KR20190100506A - Method for forecasting of promising patents and apparatus thereof - Google Patents

Abstract

Disclosed are a method for forecasting promising patents and an apparatus thereof. The method for forecasting promising patents to forecast promising of an analysis target patent by a promising patent predicting device comprises the following steps: obtaining bibliographic information of the analysis target patent; calculating a first variable value and a second variable value based on detailed technology elements of technology classification in the bibliographic information; and predicting the number of citations in the future for the analysis target patent by applying deep learning to the calculated first variable value and second variable value.

Description

Method for forecasting of promising patents and apparatus

TECHNICAL FIELD The present invention relates to a method for predicting a patent and a device thereof, and more particularly, to a method for predicting a patent and a device for predicting the promising potential of a patent to be analyzed using prior indicators capable of confirming the invention and technical features of the patent.

Technology forecasts enable the prospect of future technology changes. While the impact of technological change varies from country to country, companies typically create opportunities for profit by identifying and leading technology changes in their competition.

On the other hand, as evidenced by many studies, improperly responding to technological changes can cause long-lasting companies to decline. Appropriate forecasts and forecasts for technological change are therefore critical to aid decision making in management issues such as prioritization, resource allocation and risk reduction.

In recent years, as technology advances faster and more complex and competition intensifies, technology prediction has become more strategically important. As a result, various attempts have been made using appropriate models, methods and tools to formulate the technology prediction process.

The process of technological change and innovation is seen as a process based on the mechanism of combining conventional and new technologies. It depends on the complex functions of interconnected technologies rather than separate functions, and it is very important to predict which of the current technologies will drive future technological change.

There are two major assumptions in technology forecasting: current technology growth is a precursor to upcoming technology change and that technology change will be affected by the technology involved. Understanding how current technology growth will affect future technology is one of the most important tasks in technology forecasting because it directly affects other activities. Under this background, developing a systematic approach to assessing future technological impacts is essential to providing important information to aid strategic decision making.

Evaluating future technology impacts is inherently difficult to handle because of difficulties such as uncertainty about the future, unreliable data, and the complexity of realistic feedback. While some techniques have been devised to reduce this uncertainty and support the decision making process, conventional methods are useful for different purposes, but have the limitation that analysis results are strongly dependent on the data source.

To overcome these limitations, technical databases and innovative calculation algorithms have been introduced. This has facilitated the development of very powerful and compelling technical predictions, such as environmental research, technical roadmaps, quantitative bibliographic analysis, and creative problem solving theory, of which patent citation analysis is the most adopted for conceptual and empirical analysis of technological impact. Is one of the ways.

The basic concept of patent citation analysis is that when a patent cites another patent, there is a technical link between the two patents. Traditionally, patent citation analysis has been recognized not only to provide the only opportunity to meet the conceptual and qualitative demand for analysis of technological change, but to empirically explain most of the technological innovation. That is, the patent citation information is a representative method for analyzing a coupling mechanism between technologies, and thus, various technical opportunities can be obtained by estimating future technical knowledge flows and predicting future technical relationships.

This patent citation analysis has strengths in terms of enabling a systematic and empirical analysis of technology predictions, but has the limitation of assuming that past trends also appear in the future. It does not adequately reflect changes in technological impact.

In addition, the prior art focuses only on researches for estimating future promising patents by analyzing citation information of all patents, and the technology of the patent for one patent (for example, a patent at the time of filing and a patent without a cited number). There is insufficient research to predict the prospects using the characteristics of the invention.

In this regard, there is a Korean Patent Publication No. 2014-0146437 entitled "Business performance prediction apparatus and method using patent information".

The technical problem to be solved by the present invention is to provide a patent promising predictive method and apparatus for predicting the promising of the post-patent using a prior indicator that can confirm the invention and technical features of the patent.

Another technical problem to be solved by the present invention is to provide a patent promising prediction method and apparatus for predicting the number of citations, standard patents, and technology transfer at a future point of time using the technical classification of the patent.

The technical problems of the present invention are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those skilled in the art from the following description.

In accordance with an aspect of the present invention, there is provided a method for predicting patent promising, in which a method for predicting promising a patent of an analysis target patent is obtained by a device for predicting a patent, wherein the bibliographic information of the analysis target patent is obtained; Calculating a first variable value and a second variable value based on detailed description elements of the technical classification in the bibliographic information, and applying deep learning to the calculated first variable value and the second variable value, Predicting the number of citations in the future.

Preferably, the description classification is a structure in which descriptions are classified hierarchically in the order of section, class, subclass, main group, and subgroup, and the detailed description element may be a subclass.

