Exploring the practice of integrating innovation and entrepreneurship teaching in visual communication design professional courses using hybrid optimization based ANFIS

3.1Automatic organization method for graphic design language in visual communication

In visual communication design, this is accomplished through three links: determining the dimensions and placement of the graphic display, explaining the graphic language, and creating the control flow for the automatic arrangement of the graphic language. Inaccurate results are obtained from the typical algorithm’s computation of the graphic display size and location since it uses irrational approaches. In the graphics language description step, which is usually carried out by a meta-model, the lack of a strict standard leads to imprecise model description outcomes and complicates the effective support of graphics language choreography management. A revolutionary automatic graphic language arrangement method in visual communication design is proposed to accomplish this goal with improved outcomes and high user satisfaction.

To achieve automatic visual language orchestration, specifics on the size and placement of the visual exhibit, the explanation of the graphical language acquisition, the design and implementation of its control flow, and the process of writing and executing a program in the language are all provided.

The fixed value approach determines the display size of the buffer picture in the visual communication design. The procedure is as follows.

In visual communication design, when the required number of layout pages is decreased and surface usage is raised, the layout target function can be expressed as follows:

Where i represents the height of the most recent layout in use and n the number of pages.

The most optimal approach or the most effective exhibition positioning for graphical in visual design is found using Equation (1). This formula is also used as the objective function to determine the best display position for different images.

The rules-based grammatical representation methodology establishes the mathematical relationships between graph elements and provides information about the graph elements’ positions. One way to summarize this descriptive method is as follows:

The definition of the term is as follows, where H stands for the grammar of the graphic language, R for the set of infinite graph parts, and BH for the set of abstraction grammar rules.

When r ₁ is the current stage graph element, D _from is used to express the connection from r ₂ to r ₁, and t is used to explain the connection relationship between graph elements, including D _from and D _to . A natural integer called o is used to express how many connections there are. The following phrase uses DH to define a complete set of syntactic rules:

Where as render describes the relationship between graphical elements, layoutdescribes the various graphical element classes, q describes the graphical element symbol system, S describes the set of graphical element appearance types, and M describes the set of graphical element location relations.

ASM design in visual communication realizes the graphic language’s semantic description, which serves as the foundation for the automatic arrangement of the graphic language. ASM belongs to the dynamic algebra, which has the following definition on a particular alphabet summation:

The M-element function symbols, which comprise both static and dynamic function symbols, are described by g ⁰ in this instance. Program variables are the name for 0-element dynamic function symbols, whereas constants are the name for 0-element static function symbols. The algebra’s current state is indicated by the subscript, the static constants are denoted by M, and the dynamic constants are denoted by lM. In order to make the description of some functions easier, it is assumed that every letter in the alphabet that contains static constants is a Boolean true value.

ANAS converts the image style transfer problem into a problem of minimizing the loss function by using trained extracted content characteristics and style aspects of the image to create a loss function:

The content loss term (represented by M) between the target image J and the resultant image O is the total of the content feature errors between O and J across all convolutional layers. M is the style loss term, which is the total of the style feature errors between J and S in each convolution layer, between the final picture O and the reference image T. Grammatical matrix:

Where j is the feature matrix’s subscript and indicates the style feature matrix of layers m and j, k. The LoG operator is used to extract the edge feature matrices of the three color channels from both the target and resultant images. Next, the disparities between the target image and the resultant image’s edge feature matrices for each of the three color channels are computed. Lastly, the subject structure loss term is represented by the error total of the three color channels. The following is the subject structure loss term:

Where the d-nd color channel is indicated by d. The components of the target image’s d-th color channel are represented by J and O, respectively, in the resultant image.

Assuming that X appropriately captures the fundamental connection between cognition category and concentration intensity and that CI _yz and CD _yz , in turn, characterize the amount of gaze and gaze duration, the model used for testing is, in overall, the following, taking into account the results of user evaluations of many techniques:

The letter W ₁ stands for each of the five aspects of cognitive style: active, sensory, regulation, attraction, and anxiety. All four of the visual communication design styles complex, indirect, mainstream, and following are referred to as y. The letters y = 1, 2, 3 stand for multimedia integration, graphic use configuration, and sequential text layout. Γ₁ and Γ₂ are the control variables.

