Inferiority complex of higher vocational students based on improved random sampling

Abstract

Some factors affecting the physical and mental health of vocational college students, the sense of inferiority plays a very important role in cultivating students with physical and mental health. Inverse random under sampling algorithm is improved based on integrated learning, which can improve the performance of the classifier. Stacking integrated learning and flip random sampling reduction algorithm SIRUS is proposed. Select the individual subjective factors studied in this paper is important in self-attribution and social objective factors are important social support factors, and the only demographic variables is a significant difference.

Keywords

Vocational students inferiority feeling attribution social support sampling algorithm random sampling

1. Introduction

Inferiority is a common psychological phenomenon in human society. Before it became the research object of psychology, it first appeared in the works of Stendhal, a writer, who believed that “inferiority is an outstanding human soul disease” [1]. French psychopathologist and psychologist Janet (1859–1947) was the first person to give a scientific explanation of inferiority. He believed that inferiority was dominated by the stubborn idea that everything goes wrong [2, 3, 4, 5, 6, 7].

Experimental interventions for low self-esteem have also been relatively rare. Existing studies mainly include: Hong using standardized inferiority scale (FIS) [8, 9, 10, 11, 12] randomly surveyed 567 full-time undergraduate students, Yangzhou university, and screening out the inferiority problems of college students, a total of 88 (FIS divide, 4) [13, 14, 15, 16, 17], through the statistical analysis, reveals the characteristics of college students’ inferiority complex, and recruit were screened out, will be randomly divided into experimental group and control group, the subjects were accepted for 9 weeks psychological intervention, the experimental intervention group hypnosis guidance, and in the same condition of the control group only group guidance, results show that the long-term effect of group hypnosis guidance is superior to the general group guidance, it is very applicable to the general psychological problems of college students.

2. Methods

This study randomly selected 100 students from three vocational and technical colleges of Yantai vocational college, Yantai business vocational college and Yantai engineering vocational and technical college. A total of 280 questionnaires were distributed and 278 questionnaires were recovered, with the recovery rate of 99.2%. The incomplete and invalid questionnaires were eliminated, leaving 277 valid questionnaires with an effective rate of 99%. The basic information is shown in Table 1.

Table 1
Basic information of subjects

Basic characteristic	Variable number (person)	Percentage (%)
Gender
Male	108	36.4
Female	189	63.6
Subject
Arts	142	47.8
Science	155	52.2
Registered
City	65	21.8
Rural	232	78.1
School
1	97	32.7
2	100	33.7
3	100	33.7
Only child
Yes	113	38.0
No	184	62.0
Grade
Freshman	100	33.7
Sophomore	87	29.3
Junior	110	37.0

School: 1. Yantai engineering college; 2. Yantai business vocational college; 3. Yantai vocational college.

Firstly, the concept of confusion matrix is defined: suppose that in the original data set, the number of a few class samples (true class) is $p$ , and the number of a majority class samples (negative class) is $n$ . F_P refers to the number of samples mistakenly classified into a few categories; F_N refers to the number of samples mistakenly classified into a majority category; T_P refers to the number of samples correctly classified into a minority category; T_N refers to the number of samples correctly classified into a majority category [2], and the confusion matrix shown in Table 2 can be obtained.

Table 2

Confusion matrix

	The prediction is true	The prediction is negative
The true class is true class	T_P	F_N
The true category is false	F_P	T_N

The following formula can be calculated and obtained [2, 3, 18, 19, 20, 21, 22, 23, 24, 25, 26]:

$\displaystyle\textit{False positive rate}(\textit{fpr})=\frac{F-P}{F-P+T-N}$ (1) $\displaystyle\textit{Specificity}(\textit{tnr})=\frac{T-N}{T-N+F-P}$ (2) $\displaystyle\textit{Recall}(\textit{tpr})=\frac{T-P}{T-P+F-N}$ (3) $\displaystyle\textit{Precision}=\frac{T-P}{T-P+F-P}$ (4) $\displaystyle\textit{Globl-mean}=\sqrt{\textit{tnr}\times\textit{tpr}}$ (5) $\displaystyle\textit{F-measure}=\frac{2\times\textit{Precision}\times\textit{% Recallback}}{\textit{Precision}+\textit{Recallback}}$ (6)

Their meanings are: False true rate is called False true rate, which means the proportion of the samples of most classes that are misclassified as a few classes in all the classes. Specificity is designated as Specificity, indicating the correct proportion of specific samples in most samples. Recall, also known as true class rate, refers to the proportion of correctly judged samples in a small number of class samples. Precision is called Precision rate [27], which represents the proportion of correctly classified samples among all samples classified as a few classes. Global-mean is the geometric average of specificity and recall rate, indicating the comprehensive performance of accuracy of the two categories. Recall or Precision alone cannot accurately evaluate the performance of the classifier. For example, if there are a total of ( $x+y$ ) data samples in the test set, including m for most classes and n for a few classes ( $x>y$ ), if all samples are classified as a few classes, the confusion matrix can be obtained as shown in Table 3.

