Abstract
The Internet is a tool for free expression of will, primarily reflecting the public’s willingness to pay attention. Therefore, it is of great significance to use network attention to guide the implementation of “Artificial Intelligence (AI) + Education.” First, this study takes the “AI + Education” network attention of 31 provinces and cities in China as the research object and selects the relevant data from the Baidu Index and the National Bureau of statistics from 2012 to 2020. Then, the study uses the methods of elasticity coefficient, geographical concentration index, and panel model to analyze the spatiotemporal characteristics and influencing factors of “AI + Education.” Finally, the future development trends in “AI + Education” is predicted. The results show that the time characteristics of “AI + Education” are apparent, and there are specific interannual differences. The spatial difference between “AI + Education” attention is narrowing, and the spatial balance is gradually improving. The Internet, level of economic development, education funding, and vocational education are the main factors influencing the attention of “AI + Education.” According to the forecast results, the attention to “AI + Education” in eastern and central China will generally rise in the next 2 years, while some parts of western China will slightly decline. Therefore, in the future development, national and regional governments should pay attention to the policy guidance of regional differences, strengthen the promotion of new teaching methods, and attach importance to the intelligent construction of vocational education, to promote the integrated development of AI and Education.
1. Introduction
Artificial intelligence has received significant attention in China in recent years. To cultivate new compound talents, the government gradually put many plans such as intelligent education and programming education in 2017. In February 2019, China’s Education Modernization 2035 was released, which proposed accelerating the educational reform in the information age; building intelligent campuses; coordinating the construction of brilliant integrated teaching, management, and service platforms; and accelerating the reform of talent training mode by using modern technology [1]. Talent is critical to technology, and education is an important way of talent cultivation. The reform and development of the education model benefit from the iterative renewal of technology. Artificial intelligence provides technical support for the development of a new education model. Similarly, education gives human capital to the revival of technology. The iteration and application of technology are gradually integrating AI and education. Today, we should actively face the changes in educational methods caused by technology, innovate models, promote intelligent technology in teaching, and help the sustainable development of education.
The integration of intelligent technology and education aims to give play to the respective advantages of technology and talents and provide a reference for the problems existing in the current education field. It is beneficial to cultivating students’ accomplishments, implementing personalized education, and the sharing of resource allocation. Many scholars have conducted relevant studies based on “AI + Education” from different aspects. For example, some scholars have studied the application effects of AI in teaching various subjects, such as music teaching, English teaching, art teaching, etc., and found that AI is conducive to the management and implementation of classroom teaching [2–4]. Some scholars have applied AI technology in different education modes, such as flipped classrooms, online education, distance education, etc., and found that intelligent network facilities and technologies are conducive to the interaction of the teaching process and the improvement in existing teaching modes [5–7]. Through the experimental procedure, some scholars analyzed the application of intelligent technology in machine teaching, deep learning, intellectual learning, and other aspects and found that academic education can help students solve problems in online learning, provide personalized teaching implementation for learners, and facilitate the practical construction of knowledge [8–10]. Other scholars have discussed the future educational prospects of AI and the building of teacher-student relationships [11, 12]. The existing research provides a theoretical reference for this study, but there are some deficiencies in the existing research. First, most of the current analyses are based on the era of AI, analyze and discuss the impact of AI on education from a macro perspective, and put forward reasonable countermeasures. However, most existing studies focus more on literature review at the macro level, lacking scientific and proper data support. Second, most of the current literature is based on the educational reform caused by the change in times, without considering whether the public’s willingness to pay attention to AI will affect the educational reform in the network age. In the Internet era, the value of big massive data is priceless, and its influence cannot be ignored. Given the shortcomings and dilemmas of the existing research, this study plans to analyze the internal characteristics of “AI + Education” through data collection, data processing, experimental analysis, and other processes from the micro perspective of AI and Education development, considering the public’s willingness to pay attention to it. At the same time, based on the analysis of the influencing factors of “AI + Education,” the study predicts the development trends in “AI + Education” in the future to provide reference and suggestions for the development of education informatization and intelligence in China.
The main contributions of this study are as follows:(1)This study uses a large amount of data to carry out experimental analysis, which provides a theoretical basis and data support for the current literature research on the integrated development of AI and Education.(2)This study helps understand the influence mechanism of “AI + Education” and provides a reference basis for implementing intelligent education and teaching in various regions of China.(3)This study provides an additional reference for other scholars to conduct research and analysis at the micro-level.(4)The data set provided by this study is helpful for subsequent studies by other scholars.