Preferably, the calculating of the first variable value comprises: generating a subclass pair by combining subclasses in the technical classification of the analysis target patent, and technical classification of past patents filed before the filing date of the analysis target patent. Generating a subclass pair by combining the subclasses of the subclass, comparing the subclass pair of the target patent to a subclass pair of the past patents, and extracting a subclass pair only in the target patent; Calculating a first variable value based on the number of subclass pairs existing only in the target patent and the total number of subclass pairs of the analysis target patent.

Preferably, the calculating of the second variable value may include identifying the same technology classification patents having the same technology classification as that of the analysis target patent among the past patents applied before the application date of the analysis target patent, Calculating a difference in the filing date between the analysis target patent and each same technology classification patent, calculating a second variable value using the number of subclasses of the analysis target patent, the number of the same technology classification patent, and the difference in the filing date. It may include.

Preferably, after calculating the first variable value and the second variable value, based on at least one of the number of technical classification of the analysis target patent, the number of back citations, the number of claims, the number of inventors, Calculating a value, applying at least one or more of the first variable value, the second variable value, and values of predefined variables to a logistic regression model to predict whether the analysis target patent is standard technology or technology transfer It may further include.

Preferably, the method may further include determining a prospect of the analysis target patent based on at least one of the predicted number of citations, whether the standard technology patent is valid or whether the technology is transferred.

According to another aspect of the present invention, there is provided an apparatus for predicting patent prospects for analyzing a patent to analyze analytical targets to obtain bibliographic information, and a first variable based on detailed technical elements of the technical classification in the bibliographic information. A citation count prediction unit configured to calculate a value and a second variable value, and predict a citation count of a future time point for the target patent by applying deep learning to the calculated first variable value and the second variable value, the analysis target Calculating the values of the predefined variables based on at least one of the number of technical classifications of the patent, the number of backward citations, the number of claims, and the number of inventors; and among the values of the first variable, the second variable, and the predefined variables Standard technology / technology transfer prediction unit for predicting whether the analysis target patent standard technology or technology transfer by applying at least one or more logistic regression model, the number of citations, Includes standard techniques or not, whether the previous technology based on at least one promise for determining whether the prospective patents for the target patent prediction of the unit.

According to the present invention, it is possible to easily predict the prospect of the analysis target patent by using only prior indicators that can confirm the invention and technical features of the patent.

In addition, by using the technical classification of the analysis target patent, it is possible to predict the number of citations in the future, whether the standard patent and technology transfer.

Effects of the present invention are not limited to the above-mentioned effects, and other effects not mentioned will be clearly understood by those skilled in the art from the following description.

1 is a view for explaining a patent prospect prediction method according to an embodiment of the present invention.
2 is a block diagram illustrating a patent prospective prediction apparatus according to an embodiment of the present invention.
3 is a diagram for explaining an IPC code.
4 is a flowchart illustrating a method of predicting the number of citations of an analysis target patent by a patent prospect prediction apparatus according to an embodiment of the present invention.
5 is a flowchart illustrating a method of predicting whether a patent prospective predicting device according to an embodiment of the present invention predicts whether an analysis target patent is a standard technology patent or a technology transfer.

As the invention allows for various changes and numerous embodiments, particular embodiments will be illustrated in the drawings and described in detail in the written description. However, this is not intended to limit the present invention to specific embodiments, it should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention. In describing the drawings, similar reference numerals are used for similar elements.

Terms such as first, second, A, and B may be used to describe various components, but the components should not be limited by the terms. The terms are used only for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be referred to as the second component, and similarly, the second component may also be referred to as the first component. The term and / or includes a combination of a plurality of related items or any item of a plurality of related items.

When a component is referred to as being "connected" or "connected" to another component, it may be directly connected to or connected to that other component, but it may be understood that other components may be present in between. Should be. On the other hand, when a component is said to be "directly connected" or "directly connected" to another component, it should be understood that there is no other component in between.

The terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting of the present invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this application, the terms "comprise" or "have" are intended to indicate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, and one or more other features. It is to be understood that the present invention does not exclude the possibility of the presence or the addition of numbers, steps, operations, components, components, or a combination thereof.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art. Terms such as those defined in the commonly used dictionaries should be construed as having meanings consistent with the meanings in the context of the related art and shall not be construed in ideal or excessively formal meanings unless expressly defined in this application. Do not.

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in detail.

1 is a view for explaining a patent prospect prediction method according to an embodiment of the present invention.

Referring to FIG. 1, the patent prospective prediction apparatus 200 may be connected to the user terminal 100 through a network. The network network refers to a network that communicates through a medium, and may include at least one of a wired communication network, a computer network, and a wireless communication network.

The patent prospective prediction apparatus 200 may extract data stored in the patent DB 300. The patent DB 300 may include a patent database or a specific database linked with the KIPO database.

The patent promising predicting device 200 predicts the promising of the patent as an index for confirming the invention and technical characteristics when the patent or the analysis target patent is applied or registered.

