Abstract
It is difficult for college students to find jobs after graduation, which is the most important problem to be solved now. This paper chooses the statistical analysis method to analyze the career planning of college students under different circumstances. Four aspects are analyzed, which are decision-making action, current situation evaluation, career exploration, and self-understanding level. The main conclusions of this paper are as follows. In this study, the gender differences of college students have a certain impact on their career. Generally speaking, the career planning level of boys is higher than that of girls. The job-hunting needs of college graduates are students who enter social work. Family factors affect the level of college students’ career planning. It is found that students’ school experience is the most important factor affecting the level of career planning, and school experience is also reflected in whether students have class committee experience.
1. Introduction
Deep learning is an advanced abstraction of modeling data, which is a branch of machine learning [1]. Deep learning uses a backpropagation algorithm to command machines to change their internal entries, thus discovering complex structures in large data sets [2]. This paper shows how to solve the research problems by training the deep network of learning features and how to share the representation learning patterns and evaluate them on a specific task [3]. In this paper, through a series of experimental studies, it is proved that deep learning can solve the problem of face recognition well. Face recognition task increases the difference between people by extracting verification features from different identities, while face verification task reduces the difference between people by combining verification features extracted from the same identity, which are necessary for face recognition [4]. Through a large number of systematic experiments, this paper shows why traditional methods cannot explain the generalization ability of large neural networks [5]. This paper briefly describes the application research of deep learning in each field. The most advanced technology at present is summarized, and the future research direction is given [6]. Although deep learning technology is widely used in various fields, it cannot capture the uncertainty of the model. Although the Bayesian model can reason the uncertainty of the model, its calculation cost is very high [7]. The specificity of DNA binding to protein can be determined by deep learning technology. Experiments show that this algorithm is more excellent than other algorithms [8]. In this paper, an algorithm is created by deep learning, which can be applied to the field of medical imaging [9]. This paper analyzes the trend and key points of college students’ career development and provides guiding principles for college students’ career planning [10]. Career planning is a compulsory course for contemporary college students, which can avoid blind employment. Therefore, the school helps college students to plan their internship career through system and reform [11]. Nowadays, there are many problems in college students’ career planning. If we can make correct guidance for students’ career planning, it will help students and the development of society. In this paper, PDCA theory is used to improve and upgrade it, so that college students’ career planning is optimized [12]. Nowadays, the employment difficulties of college students are becoming more and more obvious. In order to improve the education of college students’ career planning, a series of measures should be taken [13]. Nowadays, more and more attention has been paid to the employment of college students. In view of this problem, this paper analyzes the existing problems in college students’ career planning and puts forward some suggestions and countermeasures [14]. Learning to plan one’s career after graduation is a compulsory course for college students. It can help students better understand their own situation and then make their own employment plan [15].
2. An Empirical Study on Career Planning
2.1. Gender and Career Planning of College Students
There are many differences caused by gender, which are also reflected in graduates’ career planning. Most scholars believe that there are significant differences between different genders in many aspects, including career awakening, career awareness, career experience, and career development. However, Chinese scholar Chen Lijuan found through research that girls are more mature than boys in physiological development, but in career planning, boys’ career maturity is higher than girls’ in college and beyond. However, there are many influencing factors, and only gender has differences in the two dimensions of goal planning and interpersonal relationship.
2.2. Major and Career Planning of College Students
The difference between occupation and social life is significant, and the difference between occupation experience and occupation attitude is extremely significant. Some scholars have done relevant research and found through empirical research that when students learn different subjects, their factors are not significant in terms of their understanding, status, and need for career planning.
2.3. Grade
There are some differences among college students in different grades, and their understanding of occupation will increase with the increase of grades. Domestic scholars take the college students trained by exam-oriented education under China’s national conditions as the research object. Wang Shengnan found that the differences in professional grades are significant in different grades, while grades 1 and 4 are significantly higher than grades 2 and 3, especially grades 1 and 4.
2.4. Career Counseling Experience and College Students’ Career Planning
Not only experts or scholars but also general education practitioners or most ordinary people believe that career counseling is of positive significance to college students in all aspects. Therefore, through practical experiments, it can be concluded that if a group of students who have received relevant vocational counseling are compared with the same number of students who have not received counseling help, the former will be found to be better. Therefore, schools should increase counseling centers and train professional tutors.
2.5. Career Planning of College Students Experienced by Student Cadres
It is a manifestation of students’ ability and an opportunity to exercise. Therefore, the work unit recognizes student cadres. However, some scholars have found that this is not the case. For example, Zhang’s survey results show that there is no significant difference between the size of all career planning and that of nonstudent cadres, but the report shows that, in the future life, students who have been student cadres will pay more attention to exploring their careers, have stronger self-awareness, and have better planning and better interpersonal coordination ability.