Table 3

Confusion matrix

	The prediction is true	The prediction is false
True class is true class	T_P: n	F_N: 0
The true category is false	F_P: m-n	T_N: 0

3. The social support rating scale

The social support rating scale (SSRS) [7, 11, 16, 23, 26, 27] of Xiao shuiyuan was used to test the status of social support. The scale consists of 10 items, including three dimensions of objective support (3), subjective support (4) and the utilization of social support (3). The consistency coefficient of the two-month retest total score of the scale is 0.92, and the internal consistency of each item is between 0.89 and 0.93, showing that the social support scale has a more retest reliability [9]. In this measurement, the consistent Aplpha coefficient of the total score was 0.90, and the Aplpha coefficients of objective support, subjective support and support utilization were 0.87, 0.89 and 0.83.

According to the actual situation of higher vocational students, some items in the scale are modified. In question 2, remove the word “colleague” from option 3. In the third question, “neighbor” is replaced by “roommate”; In question 4, replace “colleague” with “classmate”; The “husband and wife” in the fifth question is replaced by “lovers”, and remove the “children” column; The 6, 7 in “spouse” to “lover”, “colleague” to “classmate”, “work unit” to “teacher”; In question 10, “labor union” is replaced by “class committee”.

The overall situation of inferiority feeling of students in vocational colleges. The scores of each dimension of inferiority were processed to obtain the mean and standard deviation of each dimension. The results are expressed in Table 4.

Table 4
Standard and mean deviation of each dimension of students’ inferiority

	N	M	SD
Developmental inferiority	297	1.1422	0.71234
Inferiority of family love	297	0.9796	0.64359
Looks inferior	297	1.1532	0.68753
Learning experience inferiority	297	1.0741	0.67274
Self-abased total score	297	1.0586	0.60101

From the data shown in Table 4, it can be seen that the students of vocational colleges in the various dimensions of the sense of inferiority and the overall score are in the points (2 points) below, that is, there is no obvious sense of inferiority. This is consistent with the existing research conclusions on the inferiority of college students in China, and also in line with the current situation of college students in Shandong. The scores of each dimension were ranked from high to low in the order of personality inferiority, appearance inferiority, communication skills inferiority, learning experience inferiority, developmental inferiority and family love inferiority. The higher the score, the stronger the inferiority feeling in this dimension. On the contrary, the lower the score, the weaker the inferiority feeling in this dimension.

In order to test whether there is any difference in subjects among various dimensions of inferiority of vocational college students, independent sample T test is conducted on the data, and the mean, standard deviation and T score obtained are shown in Table 5.

Table 5

Independent sample $t$ -test of inferiority in higher vocational college students

	Arts M $+$ SD	Science M $+$ SD	T
Communication skills low self-esteem	1.19 $\pm$ 0.72	1.10 $\pm$ 0.70	1.076
Developmental inferiority complex	1.04 $\pm$ 0.66	0.92 $\pm$ 0.62	1.690
Low self-esteem in family love	0.87 $\pm$ 0.73	0.80 $\pm$ 0.64	1.094
Inferiority of appearance	1.21 $\pm$ 0.67	1.09 $\pm$ 0.70	1.484
Low self-esteem	1.42 $\pm$ 0.84	1.17 $\pm$ 0.67	2.927 ${}^{**}$
Learning experience low self-esteem	1.12 $\pm$ 0.68	1.03 $\pm$ 0.67	1.070
Inferiority score	1.12 $\pm$ 0.61	1.00 $\pm$ 0.59	1.753

Note: ${}^{*}$ Precise $<$ 0.05, ${}^{**}$ Precise $<$ 0.01, ${}^{***}$ Precise $<$ 0.001.

There are significant differences in the dimension of self-abased personality among vocational students of different disciplines ( $P=$ 0.004 $<$ 0.01). The analysis shows that the personal inferiority feeling of students of liberal arts in higher vocational college is higher than that of students of science in higher vocational college (article M $=$ 1.42 $>$ theory M $=$ 1.17). The other dimensions of inferiority are not significantly different in discipline.