Promising patents with high technical impacts have a higher number of citations than non-promising patents, and promising patents with high socioeconomic impacts generate patent licenses or proprietary rights transactions. Therefore, the promising patent may be at least one of a patent, a technology standard patent, and a right transfer patent, in which the number of citations is frequently generated. First, the citation information of a patent (ie, citation information) can confirm the influence of the patent on the subsequent invention. A patent with a large number of forward citations plays a key role in implementing an original or specific technical function in the technical field. It can be understood that you are performing. In addition, patent ownership change information, including the sale and licensing of patents, may highlight more direct technological promising in that the invention itself generates substantial profits.

Thus, the patent prospect prediction apparatus 200 may predict the patent prospect using at least one of forward citation counts, technical standards, and technology transfers. Preliminary indicators for determining patent promising are indicators that can confirm invention and technical characteristics when applying or registering an analysis target patent and include invention characteristic indexes and technical characteristic indexes. Invention characteristic indicators include recombinant novelty and component familiarity, and the technical characteristic indicators are based on the technical generality (TG) based on the number of technical classifications and technical based on the number of backward citations. It may include a technical basis (TB), a protection scope (PS) based on the number of claims, the number of technology inventors (IC, inventor counts), and the like.

Meanwhile, the patent prospective prediction apparatus 200 may be implemented as a single computing device or in the form of a collection device in which two or more computing devices are connected to each other. For example, the patent prospective prediction apparatus 200 may be implemented as a single server or in a form in which two or more servers are connected.

Detailed description of the patent prospective prediction apparatus 200 will be described with reference to FIG. 2.

The user terminal 100 is a terminal that can be connected to the patent prospective prediction device 200 and receive information on the patent prospect for a patent input by a user. The user terminal 100 may be a PC, a smartphone, a tablet, a laptop, a PDA, a navigation device, or the like. It may include. Here, the information on the patent promising may include information on the number of citations of the analysis target patent, information on standard technology, technology transfer, and the like.

FIG. 2 is a block diagram illustrating a patent prospective prediction apparatus according to an embodiment of the present invention, and FIG. 3 is a diagram for explaining an IPC code.

Referring to FIG. 2, the patent prospective prediction apparatus 200 according to an embodiment of the present invention includes an analysis unit 210 and a citation count prediction unit 220.

The analysis unit 210 obtains bibliographic information by analyzing the analysis target patent. In this case, when the analysis target patent is input as a patent document, the analyzer 210 may analyze the patent document by text mining or the like to obtain bibliographic information. In addition, when the analysis target patent is input as a patent-related number (eg, application number, etc.), the analysis unit 210 may obtain bibliographic information from a patent DB held in the patent office DB or a private institution. At this time, the analysis unit 210 may obtain bibliographic information using the PatentsView API. Bibliographic information may include an application number, application date, publication number, publication date, applicant name, technical classification, number of forward citations, number of back citations, number of claims, number of inventors, and the like. The technology classification is a patent classification code that classifies patent documents by technology, for example, from internationally accepted classification systems such as International Patent Classification (IPC) and Cooperative Patent Classification (CPC), and US patent classification. (USPC: United States Patent Classification), European Patent Classification (ECLA), Japanese Patent Classification (F-term), and Korean Patent Classification (KPC). In this embodiment, a case of using IPC will be described. As illustrated in FIG. 3, the IPC is composed of a section, a class, a subclass, a main group, and a subgroup. A section is a top technical classification to which a patent belongs, and a class is a sub technical classification of a section, and a technical classification indicating a detailed technical field in the section. A sub class is a sub technical classification of a class. It is a more detailed technical classification according to the viewpoint, and the main group is a technical classification which is a more detailed subclass of the subclass, and has the lowest technical classification having attribute values that describe and describe the viewpoint codes in detail. Sections, classes, subclasses, main groups, and subgroups are sequentially hierarchical. One section is subdivided into a plurality of classes, and each of the plurality of classes is subdivided into a plurality of subclasses. Are subdivided into a plurality of main groups, and each of the plurality of main groups is subdivided into a plurality of subgroups.

The citation count prediction unit 220 calculates the first variable value and the second variable value based on the detailed description elements of the technical classification in the bibliographic information collected by the analysis unit 210, and calculates the calculated first variable value and the second variable value. Deep learning is applied to variable values to predict the number of citations (forward citations) at a future point in time for the target patent. Here, the technical classification may be an IPC in which descriptions are hierarchically classified into sections, classes, subclasses, main groups, and subgroups, and detailed description elements may be subclasses. The first variable and the second variable are variables based on the technical classification of the patent to be analyzed, the first variable may be recombinant novelty (RN), and the second variable is component familiarity (CF). Can be. Recombinant novelty may be a proposed indicator utilizing subclass combinations of IPC, assuming that new combinations of sub-technical elements that did not appear in the prior art invention will affect the promising of patents. The familiarity of the technology elements (CF) may be an indicator of how often and how recently the technology elements of the patent to be analyzed have been used in the past.

