Application and Evaluation of Algorithms and Deep Learning in

Adolescent Mental Health Intervention

Yue Zuo

and Zhengkui Liu

2,*

CAS Key Laboratory of Mental Health, Institute of Psychology, Chinese Academy of Sciences, Beijing, P.R.China

Department of Psychology, University of Chinese Academy of Sciences

，

Beijing, P.R. 100101, China

Keywords: Algorithms, Deep Learning, Adolescent Mental Health Intervention.

Abstract: Adolescent mental health intervention refers to targeted mental health education and behavior modification

for adolescents based on scientific psychological theories and techniques, with the aim of reducing the

incidence and probability of mental illness in adolescents, reducing their socio-economic burden, and

promoting their healthy growth. At present, there are some problems in the mental health intervention of

adolescents in China. First, the methods and means of mental health education for adolescents lack

scientificity, effectiveness and pertinence, and cannot effectively solve the psychological problems of

adolescents. Second, there is a lack of understanding of the needs of adolescent mental health interventions,

and the lack of necessary psychological intervention techniques and implementation processes. Third, there

are problems in the use of existing psychological intervention techniques, such as unclear effect evaluation

methods, unclear technical paths, and non-standardized operation processes, which affect the popularization

and application of the intervention methods to a certain extent. As one of the core technologies of artificial

intelligence (AI), algorithm and deep learning refers to the use of large amounts of data for learning, which

can realize automatic analysis and recognition of data. Its application scenarios include autonomous driving,

medical imaging diagnosis, machine translation and other fields. Artificial intelligence technology has great

potential in adolescent mental health intervention, and on the basis of outlining the application and evaluation

of existing adolescent mental health interventions, this paper reviews the application and evaluation of

algorithms and deep learning in adolescent mental health interventions, aiming to provide new ideas for

further adolescent mental health interventions.

1 INTRODUCTION

In recent years, adolescent mental health issues have

attracted great attention from society and the

government. The Chinese government attaches great

importance to the mental health of adolescents and

lists it as an important part of China's "Healthy China"

strategy. In 2016, the State Council issued the

Guiding Opinions on Strengthening Adolescent

Mental Health Services, calling for greater

investment in adolescent mental health services and

promoting the construction of a juvenile

psychological service system (Ambikavathi, and

Arumugam, et al. 2023). In 2017, the General Office

of the State Council issued the Outline for the

Development of Children in China (2016-2020),

which stated that it is necessary to strengthen

children's mental health education and intervention,

and build a children's mental health service system

covering children aged 0-15 (Bakirarar, and Cosgun,

et al. 2023). In 2019, the Ministry of Education issued

the Guiding Outline for Mental Health Education in

Primary and Secondary Schools (for Trial

Implementation), which pointed out that it is

necessary to strengthen the construction of school

psychological counseling teams and improve the

school psychological counseling and counseling

system (Bergami, and Appleby, et al. 2023). In 2020,

the State Council issued the "Modernization of

Education in China 2035", emphasizing the need to

promote the establishment of a "one-stop" student

mental health service platform with the participation

of education, health, civil affairs and other

departments, as well as organizations such as the

Women's Federation and the Disabled Persons'

Federation (Brzychczy, and Zuber, et al. 2024). At

present, adolescents' mood swings, impulsive

behavior and poor social adaptability are common

478

Zuo, Y. and Liu, Z.

Application and Evaluation of Algorithms and Deep Learning in Adolescent Mental Health Intervention.

DOI: 10.5220/0013546100004664

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 3rd International Conference on Futuristic Technology (INCOFT 2025) - Volume 1, pages 478-484

ISBN: 978-989-758-763-4

psychological problems among adolescents.

According to statistics, there are nearly 40 million

teenagers in China who have varying degrees of

psychological problems (Goedemans, and Prokic,

2023). Adolescent mental health problems are mainly

manifested in negative emotions such as depression,

anxiety, and hostility, and positive emotions such as

compulsion and fear (Kannan, and Nandwana, 2023).

Some studies have shown that the occurrence of

depression and anxiety in adolescents is related to

factors such as study pressure, family environment,

and peer interaction (Maggi, and Marrella, et al.