2. Related Work
The development of AI and Education has become a hot topic in literature research. Through literature review, it can be seen that foreign research hotspots of AI and Education mainly focus on education data mining, intelligent assistant, deep learning model, educational intelligence, and other aspects [13–15]. The development of AI and Education in China is in the exploratory stage. Many scholars’ research on AI and Education mainly focuses on the following specific aspects. (1) In the era of intelligence, research on teaching methods innovation, reconstruction of teachers’ roles, human-machine symbiosis, and construction of learning patterns. Chen et al. [16] put forward that in the era of AI, people should be the purpose, a “soul” should be created with knowledge, and a human-machine-integrated teaching mechanism should be constructed. Based on multiple application scenarios of AI and Education, Yang and Ren [17] stressed the need for rational application and promotion of teaching innovation. By analyzing the reshaping and challenges of AI education to teachers, Sun [18] proposed that teachers should constantly innovate teaching ideas and maintain the two-way learning ability of technology and knowledge to better cope with future education challenges. (2) In the era of AI, the innovation of vocational education. Based on the construction of the vocational education system, Yang and Zhu [19] proposed to promote the reform of vocational education college entrance examination, improve social recognition, and explore the transformation path of vocational education public goods. Yan et al. [20] believed that vocational education should innovate education, adjust the strategy, seize the opportunity, meet the challenge, and promote vocational education talents to keep pace with The Times. Sun [21] indicated that a perfect and reasonable technical training system should be built to improve the technological literacy of vocational college teachers to better deal with the risks of AI in vocational education and teaching. (3) Literature analysis on the research hotspots of AI and education. Meng et al. [22] analyzed the hot spots, trends, and current status of AI and Education research through nearly 40 years of literature data. Tu et al. [23] conducted keyword clustering analysis based on the CNKI database and found four major themes and five significant clusters in “Internet + education” in China. (4) Application of AI in higher education. Based on the interaction between AI and higher education, Qian and Su [24] proposed that higher education should realize a benign transformation with the assistance of AI technology. Tan and Jan [25] constructed the teaching model of higher mathematics through the adaptive algorithm optimization of big data and improved the teaching level of higher education. Through experimental research, Ouyang et al. [26] found that AI technology can predict students’ learning status, investigate students’ learning satisfaction, recommend learning resources, and evaluate learning effects in online higher education.
The above research is rich in content and covers a wide range, which provides sufficient theoretical reference for this study. However, few existing studies consider the development of “AI + Education” from the time series and do not consider the impact of online attention on education. Based on this, this study plans to carry out experimental research around the following issues:(1)Analyze the spatiotemporal characteristics of “AI + Education.” This study selects Baidu Index and takes the China Bureau of Statistics data as an auxiliary to measure the spatiotemporal characteristics of “AI + Education” using elastic coefficient, geographic concentration index, and Gini coefficient. The elasticity coefficient can reflect the annual variation difference, and the geographical concentration index and Gini coefficient can reflect the degree and equilibrium of geographical concentration. Many scholars have used these methods to measure the spatiotemporal characteristics [27, 28]. Therefore, this study draws lessons from the practices of existing scholars to comprehensively measure the spatial and temporal characteristics of “AI + Education.”(2)Explore the influencing factors of “AI + Education.” This study uses panel data to build a model of “AI + Education” and its influencing factors, to understand the main aspects affecting the development of “AI + Education,” and then put forward reasonable development suggestions. The panel data model contains the dual information of time and space dimensions, reflecting the dual effects of time and personal effects. Considering that the experimental objects of this study are 31 provinces and cities in China from 2012 to 2020, the experimental data have the two-dimensional characteristics of time and individual, so the panel data model is selected for analysis to make the experimental results more reliable.(3)Predict the attention of “AI + Education.” This study uses the ARIMA model and regression prediction model to predict the attention of “AI + Education” in the next 2 years. ARIMA model is used for time-series prediction, and many scholars have used it. Based on the practice of existing scholars, this study predicts the development trends in “AI + Education” in the future. It provides data support for implementing intelligent education in various regions of China.
Compared with previous studies, the innovations of this study are as follows:
First, this study considers the impact of the public’s willingness on education and teaching methods in the era of big data. Based on the orientation of The Times, this study fully considers the development factors of The Times.
Second, from the micro perspective, this study intensely studied the internal of the object, analyzed the time and regional difference in the development of “AI + Education,” and provided data support for existing review literature.
Third, considering China’s vast territory, this study further clarified the effect of “AI + Education” in each region of China through regional division, providing a reference for implementing AI-assisted education and teaching.
3. Methodology
The object of this study is “AI + Education” network attention. Internet attention refers to the attention of Internet users to something in the Internet era, which is mainly measured by the Baidu Index (https://index.baidu.com/). Baidu Index is the record of massive search request data of Internet users on the Baidu search platform every day, reflecting the social hot spots, the interest, and the demand of Internet users in a certain period. For example, if you want to get “children’s programming” network attention, you need to enter the Baidu Index official website; using the keyword “children’s programming” query, you can get “children’s programming” network attention index of nearly a month.
The experimental process of this study is divided into four steps, as shown in Figure 1. A unique method was used to analyze experimental data at each stage, as detailed below. Step 1: Analyze the time characteristics of “AI + Education.” Based on the practice of Liang et al. [29], this study uses the elasticity coefficient to measure the growth rate of “AI + Education” network attention (see Section 3.1 for the exact method). Step 2: Analyze the spatial characteristics of “AI + Education.” Referring to the practice of Liu et al. [30], this study uses the geographic concentration index and Gini coefficient to measure the concentration and balance degree of “AI + Education” network attention (see Section 3.2 for detailed methods). Step 3: Analyze the influencing factors of “AI + Education.” This study considers multidimensional factors and uses a panel data model to analyze the influencing factors of “AI + Education” from different angles. The detailed modeling process is mainly reflected in the experimental analysis process. Step 4: Predict the network attention of “AI + Education.” Based on the relevant research on time series, this study uses the ARIMA model and regression prediction to predict the attention of “AI + Education” (see Section 3.3 for detailed methods).
