Green Innovation Efficiency Measurement Based on Sensor Data: Evidence from China

,


Te Relationship between Green Innovation and Economic
Growth.Since human beings entered the 21st century, the depletion of energy and increasingly severe green pollution have made green development new kinetic energy for the highquality growth of all countries [1].Green innovation is a concrete manifestation of the deep integration of green development concepts and technological innovation and is an efective focus for breaking the traditional extensive economic development model [2].In order to seize the opportunities, the developed countries have deployed the national strategy of green innovation.Both the United States' sustainable performance strategic plan and the EU's 2020 strategy have brought innovation and innovation into the national strategic position.
Since the reform and opening up, China's economic growth has achieved world-renowned achievements.However, the economic development model of "GDP" has led to problems such as green pollution and energy shortage in various regions of China [3].Te rapid growth of China's economy comes at the expense of green sacrifce and energy waste, which is unsustainable development [4].China is in a stage of rapid industrialization and urbanization, and it is a period when the contradictions between economic development, resource utilization, and green protection are most acute [5].On the other hand, the Chinese government has invested more in green pollution and energy shortages year by year.Te 19th National Congress and the 13th Five-Year Plan emphasize the promotion of green development and the emphasis on ecological civilization.Green innovation has become a vital method for China to break through the constraints of resources and the environment.Besides, it could also guide overall sustainable development.Its role in China's development is more important than ever.
Trough the above analysis, Chinese scholars and the Chinese government have reached an agreement that China's economic growth cannot be at the expense of the environment, and a balance should be sought between economic growth and green pollution.In October 2017, General Secretary Xi Jinping clearly stated in the report to the 19th National Congress of the Communist Party that China's economy has shifted from a stage of rapid growth to a stage of high-quality development.Green development is a part of high-quality economic development.Te two have a dialectical and unifed relationship, and green innovation is an important source of green development.To achieve the strategic goal of high-quality economic development, we cannot simply pursue GDP growth.We need to fully consider resource endowments and green-carrying capacity.
We believe that green innovation is an important means to balance the ecological environment and high-quality economic growth, and it is also the only way for highquality economic development: ① Green innovation is essentially the innovation of green technology, which is to follow the principles of ecology and the laws of ecological economy, save resources and energy, avoid, eliminate, or reduce pollution and damage of the ecological environment, and minimize the ecological negative efects of "no pollution" or "less pollution."It is a general term for technologies, processes, and products.Green technological innovation is a new modern technological system coordinated with the ecological environment system.② Green innovation is an important part of green development.Green development is a way of economic growth and social development that aims at efciency, harmony, and sustainability.Green development and sustainable development are ideologically inherited.Tey are not only the inheritance of sustainable development but also the theoretical innovation of sustainable development in China.③ Green innovation is an important way to protect the ecological environment.On the one hand, if the ecological environment is good and the living environment is good, the quality of human health is guaranteed.If ecological green protection is good, natural resource regeneration ability is strong, economic development is sustainable, the development space is broader, and stamina is more sufcient.On the other hand, economic development can provide a solid material guarantee for ecological compensation and ecological restoration.
1.2.Green Innovation Efciency.Green innovation efciency (GIE) is a basic indicator for measuring the innovation efciency of regional green innovation activities, which is also a comprehensive innovation capability that takes energy scarcity and green costs into full consideration [6].Green innovation mainly aims at energy conservation and green improvement from the two paths of product innovation and process innovation [7].China's vast territory, regional resource endowments, and economic development levels are varied, leading to obvious heterogeneity of regional green innovation, which not only afects the balanced development of the interregional economy but also the coordinated development of the interregional ecological environment.Terefore, the green transformation of China's economic growth mode is imminent [8].Among the process, government R&D investment and green regulations are the two main players.On the one hand, resources and environment are public goods, so many problems in the feld of green pollution and ecological destruction cannot be solved completely through market mechanisms.Green control policy work must be supplemented in addition to market mechanisms.On the other hand, since China entered the new normal, the factor endowment structure dependent on economic growth has changed.In the past, extensive economic growth driven by factors such as demographic dividend, land dividend, resource dividend, and investment dividend was unsustainable.Te important content of this paper includes how to systematically analyze and comprehensively evaluate regional green innovation performance and how to accurately grasp the evolutionary law of green innovation.Besides, the paper also discusses the importance of green innovation performance theories for China's exploration of green development models to achieve sustainable green, economic development, and practical signifcance.Tis paper sorts out recent related research and summarizes the main research work and main views of experts and scholars on green innovation efciency, as shown in Table 1.