2023). This also makes negative emotions such as

depression and anxiety a common psychological

problem among adolescents. Among them,

depression is one of the most common emotional

problems in adolescents. Studies have found that

depression is a devastating hazard to adolescents

(Petry, and Yager, 2023). According to the World

Health Organization (WHO) report, depressive

symptoms can directly affect adolescents' academic

performance and academic ability, and may lead to

social withdrawal and suicidal tendencies; at the same

time, depression can also reduce academic

performance and academic ability level; depression

can also reduce students' social adaptability;

depression may also cause adolescent anxiety

symptoms, affecting students' normal learning and

life. At present, a large number of research work has

been carried out on adolescent psychological and

behavioral problems at home and abroad, and a

variety of targeted, effective, and safe interventions

have been developed (Wentzel, and Floricel, et al.

2023). However, in practical application, there are

some problems, such as the difficulty in determining

the intervention object, the difficulty in grasping the

intervention timing, and the difficulty in sustaining

the intervention plan. In recent years, artificial

intelligence technology has developed rapidly as an

emerging technical means. It has the characteristics of

convenient data acquisition and fast processing speed,

and has been widely used in the medical field (Yang,

and Zhou, et al. 2023). In recent years, intelligent

health monitoring devices based on artificial

intelligence technology have been widely used, but

there are few studies in the field of children's mental

health. Therefore, this paper will focus on the

intelligent health monitoring equipment developed by

artificial intelligence technology in adolescent mental

health intervention, and focus on the application and

evaluation of intelligent health monitoring equipment

developed based on artificial intelligence technology

in adolescent mental health intervention, aiming to

provide new ideas and methods for further adolescent

mental health intervention.

2 RESEARCH METHODS

A total of 22 articles were included in this study by

searching databases such as PubMed, Web of

Science, Embase, and the Cochrane Library. During

the selection process, duplicate articles and articles

involving children or adolescents were excluded from

this study, and 20 articles were finally included. All

the literature was in English, and there were only 2

articles in Chinese. There are two main types of

experimental designs included in the literature: one is

the design of the intervention program, including the

design of the experimental group and the control

group, and the other is the design of the study subjects

and intervention materials, including the selection of

the experimental group and the control group, the

method of using the intervention materials, and the

type of materials. All experiments were randomized,

i.e., subjects were randomly assigned to the

experimental group or the control group, and all

experiments were single experiments and were not

repeated. Of all the included studies, one article

addressed a specific intervention or technical

approach, and the remaining 21 articles used different

types of research methods. Of these, 22 articles were

measured using a subjective scale, and 13 articles

were measured using a behavioral test method. Six

articles were measured using subjective scales,

objective scales and questionnaires. Two other

articles were measured using questionnaires and

experiments.

2.1 Subjective Scale Measurements

Nine of the 22 studies used subjective scales for

mental health, and most of them focused on

subjective scales for depression, anxiety, obsessive-

compulsive symptoms, and suicidal tendencies in

adolescents. Among them, Dunning et al. used the

Depression Self-Rating Scale (SDS) and the Anxiety

Self-Rating Scale (SAS) in their study, with the SDS

including 27 items and the SAS including 19 items.

Of the 8 suicide-related items, 6 were suicide-related

with the question "What do I often think or feel about

what happens when I die?", and the other 5 are "What

approach would I take if I were a patient?". Therefore,

the investigators believe that these two scales can be

used for assessment when performing adolescent

mental health interventions. However, there are also

individual studies that use a single self-rating scale

Application and Evaluation of Algorithms and Deep Learning in Adolescent Mental Health Intervention

479

for depression (SDS-R). Of the 12 suicide-related

entries, 7 suicide-related issues were "I always

thought I couldn't do it", "I always felt worthless", "I

felt helpless and hopeless", and "I often wanted to

die". The investigators believe that a single

depression self-rating scale or a single SDS can be

used in combination with SAS when intervening in

adolescent mental health.

2.2 Behavioral Test Measurement

Behavioral testing refers to the evaluation of the

effectiveness of an intervention through

experimentation, and is often used to evaluate the

effectiveness of an intervention. In this study,

behavioral testing was mainly used to assess the

impact of a specific technical means on the mental

health of adolescents, and a variety of intervention

techniques were used. As shown in equation (1).

(Φ) 0

−

ξφ





−



−φ φ



)

Among them, the most common are biofeedback,

mindfulness, and behavior modification. Biofeedback

refers to the use of some equipment to train the

participants, and the training content usually includes

emotion recognition, behavior modification, attention

ability training, etc., so as to improve the cognitive

level and regulation ability of the participants to their

own emotional state. Mindfulness refers to the use of

meditation, mindfulness and other mental practice

methods to make the subject feel their own physical

and mental state in a state of mental concentration, so

as to improve their self-awareness and control ability.

As shown in equation (2).

(3) 0 3

−

ξφ





′



−φ φ



)

Behavior modification refers to intervening and

changing mental health levels through behavioral

changes. All articles in this study were measured

using a behavioral test.

