Influence of Short Video Application on College Students' Mental Health under Big Data Monitoring Environment.

Short videos are increasingly being consumed by college students as crucial content in the age of big data since they are a perfect fit for this medium. Therefore, college students should place a high value on the utilization of short movies. In this study, a neural network is utilized to create a mental health prediction model for college students. The neural network is trained using its self-learning capability to map out the relationships between different elements and mental health. The enhanced algorithm minimizes the production of candidate item sets to some amount, as well as the algorithm's time and space requirements, significantly decreasing the initialization time of the transaction set. According to the research, the test sample's pattern recognition accuracy was 81.29%, whereas the training sample's accuracy for pattern recognition was 83.37%. The analysis's finding is that the enhanced mining algorithm offers a fresh approach to educating college students about their health.

1. Introduction

In the era of big data, the popularity of short video applications is constantly improving, and the development of short video industry is maturing, forming a complete production and broadcasting chain. Short video APP gives ordinary people a platform to show themselves and their personalities. Everyone can get the latest information, build a personal social network, and express personal feelings and opinions through the short video application platform [1]. Not only that, the fact that college students are frequently exposed to short videos has aroused more attention from academic circles on psychological research of short videos, but this attention also shows a negative trend. If contemporary college students do not have good psychological quality, it will not only affect their study and life in school, but also hinder their future development.

Short videos are more ideal for users to view in scattered free time because they play for a long time, have a lot of works, and have fragmented material. This increases the efficiency of people's scattered free time as well as the convenience of enjoyment. According to Nikpour's research, video content can effectively and powerfully evoke a range of emotions, and the evoked emotions can last for a while [2]. According to Arria et al., positive emotion is a special immediate reaction to personally significant things, and it is viewed as a transient pleasure [3]. According to Micin et al., good emotions are strongly associated with traits like prosociality, self-assurance, a favorable understanding of others, vitality, an effective response to pressures and challenges, immunity, and physical health [4]. College students with high levels of thankfulness and those with low levels of gratitude differ significantly in their understanding and perceptions of social support, according to research by Downs et al. People with high levels of thankfulness report receiving more social support than people with low levels of gratitude [5]. From the existing research, there are still limitations, mainly as follows: The research on emotion is rich, but short videos are not used as intervention materials, and their positive intervention effects are not exerted; what kind of short video clips can effectively affect college students' mood, individual capital, and mental health level; and whether this effect has the specificity of clip type still needs to be demonstrated by empirical research. The standard of mental health is a common trait of people with mental health, and it is the synthesis of many healthy personalities. In the sense of health, mental health and physical health are equally important. Good mental quality is the key factor to a person's success, and mental health is also an important pillar of life health.

In order to better understand the stimulating effect of short videos on positive emotions and their impact on developmental psychology, this study first discusses the stimulating effect of media on college students' happiness, hope, and calm emotions from a positive psychology perspective. Additional supplements include disciplines and theories from cognitive psychology and other related fields. Second, thankfulness and curiosity, two crucial good emotions, have not gotten the attention they deserve. In order to show thanks and admiration to the university, this study uses brief movies. The impact of brief videos on college students' positive psychology is covered in this essay. By watching brief movies online, college students can expand their personal resources and enhance their emotional well-being and educational input. Research contribution of the paper are as follows:

Based on the literature about short video and emotional effects, this study puts forward two key questions about the feasibility and effectiveness of short video in positive emotional and psychological intervention. It provides a practical guide for college students to choose appropriate media content in online life and promote their own development and growth. It has certain reference value for guiding college students to contact short videos rationally

This study also innovatively proposes to combine the machine learning algorithm with the evaluation of students' mental state and construct a classification model based on neural network, so as to achieve the purpose of quickly evaluating students with mental health and provide necessary psychological counseling, and minimize the possibility of adverse mental health outcomes for abnormal students

2. Related Work

2.1. Research on College Students' Mental Health Theory

Mental health is the expression of an individual's good psychological quality and an important part of a person's overall health. Albright et al. think that mental health refers to the normal mental activities, coordinated relationships, consistent content with reality, and relatively stable personality [6]. Hardy et al. believe that mental health should meet the following criteria [7]: attitude towards oneself; the degree of growth, development, or self-realization; autonomy or the degree of independence for various social influences; the integration degree of main mental functions; the appropriateness of perception of reality; and ability to control the environment. The research of Shenoy et al. reveals the difference of bipolar mental health level in the microstructure of college students' peer relationship. Popular students have stable emotions and good social adaptability [8]. It covers not only personality, cognition, and behavior, but also the requirements of interpersonal communication, life and work.