Te Implication and Innovation of Tis Paper
① It explores the relationship between green innovation and high-quality economic growth, taking China's economic growth as a case.② It integrates the entropy weight TOPSIS model and the spatial measurement model to explain the regional diferences between green innovation and economic growth.③ Taking a developing country such as China as an example, the conclusions obtained can provide reference for developing countries in the world.④ It constructed an green innovation evaluation index system and took China as an example.Tis evaluation index system can provide a reference for measuring the efciency of green innovation in other countries.
⑤ Te policy recommendations and enlightenment in this paper have practical guiding signifcance for the government to formulate policies.

Literature Review
So far, the academic community has achieved more research results in green innovation research, mainly from three levels: concept analysis, infuencing factors, and evaluation models.Tere is no consensus on the academic concept of green innovation.Case studies and cross-case comparative analysis show that green innovation is equivalent to green technology [9].Discussing the biofuel innovation system in the United States and Brazil suggests that eco-innovation is a development model that can solve energy shortages and green pollution problems [10].We fnd sustainable innovation based on sustainable energy technologies for consumable resources (natural gas, oil, and coal).Some people believe that green innovation is the same as sustainable innovation [11].Te main reason why scholars have a diferent understanding of green innovation is that they varied in their research perspectives, but all of their understandings refect the unifed relationship between resources, environment, and innovation.
In terms of factors afecting green innovation, the famous "Porter hypothesis" believes that green regulations will stimulate green innovation and reduce or ofset the cost of green regulations [12].Regulation is a positive alternative to green innovation by replacing markets, especially in developing countries, where regulation is an important component of competition.Technology, market demand, and green policy are the key infuencing factors of green innovation, and corporate green innovation activities mainly come from the interaction of three important factors [13].From a technical point of view, some studies have found that the introduction of foreign green technologies and the improvement of enterprises' green technology capabilities have actively promoted enterprises' independent innovation, which is also an important factor afecting green innovation.Using the stochastic frontier analysis method to study the pros and cons of green innovation, it is found that improving green innovation is conducive to improving the efciency of natural resource utilization [14].Competitive advantage is directly proportional to corporate green innovation, and government green regulation has a certain impact on green innovation.From the perspective of industrial organization [15], Peattie [16] compared and analyzed the factors afecting green innovation in various markets and believes that market demand has a signifcant impact on corporate green product development.It is believed that foreign direct investment can play an active role in green innovation by reducing the cost of green innovation in the host country [17].On the other hand, government R&D investment can promote the efciency of green innovation and has a leverage efect [18].Broekel [19] believes that government R&D investment is not conducive to the improvement of green innovation efciency and has a "crowding efect." In addition, the academic research on sensors mainly includes fve aspects: visual sensors, industrial robots, automobile manufacturing, medical and health monitoring, and food processing and packaging.First, in terms of visual sensors, in the production line of electronic manufacturing, both robot assembly and electronic component detection are inseparable from the application of visual sensor equipment.As one of the focuses of machine vision, image sensors are widely used in consumer electronics, medical electronics, avionics, and other felds [20].Second, in the aspect of industrial robots, in order to improve the adaptability of the robot and detect the working environment in time, a large number of sensing devices are applied to the robot.Tese sensors improve the working condition of the robot and enable it to complete complex work more fully.Te application of sensors in the robot industry has attracted the attention of most countries, mainly the United States and Japan.Driven by these advanced countries, the world has set of a boom in the development of "intelligent sensors" [21].Robots provide a good landing scene and higher requirements for the development of sensors.Tird, in terms of bicycle manufacturing, sensors are the information source of the vehicle electronic control system and the basic key components of the vehicle electronic control system.Traditional automobile sensors feed back information in the control process of each system to realize automatic control.Tey are the "neurons" of automobiles and are mainly used in powertrain systems, body control systems, and chassis Te green transformation of China's economic growth mode is imminent systems [22].In these systems, automobile sensors are responsible for the collection and transmission of information.
After information collected is processed by the electronic control unit, instructions sent to the actuator are formed to complete the electronic control.Fourth, in terms of medical treatment and health monitoring, sensors can enable medical devices to present more accurate images, helping doctors correctly diagnose diseases and efectively treat patients; High precision sensors can obtain accurate monitoring data, making medical staf's diagnosis and equipment monitoring patient's body data more accurate [23].Fifth, in terms of food processing and packaging, through the network function of wireless sensors, consumers can better understand the whole process of food production, storage and transportation in food processing plants, make food processing more intuitive and transparent, and efectively eliminate consumers' concerns about food safety [24].At the same time, in case of problems, wireless sensor technology can also facilitate the regulatory authorities to fnd problems in a timely manner and can be well documented to curb food safety problems from the source.Terefore, this paper uses the existing research at home and abroad for reference, applies sensor technology to the process of green innovation, and further studies the spacetime characteristics of innovation efciency on the basis of measuring China's provincial green innovation efciency.Based on the provincial data of China, the spatial structure characteristics and spatial overlap efect of provincial green innovation efciency are analyzed.Compared with previous studies, this paper has the following incremental contributions: First, it establishes a sensor data value incubation mechanism.Second, the multisource data acquisition model of sensor technology is constructed.Tird, it objectively and systematically measures the efciency of green innovation in China's provinces.Te fourth is to use a spatial econometric model to study the spatial spillover efect of green innovation efciency.

