Application of fuzzy control algorithm in college students’ personal career planning and entrepreneurship

4.1.1 Problem definition

This research aims to solve the problem of how to effectively use fuzzy control algorithm to help college students deal with uncertainty and fuzziness in personal career planning and entrepreneurial decision-making. Specifically, the question can be divided into two main parts:

(1)

In the personal career planning of college students, how to generate a fuzzy control model that can help them make better career decisions according to their personal characteristics (such as interests, abilities, values, etc.), environmental factors (such as market demand, industry trends, etc.) and their expectations for future careers?

(2)

How to build a fuzzy control model based on college students’ entrepreneurial ideas, resources (such as technology, capital, etc.), environment (such as market competition, policy support, etc.) and other factors to help them make better entrepreneurial decisions in the face of uncertainty and fuzziness?

Data source:

Personal Career Planning: Data will come from career planning questionnaires, industry reports, and online career platforms.

Startup decisions: Data will come from startup surveys, startup competition evaluation forms, and startup related databases and reports.

Treatment method:

Data cleansing: removal or correction of inaccurate, incomplete, or irrelevant data.

Data preprocessing: Standardize and normalize data for the construction of fuzzy control models.

Data consolidation: The consolidation of data from different sources into a unified database for analysis.

In order to solve the above problems, the research first needs to define a series of input variables and output variables to express the input-output relationship of the fuzzy control model. Some examples of possible input and output variables are listed in Table 1.

OPEN IN VIEWER

Table 1

Problem definition

Category	Input variable example	Output variable example
Personal career planning	Interests, abilities, values, market needs, industry trends	Optimal career choice
Entrepreneurial decision	Entrepreneurial ideas, technical resources, financial resources, market competition, policy support	Optimal entrepreneurial strategy

Please note that these input and output variables may need to be further adjusted and refined to suit actual research needs and data availability.

4.1.2 Variable selection

Selecting the appropriate input and output variables in the fuzzy control model is a crucial step. The selection of variables in this study is based on two main aspects: college students’ personal career planning and entrepreneurial decision-making. Here are the specific variables selected and their definitions.

For individual career planning, the study selected input variables as shown in Table 2.

The corresponding output variables are, as shown in Table 3.

For entrepreneurial decision-making, the research selects input variables as shown in Table 4.

The corresponding output variables are as shown in Table 5.

In actual research, these variables may need to be adjusted for the availability and quality of the data. For example, if data on market demand is difficult to obtain, it may be necessary to look for other relevant proxy variables.

Data acquisition is one of the key steps in constructing fuzzy control model. The data used in this study will be collected through multiple channels:

Questionnaire survey: Published through the online questionnaire platform, specifically for college students.

Industry reports: Cite the latest reports on career development and entrepreneurship.

Policy documents: Obtain data from policies issued by the education and labor departments.

Each student sample will include selected input variables such as interest, ability, environment, etc.

Each student sample will contain all the selected input variables. To provide an actionable framework, the study normalized all variables to a 0–10 scale, where 0 represented the lowest and 10 the highest.

Figure 1 below shows the data collection results of the “Personal career planning” part of 10 student samples.

OPEN IN VIEWER

In the data processing section, studies are cleaned, standardized, and formatted for subsequent model construction. To be specific:

(1)

Normalization: In order to achieve comparability under different metrics, all variables will be normalized to a level range of 0–10.

After all these processes are completed, the data set obtained will be used for the training and validation of the fuzzy control algorithm model. This step is to ensure the accuracy and consistency of the data, thereby improving the accuracy of the model’s predictions and decisions. Through such detailed data processing steps, the research can more accurately capture the needs and status quo of college students in personal career planning and entrepreneurship, so as to more effectively apply the fuzzy control algorithm.

4.2.1 Fuzzy set partition

The partition stage of fuzzy set is the key part of model construction, which is mainly used to convert the value of each variable into the corresponding fuzzy set, so as to make fuzzy rules and fuzzy reasoning.

triangular fuzzy numbers (TFN) were chosen to represent each input variable, and triangular fuzzy numbers could easily and effectively express the ambiguity of information. For each variable, the range of its value (0–10) is divided into three fuzzy sets of “low”, “medium” and “high”.