2.3 Questionnaire Survey and

Experimental Measurement

Behavioral testing refers to the use of certain test

materials to induce or stimulate the behavior of the

subjects to observe their reactions and emotional

states, and then to investigate the psychological

changes of the subjects before and after the

stimulation. The behavior test method mainly

performs a series of operations on the subject to

observe its behavior state, and then judge the

psychological state of the subject. The behavioral test

method used in this study was mainly measured by

two methods: indoor behavior and outdoor behavior

of the participants. As shown in equation (3).

(k)

fks

−

ξφ









−









−φ φ





)

The subjective scale mainly refers to the

measurement of the subject's indoor and outdoor

behavior and emotional state, in order to examine the

psychological state of the subject before and after

stimulation, and then judge the psychological state of

the subject before and after stimulation. Subjective

scales mainly include the Depression Self-Rating

Scale (SDS) and the Anxiety Self-Rating Scale

(SAS).

The Combination of Subjective and Objective

Scales Measurement The combination of subjective

and objective scales refers to the measurement of the

subject's indoor and outdoor behaviors and emotions

before using psychometric techniques, and then using

psychometric techniques for data analysis. As shown

in equation (4).

(xy)

−

ξφ









−









−φ φ





)

Among them, SDS is a psychometric tool widely

used in mental health assessment, educational

assessment and clinical psychological counseling,

which contains 20 items, covering 7 dimensions such

as depression and anxiety, and adopts a 5-level

scoring method; SAS is a psychometric tool used to

describe individual differences, which consists of 5

items and uses a 5-level scoring method.

INCOFT 2025 - International Conference on Futuristic Technology

480

3 RESEARCH PROCESS

3.1 Research on the Application and

Limitations of Algorithms and

Deep Learning in Adolescent

Mental Health Intervention

The application of algorithms and deep learning in

adolescent mental health intervention mainly

includes the prediction and evaluation of individual

mental health indicators, and model construction and

model evaluation are the key links. The prediction

and evaluation of individual mental health indicators

mainly include the prediction of individual mental

health indicators, the comparison between the

predicted value and the real value, and the

construction of the prediction model of individual

mental health indicators. As shown in equation (5).

()

fy y



=−





)

Specifically, one is the prediction of individual

mental health indicators. Previous studies have shown

that algorithms and deep learning can be used as

auxiliary tools to assess adolescent mental health, so

as to provide data support and technical support for

adolescent mental health intervention. As shown in

equation (6).

()

fz zy



=⋅−





)

The second is the comparison between the

predicted value of the model and the real value.

Previous studies have shown that prediction models

based on deep learning models have better accuracy

and stability. The third is the evaluation of model

construction, that is, to determine whether the

adolescent mental health intervention system can

achieve the expected effect by evaluating the effect of

model construction. In recent years, with the rapid

development of artificial intelligence technology, it

has been widely used in adolescent mental health

intervention. By analyzing the problems and

shortcomings in the existing research, this paper

further explores the new models and methods of

algorithm and deep learning in adolescent mental

health intervention, and provides reference for

subsequent related research. This study used the

literature review method to systematically sort out the

application of algorithms and deep learning in

adolescent mental health intervention through

reading and combing of relevant literature, and

summarized the main problems that have been

studied: first, whether algorithms and deep learning

algorithm models are suitable for adolescent mental

health assessment, whether there is comparability

between the predicted value and the true value of

adolescent mental health indicators, and whether the

evaluation of the effect of model construction can be

realized. As shown in the Table 1.

Table 1: Accuracy of adolescent psychological intervention

orith

Data Source Accurac

Natural

Language

Processin

Social media 80%

Sentiment

Analysis

Online forums 75%

Machine

Learnin

Electronic

health records

90%

Deep Learning Brain scans 85%

Third, the algorithms and deep learning models

currently applied to adolescent mental health

intervention mainly use two methods: self-built and

outsourced. However, due to the limitations of

algorithms and deep learning, how to combine them

with existing adolescent mental health intervention

technologies to achieve complementary advantages is

also the focus of future research as shown in the Fig.1

Figure 1: Accuracy of adolescent psychological

intervention

Application and Evaluation of Algorithms and Deep Learning in Adolescent Mental Health Intervention

481

3.2 Research on the Prediction and

Intervention effect of Adolescent

Mental Health Indicators Based on

Algorithms and Deep Learning

In this study, a middle school located in Jinan City,

Shandong Province, was selected as the research

object due to its remote geographical location, so the

seventh-grade students were selected as the research

objects, and the mental health indicators of the

seventh-grade students were predicted and evaluated.