Indicator System and Research Model
Te construction of the index system and the selection of research models are crucial to the research results.Figure 1 shows the logical relationship between the index system and various research models.First, an innovation and innovation indicator system and green regulation indicators are built, and the advance detailed analysis of indicators is carried out.Second, the entropy weight TOPSIS model is used to measure the efciency of green innovation in China's provinces and test its autocorrelation.Finally, on the basis of the frst two steps, the spatial measurement model is used to make spatial spillover recommendations and draw conclusions and policy recommendations.

Indicator System Design
3.1.1.Variable Selection and Description.At present, there is no separate indicator system for the green innovation evaluation system.Te common practice in the academic community is to use the input-output method as an idea to include green and energy indicators that refect green in the innovation evaluation system.According to the OECD's description of green innovation evaluation indicators, the green innovation evaluation indicators are mainly evaluated from the two aspects of green product innovation and green process innovation.
Tis paper measures the efciency of provincial green innovation as a whole and therefore selects the input and output indicators of the R&D and economic transformation stages of the innovation process.Te OECD evaluation system is adopted to comprehensively consider the redundancy and availability of China's provincial green innovation indicators and reshape the green innovation evaluation index system.Te specifc indicators have the following meanings: ② Innovative outputs are divided into expected output and non-expected output.Trough their methods [26] and [27], product innovation (GPTI) is used to measure expected output and high-tech new products are used to sell revenue and energy consumption.Te ratio is measured by the ratio.Te emissions of three industrial wastes (waste water, exhaust gas, solid waste, and THW) are used as a measure of poor output.In order to eliminate the infuence of different sizes, a three-waste weighted value calculation was performed.Product innovation is a positive indicator, while the three industrial wastes are a negative indicator.
③ Green regulation is a comprehensive consideration of China's green protection system, especially in the feld of green innovation, which is controlled by the government and the market and public participation.It is not easy to generalize green regulations with certain types of indicators.In [28], the green regulation is divided into the command-based green regulation (F), incentive green regulation (I), and public participation green regulation (P).Te command-based environment uses the SO 2 removal rate (SOO) as a measure.Te incentive-type environment uses the unit GDP's sewage charge (USD) as a measure.Te public-participating environment uses the total number of green letters and visits (EP) in each region as a measure.
Tis paper covers 30 provinces (including autonomous regions and municipalities) in China as the object of investigation and measures the efciency of China's provincial green innovation from 2005 to 2021, but the investigation does not include Tibet, Hong Kong, Macao, and Taiwan.In addition, according to the statistics revealed by the National Bureau of Statistics division in 2011, 30 provinces (autonomous regions and municipalities) are divided into four major economic zones: the eastern, central, western, and northeastern regions.Te indicators of innovation input and innovation output are derived from the China Statistical Yearbook, China Science and Technology Statistical Yearbook, and China Energy Statistics Yearbook.Te SO 2 removal rate and unit GDP sewage charges come from the China Green Statistics Yearbook and the Green Yearbook of China, the total number of letters is from the green letters, and visits comes from the China Green Yearbook.