The citation count predictor 220 includes a first variable value calculation module 222, a second variable value calculation module 224, and a citation count prediction module 226.

The first variable value calculating module 222 compares the subclass combination pair of the analysis target patent and the subclass combination pairs of past patents filed before the filing date of the analysis target patent, and compares the subclass combination pair existing only in the analysis target patent. The first variable value is calculated based on the total number of subclass combination pairs of the patent.

Specifically, the first variable value calculation module 222 generates a subclass combination pair in the technical classification of the patent to be analyzed. Since one or more technical classifications are selected at the time of patent application, the analysis target patent may include one or more technical classifications. Accordingly, the first variable value calculation module 222 may generate one or more subclass combination pairs by combining the subclasses of the IPC code. For example, if the technical classification of the patent to be analyzed is A01G 5/06, A01F 4/06, and A01C 2/04, it can be seen that three subclasses are G, F, and C. Accordingly, the first variable value calculating module 222 is a subclass combination pair A01G-A01G, A01G-A01F, A01G-A01C, A01F-A01F, A01F-A01C, A01C-A01C. Create a pair. Then, the first variable value calculating module 222 is 3P2 = 3! = 3 * 2 = You can create 6 subclass combination pairs.

Thereafter, the first variable value calculation module 222 confirms the filing date of the patent to be analyzed, collects past patents filed before the filing date from the patent DB held by the KIPO or a private organization, and collects each of the past patents. Create a subclass collation pair for. In this case, the first variable value calculation module 222 may collect past patents having the same class and the same class as the analysis target patent, and generate a subclass combination pair of the collected past patents.

))을 산출할 수 있다.Thereafter, the first variable value calculating module 222 compares the subclass combination pair of the analysis target patent and the subclass combination pair of the past patents to identify the subclass combination pair only in the analysis target patent, A first variable value is calculated based on the number of subclass combination pairs present and the total number of subclass combination pairs of the patent to be analyzed. When there is a newly introduced subclass combination by comparing subclass combinations of past patents and subclass combinations of the analysis target patent, the analysis target patent can be regarded as having recombination novelty. Accordingly, the first variable value calculating module 222 divides the number of newly-substantiated subclass combination pairs that are not observed in the past patents compared to the total number of subclass combination pairs of the patent to be analyzed, and divides the number of newly observed subclass combination pairs from the patent to be analyzed. A first variable value (RN score) having a value of may be calculated. That is, the first variable value calculating module 222 uses a first variable value (recombinant novelty) using Equation 1 below.

)) Can be calculated.

[Equation 1]

는 과거 특허에서 등장한 적 없는 특허 k의 기술분류 서브클래스 조합의 수,

는 특허 k의 모든 기술분류 서브클래스 조합의 수를 의미한다. Where k is the patent to be analyzed,

Is the number of combinations of technical classification subclasses of patent k that have never appeared in past patents,

Denotes the number of combinations of all technical classification subclasses of patent k.

The second variable value calculation module 224 calculates the second variable value based on the difference between the filing year of the previous patent and the filing year of the analysis target patent having the same technical classification as the analysis target patent.

Specifically, the second variable value calculation module 224 identifies the same technology classification patents having the same technology classification as that of the analysis target patent among the past patents applied before the application date of the analysis target patent. Since the analysis target patent includes one or more technology classification codes, the second variable value calculation module 224 extracts past patents having the same technology classification code of at least one of the technology classification codes of the analysis target patent. In this case, the second variable value calculation module 224 may extract the past patents that completely match the technical classification of the patent to be analyzed, but may also extract the same past patents up to the section, class, and subclass.

Thereafter, the second variable value calculation module 224 calculates the difference in the filing date between the analysis target patent and the same technology classification patent. In this case, the second variable value calculation module 224 may calculate the difference between the filing year, the difference between the filing year and the filing month, and the difference between the filing year and the filing month / day as the filing date difference. Thereafter, the second variable value calculation module 224 may calculate the second variable value using the number of subclasses of the analysis target patent, the number of the same technology classification patent, and the filing date. That is, the second variable value calculating module 224 may calculate the second variable value (technical element familiarity CF _k ) of the analysis target patent k using Equation 2 below. It assigns different scores according to the difference between the filing year of the previous patent and the filing year of the analysis target patent, which have the same technology classification as the analysis target patent, on the assumption that the knowledge of a specific technical area is lost by a certain level every year.