In information theory, entropy refers to the uncertainty of a random variable. Let R be a discrete random variable with a finite number of values, and assume that X has the following probability distribution [7, 11, 21, 22, 23, 24, 25, 26, 27]:

$\displaystyle P(R=r_{i})=p_{i},i=1,2,\ldots,n$ (7)

So the entropy of our random variable Y is going to be like this:

$\displaystyle H(Y)=-\sum{{}_{i=1}^{n}}p_{i}\log p_{i}$ (8)

The higher the entropy, the greater the uncertainty of the random variable. Let the joint probability distribution of random variables ( $M, N$ ) and ( $M, N$ ) be:

$\displaystyle P(M=x_{i},N=y_{i})=p_{ij},i=1,2,\ldots,n;j=1,2,\ldots,m$

The calculation formula of conditional entropy $H(M|N)$ is:

$\displaystyle H(M|N)=\sum{{}_{i=1}^{n}}p_{i}H(M|N=n_{i})$ (9)

where $p_{i}=P(R=r_{i}),i=1,2,\ldots,n$ .

Information gain glo ( $M, N$ ) of feature A to training set $M$ is defined as the difference between entropy $H(M)$ of data set $M$ and conditional entropy $H(M|N)$ , i.e.:

$\displaystyle\textit{glo}({M,N})=H(M)-H(M|N)$ (10)

The algorithm for calculating information gain is as follows:

(1)

calculate empirical entropy $H(M)$ of training set $M$ :

$\displaystyle H(M)=-\sum{{}_{k=1}^{K}}\frac{|{C_{k}}|}{|M|}\log_{2}\frac{|{C_{% k}}|}{|M|}$ (11)

(2)

calculate the empirical condition entropy $H(M|N)$ of feature A to training set $M$ :

$\displaystyle H(M|N)=\sum{{}_{i=1}^{n}}\frac{|{M_{i}}|}{|M|}H(M_{i})=-\sum{{}_% {i=1}^{n}}\frac{|{M_{i}}|}{|M|}\sum{{}_{k=1}^{K}}\frac{|{M_{ik}}|}{|{M_{i}}|}% \log_{2}\frac{|{M_{ik}}|}{|{M_{i}}|}$ (12)

(3)

calculate information gain:

$\displaystyle Gg({M,N})=H(M)-H(M|N)$ (13)

According to the above calculation, it can be seen that the size of information gain is relative to the training set, and the greater the empirical entropy of the training set is, the greater the information gain will be, without any absolute significance. The information gain ratio is corrected for this problem. The information gain ratio $g_{R}(M,N)$ is calculated as follows:

$\displaystyle g_{R}(M,N)=\frac{g(M,N)}{H(M)}$ (14)

ID3 algorithm USES [3, 4, 5, 6, 7, 8] the index of information gain to select features on each node of the tree and constructs the tree recursively. Therefore, ID3 is used to select attributes with more values, which often results in a huge but shallow decision tree. Moreover, ID3 algorithm only generates trees, so it is easy to generate overfitting problems. To solve the above problems, algorithm changed the feature screening criteria from information gain to information gain ratio during the training stage.

Suppose that in the original training set [26], $X_{N_{\min}}$ represents the collection of a few samples, where $N_{\min}$ represents the collection of most samples, where $X_{N_{\textit{maj}}}$ represents the number of samples from $N_{\textit{maj}}$ ; $S<N_{\min}$ , sets represents the number of data subsets; T represents the test sample. The flow of the improved algorithm is as follows:

(1)

Initialize $S=\frac{|{N_{\min}}|}{N_{\textit{maj}}}\times|{N_{\min}}|,i=1,\max=0,\textit{% Set}=\varphi$ .

(2)

Randomly samples s samples from $X_{N_{\textit{maj}}}$ and names them as $X^{\prime}_{N_{\textit{maj}}}$ .

(3)

Compose $X^{\prime}_{N_{\textit{maj}}}$ and $X_{N_{\min}}$ to a new data subset $T_{S}$ .

(4)

Using $T_{S}$ to train individual classifier $h_{i}$ .

(5)

Judge whether $I$ is less than sets, if so, $I=I+1$ , return (2).

(6)

Train the classifier on the dataset, and judge whether the performance of the classifier on the dataset is greater than max. if so, update the set to the set of current sets data subsets.

(7)

Let $S=S+S\times$ 10%, if $S<N_{\min}$ , return (1).

(8)

The data subset set in set is the final data subset.

The above algorithm sets the minimum value of the number of samples s extracted from the majority of class sample sets to $\frac{|{N_{\min}}|}{|{N_{\textit{maj}}}|}\times|{N_{\min}}|$ , that is, the number of false class samples in each data subset after flipping random down sampling, and then increases the value of S by 10% each time, and then conducts experiments on the new data set in order to select the optimal data distribution.