3.1. Time Measurement Method
This study uses the elasticity coefficient (ET) to measure the annual change trend in “AI + Education.” Elasticity coefficient refers to the ratio of growth rate between the two variables with a specific relationship in a fixed time, which can be used to compare the growth range of two variables. This study uses ET to analyze the dependence between the growth rate of “AI + Education” network attention and the growth rate of Internet users in various regions. It is calculated as follows:where S represents the Baidu search value of “AI + Education” and △S represents the change value within a period. N represents the number of Chinese Internet users and △N represents its change value. ET > 1 indicates that the increase of public attention to “AI + Education” is more significant than that of Internet users. When ET = 1, the two expansions are equal, and when ET < 1, the opposite is true.
3.2. Spatial Measurement Method
3.2.1. Geographic Concentration Index
This study uses the geographic concentration index (G) to measure the spatial concentration degree of “AI + Education.” The geographic concentration index measures the degree of regional distribution and is usually used to study spatial features. G is used in this study to analyze the degree of concentration and dispersion of “AI + Education” network attention spatial distribution, and its calculation is as follows:where Xi is the network attention value in Province i; T is the total value of network attention in 31 regions of China; n is the total number of areas. Generally speaking, the larger the G value is, the more concentrated the spatial distribution. On the contrary, the more dispersed it is
3.2.2. Gini Coefficient
This study uses the Gini coefficient (Gini) to measure the spatial equilibrium degree of “AI + Education.” Gini coefficient is an important benchmark to judge whether residents’ income and expenditure distribution are reasonable or not, which describes the discrete degree of constant distribution of data. This study uses Gini to measure the degree of spatial imbalance between AI and Education attention among provinces and three regions in China. It is calculated as follows:where n is the number of samples in the region, xi is the Baidu Index value of public attention to “AI + Education” in a specific part, and x is the average value of the sample region. The larger the Gini is, the more significant the difference of attention of “AI + Education” between areas is, the higher the non-equilibrium is; on the contrary, the smaller the difference is, the higher the equilibrium is.
3.3. Prediction Method
This study uses the ARIMA model to predict the attention of “AI + Education,” which includes the prediction of the influencing factors of “AI + Education” and the future development trends in “AI + Education.” ARIMA model has been widely used to analyze and predict various time-series data [31, 32]. It uses the historical data of time series to infer the future time-series data. This study mainly uses the ARIMA model to predict the index values of influencing factors of “AI + Education.” In its prediction model, the expected value is expressed as the linear function of the current and lag period of the lag term and random interference term, which is defined as follows:
There are four steps to establishing the ARIMA model: Step 1: Stationarity test of the sequence. In this study, the ADF unit root test is performed on the original time series to judge the stability of the series. If the original line is unstable, it is necessary to transform the difference or logarithmic difference and check the unit roots of the first-order and the second-order difference to ensure the stability of the whole time series. Second, the stationary sequence is used to construct the model. Step 2: Determination of model order. In this study, the autocorrelation coefficient (AC) and partial autocorrelation coefficient (PAC) are used to judge the form of the model. Then the best model according to AIC, BIC, and other criteria are selected. where L is the maximum likelihood function under this model, n is the number of data, and k is the number of variables of the model. Step 3: Estimation of model parameters. In this study, the model is tested by the significance of model parameters, the validity, and whether the residual sequence is a white noise sequence. If the model passes the test, the model’s parameters are reasonable. If the model fails to pass the test, it is necessary to judge the form of the model again and conduct a relevant inspection to determine the model parameters finally. Step 4: Prediction of sequence. In this study, the series is predicted according to the model parameters, and the error is analyzed by drawing the graph of the model fitting value and the original value. If the error is within the expected range, the model is reasonable and can be used for future prediction. If the error is large, the model’s parameters should be adjusted to ensure that the error is within a reasonable range.
3.4. Parameter Set
This study uses ET, G, and Gini to measure the spatiotemporal characteristics of “AI + Education” and uses Yt to predict the index variables of “AI + Education.” The parameters used in this study are sorted out (Table 1).
4. Experimental Analysis
4.1. Experimental Data Source
Baidu Index is a platform that records Internet users’ search behavior and is the basis for many enterprises to conduct market analysis. It reflects the scale of searches for a word in a related field and is often used to study overall trends in an industry.
Based on the Baidu Index platform, this study collects the Baidu Index values of “AI + Education” in 31 provinces and cities in China (excluding Hong Kong, Macao, and Taiwan) from January 1, 2012 to December 31, 2020 and takes them as sample data for experimental analysis. Second, the study analyzes the spatiotemporal characteristics and influencing factors of “AI + Education” network attention through appropriate methods and forecasts the attention situation in the next few years. In the sample data of this study, the index of “AI + Education” network attention comes from the Baidu Index platform [33]. The number of Internet users in each province comes from the Statistical Report on China’s Internet Development released by the Communications Administration [34]. National economic data and education-related data come from the Statistical Yearbook published by the Bureau of Statistics of China [35].