[Equation 2]

는 기술분류 서브클래스 j를 가진 모든 과거 특허 i의 수,

은 특허 k의 출원년도와 과거 특허 i의 출원년도 차이,

는 지식 손실 시간 상수를 의미한다. Where m is the number of all technical classification subclasses of patent k,

Is the number of all past patents i with technical subclass j,

Is the difference between the filing year of patent k and the filing year of patent i,

Means the knowledge loss time constant.

The citation count prediction module 226 is applied to an algorithm that uses the first variable value calculated by the first variable value calculation module 222 and the second variable value calculated by the second variable value calculation module 224 as input variables. Thus, the number of citations in the future for the target patent can be predicted. At this time, the citation frequency prediction module 226 may use algorithms such as deep learning, citation-lag distribution (CLD), and censored regression. Preferably, deep learning may be used, and a non-linear regression model for causality check may be used along with a forward neural network (FFNN) of deep learning algorithms. That is, the citation count prediction module 226 may use a forward neural network based regression analysis. Deep learning performs proper hyperparameter tuning through grid search because the model settings specified by the analyst have a significant impact on performance. In addition, crisis normalization is performed to prevent overfitting of FFNN, and such normalization method is a method of assigning a penalty to a loss function by L1 / L2 normalization, and dropout of randomly disconnecting neurons. dropout methods, and learning rate annealing.

As described above, the citation count prediction unit 220 may predict the number of citations of the analysis target patent using only RN and CF which are indicators reflecting the inventive characteristics of the analysis target patent.

In addition, the citation count predictor 220 may predict the number of citations by simultaneously utilizing the post-information as well as the RN and CF of the application point or current point of the analysis target patent.

On the other hand, the patent prospect prediction apparatus 200 may further include a prospect prediction unit 230 for predicting the prospect of the analysis target patent based on the number of citations predicted by the citation count prediction unit 220. The promising predictor 230 may determine that the promising number is promising when the number of times of citation of the analysis target patent is greater than or equal to a predetermined reference number. In this case, the prospective predictor 230 may derive the patent determined to be promising as a patent.

In addition, the patent prospective prediction apparatus 200 may further include a standard technology predictor 240 that predicts whether or not the standard technology of the analysis target patent.

The standard technology prediction unit 240 calculates values of preset variables in addition to the first variable value and the second variable value of the analysis target patent, and applies at least one or more of the calculated values of the variable to the logistic regression model. Predict whether a patent is a standard technology. Here, the predetermined variable is a variable according to the technical characteristic index of the patent to be analyzed, the technical generality (TG) based on the number of technical classifications, the technical basis (TB, technical basis) based on the number of backward citations, the number of claims Protection scope (PS) based on the invention, the number of technical inventors (IC, inventor counts) and the like.

The standard technology prediction unit 240 includes a standard technology related variable value calculation module 242 and a standard technology prediction module 244.

In addition to the first variable value and the second variable value, the standard technology-related variable value calculation module 242 includes, in addition to the first variable value and the second variable value, the generality of the technology (TG), the technology basis (TB), the protection scope (PS) of the technology invention, and the number of technology inventors (IC). Each variable value of) is calculated.

Since the novelty of recombination corresponding to the first variable is closely related to technological progress, a highly novel patent may have a significant impact on technical standards. In addition, patents with high familiarity of technical elements are likely to be path dependent technology inventions, and therefore can have a significant impact on technical standards where compatibility is important. Therefore, when determining whether the standard technology, recombination novelty and familiarity of the technology elements can be used as variables. Since the method of calculating the second variable value corresponding to the familiarity between the first variable value corresponding to the recombinant novelty and the technical element is the same as in Equation 1 and Equation 2, description thereof will be omitted.

)은 아래 기재된 수학식 3을 이용하여 산출할 수 있다. The generality of a technology can be defined as the number of technical classifications, and since the number of technical classifications assigned to a patent can be applied to various technologies, the number of technical classifications can be regarded as the generality of technical knowledge. In this regard, the number of IPCs can be used as an indicator for predicting patent impact. Therefore, technical generality (

) Can be calculated using Equation 3 described below.

[Equation 3]

는 분석 대상 특허 k의 기술분류일 수 있다. here,

May be the technical classification of the target patent k.

The technical basis may be a variable using the number of patent back citations to identify the technical basis of the invention. Although the number of citations of patents is regarded as an indicator of the technical basis of the invention, it can be regarded as a domestic technological basis (DTB) and a foreign technological basis (FTB) in terms of local search. Can be separated.

)은 아래 수학식 4를 이용하여 산출할 수 있고, 해외기술기반 변수값(

)은 아래 수학식 5를 이용하여 산출할 수 있다. First, the variable value of domestic technology

) Can be calculated using Equation 4 below.

) Can be calculated using Equation 5 below.

[Equation 4]

는 특허 k의 국내특허인용 수를 의미한다. here,

Denotes the number of domestic patent citations of patent k.

[Equation 5]

는 특허 k의 해외특허인용 수를 의미한다.here,

Denotes the number of foreign patent citations of patent k.