Melnyk et al. defined positive emotion as an individual's unique immediate response to something meaningful, which is a short-term pleasure. Studies have shown that comedy videos can arouse a sense of entertainment. Although comedy videos can arouse a series of positive emotions, research shows that comedy videos can arouse the sense of entertainment most [9]. Denenny et al. put forward the theory of expansion and construction of positive emotions. This theory holds that positive emotions can broaden people's instantaneous thinking and actions, and such expansion can establish physiological resources, cognitive resources, psychological resources, and social resources, thus promoting personal happiness and health [10]. Barratt et al. show that there is a positive correlation between college students' psychological capital and social support, psychological capital and campus life satisfaction, and social support and life satisfaction [11].

2.2. Research on the Evaluation Model of Mental Health Status

As a realistic and important issue, it has been highly valued in the development of the country. Since the beginning of school, children have established mental health files, a systematic and manageable psychological research system, and the research has been relatively mature. Compared with western countries, the national college students' mental health education started late. The psychological evaluation system of college students has brought great progress to the research of students' psychological analysis. One of the most popular techniques in association knowledge discovery is association rule mining [12, 13]. In order to gather knowledge and patterns from semi-structured patient records of those suffering from mental illness and investigate the impact of genetic and environmental factors on mental health issues, Shahraki et al. utilized the tree mining algorithm [14].

Based on the data mining theory, Zhen et al. examined and evaluated the mental health status of college students [15]. In order to group and study the dangerous behaviors of college students, Jorge et al. employed two clustering analytic techniques, two-step clustering and quick clustering [16]. Based on their findings, they created an intervention strategy for college students. In order to create mathematical models that address real-world issues, Chekroud et al. use neural networks [17]. Through the development of a neural network model and online learning, Rohajawati et al.'s prediction system for college students' mental health [18] provides a basis for the prediction of college students' mental health to some extent; Su et al., based on the fuzziness of mental health status and the highly nonlinear characteristics of mental status evaluation, successfully combined neural network [19] with fuzzy mathematics.

3. Methodology

3.1. Analysis of College Students' Mental Health Data

In the field of college students' mental health, emotion is still an indispensable factor. Compared with comedy videos that can create a sense of entertainment, there are few researches on inspirational and hopeful videos. Some researchers believe that hope is a positive emotional state that wants to overcome difficulties. The short video covers this topic. When health and happiness increase, more positive emotions will be generated, and then health and happiness will increase like a spiral. Online intervention is based on the Internet, with websites or webpages as the carrier and transmission channel of intervention content, and it involves the participants independently, so as to influence the participants in a certain way. The positive psychological quality intervention to be carried out in this study refers to the online positive psychological intervention with the Internet as the carrier medium and short videos as the intervention materials.

As a popular social software for college students, short video will have many influences on college students' psychological values and behaviors, both positive and negative, such as promoting their personal development. Many people are using short video software to watch or create on the platform, and this group of college students is no exception. College students are still the main force in using short videos because they have enough time and great creative enthusiasm. When college students use short video apps, the quality of the content they receive is uneven, and college students are in a critical period of value formation. If they are often exposed to inferior videos in short video applications, it will definitely have a negative impact on you to develop good behavior habits and correct values.

Faced with the new learning environment, interpersonal relationship, emotional frustration, and work and study pressure, college students are prone to mental stress, anxiety, insomnia, grumpy, and other phenomena. A large number of open information resources provided by short video applications broaden the horizons of college students' psychological knowledge, but if the information quality is uneven and the comment content is chaotic. Some students are not sociable by nature and dare not express themselves in front of others, so it is difficult to have good interpersonal relationships. Poor interpersonal communication will affect students' loneliness and depression, as well as their mental health.

This paper examines the intelligent approach of mental health analysis in an effort to raise people's attention to and awareness of mental health. Five categories are used to categorize the mental health of college students: anxiety, depression, fear, paranoia, and hostility. Others split things into groups that are mutually exclusive. Some grouping methods divide objects into groups without clear boundaries. Nevertheless, other techniques use the separation between two items as a measure of similarity. In order to better avoid the development of mental disease, we can identify changes in users' psychological states globally through the extraction of multimodal data information.