For example, logistic regression will output a real value $z$ in general. In order to produce a classification result, the real value is substituted into the following Sigmoid function [12, 19, 21, 27, 28, 29]:

$\displaystyle\textit{Siglmoid}(z)=\frac{1}{1+e^{-2}}$ (15)

The Siglmoid function image is shown in Fig. 1.

4. Conclusion

In order to test whether there is any difference in disciplines among various dimensions of self-attribution of higher vocational students, independent sample $t$ -test was carried out on standard deviation and the data, the mean and $t$ -score obtained were shown in Table 6.

Table 6

Independent sample $t$ -test of self-attribution of vocational college students in disciplines

	Arts M $+$ SD	Science M $+$ SD	T
Ability attribution	10.83 $\pm$ 3.23	9.89 $\pm$ 4.10	2.197 ${}^{*}$
Effort attribution	13.21 $\pm$ 4.04	11.79 $\pm$ 4.72	2.781 ${}^{**}$
Background attribution	9.28 $\pm$ 3.55	8.76 $\pm$ 3.88	1.195
Luck attribution	9.25 $\pm$ 3.44	8.59 $\pm$ 3.90	1.540
Successful attribution	22.80 $\pm$ 6.60	20.74 $\pm$ 7.91	2.445 ${}^{*}$
Failure attribution	19.76 $\pm$ 6.33	18.31 $\pm$ 7.73	1.769
The attribution score	5.51 $\pm$ 5.40	4.39 $\pm$ 4.90	1.882

Note: ${}^{*}$ Precise $<$ 0.05, ${}^{**}$ Precise $<$ 0.01.

Figure 1.

Siglmoid function.

Seen from Table 6, there is a difference in the discipline of students’ ability attribution in vocational colleges ( $P=$ 0.030 $<$ 0.05). There was a significant difference in effort attribution ( $P=$ 0.006 $<$ 0.01). There was also a significant difference in successful attribution ( $P=$ 0.016 $<$ 0.05). Moreover, the attribution of higher vocational liberal arts students is greater than the scores of higher vocational science students in these dimensions. There is no significant difference in the background attribution dimension, luck attribution dimension, failure attribution dimension and attribution total score.

(1)

On the whole, the inferiority level of students in vocational colleges has not reached the average of inferiority, that is, students in vocational colleges have no serious inferiority.

(2)

There are significant differences in demography variables such as gender, discipline, origin and oneness of students in the sense of inferiority, attribution and social support of vocational students.

College student inferiority look inferiority dimension grade level difference is significant; There are significant differences in the dimension of personality inferiority.

The total inferiority feeling of college students in gender, whether the only child and students on the ground, there is no significant difference, and each dimension in the above level there is no significant difference.

(3)

The relationship between inferiority, attribution style and academic performance

There is a significant true correlation between each dimension of inferiority feeling and each dimension of attribution style.

Inferiority of communication skills is significantly truly correlated with subjective support, support utilization and social support total score, while other factors are significantly falsely correlated.

Except that there is no significant correlation between effort attribution and subjective support, all the other dimensions have significant false correlation.

References

Quinlan

J.R.

, Induction on decision tree, Machine Learning 1(1) (1986), 81–106.

Kononenko

, Semi-naive bayesian classifier, Lecture Notes in Computer Science 482 (1991), 206–219.

Haykin

, Neural networks: A comprehensive foundation, Macmillan, 1998.

Maurya

C.K.

Toshniwal

and Venkoparao

G.V.

, Online Anomaly Detection Via Class-imbalance Learning, in: Eighth International Conference on Contemporary Computing, IEEE, 2015, pp. 30–35.

Phua

Alahakoon

and Lee

, Minority report in fraud detection: classification of skewed data, Acm Sigkdd Explorations Newsletter 6(1) (2014), 50–59.

Polikar

, Ensemble learning, Scholarpedia 4(1) (2009), 1–34.

Albrecht

W.S.

Albrecht

and Albrecht

C.C.

, Current trends in fraud and its detection, Information Systems Security 17(1) (2008), 2–12.

Liu

Y.Q.

Wang

and Zhang

, Decision Tree Based Predictive Models for Breast Cancer Survivability on Imbalanced Data, in: International Conference on Bioinformatics and Biomedical Engineering, IEEE, 2009, pp. 1–4.

Wei

J.J.

Cao

L.B.

Y.M.

and Chen

J.H.