3. Deep Reinforcement Learning
A complete Markov decision process consists of a quintuple : S represents the set of environmental states; A represents the set of actions; P represents the state transition matrix ; R represents the reward function, ; denotes the attenuation factor; .
Reinforcement learning agents need to use strategy to determine the behavior mechanism. Different reinforcement learning can adopt fixed strategies or unfixed strategies. Getting a perfect strategy to describe agent behavior is the ultimate goal of reinforcement learning. Formula (1) can be obtained by using strategy for functions:
Further decomposing it into the current reward and the subsequent status can result in formula (2):
Then, using strategy , formula (3) can be obtained:
Decompose the function into subsequent states and current rewards to obtain formula (4):
They can be transformed into each other, as shown in formula (5):
The median optimal value function of all strategies is found by reinforcement learning as shown in formulas (6) and (7):
In order to find the optimal value function, the recursive relationship between the optimal Q function and the optimal V function is used, as shown in formulas (8) and (9):
Through continuous iteration, the optimal function can be obtained.
3.1. Q-Learning
In Q-Learning, Q means that is the expectation that and actions can get benefits at a certain moment s. The environment gives feedback rewards to agents according to the actions taken by agents, then constructs Q-Table through states and actions to store the learned Q values, and selects the actions that can get the maximum benefits according to the Q values, as shown in Table 1.
By defining the problem as an MDP process, it can be expressed as formula (10):where the state value function of can be defined as in formulas (11) and (12):where is the total discount reward starting at time t, and , when it is closer to 1, means that it pays more attention to the value of subsequent states, and when it is closer to 0, it means that it pays more attention to the current income. The optimal value function Q can be expressed as formula (13):
Expand the desired formula as shown in formulas (14)–(16):
Then, the Q-Table is updated in a time difference manner as shown in equation (17):where is the decay factor, is the learning rate, and the next state s is selected and updated according to the corresponding position in Q-Table. Q-Learning is shown in Algorithm 1.
|
3.2. Strategy Gradient
Q-Learning cannot deal with reinforcement learning. Therefore, reinforcement learning method is produced, which is to find the optimal strategy by learning the gradient information of strategy parameters. The specific strategy can be described as a function with parameter , as shown in formula (18):
The original strategy is transformed into a continuous function to find the optimal strategy. Specifically, the optimization target is set to the expectation in the initial state, as shown by formula (19):
The strategy gradient can be expressed by multiplying the strategy by the reciprocal of a likelihood function, which becomes the score function, where the Softmax-based score function is expressed as formula (20):
The score function of the Gaussian distribution can be expressed as formula (21):
The gradient is then obtained by derivation, as shown by formulas (22) and (23):
Selecting an action to maximize the reward and selecting the optimal action depending on the current state and action, the expectation of the reward multiplies the score function to obtain the derivative of the reward function, resulting in formula (24):
Expectations in the reward function are replaced by samples, and after the end of a segment, the parameters are updated with each step in the segment. The policy gradient is shown in Algorithm 2.
|
4. Experimental Analysis
4.1. Performance Comparison
Deep learning, analytic hierarchy process, and fuzzy recognition are frequently used in daily experiments. In order to make the experiment more accurate and concise, we counted the performance capabilities of deep learning, analytic hierarchy process, and fuzzy recognition in information acquisition, model recognition, and model recognition accuracy for 10 times, and the results are shown in Figures 1–3.
By analyzing Figures 1–3, we come to the conclusion that deep learning is the best in information acquisition ability, followed by analytic hierarchy process, the worst is fuzzy recognition, the best in model recognition ability, the worst is analytic hierarchy process, the best in model recognition accuracy, the second is fuzzy recognition, and the worst is analytic hierarchy process. It is concluded that deep learning is superior to other AHP and fuzzy recognition in all aspects, so we choose deep learning for the next research.
The accuracy of Mean, PMF, AutoRec, NADE, DLTSR, REda, and DMF in five data sets was tested, and the lowest RMSE, MSE, and MAE were compared in Figures 4–8. On the whole, compared with general methods, AutoRec, NADE, DLTSR, REda, and DMF have significant improvement in accuracy compared with other methods, and DMF has achieved certain advantages in the process of comparing with other comparison methods, and DMF has a fast convergence speed in the training process. AutoRec, NADE, DLTSR, and REda all take the user’s scoring history behavior as input information, but they all use an adjacency matrix to encode, which leads to extremely high-dimensional and sparse input information of the model, which makes the number of parameters of the model extremely large. Different from these models, DMF uses the IFE model to encode historical behavior, which reduces the number of parameters of the model and improves the training efficiency of the model. It can be observed that DMF has great advantages in training efficiency compared with AutoRec, NADE, DLTSR, and REda.