Taking the variable “interest” as an example, you can define the triangular fuzzy number as follows:

The fuzzy set of “low” is represented by TFN( L 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 ) = (0, 0, 5), indicating that the student has a low interest in a certain career, and the corresponding value range is 0–5. The fuzzy set of “in” is represented by TFN( M 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 ) = (0, 5, 10), indicating that the student has a medium interest in a certain career, and the corresponding value range is 0–10. The fuzzy set of “high” is represented by TFN( H 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 ) = (5, 10, 10), indicating that the student has a high degree of interest in a certain career, and the corresponding value range is 5–10.

In the process of fuzzy set partition, some data may belong to two fuzzy sets at the same time. For example, for a student with an interest value of 5, his interest may be classified as either “low” or “high”. In this case, the study will adopt the principle of maximum membership, that is, the data will be classified into the fuzzy set with the highest membership. The fuzzy membership function can be used to calculate the membership degree.

The quantized input variables can be transformed into fuzzy language variables through the fuzzy set partition, which provides the basis for the construction of fuzzy control model.

4.2.2 Fuzzy rule making

In the fuzzy control model, the formulation of fuzzy rules is a key task, and these rules are generally expressed as “if … So … ” The form of. The input part (if part) of the rule is a description of the state of things, while the output part (then part) of the rule is a description of the action that should be taken.

In this study, the goal is to provide individualized career planning and entrepreneurial advice to students based on their individual characteristics, such as interests, abilities, and circumstances. The first step is to develop a set of vague rules that describe how these factors affect career planning and entrepreneurial decisions.

Using the two factors of interest and ability as examples, the following fuzzy rules can be formulated:

Rule 1: If the student’s interest in a certain field is “high” and his ability in that field is also “high,” then the study recommends that he choose that field for career planning or entrepreneurship. Note: IF L 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 is High AND L 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 is High THEN L 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 is Go. Rule 2: If a student’s interest in a certain field is “moderate” and his ability in that field is “high,” then the study recommends that he consider that field for career planning or entrepreneurship. Note: IF L 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 is Medium AND L 𝑎𝑏𝑖𝑙𝑖𝑡𝑦 is High THEN L 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 is Consider. Rule 3: If a student’s interest in a certain field is “low,” it is not recommended that he choose this field for career planning or entrepreneurship, regardless of his ability in this field. Note: IF L 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 is Low THEN L 𝑑𝑒𝑐𝑖𝑠𝑖𝑜𝑛 is Avoid.

In this way, the research can input fuzzy rules according to the students’ personal characteristics and get fuzzy decision results. These results can then be translated into specific career planning and entrepreneurial advice through the de-blurring process.

4.2.3 Fuzzy reasoning and decision algorithm

By adopting Mamdani type fuzzy inference system. The main steps of fuzzy inference and decision algorithm are as follows:

(1)

Fuzzification

The goal of this step is to determine the fuzzy membership of the input variable. These membership degree functions ( μ ) are a function of the interval [0, 1] for each element of the fuzzy set.

For Interest:

μ L Interest ⁢ ( x ) = 5 - min ⁡ ( x , 5 ) 5 , μ M Interest ⁢ ( x ) = 1 - | x - 5 5 | , μ H Interest ⁢ ( x ) = max ⁡ ( x , 5 ) - 5 5 .

(2)

Rule Evaluation

Use fuzzy logic “AND” and “OR” operators to evaluate individual rules.

For example, the output fuzzy set of rule 1 (if interest and ability are high, the decision is made) can be expressed as:

μ Rule ⁢ 1 ⁢ Output ⁢ ( y ) = min ⁡ ( μ H Interest ⁢ ( x ) , μ H Ability ⁢ ( x ) ) .

(3)

Aggregation

The output fuzzy sets of all rules are aggregated into a single fuzzy set. This is usually done by fuzzy “or” operations.

μ Output Aggregate ⁢ ( y ) = max ⁡ ( μ Rule ⁢ 1 ⁢ Output ⁢ ( y ) , μ Rule ⁢ 2 ⁢ Output ⁢ ( y ) , μ Rule ⁢ 3 ⁢ Output ⁢ ( y ) ) .

(4)

Defuzzification

The final step is to convert the fuzzy output set to a clear output value.

This optimized description should provide the reader with a clear, concise, and comprehensive view of the model. The mathematics and logic behind each stage are explained in detail, making the model both theoretical and practical.