Sensor Data Value Incubation Mechanism.
Te sensor data itself are not the focus of our research, the key lies in the unexplored potential value behind the sensor data, and the hidden spatial value of the sensor data is boundless [29].Sensor data, as a virtual production factor, also have a "tangible" value and an "intangible" value.Specifcally, the entity of the sensor data refers to a measurable, fxed, and diferent magnitude from other entities and the hidden space of the sensor data.It refers to infnite possibilities that are boundless, changing, and efective.Tis paper introduces the concept of biological incubation and metaphorizes the realization of the industrial big data value as the incubation process of oviparous animals shown in Figure 2.
With the in-depth integration of the new generation of information technology and the real economy, enterprises continue to accumulate production, R&D, economic management, operation and maintenance, and other data in the manufacturing process, and the accumulated quantity is huge and various.Multisensor data originate from various data integrations in various links, including informatization data, Internet of things data, and cross-border data, and have many characteristics such as complexity, multisource, and heterogeneity.Te collision and fusion of multisource heterogeneous data is valuable, and the construction of the multisensor data "resource pool" is the primary goal of realizing data.Multisource heterogeneous data utilize nutrients provided by the resource pool to release and magnify the value through the incubation mechanism, just like the hatching of oviparous animal embryos.Te resource pool is called the cradle of value breeding.Sensor data cover product research and development, production, market, customers, logistics supply chain, after-sales service, fnance, manpower, production equipment and instruments, sensors, products, environmental regulations, social economy, and other data, covering a long process, a variety of types, and a wide range.Te quality-tested multisensor data are dazzling, and the multisensor data need to be classifed, cataloged, and described in detail so that data users can better discover data and enterprise managers can efciently manage data, and it is conducive to fully mining its value.

Combined Sensor Data Collection Technology.
Single-point acquisition technology is the basis of multisensor data acquisition and has been widely used, but many limited counties are also exposed in the process of practical application.For example, the data are irrelevant, and the data collection cycle is long.In addition, data items collected by the single-point collection technology through each collection channel are discrete, and the collection and transmission of various types of data are independent, which ensures that the data will not interfere with one another, but also leads to the lack of interaction between the data.For the key relationship network, it is difcult to carry out accurate correlation analysis in the follow-up, and it is difcult to guarantee the utilization rate of the relationship between data.Te application and implementation of complete and accurate data from the sensor data in all aspects of the industrial chain such as process design, production and processing, and workshop management; repair and maintenance plays an important role in improving product quality, optimizing processes, and enhancing user experience.
Te combined technology is also suitable for data analysis applications of other digital control-based automation equipment (such as robots, laser-cutting machines, and production lines).Te combined acquisition technology has broad application prospects in the open intelligent manufacturing ecosystem.Combination technology can collect various data items of industrial equipment in combination and establish a relationship network between data.Te specifc implementation process of the combination technology is as follows: First, the user uses the acquisition parameters to confgure the interface acquisition parameters, needs to confgure the parameters of the data to be collected, and defnes the data acquisition period and the combination period of the collected data.Second, the data acquisition module continuously and periodically collects the equipment.After that, the local cached data are combined to form one or more sets of combined data; the ffth step is to cache the combined data in the database and perform it in the cloud persistent storage.Te specifc process is shown in Figure 3. Te TOPSIS model of entropy weight is the fusion of the information entropy and TOPSIS model.Specifcally, the entropy weight method is adopted to determine the index weight in the traditional TOPSIS model.Te main purpose is to prevent subjective factors from being afected when the index weights are determined during the analysis, which enhances the objectivity of the evaluation results.Te entropy-weighted TOPSIS model is a method to approximate the ideal solution.Tere is no strict restriction on the sample data.It is mainly applicable to multi-index and multischeme decision analysis system evaluation.By constructing and calculating the Euclidean distance of positive and negative ideal solutions, multiple decisions are made.Te unit is rated for superiority and superiority (better than SFA or DEA).Te main calculation steps are as follows [30]:

Te Sensor Usage in the Work
① Construct a decision matrix: Tere are m indicators participating in evaluation units, and there are n evaluation indicators for each evaluated unit.Te structural decision matrix is as follows: 6 Discrete Dynamics in Nature and Society ( ③ Te information entropy of the indicator is calculated as follows: ④ Te index entropy weight is calculated as follows: ⑤ Te weight matrix is calculated as follows: ⑥ Te optimal solution and the worst solution are calculated as follows: ⑦ Te distance between each unit is calculated, and the positive and negative ideal solutions using Euclidean distance are calculated as follows: ⑧ Te relative progress of each unit is calculated as follows: In formula (8), the greater the relative proximity C i of each unit, the closer the evaluation target i is to the ideal solution.According to relative closeness, the higher the green innovation efciency of the province, the higher the classifcation and ranking of each green innovation efciency.