4.2. Analysis of the Present Situation of College Students’ Career Planning
4.2.1. Overall Status Analysis
It can be seen from Table 2 that the average score of students’ self-status evaluation level after calculation is 31.72 points, which is higher than the average score of 30 points corresponding to “conformity,” indicating that students have a good understanding of career planning. The average score of students’ self-understanding (including interests, personalities, values, and skills) is 27.18. Although it is already in the second place, it is far below the average score of 33, which is the middle level of achieving “conformity.” It shows that although students have a better sense of career planning, there is some confusion in planning behavior, which comes from their own understanding and does not know what they really want and for. Finally, we can see that the average score of career exploration and decision-making in Table 1 is no more than 30, which is lower than the average score, indicating that students “talk but do not do,” lack action, only realize its importance, but lack decisive career line and clear career goals, so they have no concrete actions.
In addition, it can be seen from Table 2 that the distribution at each level is right-sided.
(1) Current Situation Evaluation and Analysis. Table 3 gives descriptive statistics on the evaluation of college students’ professional status. It can be seen from the table that students have the highest score of environmental awareness, which is 11.47 points, indicating that students still have a good understanding of the employment environment, and it may also be that schools and society instill more in this respect for college students. After excluding environmental cognition, the scores of the other two parts are not high, which shows that students are a little deficient in these two aspects and need to improve their clear understanding of their hobbies, personalities, and interests.
(2) Analysis of Self-Understanding Level. Table 4 shows the descriptive statistics of students’ self-understanding. Here, it is divided into two categories: “self-adaptation” and “expected goal.” The value of self-adaptation and expected goal is generally low, which indicates that students cannot determine which jobs are suitable for themselves in the process of career planning, and there is a gap between the jobs they want to engage in and their personal abilities.
(3) Career Exploration Analysis. It can be seen from Table 5 that the scores of these three parts are not much different, and they are all about 9 to 10 points, which are not high. But generally speaking, the scores of interpersonal relationship are higher than those of the other two, which shows that students think that good interpersonal relationship is very important for employment and their career planning process, such as having good teacher-student relationship and good relationship with classmates and roommates.
(4) Analysis of Decision-Making Action Level. Table 6 makes a statistical analysis of the decision-making action level of college students’ career planning. From the table, we can see that the scores of these two parts are very low, which is lower than those of the three tables analyzed above, which shows that students lack mobility and action awareness in career planning. In particular, the average score of the evaluated career goals is only 7.37, which indicates that students basically will not set the revision and action of goals, let alone make evaluation and feedback, which is consistent with the research results of career goals.
(5) Analyze by Item. Table 7 gives an analysis of some of the most representative items. According to the average value and standard deviation analysis in the table, the students’ scores are not very optimistic. In the table, there are 11 questions with an average score of more than 3 points, accounting for one-third, while the remaining 20 questions have low average scores. This result shows that the overall situation of college students’ career planning should prove the theory mentioned in the second section of this study.
It can be seen that the average score of the third question in the question is the highest, so it can be concluded that college students are still very aware of the importance and necessity of career planning, so they are still happy to participate in relevant training in the school. Another high score item is Question 27: It is believed that college students should “strive to learn professional knowledge well and improve professional quality.” This title is 3.57 points, which is higher than the above topic. It shows that students think that the most important thing in their career planning is to learn professional knowledge to enhance their competitiveness. Only in this way is an effective way to achieve high-quality employment; after analyzing the high-scoring items, let us look at the low-scoring situation. We can see that the average score of the 20th question in the table, “Often consult relevant experts about work knowledge,” is the lowest, as low as less than 2 points, which shows that there is a lack of professional exploration spirit, which is manifested in the lack of professional consultation and help from teachers in the school employment guidance center. This phenomenon deserves the double attention of universities and society.
To sum up, the high-scoring items all lie in the self-understanding and current situation cognition of career planning, and the average scores of these two levels are optimistic, while the scores of other items are not so optimistic. Especially in the decision-making action level, the average scores of basically every item are very low, which shows that college students generally feel confused when facing employment.
4.2.2. Difference Analysis
(1) Personal Characteristic Variable. As mentioned above, gender is usually regarded as a prominent variable in many human nature studies because there are obvious differences between men and women in many aspects. In this study, it can be found from Table 8 that the value of the t-test in total table is 0.005, which is a significant difference. From a numerical point of view, the average score of male students’ career planning level is higher than that of female students that is, there are significant differences between male and female college students in all aspects of career planning. Among these four levels, only the self-understanding level has little difference in scores between men and women. Except for it, there are significant differences between male students and female students in status assessment, career exploration, and decision-making actions, and this difference is that male students are generally higher than female students.