4.2. Analysis of the Spatiotemporal Characteristics of “AI + Education”
4.2.1. Time Variation Characteristics of “AI + Education”
With the promotion and implementation of education informatization 2.0, intelligent education is becoming more and more critical for the country. The application of artificial intelligence technology in education makes the public pay more attention to reforming new teaching methods. Figure 2 shows the annual proportion trend in the Baidu Index of “AI + Education” combined keywords, reflecting the evolution trend in public attention to “AI + Education” in 31 provinces and cities from 2012 to 2020. Figure 3 shows the annual growth rate and elasticity coefficient obtained according to formula (1), reflecting the yearly change in network attention.
As can be seen from Figures 2 and 3, the annual change in “AI + Education” network attention in China has the following characteristics:(1)From the index’s annual proportion trend and growth rate, the “AI + Education” network attention showed a gradual upward trend from 2012 to 2017, increasing from 36% in 2012 to 86% in 2017. Among them, the index curve from 2016 to 2017 was steep, with the highest growth rate reaching 63.01%, indicating that the development of intelligent education in China is in a rapid development stage. This phenomenon is because, in 2016, the “Internet plus” brilliant development plan proposed to form an AI application market with 100 billion users in 2 years. In 2017, artificial intelligence was written into the report of the 19th National Congress of the Communist Party of China. With the rapid development of technology, cloud computing, big data, and mobile Internet are gradually applied to all aspects of production and life. Education and teaching methods are constantly updated and iterated. The 2017 horizon report (Basic Education Edition) points out that the application of AI in teaching positively impacts students’ metacognitive ability, provides insights for effective teaching, and reduces teachers’ tedious work. With the continuous development and popularization of essential technologies of artificial intelligence, educators must expose students to artificial intelligence, to cope with future changes in the working environment [36]. The issuance of relevant policies has intensified the public’s attention to applying artificial intelligence in education.(2)It can be seen from Figure 2 that the attention to “AI + Education” shows a declining trend from 2017 to 2020, which reflects that the reform of AI-assisted education and teaching in China has entered a stable period and showed a gradually decreasing trend. The mobile search index keeps rising slowly, indicating that the broad application of intelligent terminal devices makes more and more people study in different ways through mobile phones, iPad, etc. At the same time, with the gradual promotion of artificial intelligence in the field of teaching, technologies such as voice evaluation, automatic correction, and photo search are promoting the rapid development of educational informatization.(3)It can be seen from the elasticity coefficient in Figure 3 that the attention of the “AI + Education” network shows a continuous fluctuation trend. The elasticity coefficient in 2014, 2016, and 2017 is greater than 1, indicating that the attention of “AI + Education” increases with the increase of the number of Internet users, which may have a great relationship with the current national policy implementation. Since 2017, the elasticity coefficient has been decreasing, indicating that the growth rate of public attention to “AI + Education” has slowed down, and it may recover after 2020.(4)According to comprehensive Figures 2 and 3, there is a small peak in network attention from 2017 to 2018. On the one hand, the policies issued by the state have a particular impact on its concentration. On the other hand, various new education methods such as network teaching, micro-class, flipped classrooms, and distance teaching makes the public pay more attention to intelligent education and teaching. However, there is a flattening state in 2019-2020, indicating that the product of AI-assisted teaching has been relatively flat in recent years.
4.2.2. Spatial Distribution Characteristics of “AI + Education”
(1) Evolution Trend in “AI + Education” Spatial Distribution. The average proportion of AI and Education network attention in provinces and cities from 2012 to 2020 was obtained (Figures 4 and 5). According to previous studies on regional distribution, the eastern part includes Beijing, Shanghai, Guangdong, Tianjin, Jiangsu, Zhejiang, Fujian, Shandong, Hebei, Liaoning, and Hainan. The central region comprises Hubei, Hunan, Henan, Shanxi, Heilongjiang, Anhui, Jiangxi, and Jilin. The western part includes Xinjiang, Tibet, Qinghai, Guizhou, Yunnan, Inner Mongolia, Guangxi, Ningxia, Shaanxi, Gansu, Sichuan, and Chongqing.
It can be seen from Figures 4 and 5: (1) 2012-2020, the regions with a high proportion of “AI + Education” network attention are Beijing and eastern coastal areas, such as Guangdong, Shanghai, Zhejiang, Jiangsu, and Shandong. These regions have more advanced economic and information levels, so that they may pay more attention to the intellectual development of education and teaching than other places. (2) The network attention of “AI + Education” in the three regions shows an increasing trend. Still, in recent years, the attention of the eastern region has decreased, the growth of attention in the central and western areas has slowed down, and there is a decreasing trend year by year among the three regions. (3) Generally speaking, southern China pays more attention to the “AI + Education” network than northern China.
(2) Concentration Degree of Spatial Distribution of “AI + Education.” The study first obtains the daily average of “AI + Education” network attention in all provinces and cities from 2012 to 2020. Second, the study calculates the geographical concentration index of each year according to Formula (2). It calculates the Gini coefficient of provinces and cities and three regions in China according to Formula (3). The calculation results are shown in Table 2.