Exploratory Spatial Data Analysis.
We use exploratory spatial data analysis (ESDA) to analyze the spatial and spatial relevance of green innovation efciency.In the research process, the spatial weight matrix is generated to determine the weight of each spatial unit, and the spatial correlation analysis is performed according to the economic attributes of each unit.Te spatial autocorrelation test determines whether the samples are spatially related, the correlation between them, and the spatial correlation of the description object.ESA has two types of analysis methods: global statistics and local statistics.
In this paper, the global spatial autocorrelation index Moran's I is used to measure the spatial correlation of the evaluation units in the province.Te Moran index is an important indicator of the similarity of spatial neighboring unit elements.Te Moran index is calculated as follows [31]: In formula (9), S 2 � 1/n n i�1 (x i − x) 2 , x � 1/n n i�1 x i S 2 is the sample variance, n represents the number of spatial units, x i represents the attribute value in the i area, W ij represents the spatial unit neighbor weight, and G(d) represents the global G coefcient.Te Moran index is generally between − 1 and 1. Greater than 0 indicates the positive autocorrelation, and the larger the data, the more obvious the spatial distribution agglomeration; smaller than 0 indicates the negative correlation; the smaller the data, the stronger the spatial negative correlation.Te Moran index can be regarded as the correlation coefcient between the observed value and its spatial lag.Te observation value and its spatial lag are drawn as a scatterplot, called Moran scatterplot, and Moran's I is the slope of the retracement regression line.In this formula, the global

Spatial Econometrics Model.
Maintaining the optimal allocation of government R&D investment and green supervision is not only a key factor for improving the efciency of green innovation in China's provinces but also a key factor for achieving high-quality economic growth and the coordinated development of resources and the environment.Te internal unity is Qingshan and Jinshan Yinshan.We make full use of previous research results to incorporate government R&D investment, green supervision, and regional green innovation capabilities into the same research framework and build a spatial measurement model of provincial green innovation efciency based on the traditional Cobb-Douglas production function.
Moderately intensive government R&D investment has a positive impact on green innovation, which can reduce innovation costs and risks and drive local R&D investment with leverage.Appropriate types of green regulations can stimulate green innovation and produce compensation, thereby reducing energy consumption and improving the level of technology.Safeguarding the optimal allocation of government R&D investment and the green regulation is not only a key factor in improving the efciency of green innovation in China's provinces also in achieving the highquality growth of the economy and the coordinated deof resources and environment.Lucid waters and lush mountains are invaluable assets.We fully draw on previous research results, incorporate government R&D input, green regulation, and regional green innovation capabilities into the same research framework, and build a spatial measurement model of provincial green innovation efciency based on the traditional Cobb-Douglas production function.Te specifc model [33] is as follows: In order to eliminate the heteroscedasticity and the infuence of diferent dimensions, the logarithm of each side of ( 11) is processed, and an econometric model is constructed as follows: In the formula, the variable EIE represents the green innovation efciency, the variable GRD represents the government R&D input cost, the variable ENR represents the green regulation, A represents a constant term, α and β 1 or β 2 , respectively, represent the government R&D input and green regulation elasticity coefcient, and ε i,t represents the random error [34].
Spatial econometrics was originally derived from the statistical analysis of spatial data.Te integration of spatial statistics and econometrics not only changes the classical assumptions of traditional econometrics but also promotes spatial econometrics as an independent discipline and is widely used in many felds of natural sciences and social sciences.Spatial econometrics research focuses on the issue of spatial self-frst.Tere are four main reasons for the source of spatial autocorrelation, which is also an important area for the application of spatial metrology analysis.Te frst is externality.For example, in economic felds such as endogenous economic growth theory and new economic geography theory, the analysis is concentrated on the infuence of changes in the characteristics of related units of a given unit.Te second one is the spillover efect.For example, the behavior of the interpreted variable is also afected by the change of the explanatory variable of the adjacent observation unit.Te third reason is to ignore important variables.For example, there is a lack of important spatial structure latent variables, which will have an impact on diferent spatial observation units, and the spatial measurement model needs to be analyzed.Te last one is spatial heterogeneity and mixing efects.
Te spatial econometrics was frst proposed by some scholars.It is widely used in various disciplines and has been recognized by the academic community.In the nearly 40 years of the development of spatial econometrics, a variety of spatial econometric models have emerged.Among them, the spatial error model (SEM) and the spatial lag model (SLM) are the two most used spatial measurement models in the empirical analysis.Te former could be applied to the spatial correlation of error terms, and the latter is applicable that there is a spatial lag that is interpreted as a variable.Tey proposed a spatial Durbin model (SDM) with both SEM and SLM properties, which greatly enriched SEM and SLM.Tis article builds three spatial econometric models of SDM, SLM, and SEM based on the basic econometric model [35].Te specifc model is as follows.

