect industry trends. For example, in the field of electrification, the transition from "ternary lithium batteries" to "cobalt-free batteries" can drive an increase in correlation even if the historical co-occurrence counts are low, thereby accurately identifying the synergistic value of emerging technologies.

[0068] In one embodiment, in S32, the time increase / decrease weight corresponding to each implemented automobile project is obtained according to the time node of each implemented automobile project and is expressed as:

[0069] ;in,

[0070] is the time increase or decrease weight corresponding to the k-th implemented automobile project, The time node for automotive innovation projects, is the time node of the kth implemented automobile project, is the technology decay cycle coefficient.

[0071] In this embodiment, it should be noted that Among them Indicates the time node of the current automotive innovation project The time node of the kth implemented automotive project in history The time difference (in years) increases with the time difference. The value of grows linearly, causing the denominator to expand rapidly, thus making the weight In this way, the weight of recent projects (with a small time difference) is significantly higher than that of long-term projects. For example, if - =0 (i.e. historical projects and innovation projects are in the same year), weight =2 (maximum); if the time difference is 10 years ( = 0.1), the weight decays to 0.736. More importantly, The decay rate can be controlled. The general technology cycle is 7 years. It can be 0.1, which means that the half-life of technology correlation is about 7 years (the time difference when the weight drops to 1 is 7 years), which is consistent with the long iteration cycle of automobile technology.

[0072] In summary, the technology combinations that have been eliminated in early projects (such as "carburetor-mechanical supercharger") have a high correlation in the static model due to their high historical co-occurrence frequency. This implementation method can suppress outdated technologies through time decay; it can also solve the problem that emerging technology combinations (such as "solid-state battery-thermal management") do not exist or appear infrequently in early projects, resulting in underestimated static correlation.

[0073] In one embodiment, in S33, the second correlation degree between each two benchmark words is obtained based on the time increase / decrease weight corresponding to each implemented automobile project and the multiple correlation times corresponding to each two benchmark words, and is expressed as:

[0074] , ;in,

[0075] is the second correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project, The time increase or decrease weight corresponding to the k-th implemented automobile project.

[0076] In this embodiment, it should be noted that the expression for obtaining the first degree of association is substantially the same, except that In the example, , the time-weighted average of the co-occurrences of each pair of benchmark words in all historical projects is calculated. For example, if the benchmark word pair ("solid-state battery", "thermal management") appears 10 times in the 2023 project (weight 2.0) and 5 times in the 2020 project (weight 1.48), the weighted average is 13.7. This way, recent technology trends can be strengthened: through time weighting , the co-occurrence times of recent projects are amplified. For example, the weight of the 2023 project is 1.35 times that of the 2020 project (2.0 / 1.48), and its co-occurrence contribution accounts for a higher proportion; it can also suppress the interference of historical data. The weight of early projects (such as 2010) may be as low as below 0.5. Even if their co-occurrence times are high, their contribution is significantly weakened after weighting. For example, the 2010 project co-occurs 20 times (weight 0.5), and its effective contribution is only 10, while the 2023 project co-occurs 20 times (weight 2.0) contributes 40.

[0077] In one embodiment, obtaining the target credibility according to the first reference credibility and the second reference credibility in S4 is expressed as:

[0078] , ;in,

[0079] is the target credibility, is the first benchmark credibility, is the second benchmark credibility, is the basic weight, is the attenuation weight.

[0080] In this embodiment, it should be noted that the dual-dimensional weight distribution mechanism

[0081] The expression is expressed by weight coefficient (base weight) and (Decrease weight) Dynamically balance technology maturity and trend adaptability. Basic weight : reflects the static correlation ( ) and focuses on the stability of historical technology synergy; for example, for improvement projects with mature technologies but low popularity in the near future (such as traditional engine energy efficiency optimization), =0.8, emphasizing the reliability of historical data. Reflecting dynamic correlation ( ) and focuses on the foresight of technology trends; for example, for innovative projects with immature technology and low recent popularity (such as all-solid-state battery research and development), =0.9, even though its historical correlation is low, it can still be trusted by recent trends.