The connection function’s independent factor is the participant’s adherence to a specific cognitive approach in communication through visual design, as defined by υ _s , υ ∈ (0, 1). The initial compensation quantities CI _yz and CD _yz are obtained by dividing the user’s attention to the visual area I _yz and their cognitive style affiliation D _yz by the corresponding area adjustment factor ζ _yz . These values can then be divided by CI _yz and CD _yz to yield the values of BI _yz and BD _yz , as shown in the expression below:

BD _yz describes the average gaze volume in each zone and BI _yz , consequently, the average glance time in each location. WI _y describes the whole glance time, and WD _y describes the entire gaze region.

In the instance of y = 0, BI _yz , and BD _yz , in turn, provide the lowest values for the average quantity and duration of attention in each region. When y ≠ 0occurs, meaning that the two variables average both the quantity and duration of gazes over multiple communities, then get:

To implement the AI-based visual communication approach the visual features of the AI interface are sampled using a 4×4 chunked combination model; face image acquisition and 2D feature reconstruction are done using the frame scanning technique; the vector-based weighing of the AI interaction the visual communication image retrieval is represented by the target template matching matrix υ₁, υ₂, . . . , υ _ej ; and the edge contouring characteristics of the AI interface are computed using the adaptive hierarchical analysis approach. Through the use of adaptive hierarchical analysis, the set of edge contour features of the AI interface is determined. In accordance with the observation that the AI interface’s depending system connection the distribution is represented visually with a length of M = y _max - y _min, width of X = y _max - y _min, and sparse joint features of I = a _max - a _min, the 3D spatial rebuilding method is utilized to obtain the topology of the reconstructed image of the AI interface. The network topology of this graphic is portrayed through all four a reference templates vectors, y ₁, y ₂, y ₃ and y ₄, that follow:

Where n the referencing template vector instances of one of the frames of the interactive user interface J (y, z) in the entire reconstructed geometric space are calculated, and N is the edge division scale of the visually conveyed image of the AI interface in the neighborhood. The regional distribution basis function of the visually transmitted image of the AI interface is derived as follows when combined with the sparse feature fusion:

g (y, z) , h (y, z) , δ (y, z) among them stands for, accordingly, the pixel intensity of the AI interfaces visually conveyed image. In order to achieve the visual communication information improvement processing of the AI interface, it thus merged with the cross-sectional fusion of information technologies. Following that, enhance the AI technology’s visually communicated image’s edge envelope feature and strengthen its capacity for innovative visual communication.

With the assistance of the hybrid optimisation-based adaptive network-based fuzzy inference system (ANFIS) depicted in Fig. 2, the goal is to provide students with theoretical background topics and entrepreneurial intentions, as well as the values of entrepreneurial culture, and assist them in obtaining the education required to enable them to materialise their ideas. The suggested method aids in providing high-relevance visualisation of entrepreneurial and innovative courses that cover a variety of topics, including the fundamentals of starting a business, investigating financing, trying to manage, business management topics, technological tools (ICT), developing business plans, SWAT analyses, marketing, and many others. The students are also given examples of successful businesses, applications, products, and entrepreneurs that have been examined.

3.2Questionnaire based data collection

Evaluation of quality in institutions of higher education is the goal. The QS based on classroom teaching process of college teachers and students is created based just on QS of ANFIS subscale against the backdrop of ANFIS. The QS has two components, as depicted in Fig. 3.

Fig. 3

QS Components.

Figure 3 shows that perhaps the QS are divided into two sections: teaching satisfaction and fundamental data about teachers and students. There are 42 total questions, with 3 and 39 in each of the first and next portions, correspondingly. The second component also includes nine dimensions. The QS is scored using a Likert’s 5-point system, with 1–5 denoting “totally inconsistent” to “entirely consistent,” accordingly.

3.3Reliability Analysis of Innovation and Entrepreneurship teaching for Teachers and Students Based on ANFIS

As a general rule, the reasoning system, rationale, and perception of the general climate are regularly unclear and questionable. Hence, the customary insightful strategies that offered fresh worth outcomes frequently neglect sufficient for genuine situations. The fuzzy set hypothesis can be productively applied to determine equivocal and unsure information articulation and transmission. The critical thought behind the fluffy set hypothesis is that a component has a level of participation in a fluffy set that is characterized by enrollment capability. The most generally involved range for communicating the level of enrollment capability is the unit stretch [0, 1]. A fluffy set envelopes components that have various levels of enrollment in them, as shown in Fig. 3.