Model 1. Te spatial
Te choice of diferent models is mainly based on judgment rules [36] It is neither the beneft obtained within the economic activity itself nor the beneft obtained by the user of the product of the activity.In other words, this kind of interest is external to the economic activity itself and produces an external economy to society.

Analysis of the Spatial Spillover Efect Mechanism.
Te so-called spillover efect refers to when an organization conducts an activity.It will not only produce the expected efect of the activity but also afect people or society outside the organization.Spillover efects are divided into economic beneft efects and technology spillover efects: ① Arrow frst explained the role of spillover efects in economic growth with externalities.He believes that new investment has a spillover efect.Companies that invest in not only can increase productivity by accumulating production experience, but other companies can also increase productivity by learning from those companies that invest.② Paul Romer proposed a knowledge spillover model.
Knowledge is diferent from ordinary commodities in that knowledge has spillover efects.Tis enables the knowledge produced by any manufacturer to increase the productivity of the whole society."Endogenous technological progress" is the driving force of economic growth.In Romer's model, the total production function describes the stock of capital, labor, and the stock and output of creative technology and the relationship between.③ Palente studied the relationship between technology difusion, learning-by-doing, and economic growth.He designed a learning-by-doing model for a specifc manufacturer to select technology and absorb time.
He believes that before and after absorbing various technologies, the proprietary technical knowledge accumulated by manufacturers through learning-bydoing is ready for further introduction of technologies.even more serious although the northeast has strong R&D personnel and high-level infrastructure, and most of the northeast manufacturing companies are old companies.It is still difcult to form new growth poles by using traditional technologies.Due to economic strength, historical reasons, resource endowments, and other reasons, the western region's green innovation research and development lags.Te weak technological transformation capacity of the western region is a bottleneck, restricting the development of green innovation.

Spatial Autocorrelation Test.
We use the exploratory spatial data analysis method to calculate the global Moran index of China's provincial green innovation efciency through Stata 15.0 software, and the Monte Carlo simulation method is used to test the signifcance of Moran's I. Te results are shown in Table 3. Te analysis of the form is as follows.
Moran's I fuctuated between 0.263 and 0.419, and both were signifcant at the 1% level, rising frst and then rising and rising (N-type), indicating that there is a signifcant positive spatial correlation in regional green innovation.In order to further show a spatial correlation, Moran scatterplots were drawn for 2005, 2010, 2014, 2018, and 2021.Tere are obviously four quadrants in the Moran scatterplots: the frst quadrant is high-value clustering (H-H), the second quadrant is a low value surrounded by a high value (L-H), the third quadrant is low-value clustering (L-L), and the fourth quadrant is surrounded by a low value (H-L).Most of the provinces fall in the frst and third quadrants.Te result rejects the hypothesis that green innovation efciency is spatially randomized, which further confrms the agglomeration of China's provincial green innovation efciency in the geospatial space.