[0082] Furthermore, the weight coefficient clearly divides the contribution ratio of the two types of credibility, which is convenient for flexible adjustment according to the evaluation needs, and supports the introduction of more evaluation dimensions (such as market risk, patent barriers) to maintain the consistency of the model framework. At the same time, the weight coefficient can be dynamically configured according to the technology life cycle (introduction stage, growth stage, maturity stage), for example: in the introduction stage =0.9: Strengthen emerging technology trends. Mature stage =0.8: Rely on historical technical stability.

[0083] A knowledge graph-based automobile innovation project credibility assessment system is also provided. The system is used to implement the above-mentioned knowledge graph-based automobile innovation project credibility assessment method. The system includes:

[0084] An acquisition module is used to acquire implemented automotive projects based on the knowledge graph, extract multiple automotive keywords from the implemented automotive projects and form an automotive keyword library, and acquire text information of automotive innovation projects. Based on the text information and the automotive keyword library, the automotive keywords appearing in the text information are acquired and used as benchmark words, and the number of times each benchmark word appears in the text information is acquired and used as an independent frequency.

[0085] A first evaluation module is configured to obtain a first correlation between each two benchmark words in implemented automotive projects, and obtain a first benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the first correlation, and a plurality of independent frequencies;

[0086] A second evaluation module is configured to obtain a time node for each implemented automotive project, obtain a second correlation between each two benchmark words in the implemented automotive project based on the time node for each implemented automotive project, and obtain a second benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the second correlation, and a plurality of independent frequencies;

[0087] The evaluation decision module is used to obtain target credibility according to the first benchmark credibility and the second benchmark credibility, and associate the target credibility with the automobile innovation project.

[0088] In one embodiment, the first evaluation module is further configured to: obtain the number of simultaneous occurrences of each two benchmark words in each implemented automotive project and record the number of associations; and obtain a first degree of association corresponding to each two benchmark words based on the multiple number of associations corresponding to each two benchmark words.

[0089] In one embodiment, the second evaluation module is also used to: obtain the number of times each two benchmark words appear at the same time in each implemented automobile project and record it as the number of associations; obtain the time increase or decrease weight corresponding to each implemented automobile project based on the time node of each implemented automobile project; obtain the second correlation degree corresponding to each two benchmark words based on the time increase or decrease weight corresponding to each implemented automobile project and the multiple correlation times corresponding to each two benchmark words.

[0090] In this embodiment, it should be noted that, regarding the above-mentioned automobile innovation project credibility assessment system based on knowledge graph, the specific method of performing operations has been described in detail in the implementation of the automobile innovation project credibility assessment method based on knowledge graph, and will not be elaborated here.

[0091] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings. However, the present disclosure is not limited to the specific details of the above embodiments. Within the technical concept of the present disclosure, various simple modifications can be made to the technical solutions of the present disclosure, and these simple modifications all fall within the scope of protection of the present disclosure.

[0092] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0093] In addition, the various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

[0094] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and description of the present invention.

Claims

1. A method for evaluating the credibility of automotive innovation projects based on knowledge graph, characterized by: include: Based on the knowledge graph, we obtain implemented automotive projects, extract multiple automotive keywords from the implemented automotive projects and form an automotive keyword library. We also obtain text information about automotive innovation projects. Based on the text information and the automotive keyword library, we obtain automotive keywords that appear in the text information and use them as benchmark words. We also obtain the number of times each benchmark word appears in the text information and use it as an independent frequency. Obtaining a first correlation between every two benchmark words in implemented automotive projects, and obtaining a first benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the first correlation, and a plurality of independent frequencies; Wherein, obtaining the first correlation degree is expressed as: , ; is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project; Among them, the baseline model is expressed as: ; is the first benchmark credibility, is the number of first degree of association, is the independent frequency corresponding to the P-th benchmark word, is the number of benchmark words; Obtaining a time node for each implemented automobile project, and based on the time node for each implemented automobile project, obtaining a second correlation between every two benchmark words in the implemented automobile project, and obtaining a second benchmark credibility corresponding to the automobile innovation project based on the benchmark model, the second correlation, and a plurality of independent frequencies; The obtaining of the second correlation degree includes: obtaining the number of times each two benchmark words appear simultaneously in each implemented automobile project and recording the number of correlations; obtaining a time increase or decrease weight corresponding to each implemented automobile project based on a time node of each implemented automobile project; and obtaining a second correlation degree corresponding to each two benchmark words based on the time increase or decrease weight corresponding to each implemented automobile project and a plurality of correlation numbers corresponding to each two benchmark words. The acquisition time increase or decrease weight is expressed as: ; is the time increase or decrease weight corresponding to the k-th implemented automobile project, The time node for automotive innovation projects, is the time node of the kth implemented automobile project, is the technology decay cycle coefficient; Wherein, obtaining the second correlation degree is expressed as: , ; is the second correlation degree corresponding to the i-th benchmark word and the j-th benchmark word; The target credibility is obtained according to the first benchmark credibility and the second benchmark credibility, and the target credibility is associated with the automobile innovation project.