From Table 2, it can be seen that(1)From 2013 to 2017, the G value increased year by year, indicating that the concentration degree of Internet attention among provinces increased. Due to the rapid development of new teaching methods, the relevant policies of AI-assisted education reform have been implemented in specific regions. The attention has increased, resulting in an imbalance of spatial attention. From 2018 to 2020, the G value decreases continuously, and the concentration degree of provincial network attention decreases year by year. That is to say, with the full implementation of informatization 2.0, the impact of artificial intelligence on teaching is becoming wider and wider and gradually spreading to all regions. The spatial balance of attention has improved, and the integration of artificial intelligence and education has been a broad concern nationwide.(2)Gini values at the provincial level show that the coefficient kept rising from 2013 to 2017 and reached its peak in 2017, indicating that the network attention difference between AI and education among provinces was continuously expanding before 2017. From 2018 to 2020, the coefficient value decreases. With the nationwide popularization, secondary and higher education institutions in all areas and regions gradually implement AI technology in education and teaching and vigorously promote the introduction of AI into the classroom. Scratch learning children’s programming language and intelligent teaching systems have gradually narrowed the differences between AI and education networks in different regions.(3)From the perspective of regional Gini values, the coefficient values of the eastern and western regions are higher, indicating that there is a significant difference in public attention to AI and education networks between the two regions. Combined with Figure 4, Guangdong, Beijing, and other places in the eastern region are more concerned, while Hebei and Fujian are less concerned. Shaanxi and Sichuan in the west are relatively high, while the indexes of Qinghai and Tibet are low. The central region is more stable.(4)The G and Gini of provinces and three regions decreased in 2019–2020 compared with 2018. It indicates that the degree of disequilibrium among the three areas tends to weaken with the development of the economy and technology. The public’s attention to AI and education in each region has gradually generalized.
4.3. Analysis of Influencing Factors of “AI + Education”
4.3.1. Overview of Influencing Factors and Index Selection
Existing studies show that network attention is the record of users’ search information in relevant fields, and it is one of the applications of big network data. Regional economic, network development, education, population size, and time will affect Internet attention. Given data availability, this study takes 31 provinces and cities as statistical analysis samples to analyze the main factors contributing to the attention difference in the integration development of AI and education from the following aspects:(1)Level of economic development: According to the GDP statistics in recent years, Guangdong, Jiangsu, and Shandong rank the top three in the comprehensive ranking, respectively [35], while Beijing and Shanghai rank the top two in per capita GDP. Combined to Figure 4, Guangdong, Beijing, and Jiangsu account for the highest proportion, indicating that Internet attention may be related to the local economic level. The higher the local financial status, the more ways and conditions for the public to obtain network resources. The more likely they are to search for relevant information differently, the higher their attention will be on the Internet. Guobin et al. [37] show that the network attention of the central and western regions with relatively backward economic development is lower than that of the economically developed areas. In this study, the per capita GDP of each province is used as the index of the economic-level factor.(2)Network development degree: Network attention itself embodies the integration of users and information technology. Generally speaking, the more developed the Internet is, the more ways users can obtain information, the more likely they are to search for relevant information through the Internet, and the higher the correlation index is. Xu et al. [38] believe that network attention is affected by the rapid development of modern communication network technology. In this study, the Internet penetration rate in each region indicates the network development factor.(3)Education funds: As the primary external condition of rejuvenating the country through science and education, education funds affect the reform and implementation of education and teaching to a certain extent. According to the Yearbook statistics in recent 5 years [35], the areas with high education funds are Guangdong, Jiangsu, and Shandong. As the foundation of education development, funds play a positive role in education. This study assumes that education funds positively affect network attention and uses the education funds of various provinces as factor indicators.(4)College teachers: College teachers are the necessary resources for constructing various disciplines, the backbone of higher talent cultivation, and the leaders and leading talents in multiple fields. Higher education is the fundamental way to cultivate high-tech skills. In the era of intelligence, we should give full play to the advantages of higher education and develop more multi-type talents with innovation and practice to drive the development of science and technology. This study assumes that college teachers positively affect network attention and selects full-time college teachers in each province as indicators for analysis.(5)Vocational education: Vocational education emphasizes the application of talents, that is unique talents with various skills and technologies in social production and service. With the development of artificial intelligence technology, there are new changes in the human resources requirements of employment positions. This structural change has a specific impact on the system and content of vocational education. Zhu [39] believes that higher vocational education driven by artificial intelligence is a challenge and a shock. This study assumes that the development of vocational education is negatively correlated with the network attention of “AI + Education.” It selects the number of vocational education students in each province as an indicator to measure.
4.3.2. Model Building
(1) Model Building. First, the selected indicators are defined as shown in Table 3.
Second, to verify the rationality of the above factors, this study selects the regression model of panel data in econometric research. It collects values with multiple cross-sections as sample data. The general form is as follows:where yit represents the Baidu search value of “AI + Education” in t years in province i, and c represents the constant term. x1it∼x5it represents Internet penetration rate, per capita GDP, education funds, number of full-time teachers in colleges and universities, number of students in vocational education in t years in province i. a1∼a5 represents the influence parameters of each factor, and uit is the error term. Due to the significant difference between the Internet penetration rate and the education expenditure, regression analysis was conducted after taking natural logarithms of all dependent variables and independent variables in the calculation process to reduce the error caused by inconsistent data dimensions.