K-means clustering is an unsupervised hard partition clustering method [20]. Is the objective function J, and k clusters are found from the data. The specific definition is as follows: where d2(C − X) is the square of the Euclidean distance between the centroid of the ith cluster and the jth data point. N is the total number of data points. According to the obtained distance, the points are assigned to the cluster with the smallest distance from the centroid.

In order to reduce the workload on mental health experts and increase productivity, a comprehensive intelligent mental health analytic system is developed with a focus on the mental health of college students. Building an intelligent mental health analysis model is the main objective of mental health analysis methodology. The framework of this concept is shown in Figure 1.

Figure 1

Intelligent mental health analysis model framework.

It can be divided into three modules: The first module is data processing module, the second module is data extraction module, and the third module is psychological testing module. The data processing part includes data acquisition, data extraction, data cleaning, and data specification. The data extraction module includes anomaly extraction module and association rule module, and the psychological test module provides a relatively reliable psychological test platform for students. Reasonable data collection can provide reliable sources for data mining, and only sufficient data sets can improve the reliability and persuasiveness of data mining.

Combining the idea of parallelization and sorting of item sets, this paper proposes an improved Apriori algorithm based on multithreading and compressed matrix. This algorithm solves the shortcomings of Apriori algorithm based on array, mainly multi-threading, item set sorting and array storage. It optimizes and improves the Apriori algorithm from array compression and ending conditions. The generation of infrequent candidate item sets is more and more reduced than that without sorting, and the time consumed by quick sorting is shorter, and the advantages of sorting are more and more obvious, so sorting can reduce the iteration time.

Let the clustering target X = {x, x2, ⋯, x} be a finite sample data set in real space, and n be the number of elements in the data set. Membership U = {u} is a n × c-dimensional matrix, which satisfies the following formula:

The process of c-means clustering algorithm is to find the solution when the second formula takes the minimum value under the constraint condition that the first formula of formula is established, where the class number c is given in advance.

By judging which category of dictionary the word belongs to, we can adjust the weight of the word in identifying the emotional tendency of the text. The specific formula is shown in the following formula:

For the extraction of word vectors and the recognition of emotional tendency of text information, the recognition model of emotional tendency of text information based on random forest and convolutional neural network is adopted.

3.2. An Analysis Model of the Influence of Short Video Application on College Students' Mental Health Is Established

They have complex emotions, are easily swayed by other people's words and actions, feel the need to express themselves, and anticipate respect and recognition from others. A new social platform called a short video APP gives college students a new platform to showcase their personalities. College students enjoy watching short films occasionally because the information is interesting and colorful, but they can become engrossed in the captivating material to the point that they find it difficult to break free, which also serves as a form of social entertainment. If a student has mental health issues in college and does not address and treat them in a timely manner, it will probably have an impact on his life's trajectory and, to a certain extent, his overall health. Any excellent data model must ultimately demonstrate its worth in use, mostly by extracting fresh data sets using the finished model to offer insightful information for making decisions. A particular model, such as a distribution model and probability model, can be simulated for a specific data collection as part of data mining. Specific techniques for data mining can be utilized based on this paradigm. This theory has been applied broadly in numerous data mining disciplines.

College students are in adolescence, and because they lack certain life experience and social experience, they are easily stimulated and influenced by various external environmental factors, such as excessive learning burden, insufficient learning methods, or aversion to race. Show pain and anxiety, and in the long run, there will be fatigue and truancy. Understand the general mental health status of freshmen, explore the changing trend of freshmen's mental health status, and carry out college students' mental health education in a timely and effective manner.

The original purpose of the short video of the audience is to relax and decompress, seeking to meet their emotional needs. Therefore, most short videos are mainly designed to make people laugh or relax and give them a good feeling. The facility combines “slow” and “fast” rhythms. It is in line with the fragmented information reception psychology of today's audience, and it also meets the needs of people to make use of fragmented time for self-improvement.

The two-step clustering algorithm is used to analyze the database and generate clustering results for the analysis of college students' mental health. First, data on college students' health condition are collected. Figure 2 displays the data mining methodology used to analyze the mental health of college students.

Figure 2

Data mining model of college students' mental health analysis.

The investigation of mental health problems is very complicated, involving many factors, and there is no correlation among them. Solving this nonlinear relationship has become the central issue. Because of the complex relationship between various influencing factors and their psychological conditions, this paper chooses an artificial neural network algorithm to form a neural system; then, the behaviors that this neural network can realize will be colorful, and its functions will be unlimited. In a genetic algorithm, an individual's fitness, which is the degree of conformity to the issue goal, describes the individual's capacity for survival. The person with the highest fitness value emerges as the best option from the algorithm's iterative procedure. If the minimum issue is the objective function, then

If the objective function is the maximum problem, then

C min is a given smaller number, and Cmax is a given larger number. According to the specific problems, choose the appropriate fitness function.