There are 416 students in the seventh grade, including

182 boys and 233 girls, and 595 boys and 617 girls in

the junior high school. The junior high school stage is

a key middle school, and the respondents are all

students in the key class of the middle school. In order

to better understand the mental health status of the

respondents and the application effect of algorithms

and deep learning in mental health intervention, this

study predicted and evaluated the mental health

indicators of the respondents on the basis of

questionnaires and psychological tests. As shown in

the Fig.2 .

Figure 2: Intervention results and expectations

The mental health indicators of the respondents

included: whether there was depression and anxiety,

whether there were obsessive thoughts and

compulsive behaviors, whether there were suicidal

tendencies, whether there were problems such as

Internet addiction and sleep disorders, and whether

there were social phobia. In order to investigate the

application effect of algorithms and deep learning in

adolescent mental health intervention, the

respondents were divided into two groups. One group

is the control group, which mainly includes school

and home, and the other group is the experimental

group, which mainly includes algorithms and deep

learning. Questionnaire method and experimental

method were used to predict and evaluate mental

health indicators in both groups.

3.3 Research on Adolescent Mental

Health Indicator Prediction and

Intervention System Optimization

This study uses experimental research methods to

focus on the empirical discussion of "the construction

and evaluation of adolescent mental health

intervention system". The detailed research process is

as follows:

This paper deeply analyzes the characteristics and

influencing factors of adolescent mental health

indicators, and constructs a comprehensive index

system suitable for this study. The system covers

multiple dimensions, including but not limited to

emotional state, social adjustment, learning pressure,

etc., and each dimension is composed of a series of

specific indicators.

In the data pre-processing phase, a variety of

statistical techniques are employed. Data

normalization is processed using the Z-score

normalization method, and its formula is: 𝑧=





where (x) is the raw data, is the mean, and is the

standard deviation. For the filling of missing values,

a variety of methods such as mean imputation and

regression interpolation were adopted. Outliers are

identified and handled appropriately using the

interquartile range (IQR) rule. μσ

Then, the Support Vector Machine (SVM)

algorithm was used to construct a prediction model

for adolescent mental health indicators. The core idea

of SVM is to find a hyperplane that maximizes the

spacing between positive and negative samples. The

decision function is in the form of: where is the kernel

function, is the Lagrangian multiplier, and b is the

bias term. 𝑓



𝑥



= sign

∑



𝑦



𝐾



𝑥,𝑥









𝑏



𝐾



𝑥,𝑥







In the empirical study of "Construction of

Adolescent Mental Health Indicator Prediction

Model", a variety of statistical methods were used for

analysis. One-way ANOVA is used to explore the

impact of a single factor on mental health indicators,

and its F-statistic formula is: 𝐹=

MSbetween

MSwithin

where is

the mean square between groups and the mean square

within groups. Multiple comparisons are used to

compare differences between different groups, such

as t-test, ANOVA, etc. Correlation analysis is used to

explore the degree of correlation between indicators,

and the MSbetweenMSwithin Pearson correlation

coefficient (r) is commonly used, and its formula is:

（7）.

INCOFT 2025 - International Conference on Futuristic Technology

482

𝑟=

∑

𝑥



−𝑥



𝑦



−𝑦









∑

𝑥



−𝑥





∑

𝑦



−𝑦













(7)

Subsequently, the "Evaluation of Adolescent

Mental Health Indicator Prediction Model" was

conducted. By comparing the prediction results with

the real values, the accuracy, recall, F1 value and

other indicators of the model are calculated, and the

performance and stability of the model are evaluated.

In addition, the ROC curve was plotted and the AUC

value was calculated to fully evaluate the predictive

power of the model.

Based on the above analysis results, the optimal

design and evaluation of the adolescent mental health

intervention system were carried out. According to

the prediction results of the model, targeted

interventions were formulated, and the effectiveness

of the intervention effect was verified by statistical

methods such as multi-group independent sample t-

test and paired-sample t-test.

Through the above complex statistical analysis

and model construction, this study ensured the

credibility and applicability of the research results,

and provided a scientific basis for the construction

and evaluation of adolescent mental health

intervention system.