)는 아래 수학식 6을 이용하여 산출할 수 있다. The protection scope (PS) of the present invention can be calculated using the number of claims of the patent. Since the legal rights of a patent are determined through the claims of the technical invention, the number of claims of the patent is frequently used as an indicator to confirm the protection scope of the technical invention. In general, even if the scope of protection is increased, the possibility of infringement litigation increases, but in practice, litigation only occurs when technical necessity is high due to cost, and the number of patent claims that the original patent holder wins is lost. Statistically significant compared to. Also, in view of the fact that each claim of the patent can be regarded as a separate and independent invention, the number of claims of the patent has the potential to evaluate the technical capability. Therefore, since the protection scope of the technical invention is considered to have the potential to evaluate the influence of the patent, the number of claims of the patent can be used as an indicator representing the protection scope of the technical invention. Protection range variable value

) Can be calculated using Equation 6 below.

[Equation 6]

는 특허 k의 청구항 수를 의미할 수 있다. here,

May refer to the number of claims in patent k.

)는 아래 기재된 수학식 7을 이용하여 산출할 수 있다. The inventor counts (ICs) can be used as one of the prior indicators for predicting patent impact. A large number of inventor pools can bring together a variety of experiences, knowledge, and creativity, so you can expect great benefits in technology invention. From this point of view, the technical invention developed by many inventors argued that high creative diversity not only affects knowledge creation but also makes a difference in the generation of future forward citations. In addition, in the aspect of research and development, the inclusion of a large number of inventors in a single patent may represent that the efforts of the companies invested in the technology invention are large and interesting. Can be. Under this assumption, there is a significant relationship between the value of a patent and the number of patent inventors. Therefore, the number of inventors can be utilized as one of the prior indicators for predicting patent influence, and the number of technical inventors (

) Can be calculated using Equation 7 described below.

[Equation 7]

는 특허 k의 발명자를 의미한다.

Means inventor of patent k.

As described above, the standard technology-related variable value calculation module 242 calculates the variable value according to the invention characteristic index of the patent to be analyzed and the variable value according to the technical characteristic index, respectively.

The standard technology prediction module 244 may predict whether the analysis target patent is a standard technology by applying at least one or more of the values of the variables calculated by the standard technology related variable value calculation module 242 to the logistic regression model. That is, the standard technology prediction module 224 may predict whether the analysis target patent is the standard technology by using Equation 8 below.

[Equation 8]

는 i번째 종속변수,

은 이분변수 Y가 발생할 확률(표준기술 여부),

는 상수,

는 i번째 독립변수에 대한 전체 계수 벡터,

는 i번째 독립변수 벡터(표준기술 관련 변수, 즉, 7개의 변수를 의미함)를 의미한다. 계수(

)는 데이터의 독립변인 값이 들어갔을 때 전체 데이터(즉 모형) 측면에서 가장 오차 값이 적은 계수 값으로 결정된다. 따라서, 회귀분석에서는 비용함수(cost function)을 통해 이를 최소화하는 방향으로 계수값을 산출하는데, 예컨대, 경사하강법(gradient descent) 등을 이용할 수 있다. here,

Is the i dependent variable,

Is the probability of occurrence of the binary variable Y (standard or not),

Is a constant,

Is the full coefficient vector for the i independent variable,

Denotes the i-th independent variable vector (meaning seven standard variables). Coefficient(

) Is determined as the coefficient value with the least error in terms of the overall data (ie model) when the independent variable value of the data is entered. Therefore, in the regression analysis, the coefficient value is calculated in a direction to minimize the cost through a cost function, for example, a gradient descent method may be used.

The standard technology prediction module 244 may predict whether the analysis target patent is standard technology by multiplying the calculated coefficient values by the values of the independent variables. Here, the regression analysis is used to predict whether the standard technology is used. However, the standard technology prediction module 224 may predict whether the standard technology of the target patent is analyzed using various algorithms such as deep learning.

When the analysis target patent is predicted as the standard technology by the standard technology prediction unit 230, the prospective prediction unit 230 may determine the analysis target patent as the promising patent.

In addition, the patent prospective prediction apparatus 200 may further include a technology transfer predictor 250 predicting whether the analysis target patent transfers technology. The technology transfer prediction unit 250 calculates values of preset variables in addition to the first variable value and the second variable value of the analysis target patent, and applies at least one or more of the calculated variables to the logistic regression model. Predict whether or not a patent transfers technology.

The technology transfer prediction unit 250 includes a technology transfer related variable value calculation module 252 and a standard technology prediction module 254. The method of the technology transfer prediction unit 250 predicting technology transfer is a standard technology prediction. Since the operation of the unit 240 is the same, a detailed description thereof will be omitted.