Using calculations that assume that all of the feature values in this chapter follow the normal distribution, the Pearson correlation coefficient is used to investigate the link between the feature values and the scores on the users' self-rating depression scale (6).

When the absolute value of the Pearson correlation coefficient is higher, the association between characteristics and depression is stronger and more features can reflect users' depression. When the correlation coefficient is negative, it indicates that the characteristic is linked to a higher risk of depression; when it is positive, it indicates that the characteristic is linked to a lower risk of depression, and the characteristic with the highest absolute correlation coefficient is retained.

TF-IDF just makes up for this defect. It combines the frequency f(t, d) of feature item t in document d with the distribution of feature items in the whole document set. The calculation formula of TF-IDF is as follows: where |D| is the total number of documents in the document set D and df(t, d) is the total number of documents containing feature items t.

The average distance of the network is the average distance between all nodes in the network, which describes the average dispersion degree of nodes in the network. The formula is as follows: where N is the number of nodes. The average distance L is also called the characteristic path length of the network.

If the center of a node's betweenness is large, it means that the information flow through the node is large in the process of information dissemination. Intermediate centrality is mainly used to measure the degree of some nodes, but it connects two unrelated communities, so it has high importance, which is defined as follows:

g is the number of shortest paths of user nodes V to V and g is the number of shortest paths passing through V among the shortest paths of user nodes V to V. Therefore, the betweenness of node V is the proportion of the number of all shortest paths passing through this node in the network.

The optimization algorithm flow is shown in Figure 3 below.

Figure 3

Optimization algorithm flow chart.

The following is a description of the particular algorithmic steps: (1) Binary coding: in other words, the variable is encoded as a string of values ranging from 0 to 1, and a constant is used to calculate its length. (2) Building a fitness function: the person with the highest fitness value emerges as the best option from the algorithm's iterative procedure. (3) The selection operator's design: the selection operator is the process of screening each individual population, which efficiently decreases calculation time, enhances global convergence, and prevents the occurrence of significant gene loss. Design of the mutation operator: in a range, a chromosome is chosen at random as the mutation point, and the original gene must be changed with a low probability.

4. Experiment and Results

In order to verify the correctness of the method proposed in this paper, 85% of the real data sets of college students, undergraduates, and doctoral students were randomly selected as model training data, and the rest were used as perception data. By establishing a control group, you can evaluate how well the protocol outlined in the text performed. The comparison of the experimental group's and control group's recognition scores for five mental health dimensions is shown in Figure 4. In Table 1, the statistical findings are displayed.

Figure 4

Comparison of recognition results.

Table 1

Statistical results.

Algorithm	Training sample accuracy (%)	Test sample accuracy (%)
Algorithm in this paper	83.37	81.29
Apriori	77.24	79.01
Ref [14]	80.26	76.24

It can be clearly seen that the ranking result of this algorithm is closer to the expert judgment. The recognition accuracy of the scheme described in this paper is 83.37% for training samples and 81.29% for testing samples. Both values are higher than those of Apriori and ref [14] algorithms.

By using abnormal data mining, we can determine each college student's scores in a variety of areas, including their overall mental health status and the projection value of each data object across ten factors. 15 factors can currently be assigned as class labels, and the corresponding rules can be extracted. Scores on all 15 factors may, however, only be a temporary indicator of a college student's performance. Different support thresholds and minimum confidence thresholds are used in the mining in Figure 5.

Figure 5

Mining conditions under different parameters.

Communication ability also has a great influence on students' psychological state, and communication is the guarantee of people's physical and mental health. Effective communication with family and friends can broaden students' horizons and dispel negative emotions. Through correlation analysis, it is also confirmed that this factor will have a great negative impact on college students' psychology, which will also be one of the main factors in the follow-up modeling analysis. In addition, the database can be maintained accurately, such as adding data items, deleting data items, querying data items, and updating data items. More importantly, the program that facilitates the system simplifies the operation of the database. Therefore, it is very important to standardize the database.