While applying ANFIS, too many information factors cause numerous boundaries for preparation, and this makes the framework convoluted, reducing its materiality. To tackle this issue, the primary predictions of factors such as individual qualities (IC), individual inspirations (IM), individual hindrances (IB), visual design inspirations (VDI), visual design boundaries (VDB), ecological inspirations (EM), and natural boundaries (EB) are utilised as information factors. The result of the ANFIS model is the degree of achieving a better exploration of entrepreneurship and innovation via visual design courses.

The observational information was gathered by a survey, as shown in Fig. 4. Polls are productive information-gathering devices in terms of specialist time, energy, and expenses. Likewise, polls are a productive information assortment system in situations where specialists know precisely what is required and how to quantify the significant factors. The majority of surveys were finished by email and the web, and some of them were finished by mail.

Fig. 4

ANFIS Architecture.

Circle hub on the ANFIS layer addresses the non-versatile hub, while the square addresses versatile hub, and that implies the worth of the boundary can be changed by the learning gave. A full depiction of the construction of ANFIS the layers are as per the following. a. b. c. d. The main layer is the versatile fluffy guidelines; every boundary can be changed by the capability hub, as displayed in condition (24).

A condition (20) show that X is the contribution to hub i, and A _i simulated intelligence is the etymological mark. L _1,i is a participation capability that decides the level of enrollment as per the given X. There are multiple ways of deciding participation capability, for instance with a trapezoidal capability. The subsequent layer is a no versatile layer, the capability of this layer is to duplicate all approaching qualities to deliver a result esteem, as displayed in condition (21).

Each result hub addresses the terminating strength of a standard. The third layer is a no versatile layer, use to standardize the terminating strength. Standardization IS finished as displayed in condition (26).

The fourth layer is a versatile layer, which worth can be tweaked by ensuing boundaries. The capability of this layer displayed in condition (27),

Where w¯i is the boundary set and is the third layer’s result s _i , r _i q _i . The fifth layer is made up of a flexible hub that aggregates data from all sources and shows the results in condition (28).

Preparing set is utilized to fix the boundaries of the versatile layer. Preparing in ANFIS utilizes back engendering calculation. On the back engendering learning, the ensuing boundaries are made super durable. Mistakes that happen between versatile organization yield and the real result engendered back utilizing slope drop to fix the boundaries of the reason. The learning stage is called epoch.

4.1Data collection

In light of ANFIS, the QS based on college instructors’ and students’ classroom instruction is developed using only the ANFIS subscale’s QS [22]. Teaching satisfaction and basic information about teachers and students make up the two components of the QS. There are 42 questions in all, with 3 and 39 questions in the first and subsequent sections, respectively. There are nine dimensions in the second component as well. Likert’s 5-point rating system is used to rate the QS; 1–5 represents “totally inconsistent” to “entirely consistent,” respectively.

4.2Discussion

An ellipsoid model is built based on ANFIS, and a theoretical model is designed using DL. Reliability research reveals that, on general, UGs lack critical and innovative thinking. The development of students’ critical and innovative thinking is mostly dependent on their willingness to learn and their capacity for cooperation. Thus, educators should focus on developing students’ critical thinking skills as well as their motivation to learn and ability to collaborate with others. They should also encourage parents and other relevant social departments to support entrepreneurial education and give students access to a real entrepreneurial environment. In addition to improving students’ capacity for innovation, seeing business opportunities, and engaging in entrepreneurial activity, a more diverse education system that integrates IEE with professional education must be established. IEE’s foundational courses ought to be more comprehensive, focused, and useful. The ellipsoid model can be used to assess the dependability of entrepreneurial classroom instruction, and the theoretical model developed can be applied to practise through research and analysis. The findings demonstrate that students’ cooperative ability is crucial to classroom instruction, with path coefficients of 0.9 and 0.66 for cooperative ability to critical thinking and inventive thinking, respectively.