(2) Relative Error. Referring to the practice of existing scholars [40], this study uses relative error to test the model’s accuracy to ensure the rationality and reliability of the constructed model. The relative error is calculated as follows:where represents relative error, lnXactual represents the actual value of the selected indicator, and lnXpredict represents the estimated value of the selected indicator.
4.3.3. Regression Analysis
With the help of SPSSAU and according to Formula (7), this study conducts panel data regression analysis on the influencing factors of AI + Education network attention in China and the three regions of the east, central, and west China. Meanwhile, the optimal model is selected by the F test, BP test, and Hausman test (Table 4).
Eastern region:
Central region:
Western region:
It can be seen in Table 4: (1)Nationally, the degree of network development (x1), level of economic development (x2), education funds (x3), college teachers (x4), and vocational Education (x5) all significantly affect the network attention of AI + Education. Among them, except for vocational Education, the other four factors can promote the attention of “AI + Education,” while vocational Education shows a negative significance of 1%. This result is consistent with the hypothesis in this study. As the development of the Internet benefits from the economic level of the region, the Internet is also relatively developed in cities with fast economic growth. The more developed the network, the higher is its online attention index.(2)From the perspective of regional distribution, the influence degree of each factor is different. ① The level of economic development has a positive effect on the network attention of “AI + Education” in the three regions, especially in the western region. The eastern and central areas have natural geographical advantages, and their economic development has been in a highly deepened stage, which does not have a noticeable impact on attention. The economic development of the western region can promote the significant improvement in infrastructures such as transportation, network, and technology and then encourage the substantial improvement in the network attention index. Therefore, the government should implement policies such as urban assistance, industrial undertaking, and resource complementarity to promote the development of the economy and education in western China. ② Education funds positively promote the “AI + Education” network attention in the eastern, central, and western regions, with a more significant impact on the western area. It shows that the west should give full play to the role of education funds, strengthen teaching research, and promote AI in the classroom. At the same time, the state should appropriately increase the investment proportion of western education funds and improve the investment output ratio of western education. ③ The Internet penetration rate only promotes the attention of “AI + Education” in the central region, but not significantly. This means that with the development of the Internet, the central area can make full use of the advantages of the Internet to promote the use of technologies such as distance education, online teaching, micro-class, and intelligent assessment in teaching and freeing teachers and students. ④ Vocational education has a pronounced negative effect on the eastern, central, and western regions. This phenomenon may be that the impact of AI on vocational education is reflected in the concept and means, and the core and essence change [41]. Therefore, with the continuous updating of intelligent technology, vocational colleges in various regions must follow the trend, comply with the development requirements of vocational education and the application trend in AI, take the initiative to change, and move according to the times, to meet the challenges of the new vocational era.
4.3.4. Error Analysis
To better predict the attention of “AI + Education” in the future, error analysis of the panel regression model is required to ensure the reliability of the regression prediction model. Therefore, this study first selects some cities from the eastern, central, and western regions as the test object. Second, the independent variable index values of these cities in 2018 are selected and substituted into formulas (9)–(11) in sequence according to the region of the city to obtain the predicted value of the dependent variable of each city. Finally, the actual and expected values of the dependent variables of each city are substituted into Formula (8), and the relative errors of each region can be obtained (Table 5). According to the statistical error results, the relative error of the model in the eastern part is less than 2% and about 1% in the central and western regions. The model error of each area is small, and the model is reliable.
4.4. Prediction of the Future Development Trends in “AI + Education”
Considering that “AI + Education” attention is affected by many factors, this study uses the ARIMA model and regression model to predict the future development trends in “AI + Education” attention. This study explores the development trends in “AI + Education” in various regions of China under the influence of multiple factors. It provides reference suggestions for the development of AI-assisted education and teaching.
Since the data of each influencing factor of “AI + Education” come from the China Statistical Yearbook, and the data for 2021 have not been released yet, the study first needs to predict the data values (independent variables) of each influencing factor. Second, after obtaining independent variables, the future attention of “AI + Education” is further predicted according to the regression model of influencing factors of “AI + Education.” The specific steps of the prediction process are as follows:(1)Prediction of influencing factors in each region: Take the education funds in Guangdong as an example for a detailed explanation, and other factors are similar. This study first obtains the historical data of education funds in Guangdong Province from 2012 to 2020. It then predicts the data of education funds in this region from 2021 to 2023 according to the ARIMA model (formulas (4)–(6)), which is taken as one of the independent variables of the regression prediction model. The same applies to other independent variables in the region.(2)Prediction of “AI + Education” attention in various regions: According to the multiple linear regression model of influencing factors of “AI + Education,” and combined with the independent variables in step (1), the attention of “AI + Education” from 2021 to 2023 is predicted. Considering the characteristics of the ARIMA model and regression prediction model, there may be significant errors in long-term prediction. Therefore, this study only forecasts the development trends in “AI + Education” network concerns in each region in the next 2 years (2022-2023) to ensure the stability of short-term forecast data.