Spatial Spillover Efect.
After reshaping the indicators of the green evaluation system, green innovation efciency (EIE) was selected as the explanatory variable and government R&D investment and green regulations were used as the explanatory variable to test the spatial spillover efect of green innovation efciency in the eastern, central, northeastern, and western regions of China.It also analyzes the spatial and temporal diferentiation characteristics of green innovation efciency in the four major economic regions.
Te Moran index can test whether the sample data have spatial autocorrelation but cannot determine the specifc form of the spatial model.Terefore, it is necessary to select the appropriate model through the spatial measurement model screening rule.According to Elhorst et al and Anselin et al. judgment rules, the LM test and the Hausman test were  Discrete Dynamics in Nature and Society , and model ( 32) is stronger.To further illustrate the interaction mechanism between the explanatory variables and the explained variables, the direct and indirect efect coefcients were calculated using the three selected models, as shown in Table 6.Te analysis of the form could be conducted according to the following perspectives: ① Te perspective of government R&D investment has an enormous impact on the efciency of green innovation and has signifcant spatial spillover efects.Nonetheless, the spatial spillover efects between varied regions are quite diferent, and infuence strength varies.For example, for the eastern region, government R&D investment inhibits green innovation (− 0.410), while central and northeastern and western government R&D inputs will promote green innovation efciency, with correlation coefcients of 0.553 and 0.594, respectively.Te spatial spillover efect of government R&D investment shows the same pattern.For the eastern region to improve the R&D investment of the provincial government, it will inhibit the green innovation effciency of neighboring provinces and cities and promote the R&D investment of the provincial and municipal governments in the central, northeast, and western regions.Te efciency of green innovation in neighboring provinces and cities has surged.On the other hand, the infuence of R&D investment from the eastern region to the central region to the western region on the efciency of green innovation has gradually increased, and the spatial spillover efect has gradually increased.Te main reason that could explain for the fact is that the economic base and innovation resources in the eastern region are relatively sufcient and that enhancing government R&D investment will not signifcantly promote the efciency of green innovation.On the contrary, it may cause a waste of resources and corporate speculation.② Te perspective of green supervision: It can be seen from the results of the spatial measurement model test that diferent types of green supervision have diferent mechanisms for green innovation efciency.Te command-based green regulation and the public-participating green regulation have a signifcant impact on green innovation efciency, both at a level of 1%.However, the impact of diferent regional green regulations on the efciency of green innovation is diametrically opposed.For example, the directive green regulations (0.165) in the eastern region promoted green innovation, while the central, northeastern, and western regions did inhibit green innovation, with correlation coefcients of − 0.228 and − 0.360, respectively.From the perspective of spatial spillover efects, R&D investment and mandatory green regulations in the eastern region have negative spillover efects.Te command-type green regulation and the incentive-type green regulation in the central region have positive spillover efects and incentive green regulations in the northeast and western regions, and public participation in green regulations has a positive spillover efect.From the eastern region to the central region to the northeast and western regions, the spatial spillover efect intensity gradually weakened.③ Te perspective of direct and indirect efects: Te direct efect value and signifcance refect the relationship between each explanatory variable and the regional innovation efciency, and the indirect efect refects whether the variable has a spatial spillover efect.Trough the direct efect, it is found that the R&D input and the command-type green regulation coefcient of the eastern region are negative, indicating that it has a negative direct efect on the efciency of green innovation.Te R&D input coefcients of the central, northeastern, and western regions are positive, indicating that they have a positive direct efect on the efciency of green innovation, and the command-type green regulation  ② Te green innovation efciency in the northeast and western regions is relatively low, but most of them belong to China's key development areas and have a strong resource and green-carrying capacity.To improve the efciency of green innovation in the northeast and the west, the government should support it from the policy level of R&D capital investment and green innovation subsidies, reduce the burden of green innovation, and stimulate the vitality of enterprise innovation.On the other hand, it is necessary to establish a green and low-carbon development concept and achieve pollution reduction and emission promotion.Te development of the central region is in full swing, and the economic foundation is strong.Te government should give guidance from the policy level.Te western region is a region of innovation and backwardness.We should bear in mind that making rapid progress while avoiding the old road of "the frst pollution after treatment" in developed areas.Te eastern coastal areas are economically developed and have high efciency in green innovation.Teir pressure to undertake green innovation costs is relatively small, but their resource and green-carrying capacity have begun to weaken.Te government should strengthen the economic structure and resource consumption, etc., by building an open and innovative ecological environment.Te government should create a good atmosphere for innovation and encourage enterprises to carry out more green innovation activities.

Conclusion and Measures
③ A coordinated and open economic system should be established to break the administrative barriers among provinces.Green innovation efciency has a positive spatial spillover efect, and the existence of administrative barriers among provinces hinders the spatial spillover of green innovation.A coordinated and open economic system not only promotes the spatial balance of population, economy, resources, and environment but also promotes the fow and sharing of innovation factors among provinces and contributes to the "strong alliance" of green innovation among provinces.Te continuous spatial spillover efect produces positive radiation, which drives the provinces with low green innovation effciency to improve together.④ A multigreen regulation policy should be implemented, combining appropriate provincial and regional innovations and formulating appropriate regulatory combinations.Command-based green regulations can stimulate green product innovation more than market-incentive green regulations.For green process innovation, the incentive efect of the market-incentive green regulation is relatively better, because the incentive green regulation has greater fexibility and stability, which enables enterprises to have a certain degree of freedom of choice and provides enterprises with green process innovation and strong external economy incentives.In addition, the public should be encouraged to participate in the formulation of green regulations and become implementers and supervisors of regulations and policies.Green laws and regulations of public participation are an efective incentive for green innovation.