4 FINDINGS

According to the literature selected in this paper, there

are three main forms of application of algorithms and

deep learning in adolescent mental health

intervention: one is the psychological intervention

system based on artificial intelligence algorithms,

which mainly establishes an algorithm model for

adolescent psychological problems through learning

from a large amount of data, the second is a

psychological intervention system based on artificial

neural network algorithms, which mainly uses neural

network algorithms to establish neural network

models for adolescent psychological problems, and

the third is a psychological intervention system based

on machine learning algorithms and deep learning

technology, which mainly establishes a neural

network model for adolescent psychological

problems through machine learning algorithms and

deep learning technology。 Among them, the

psychological intervention system based on artificial

neural network is the most widely used, followed by

the psychological intervention system based on deep

learning technology. At present, some studies have

verified the effectiveness of psychological

intervention systems based on artificial neural

networks in adolescent mental health intervention.

For example, some studies have evaluated the

treatment effect of depressed patients by comparing

the artificial neural network algorithm and the

psychological intervention system based on deep

learning technology, and the results show that the

artificial intelligence intervention system based on

deep learning technology is more effective than the

artificial neural network method in the treatment of

depressed patients. However, there are still some

problems and challenges in the intervention of

adolescent mental health using artificial intelligence

technology: First, the amount of data is small, and the

amount of data used for adolescent mental health

intervention in existing studies is small. Second, most

of them are experimental verification studies, and

there is a lack of empirical studies supported by large-

scale sample sizes. Third, there are privacy and

ethical issues in the intervention of adolescent mental

health with artificial intelligence technology, such as

the need to protect personal privacy data. In

conclusion, future research should focus on the

problems and challenges in the application and

evaluation of algorithms and deep learning in the field

of adolescent mental health intervention. In order to

better carry out adolescent mental health intervention,

it is necessary to establish an effective, standardized

and quantifiable adolescent mental health

intervention system and process from various aspects.

5 CONCLUSIONS

Algorithms and deep learning have great potential in

adolescent mental health intervention, which can

provide personalized mental health intervention

programs for adolescents. However, there are still

some problems in this technology, which are mainly

manifested in: first, there are limitations in the

application of algorithms and deep learning, and

second, there is a lack of generalizability and

universality in adolescent mental health intervention.

Future research can be carried out in the following

aspects: first, to further explore the potential of

algorithms and deep learning in theoretical research

to improve their effectiveness and applicability in

adolescent mental health interventions, second, to

establish unified data standards and operational

norms to realize the standardization, normalization

and standardized application of algorithms and deep

learning, and third, to strengthen the evaluation and

verification of the application effect of algorithms and

deep learning in adolescent mental health

Application and Evaluation of Algorithms and Deep Learning in Adolescent Mental Health Intervention

483

intervention, and to verify their popularization and

application effect through practice.

REFERENCES

Ambikavathi, V., Arumugam, P., & Jose, P. (2023).

Diabetes detection by data mining methods. Wireless

Personal Communications, 133(4), 2087-2104.

Bakirarar, B., Cosgun, E., & Elhan, A. H. (2023). Hybrid

model approach in data mining. Communications in

Statistics-Simulation and Computation

Bergami, G., Appleby, S., & Morgan, G. (2023).

Specification mining over temporal data. Computers,

12(9)

Brzychczy, E., Zuber, A., & Aalst, W. V. (2024). Process

mining of mining processes: Analyzing longwall coal

excavation using event data. Ieee Transactions on

Systems Man Cybernetics-Systems

Goedemans, R., & Prokic, J. (2023). Mining metrical data.

In J. VanDeWeijer (Ed.), Representing phonological

detail, pt 2: Syllable, stress, and sign (Vol. 33, pp. 133-

150).

Kannan, R., & Nandwana, P. (2023). Accelerated alloy

discovery using synthetic data generation and data

mining. Scripta Materialia, 228

Maggi, F. M., Marrella, A., Patrizi, F., & Skydanienko, V.

(2023). Data-aware declarative process mining with sat.

Acm Transactions on Intelligent Systems and

Technology, 14(4)

Petry, F., & Yager, R. (2023). Data mining using

association rules for intuitionistic fuzzy data.

Information, 14(7)

Wentzel, A., Floricel, C., Canahuate, G., Naser, M. A.,

Mohamed, A. S., Fuller, C. D., . . . Marai, G. E. (2023).

Dass good: Explainable data mining of spatial cohort

data. Computer Graphics Forum, 42(3), 283-295.

Yang, H. F., Zhou, L. C., Cai, J. H., Shi, C. H., Yang, Y.

Q., Zhao, X. J., . . . Yin, X. N. (2023). Data mining

techniques on astronomical spectra data - ii.

Classification analysis. Monthly Notices of the Royal

Astronomical Society, 518(4), 5904-5928.

INCOFT 2025 - International Conference on Futuristic Technology

484