In this experiment, by setting the parameter values, the databases with different transaction numbers are generated, respectively, and a comparative experiment is conducted for different transaction numbers. In the experiment, other parameters are the same except the number of transactions. We set the number of articles to 30, the average transaction time to 20, and the support rate to 30%, and then change the number of transactions. The original transaction set required for the experiment is generated by the data generator.

During the experiment, we printed the number of transactions after merging and repeating the original transaction sets, the number of frequent item sets in the iteration process, the maximum frequent item sets, and the time spent in the whole operation process. The performance of the algorithm is measured by time, and the experimental results of different algorithms are shown in Figure 6.

Figure 6

Experimental comparison of algorithms.

It is obvious that the original Apriori algorithm's execution time increases linearly with the number of transactions, whereas the execution times of the other three enhanced array-based Apriori algorithms are significantly faster than the original algorithm. The iterative process increases the ordering of item sets while decreasing the generation of candidate item sets of many rare items when seeking out frequent item sets. The benefits of parallel operation and sorting become increasingly clear as the number of transactions rises.

Add the emotional and media feature scores of different subjects to the same media segment to obtain the total sample of emotional and media feature scores of each segment. Calculate the correlation between each emotional score and each media function score, and the results show that there is a more significant correlation between emotional elements and media function elements. This study combined the original scores of each person's two videos in the same scene, calculated the mean, and got the objective emotional score of each subject in each scene. Descriptive statistics are made on the processed data, and the results are shown in Table 2.

Table 2

Describe the statistical results.

Emotion category	Minimum	Maximum	Average	Standard deviation
Inspirational class	1	6.33	3.25	0.52
Nature	1	7.85	4.02	0.41
Comedy class	1	6.01	3.15	0.36

The analysis of emotional response in the previous section demonstrates that media content can successfully arouse emotions in each situation and that the three emotions are each arouse in a different way by different media content. There are significant differences between the interesting features of the comedy and the other two types of video clips, according to the findings of a one-way analysis of variance in the score of interesting features of the three different kinds of short films. No discernible difference existed between inspirational and natural clips in terms of interesting feature scores. Figure 7 displays the results of counting the mean and variance according to the information of men and women in order to clearly display the mean and standard deviation of each variable in each cluster.

Figure 7

Differences of variables between male and female students.

As can be seen, these four areas significantly distinguish boys and girls. Girls excel over guys in terms of flexibility, extroversion, aggression, environmental adaptation, and decisiveness, from a comparison of the male and female groupings. Enter the data from the constructed neural network model and simulate test data, as illustrated in Figure 8. The fitting degree of the input data largely satisfies the requirements, and it can be observed through the simulation test of each data group that the error groups and real values among the prediction results of each group are extremely small. The model is effective at forecasting the mental health of college students.

Figure 8

Result analysis of sample data.

The enhanced neural network's capacity for data fitting has greatly increased. The evolutionary algorithm's discovery of the ideal weight and threshold through the fitness function is the cause, which significantly reduces the weight. The improved neural network's iterative steps are obviously cut down, and the predetermined objective can be reached more quickly. The findings demonstrate that the prediction model may accurately anticipate pupils who have psychological issues and, in general, achieve the desired outcome.

Through lectures, college students can have a deeper understanding of short videos and fully understand the impact of excessive use of short videos. Watching short videos is beneficial, but it should not be overused. At the same time, ensure that the push content is rich and diverse, giving consideration to professionalism, practicality, and interest, guiding college students to read more books and improving their reading mental health.

Short videos that can induce positive emotions can be introduced, which can easily arouse students' interest on the one hand, and students' interest on the other. It is also helpful to improve students' positive psychological state, and ultimately conducive to students' growth and development. At the same time, colleges and universities can carry out relevant love education, organically combine students' concern for themselves with their concern for others, organically combine cultivating students' personal development with caring for others' progress, and create positive support for college students with psychological problems to face their problems correctly.

5. Conclusion

Big data has given rise to an abundance of content and a growing number of methods for implementing mental health education. College students' lives should not be hampered by a quick video; instead, it should foster their development. Short movies are simple to use, which helps college students develop as individuals. The network prediction model has essentially been finished, and simulation testing has been used to confirm the model's effectiveness. When using training samples, the solution recognition results reported in this research had an accuracy of 83.37%; when using test samples, it had an accuracy of 81.29%. The outcomes demonstrate that the model may, to some extent, predict college students' mental health. The paths taken by personal resources, mental health, and educational input are not entirely parallel. The learning and mental health of college students can be improved more by inspirational videos than by comedic ones, which are better for building up their psychological resources.