When the analysis target patent is predicted as the technology transfer patent by the technology transfer predicting unit 230, the prospective predicting unit 230 may determine the analysis target patent as the promising patent.

The promising prediction unit 230 may determine the analysis target patent as a promising patent based on at least one of the number of citations of the analysis target patent, whether the standard technology is present or whether the technology is transferred. Promising patents mean patents with many forward citations, standard technology patents, and patents with technology transfers, and thus, the prospective prediction unit 230 uses the number of citations of the analysis target patent, whether it is a standard technology, whether a technology is transferred, etc. It is possible to determine whether the patent to be analyzed is a promising patent.

In addition, the patent prospective prediction apparatus 200 may predict whether or not there is a dispute of an analysis target patent by using an advance index that can confirm the invention and technical features of the patent.

Meanwhile, each of the analyzer 210, the citation count predictor 220, the standard technology predictor 230, and the technology transfer predictor 240 may be implemented by a processor required to execute a program on a computing device. . As such, each of the analysis unit 210, the citation count prediction unit 220, the standard technology prediction unit 230, and the technology transfer prediction unit 240 may be implemented by physically independent components, or within one processor. It may be implemented in a functionally distinct form.

The controller 260 operates operations of various components of the patent prospect prediction apparatus 300 including an analyzer 210, a citation count predictor 220, a standard technology predictor 230, and a technology transfer predictor 240. To control the configuration. The controller 260 may include at least one arithmetic unit, wherein the arithmetic unit is a general purpose central arithmetic unit (CPU), programmable device elements (CPLDs, FPGAs), and on-demand semiconductor arithmetics suited for specific purposes. It may be an apparatus (ASIC) or a microcontroller chip.

4 is a flowchart illustrating a method of predicting the number of citations of an analysis target patent by a patent prospect prediction apparatus according to an embodiment of the present invention.

Referring to FIG. 4, the patent prospect prediction apparatus analyzes an analysis target patent to obtain bibliographic information (S410). That is, the patent promising prediction device analyzes an analysis target patent to obtain bibliographic information including an application number, an application date, a publication number, a publication date, an applicant's name, a technology classification, a number of forward citations, a number of back citations, a number of claims, and a number of inventors. Acquire.

After performing step S410, the patent prospective prediction apparatus calculates a first variable value based on the subclass of the technical classification in the bibliographic information (S420). That is, the patent prospect prediction apparatus compares a subclass combination pair of an analysis target patent and a subclass combination pair of past patents filed before the filing date of the analysis target patent, and compares the entire subclass combination pair and only the analysis target patent. A first variable value is calculated based on the number of subclass combination pairs.

When step S420 is performed, the patent prospect prediction apparatus calculates a second variable value by using past patents having the same technology classification as the analysis target patent (S430). That is, the patent prospective prediction apparatus calculates a second variable value based on the difference between the filing year of the past patent and the filing year of the analysis target patent having the same technology classification as the analysis target patent.

When step S430 is performed, the apparatus for predicting patent prospects is applied to an algorithm using the first variable value and the second variable value as input variables to predict the number of citations of a future time point for the analysis target patent (S440). In this case, the patent prospective prediction apparatus may predict the number of citations of the analysis target patent by using algorithms such as deep learning and regression analysis.

The patent promising prediction device may determine whether a patent to be analyzed is a promising patent based on the number of citations.

5 is a flowchart illustrating a method of predicting whether a patent prospective predicting device according to an embodiment of the present invention predicts whether an analysis target patent is a standard technology patent or a technology transfer.

Referring to FIG. 5, the patent prospect prediction apparatus analyzes an analysis target patent to obtain bibliographic information (S510). That is, the patent promising prediction device analyzes an analysis target patent to obtain bibliographic information including an application number, an application date, a publication number, a publication date, an applicant's name, a technology classification, a number of forward citations, a number of back citations, a number of claims, and a number of inventors. Acquire.

After performing step S510, the patent prospective prediction apparatus calculates a first variable value based on the subclass of the technical classification in the bibliographic information (S520). That is, the patent prospect prediction apparatus compares a subclass combination pair of an analysis target patent and a subclass combination pair of past patents filed before the filing date of the analysis target patent, and compares the entire subclass combination pair and only the analysis target patent. A first variable value is calculated based on the number of subclass combination pairs.

When step S520 is performed, the patent prospect prediction apparatus calculates a second variable value by using past patents having the same technology classification as the analysis target patent (S530). That is, the patent prospective prediction apparatus calculates a second variable value based on the difference between the filing year of the past patent and the filing year of the analysis target patent having the same technology classification as the analysis target patent.

When step S530 is performed, the patent prospect prediction apparatus calculates variable values according to technical characteristic indexes of the analysis target patent, respectively (S540). Here, the variable values according to the technical characteristic index may be values of variables of generality of technology (TG), technical basis (TB), protection scope (PS) of technology invention, and number of technology inventors (IC).