4.4.1. Application of Prediction Model
When analyzing the influencing factors of “AI + Education,” this study establishes models of “AI + Education” attention and influencing factors in eastern, central, and western China. Therefore, in the prediction process, this study also selects objects according to the three regional dimensions of China. Considering many cities in China, the space of this study is limited, and it is impossible to predict each city one by one. Therefore, this study plans to select three eastern, central, and western China cities, respectively, for prediction and analysis. The selected objects of each region are East: Guangdong, Zhejiang, and Shandong; Central: Anhui, Hubei, and Hunan; West: Guizhou, Yunnan, and Gansu. Based on the above-selected objects, the future development of “AI + Education” in each region is predicted.
Taking Guangdong in the eastern region as an example, the ARIMA model predicts and verifies education funds. First, the study adopts the ADF test method to test the stationarity of the original sequence and the different sequences of the selected indicators to judge whether they meet the requirements of the ARIMA model. The test results are shown in Table 6. By observing the test results of the sequence, it can be seen that the significant -value of the second-order difference sequence is 0.000, presenting significance at the level. In other words, the original hypothesis is rejected, and the series is a stationary time series.
Second, autocorrelation and partial autocorrelation analyses are carried out for the time series after stabilization to judge the model’s parameter values (Figure 6). Considering that the parameters of the model determined by autocorrelation and partial autocorrelation analyses are subjective, the study uses Python and AIC criteria to find the optimal parameters automatically. It then determines the model parameters (Table 7). As seen from the ARIMA model (0,2,1) test table, Q6 is not significant at the level. The hypothesis that the model residual is a white noise sequence cannot be rejected (p value more than 0.1 is white noise). Meanwhile, the goodness of fit R2 of the model was 0.964, indicating that the model performed well and met the requirements.
Table 8 shows the model test results based on the above series of tests. The test results include coefficient, standard deviation, and T-test results of the model. The test results show that the model result is the ARIMA model (0,2,1) based on second-order difference data. The prediction model is expressed as follows:
Finally, according to the prediction model (formula (12)), the data for education funds in 2021–2023 are predicted. Figure 7 shows the expected data for education funds in Guangdong Province in the next 2 years. By calculating the error between the actual value and the fitted value from 2016 to 2020, the relative error of the prediction model is about 5%. Considering the large cardinality of the sequence, the model’s error is relatively small, and the model’s prediction is reasonable.
Following the same ideas and methods, other variables in Guangdong are predicted. Figure 8 shows the predicted values of Internet penetration rate, vocational education students, and full-time teachers in colleges and universities in Guangdong in recent 2 years. The prediction process of other variables in this region is similar to that above.
4.4.2. Prediction of Independent Variable Indicators
According to the historical data of each variable index in the selected region, this study uses the ARIMA model to predict the independent variable index values of the other eight areas in the east, central, and west. The prediction process is based on the model application process for the influencing factor of education funds in Guangdong Province. Figures 9 and 10 show the prediction results of some independent variables in central Anhui and Western Guizhou, respectively. Due to limited space, other areas will not be displayed.
4.4.3. Prediction of “AI + Education” Attention
According to the predicted values of independent variables in each region from 2021 to 2023, the future trends in “AI + Education” attention can be further expected by combining with the prediction model of “AI + Education” attention and its influencing factors constructed in Section 4.3. The prediction of “AI + Education” attention follows the following ideas:(a)By substituting the independent variables (influencing factors) of the selected eastern regions (Guangdong, Zhejiang, and Shandong) into the eastern region model (formula (9)), the attention of “AI + Education” in the eastern regions can be obtained.(b)By substituting the independent variables (influencing factors) of the selected central regions (Anhui, Hubei, and Hunan) into the central region model (formula (10)), the attention of “AI + Education” in the central regions can be obtained.(c)By substituting the independent variables (influencing factors) of the selected western regions (Guizhou, Yunnan, and Gansu) into the western region model (formula (11)), the attention of “AI + Education” in the western areas can be obtained.
Since the models of the three regions are constructed based on logarithmic data, after the results are obtained, the results need to be logarithmically restored to get the original value of “AI + Education” attention. Figure 11 shows the future development trends in “AI + Education” attention in the eastern, central, and western regions.
As can be seen from Figure 11, “AI + Education” attention generally shows an upward trend in the next 2 years, except for some parts of western China. It shows that the development of AI-assisted education and teaching has received widespread attention in all regions of China. By 2023, the attention index for AI + Education will exceed 1,100 in the eastern area, 900 in the central part, and 300 in the western region. In the future, the development of high and new technology will break through the restrictions of regional barriers and fully integrate the development of the eastern, central, and western regions. For the west of regions, related industries, technology policy should be inclined to this region, to drive education with science and technology and make teaching more effective. At the same time, the western part should keep up with the trend of The Times and make full use of the products of the development of The Times to promote the improvement in education and teaching methods.