Because the topic of this study is college students’ career planning, majors are naturally worth discussing for college students. In order to study the influence of majors on them, it is concluded in Table 9 that the value of the t-test in the total table is 0.692, which is not significant, indicating that professional differences (only divided into arts and sciences to discuss here) do not lead to various differences in college students’ career planning level.
When students are discussed by grades, it can be known from Table 9 that the value of the t-test of total scale is significant; that is to say, there are significant differences in grades. In addition, from the numerical point of view, the F test values of status assessment and self-understanding level are 0.015 and 0.002, respectively, which is very significant, indicating that there are significant differences in students’ cognitive level of themselves in different grades. According to psychological research, this is because students’ cognitive level is gradually rising with age.
Through Table 10 and Table 11, multiple comparison tables of different grades, it is found that the evaluation level of sophomores at all levels is significantly lower than that of graduating class. For example, in the single level of self-understanding, the level of graduating class students is significantly higher than that of freshmen, sophomores, and juniors. On the overall level, sophomores and juniors are still obviously lower than graduating class students. Generally speaking, senior students in graduating class have higher career planning level than other grades, but sophomores and juniors in middle grades have the lowest career planning level. In addition, there are significant differences in four aspects of career planning, namely, current situation evaluation, self-understanding, career exploration, and decision-making action. The values of t-test are 0.007 < 0.01, 0.002 < 0.01, 0.0005 < 0.001, and 0.0009 < 0.001, respectively.
(2) Family Variable. Family has a great influence on children, especially on adolescent children. In order to verify this point, in Table 12, the value of the t-test of the total size of college students from different families is 0.009, which is significant, indicating that there are significant differences in the level of career planning between college students from cities and rural areas; that is to say, the level of career planning of college students from cities is higher than that of students from rural areas. It is possible that this is because students who grew up in cities have more knowledge and opportunities, which will naturally have a positive impact on their career planning. In addition, there are significant differences between urban and rural students in terms of current situation assessment and career exploration; specifically, urban students are generally higher than rural students.
(3) School Experience Variable. The student cadres mentioned above are a manifestation of students’ ability in school, which can be obtained from Table 13. The t-test value of the total table is 0.002, which is significant, indicating that there is a very significant difference. It shows that the level of students who have experienced student cadres is higher than that of students who have not been student cadres. This conclusion supports the previous hypothesis H2.5; that is to say, there are significant differences in all levels of career planning among those who have work experience as student cadres. Outstanding in the current situation assessment and career exploration level, there are obvious differences in the work experience of student cadres, and the values of the t-test are 0.048 < 0.05 and 0.000 < 0.001, respectively.
Part-time job or internship is a way for students to contact the society, and it is also necessary. On the premise of ensuring safety, it should be advocated in large quantities. I believe that there should be differences in career planning level with or without part-time internship experience, which is also confirmed by Table 14. The t-test value of the total table is 0.009, which is significantly different. In addition, the values of the t-test are 0.009 < 0.01, 0.003 < 0.01, and 0.013 < 0.05, respectively, which shows that students will take part-time jobs and internships, which is very helpful to students’ personal employment and career planning. Therefore, students with internship and part-time jobs must be stronger than students without these experiences in many aspects.
Very few students spontaneously have a higher level of career planning, so students must have external forces such as career counseling to promote. From Table 15, the t-test value of the total scale is 0.000, which shows that there is a very significant difference; that is, the level of students who consult employment information and have relevant career counseling experience is significantly higher than that of students who have not experienced these that is, there are significant differences among students with different career counseling experiences at all levels. Similarly, students with career counseling experience are higher than those without.
First, it is very important for students to understand the social employment environment in which they are currently located. They should always understand that there are still many aspects that need self-improvement and maintain a positive and progressive attitude. Second, as a university, it is necessary to understand the particularity and difference of career planning guidance, and functional departments should change their working methods and try their best to ensure that employment services and guidance can follow up with every student.
5. Conclusion
Generally speaking, there are many successful places in China’s university education, and the enrollment data can be enough to show that China’s university education has made great achievements; at least the public enjoys the right to receive higher education. However, China’s total population ranks among the best in the world, with a large population and surplus labor force, so people’s employment pressure can be imagined. Graduates from good universities are better at finding jobs, while students from ordinary universities are relatively difficult to find jobs. For the development of our society, we must pay attention to the disadvantages of college students’ lack of career planning. From the above analysis, it can be concluded that decision-making action is very lacking in college students nowadays, and it is the last step of personal career planning, which is very important. In the process of visiting and investigating some students offline, it is found that most students do not know what decision-making actions are, are too lazy to act, or are confused and indecisive when acting. Therefore, under the current situation of no action, the overall level of college students’ career planning is naturally not high, which deserves great attention.
Data Availability
The experimental data used to support the findings of this study are available from the corresponding author upon request.
Conflicts of Interest
The author declares that there are no conflicts of interest regarding this work.