Figure 1 :
Figure 1: Te logical route of this article.

Figure 4 :
Figure 4: Trends of green innovation efciency in the four major economic regions.
Combine the northeast region with the western region due to the number of samples.12Discrete Dynamics in Nature and Society carried out on the indicators of green innovation efciency in diferent regions (

Table 1 :
Summarization of the recent works.
. Te SLM model and the SEM model are screened using the LM-error and LM-lag, robust LM-error test, and robust LM-lag test.If both models are applicable, the corresponding Wald test and the LR test are carried out to determine whether the SDM model can be simplifed into 8 Discrete Dynamics in Nature and Society an SLM model or an SEM model.Finally, the Hausman test is used to determine whether the fxed or random efect is used to determine the most superior spatial econometric model.

Table 3 :
China's provincial green innovation efciency global Moran index.

Table 4 :
LM test and Hausman test.
Note. "-" means no inspection is required.

Table 5 :
Estimation results of spatial measurement models for fxed efects.

Table 4 )
. According to the Hausman test, the spatial spillover efect of green innovation efciency needs to adopt the fxed-efect model.Which fxed efect model is used?It can be seen from the LM test that the SLM, SLM, and SEM models are more advantageous as the spatial measurement model in the eastern, central, northeast, and western regions.Mixed regression efects, spatially fxed efects, time-fxed efects, and double-fxed efects tests were performed on the selected models, see Table5.It is judged by combining the goodness of ft (R2) and the log-likelihood value.It can be seen in Table5that relatively high R2 and log-L are the double fxed-efect model, the mixed fxedefect model, and the double fxed-efect model and that the goodness of ft and natural log-likelihood function values of these three models are 0.868 and − 36.882,0.876and− 110.078, and 0.886, and − 146.98, respectively, indicating that the overall interpretation ability of model (14), model
In diferent regions, the spatial spillover efects and impact mechanisms of government R&D investment, green regulations, and green innovation are quite diferent.From the eastern region to central region to northeast and western regions, the impact of government R&D investment on green innovation has gradually increased and the impact of green regulations on green innovation has gradually weakened, so the spatial spillover efect has gradually increased.5.2.Future Research① Te research in this paper does not involve the analysis of infuencing factors.Future research can analyze the infuence mechanism of green innovation efciency through models from diferent perspectives of infuencing factors.② Te sensor data collection model proposed in this paper is relatively simple, but in reality, it is often more complex.Future research can collect multisource data and integrate it more maturely through the Internet of things technology.③ Tis paper takes China's regional green innovation efciency as the research object and draws the phenomenon of spatial aggregation of green innovation efciency.Future research can choose diferent research objects to demonstrate the conclusions of this research.5.3.Measures.Based on the above conclusions, the following measures can be drawn: ① At present, the manufacturing enterprises of sensors are mainly concentrated in the Yangtze River Delta and gradually form a regional spatial layout dominated by central cities such as Beijing, Shanghai, Nanjing, Shenzhen, Shenyang, and Xi'an.Among them, nearly half of the major sensor enterprises are located in the Yangtze River Delta region, and the others are in turn in the Pearl River Delta, Beijing-Tianjin region, central region, and northeast region.Te government should speed up the standardization, performance normalization, function integration, and structure standardization of sensor products, accelerate the formulation of relevant standards and specifcations, and improve the product quality control capability with standardization.Te government should strengthen technological innovation in sensor material preparation and special equipment, create a "diamond" for sensor R&D and manufacturing, and provide tamp tool support for improving the quality of sensor products.Te government should accelerate the research and development of new sensor materials, new technologies, new processes, and new tools, strengthen systematic management, and improve the product quality control ability with refned management.Te government should strengthen the development of special sensors under complex environmental conditions, enhance stability, reliability, and durability, and improve the sensor guarantee level under harsh conditions and high-intensity operation conditions.