4.3Performance analysis

Based on the sample questionnaire data various analyses have been made with respect to various metrics. Standard deviation (SD), mean, standard error mean (SEM), t test, MAE, MAPE, RMSE, R2, and correlation are the performance metrics. The high-level parameters are represented by the stated parameters. Parametric and non-parametric statistical approaches are the two most used categories. Parametric statistical analysis is t-test and Non-Parametric fried man test. Each of them has its own presumptions, characteristics, goals, and methods. Samples of questionnaires are taken for performance analysis and based on the questions Standard deviation (SD), Mean, Standard error Mean (SEM) and t test is validated in Table 1.

Table 1 illustrates the statistical validation of the extracted features via several parameters. From the observation, it can be said that the questionnaire collection is highly relevant to analyzing the E&I via VDC. As it can be seen, the spread of the data is forming a normal distribution, and the values are not that skewed, indicating high creativity. It can be stated that the data obtains an overall mean of between 250 and 400, and the data deviation is the maximum achieved between the ranges of 280 and 3348. The data maximum falls within the 2nd-order standard deviation. The obtained In that they allow the investigator to create confidence intervals around the sample statistic collected, standard error statistics are a form of inferential statistics that operate rather like descriptive statistics. The interval in which the population parameter will fall is estimated using the confidence interval that has been thusly created. This parameter estimation guide helps prepare appropriate data samples to analyse the work over VDC. Finally, the t test states the uniqueness of the data by saying that one questionnaire mean group is different from the other. All the questions obtain a t-value of 85, which states that the groups are highly independent of each other and are unique to the work.

The Table 2 shows Inferential statistics, in particular the non-parametric Friedman test, were used to statistically analyse the data gathered at each of the intervention’s four phases. The training had a statistically significant (p <  0.01) impact on the participant instructors’ English language pedagogical competence, according to the results.

Figure 6 graphically demonstrates the data sample distribution. From figure it can be said that the data distributed is two tailed skewed normal distribution that illustrates a highly independent data understanding.

Fig. 5

Questionnaire data samples.

Fig. 6

Graphical demonstration of SD, Mean and SEM analysis.

Table 3 gives the ANFIS-based regression analysis of the data. The primary metrics tested during hypothesis testing are the hypothetical model’s MAE, MAPE, RMSE, R2, and correlation. 302 QSs were gathered and analysed using AMOS 21.0. Figure 6 displays the model fitting degree’s assessment criteria and fitting outcomes. MAE elaborates the mean square sum of progressive residuals and shows how to assign less importance to outliers that don’t make sense for the job; MAPE provides the good fit index; RMSE gives more priority to the highest error; R2 and correlation perform the comparative fitting index.

The value of MAE lies in the range of 16 to 16.88, which indicates that the ANFIS is robust against the outliers. Thereafter, the work obtains an MAPE, RMSE, and R2 overall ranging between 2.7 and 4, 12 to 14.77, and 0.1 to 0.5, demonstrating a high level of fitness. Values greater than 0.8 show a highly reliant model that analyses I&E improvements through VDC. The information above demonstrates that the theoretical model based on the ANFIS process fits the facts from the actual research and the degree of uncertainty well, achieving a better exploration of entrepreneurship and innovation via visual design courses.

The accuracy performance comparison of several machine learning models with the suggested model is displayed in Table 4.

Figure 7 shows that every path in the new structural model’s significance test and rationality test meets the required number, and the test result is legitimate.

Fig. 7

Graphical analysis based on Regression parameters.

Figure 8 demonstrates that while learning routes have not enhanced students’ critical and inventive thinking, the process of I&E instruction in higher education institutions will greatly alter learners’ problem-solving capacity and evaluation and reflection ability. For learners to learn superficially, learning motivation is essential to their learning process. According to the aforementioned computation, learners’ cooperative ability fosters both critical and inventive thinking, as seen by the route coefficients of cooperative ability for those two types of thinking.

Fig. 8

Membership Functions for Each Question.

The suggested algorithm has a mean user satisfaction value of 7.65, which is much higher than the scores of the computer vision algorithm and the linear discriminant algorithm, as indicated in Table 5. This result further validates the dependability of the suggested algorithm by showing that people are satisfied with its impact on the visual communication design style.

In conclusion, UGs typically lack critical and creative thinking when learning I and E. The keys to fostering the growth of students’ critical and inventive thinking are, concurrently, their learning drive and cooperative skills. Overall, the designed questionnaires have good reliability, validity, and fitness.