2. The method for evaluating the credibility of automobile innovation projects based on knowledge graph according to claim 1 is characterized in that: The obtaining of a first correlation between each two reference words in the implemented automobile project includes: Obtain the number of times each two benchmark words appear together in each implemented automotive project and record it as the number of associations; A first correlation degree corresponding to each two reference words is obtained according to a plurality of correlation times corresponding to each two reference words.

3. The method for evaluating the credibility of automobile innovation projects based on knowledge graph according to claim 1 is characterized in that: The target credibility is obtained according to the first reference credibility and the second reference credibility as follows: , ;in, is the target credibility, is the first benchmark credibility, is the second benchmark credibility, is the basic weight, is the attenuation weight.

4. A knowledge graph-based automotive innovation project credibility assessment system, characterized by: The system comprises: An acquisition module is used to acquire implemented automotive projects based on the knowledge graph, extract multiple automotive keywords from the implemented automotive projects and form an automotive keyword library, and acquire text information of automotive innovation projects. Based on the text information and the automotive keyword library, the automotive keywords appearing in the text information are acquired and used as benchmark words, and the number of times each benchmark word appears in the text information is acquired and used as an independent frequency. A first evaluation module is configured to obtain a first correlation between each two benchmark words in implemented automotive projects, and obtain a first benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the first correlation, and a plurality of independent frequencies; Wherein, obtaining the first correlation degree is expressed as: , ; is the first correlation between the i-th benchmark word and the j-th benchmark word, is the number of implemented automotive projects, is the number of occurrences of the i-th benchmark word and the j-th benchmark word in the k-th implemented automotive project; Among them, the baseline model is expressed as: ; is the first benchmark credibility, is the number of first degree of association, is the independent frequency corresponding to the P-th benchmark word, is the number of benchmark words; A second evaluation module is configured to obtain a time node for each implemented automotive project, obtain a second correlation between each two benchmark words in the implemented automotive project based on the time node for each implemented automotive project, and obtain a second benchmark credibility corresponding to the automotive innovation project based on the benchmark model, the second correlation, and a plurality of independent frequencies; The obtaining of the second correlation degree includes: obtaining the number of times each two benchmark words appear simultaneously in each implemented automobile project and recording the number of correlations; obtaining a time increase or decrease weight corresponding to each implemented automobile project based on a time node of each implemented automobile project; and obtaining a second correlation degree corresponding to each two benchmark words based on the time increase or decrease weight corresponding to each implemented automobile project and a plurality of correlation numbers corresponding to each two benchmark words. The acquisition time increase or decrease weight is expressed as: ; is the time increase or decrease weight corresponding to the k-th implemented automobile project, The time node for automotive innovation projects, is the time node of the kth implemented automobile project, is the technology decay cycle coefficient; Wherein, obtaining the second correlation degree is expressed as: , ; is the second correlation degree corresponding to the i-th benchmark word and the j-th benchmark word; The evaluation decision module is used to obtain target credibility according to the first benchmark credibility and the second benchmark credibility, and associate the target credibility with the automobile innovation project.

5. The automobile innovation project credibility assessment system based on knowledge graph according to claim 4 is characterized in that: The first evaluation module is further configured to: Obtain the number of times each two benchmark words appear together in each implemented automotive project and record it as the number of associations; A first correlation degree corresponding to each two reference words is obtained according to a plurality of correlation times corresponding to each two reference words.

Images