5. Conclusions
Based on the online attention of “AI + Education” in 31 provinces and cities in China from 2012 to 2020, this study analyzes the spatiotemporal characteristics and influencing factors of “AI + Education” online attention by using elastic coefficient, geographic concentration index, panel regression analysis, and ARIMA model. It also predicts the “AI + Education” development trends in the next 2 years. The main conclusions of this study are as follows:(1)The attention to AI and education networks in China fluctuates and increases in time, but it has decreased in the 2 years—2019-2020. The development of AI-assisted teaching has entered a stable period. Due to the influence of relevant national documents and policies, attention reached the highest level in 2018. With the promotion of time, more and more people search for relevant information through intelligent terminals and other devices and diversified ways of teaching and learning.(2)The spatial concentration of artificial intelligence and education network attention in China decreases year by year, and the balance of spatial attention keeps improving. South China has attracted more attention than north China. The difference in attention between the east and the west is significant. In contrast, the difference between the central region is slight, and the difference in attention among different regions decreases. The application and development of AI in education and teaching have attracted the public’s attention in all regions.(3)The Internet, the level of economic development, educational funds, teachers in universities, and vocational education are the main factors influencing the AI and Education attention network. The Internet, economic development, and education funds can promote the online attention of “AI + Education” to a certain extent. In contrast, vocational education negatively affects the attention of “AI + Education” in various regions of China. The influencing factor of college teachers is not significant in some areas.(4)According to the prediction results of attention, “AI + Education” attention in all regions of China will generally rise in the next 2 years. Among them, the attention index of eastern, central, and western areas will exceed 1100,900,300, respectively, in 2023. AI contributes to the long-term sustainable development of education in all regions of China.
6. Discussion
This study analyzes the spatiotemporal characteristics and influencing factors of “AI + Education” through experimental research. It predicts the future development trends in “AI + Education” based on its influencing factors. This study provides a theoretical basis and data reference for existing research. In addition, this study also mentions the measures and development direction of AI + Education in China in the future, which are as follows:(1)Adhere to the introduction of first-line Internet technology. The imbalance of industry and resources leads to the inevitable gap between the regions. However, artificial intelligence is built based on the Internet, so there will be no vast time difference in acquiring information and disseminating knowledge. Under the condition of financial conditions, the western region should adhere to the introduction of technology, organize teacher training, establish the urban counterpart assistance, form the mutual aid of colleges and universities, and break through the information barriers in the eastern and western regions to drive the development of the western region.(2)Promote the application and innovation of new educational methods. The focus of AI to boost education lies in education reform. The intelligent teaching environment constructed by the new generation of information technology has the characteristics of intellectual perception, decision-making data, resource integration, real-time evaluation, accurate recommendation, and three-dimensional interaction [42]. Therefore, all regions should actively promote the application of big data technology in education and teaching, such as voice tutoring, online correction, intelligent teaching system, children’s programming, etc. This is helpful for teachers’ accurate lesson preparation and differential teaching. At the same time, it can provide intelligent guidance for students’ learning situation analysis, learning progress planning, homework feedback, etc. In the intelligent post-epidemic era, education will take on a new form: online and offline teaching [43]. Colleges and universities in all regions should provide University assistance, establish an in-service exchange and learning mechanisms for teachers, actively use information-based teaching methods, and promote big data intelligent technology in teaching management, examination, and evaluation.(3)Build an innovative road of intelligent vocational education. First of all, it is necessary to promote the deep integration of industry, learning, and research; flexibly adjust the mechanism; cultivate talents needed by the market; and alleviate the career impact caused by the change in talent demand. Then, vocational education courses should be offered based on complete market research. What kind of courses should be provided should be found on the orders of new occupations and posts needs of society. The teaching materials should also be carefully considered to avoid profound disconnection with the market. Traditional majors should be introduced realistically in the teaching process to promote the enabling effect of “AI + Education.” Finally, vocational education should strengthen “soft power,” cultivate students’ ability of technology transfer, integrate social resources in education and industry, and improve the applicability of skilled talent to develop AI.
This study puts forward some policy development suggestions through experimental research, but still has some limitations. First of all, considering the intelligent era, many users may search for information through different platforms, Applets, APPs, and other search engines. This study only collects the index of the Baidu platform for research, and there are certain limitations in data collection. Therefore, subsequent research can further increase data collection channels and expand the breadth of data collection, such as the Tencent Browsing Index and 360 Index. Second, given the data availability, the influencing factors selected in this study mainly come from the socioeconomic field. Factors in other areas can be further expanded and compensated for in subsequent analyses. Finally, considering the vast territory of China, there are many research objects and a large amount of data in this study. Due to the limited space, this study is difficult to analyze a single region and a single city. All acronyms used in this article are shown in Table 9.
Data Availability
All data generated or analyzed during this study are included within the article. Further details can be obtained from the corresponding author on request.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This work was supported in part by the National Natural Science Foundation of China under Grant no. 61862051; the Science and Technology Foundation of Guizhou Province under Grant no. [2019]1447; the Top-Notch Talent Program of Guizhou Province under Grant no. KY[2018]080; the Philosophy and Social Science Planning Youth Project of Guizhou Province (18GZQN36); the Nature Science Foundation of Educational Department under Grant nos. [2022]094 and [2022]100; the Nature Science Foundation of Qiannan Normal University for Nationalities under Grant nos. 2020qnsyzd03, QNSYRC201714, QNSY2018JS013, and QNSYRC201715; and the Postgraduate Project of Qiannan Normal University for Nationalities under Grant no. 21yjszz013.