4.2.4 Defuzzification

Defuzzification is the last step of fuzzy reasoning, which aims to transform fuzzy output into concrete and precise output. The commonly used defuzzification methods include Centroid Method, Maximum Membership Principle, Mean of Maximum (MoM) and so on. In this study, the centroid method is used, which assumes that the output range of the result is continuous, and the geometric center of this range can be found. The calculation formula of the centroid method is shown in Eq. (1):

Y = frac int mu A ⁢ cdotx(x), dx int mu A ⁢ ( x ) , 𝑑𝑥(1)

Where μ A ⁢ ( x ) is the membership function of the fuzzy output set A at x .

The membership function of a fuzzy output is studied as shown in Fig. 2.

OPEN IN VIEWER

The output of specific career decisions can be calculated using the centroid method. First, you need to find the central value of each career decision, where the central value of “Go” is 1 and the central value of “Consider” is 0.5. Then, using the centroid formula:

Y = ( 1 × 0.7 ) + ( 0.5 × 0.3 ) 0.7 + 0.3 = 0.85

This result indicates that between the two decisions of “Go” and “Consider”, the de-fuzzification result is more inclined to “Go”, that is, according to the inference of the model, the student’s career decision should be more inclined to positive action. Therefore, through the fuzzy control algorithm, the research can provide personalized career planning and entrepreneurship suggestions for college students.

4.3.1 Model verification

After the model is constructed, the validity of the model needs to be tested by verification. The model will be tested by in-sample verification and out-of-sample verification. In-sample verification is mainly to check the model’s predictive performance on the training set, while out-of-sample verification is to verify the model’s generalization ability to unknown samples.

For in-sample validation, Mean Squared Error (MSE) is mainly used for evaluation. This indicator reflects the difference between the predicted value of the model and the actual value. Its calculation formula is as follows Eq. (2):

𝑀𝑆𝐸 = 1 n ⁢ ∑ i = 1 n ( y i - y i ^ ) 2(2)

Where y i is the actual value of the i th observation, y i ^ is the predicted value of the model for the i-th observation, and n is the number of samples.

For out-of-sample verification, the method of 10-fold cross-validation was adopted in this study. The sample was randomly divided into 10 parts, one of which was taken as the test set and the rest as the training set, and repeated 10 times to get the average prediction accuracy of the model.

The study selected 10 students’ data as a sample set, and the data of each student included the score of various indicators and the final career decision. In the process of model validation, each student’s data is taken as a test set and the rest data is taken as a training set. The predictive performance of the model can be evaluated through the results of mean square error and cross-validation. As shown in Fig. 3.

OPEN IN VIEWER

According to the above data, the mean square error of the model can be calculated, and the prediction accuracy of the model can be obtained through the method of 10-fold cross-validation. The prediction result of the model is basically consistent with the actual career decision, which proves that the model has high prediction accuracy. However, the research also noted that for some students, there is a certain gap between the prediction results of the model and the actual decision, which suggests that it may be necessary to further adjust the fuzzy control rules or fuzzy set partition in the process of model construction to improve the prediction accuracy of the model.

Based on the results of model verification, it is found that in some cases, there is a certain gap between the prediction results of the model and the actual decision. This is mainly due to the shortcomings in fuzzy control rule making or fuzzy set partition. In order to improve the predictive performance of the model, it is necessary to adjust and optimize the model.

Firstly, the formulation of fuzzy control rules is re-examined. In the previous model, let all rules have the same weight. In practice, however, this assumption may not hold true. The impact may be different for different rules. Therefore, weight factor w i is introduced in this study to carry out weighting processing for different rules. The weight factor of each rule is determined by minimizing the mean square error of the model. The following Eq. (3) is shown:

w i = min ⁢ ∑ i = 1 n ( y i - w i ⁢ y i ^ ) 2(3)

Where y i is the actual value of the i-th observation, y i ^ is the predicted value of the model for the i-th observation, n is the number of samples, and w i is the weight factor of the i-th rule.

Secondly, the partition of fuzzy sets is adjusted. In the previous model, fuzzy sets are divided according to experience. However, this division may not be optimal. Optimization algorithms, such as genetic algorithms, can be used to find the optimal fuzzy set partitioning. In this process, we need to define the fitness function, such as the mean square error of the model, and then find the fuzzy set partition corresponding to the minimum mean square error by genetic algorithm.

Through the above adjustment and optimization, the prediction performance of the model can be improved. The research conducted model validation again, and the results were shown in Fig. 4.

OPEN IN VIEWER

After adjustment and optimization, the prediction result of the model is closer to the actual career decision, and the mean square error of the model is significantly reduced. This shows that the adjustment and optimization of the study can effectively improve the prediction performance of the model.