When step S540 is performed, the apparatus for predicting patent promising is applied to an algorithm having variable values according to technical characteristic indexes as input variables in addition to the first variable value and the second variable value to determine whether the standard technology or the technology transfer for the target patent is analyzed. It is predicted (S550). In this case, the patent prospective prediction apparatus may predict whether standard technology or technology transfer is performed by using algorithms such as deep learning and regression analysis.

In addition to the first variable value and the second variable value of the patent to be analyzed, a value of preset variables may be calculated, and at least one or more of the calculated values of the variables may be applied to a logistic regression model to predict whether or not the standard technology of the patent to be analyzed is standard technology. have

The patent promising prediction device may determine whether a patent to be analyzed is a promising patent based on the number of citations.

The above-described patent prospect prediction method is implemented in the form of program instructions that can be executed by various computer means, and can be recorded in a computer-readable recording medium. In this case, the computer-readable recording medium may include program instructions, data files, data structures, and the like, alone or in combination. On the other hand, program instructions recorded on the recording medium may be those specially designed and constructed for the present invention, or may be known to those skilled in computer software.

Computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tape, optical media such as CD-ROMs, DVDs, and magnetic disks such as floppy disks. Included are hardware devices specifically configured to store and execute program instructions, such as magneto-optical media, and memory storage devices such as ROM, RAM, flash memory, and solid state drives (SSD).

So far I looked at the center of the preferred embodiment for the present invention. Those skilled in the art will appreciate that the present invention can be implemented in a modified form without departing from the essential features of the present invention. Therefore, the disclosed embodiments should be considered in descriptive sense only and not for purposes of limitation. The scope of the present invention is shown in the claims rather than the foregoing description, and all differences within the scope will be construed as being included in the present invention.

100: user terminal
200: patent prospect prediction device
210: analysis unit
220: citation count prediction unit
230: standard technology prediction unit
240: technology transfer prediction unit

Claims (7)

In the method of predicting the prospect of the patent of the analysis target patent patent prediction apparatus,
Obtaining bibliographic information of the analysis target patent;
Calculating a first variable value and a second variable value based on detailed description elements of the technical classification in the bibliographic information; And
Predicting the number of citations of a future time point for the analysis target patent by applying deep learning to the calculated first and second variable values
Patent prospective prediction method comprising a. The method of claim 1,
The technical classification is a hierarchical structure in which descriptions are classified into sections, classes, subclasses, main groups, and subgroups.
The detailed description element patent prediction method, characterized in that the subclass. The method of claim 2,
Computing the first variable value,
Generating a subclass pair by combining subclasses in the technical classification of the target patent;
Generating a subclass pair by combining subclasses in the technical classification of past patents filed before the filing date of the patent to be analyzed;
Comparing the subclass pair of the patent to be analyzed with the subclass pair of the past patents to extract a subclass pair only in the patent to be analyzed; And
Calculating a first variable value based on the number of subclass pairs existing only in the analysis target patent and the total number of subclass pairs of the analysis target patent. The method of claim 3,
Computing the second variable value,
Identifying identical technical classification patents having the same technical classification as the technical classification of the analysis target patent among past patents applied before the application date of the analysis target patent;
Calculating a difference of an application date between the analysis target patent and each same technology classification patent; And
Calculating a second variable value using the number of subclasses of the analysis target patent, the number of the same technology classification patent, and the filing date difference. The method of claim 4, wherein
After calculating the first variable value and the second variable value,
Calculating values of pre-defined variables based on at least one of the number of technical classifications, backward citations, claims, and inventors of the target patent; And
Applying the at least one or more of the first variable value, the second variable value, and values of predefined variables to a logistic regression model to predict whether the analysis target patent is standard technology or technology transfer. Forecast method. The method of claim 5,
And determining the prospect of the analysis target patent based on at least one of the predicted number of citations, standard technology patent, and technology transfer. An analysis unit for obtaining bibliographic information by analyzing the analysis target patent;
The first variable value and the second variable value are calculated based on the detailed description elements of the technical classification in the bibliographic information, and deep learning is applied to the calculated first variable value and the second variable value to determine the patent. A citation count prediction unit for predicting a citation count at a future time;
Calculating values of the predefined variables based on at least one of the number of technical classifications of the analysis target patent, the number of back citations, the number of claims, and the number of inventors; and the first variable value, the second variable value, and the predefined variable. A standard technology / technology transfer prediction unit for predicting whether the analysis target patent is standard technology or technology transfer by applying at least one of the values to the logistic regression model; And
Prospective prediction unit for determining whether the patent is a promising patent for the analysis target patent based on at least one of the number of citations, whether the standard technology, technology transfer
Patent prospective prediction device comprising a.

Images