, Effective detection of sophisticated online banking fraud on extremely imbalanced data, World Wide Web 16(4) (2013), 449–475.

10.

and Garcia

E.A.

, Learning from imbalanced data, IEEE Transactions on Knowledge & Data Engineering 21(8) (2009), 1263–1284.

11.

Drummond

and Holte

R.C.

, C4.5, class imbalance, and cost sensitivity: why under-sampling beats over-sampling, workshop on learning from imbalanced datasets II, Washington DC: Citeseer 11 (2003), 1–8.

12.

Tomek

, Two modifications of CNN, IEEE Transactions on Systems Man & Cybernetics SMC-6(10) (1976), 769–772.

13.

Kubat

and Matwin

, Addressing the curse of imbalanced training sets: one-sided selection, ICML 97 (1997), 179–186.

14.

Wilson

D.L.

, Asymptotic properties of nearest neighbor rules using edited data, IEEE Transactions on Systems Man and Cybernetics 2(3) (2007), 408–421.

15.

Chawla

N.V.

Bowyer

K.W.

Hall

L.O.

and Kegelmeyer

W.P.

, SMOTE: synthetic minority over-sampling technique, Journal of Artificial Intelligence Research 16(1) (2002), 321–357.

16.

Torres

F.R.

Carrasco-Ochoa

J.A.

and Martínez-Trinidad

J.F.

, SMOTE-D a Deterministic Version of SMOTE, in: Martínez-Trinidad

Carrasco-Ochoa

Ayala Ramirez

Olvera-López

Jiang

, (eds), Pattern Recognition, MCPR 2016, Lecture Notes in Computer Science, Vol. 9703, Springer, Cham, 2016, pp. 177–188.

17.

Han

Wang

W.Y.

and Mao

B.H.

, Borderline-SMOTE: a new over-sampling method in imbalanced data sets learning, in: Huang

D.S.

Zhang

X.P.

Huang

G.B.

, (eds), Advances in Intelligent Computing, ICIC 2005, Lecture Notes in Computer Science, Vol. 3644, Springer, Berlin, Heidelberg, 2005, pp. 878–887.

18.

Batista

G.E.A.P.A.

Prati

R.C.

and Monard

M.C.

, A study of the behavior of several methods for balancing machine learning training data, Acm Sigkdd Explorations Newsletter 6(1) (2004), 20–29.

19.

Batista

G.E.

Bazzan

A.L.C.

and Monard

M.C.

, Balancing Training Data for Automated Annotation of Keywords: A Case Study, II Brazilian Workshop on Bioinformatics, DBLP, 2003, 10–18.

20.

Japkowicz

and Stephen

, The class imbalance problem: a systematic study, Intelligent Data Analysis 6(5) (2002), 429–449.

21.

Polikar

, Ensemble based systems in decision making, IEEE Circuits and Systems Magazine 6(3) (2006), 21–45.

22.

Freund

and Schapire

R.E.

, A decision-theoretic generalization of on-line learning and an application to boosting, Journal of Computer and System Sciences 55(1) (1997), 119–139.

23.

Breiman

, Bagging predictors, Machine Learning 24(2) (1996), 123–140.

24.

Wolpert

D.H.

, Stacked generalization, Neural Networks 5(2) (1992), 241–259.

25.

Nanni

and Lumini

, Fuzzy bagging: a novel ensemble of classifiers, Pattern Recognition 39(3) (2006), 488–490.

26.

Huang

Z.H.

Tang

J.Y.

Shan

G.X.

Chen

Y.W.

and Wang

, An efficient passenger-hunting recommendation framework with multi-task deep learning, IEEE Internet of Things Journal 6(5) (2019), 7713–7721.

27.

Huang

Z.H.

Zhu

H.H.

and Zhou

M.C.

, An efficient group recommendation model with multiattention-based neural networks, IEEE Transactions on Neural Networks and Learning Systems (99) (2020), 1–14.

28.

Huang

Z.H.

Zhu

and Wang

, Multimodal representation learning for recommendation in internet of things, IEEE Internet of Things Journal 6(6) (2019), 10675–10685.

29.

Huang

Z.H.

and Wang

, Comparative convolutional dynamic multi-attention recommendation model, IEEE Transactions on Neural Networks and Learning Systems, 2021, 99.

Inferiority complex of higher vocational students based on improved random sampling

Abstract

Keywords

1. Introduction

2. Methods

Table 1 Basic information of subjects

Table 4 Standard and mean deviation of each dimension of students’ inferiority

References

Table 1
Basic information of subjects

Table 4
Standard and mean deviation of each dimension of students’ inferiority