Machine learning and Deep learning Techniques in Mental Health

Diagnosis: A Survey
Abstract
In Today’s daily life, AI technology has widespread in many fields, where it plays an
increasingly important role in Health diagnosis. In Recent Years, Researches were done
regarding mental health diagnosis. A person’s mental well-being is his or her mental condition,
as well as an overview of his or her general environment. This Study identified various machine
learning and deep learning techniques to predict various mental disorders like OCD (Obsessive
Compulsive Disorder), PTSD(Post Traumatic Disorder), Schizophrenia, Depression. 15
articles on the diagnosis of schizophrenia, depression, anxiety, bipolar disorder, post-traumatic
stress disorder (PTSD), anorexia nervosa, and attention deficit hyperactivity disorder (ADHD)
were retrieved from various search databases.

Keywords
Machine learning, deep learning, mental health, CNN, SVM, Natural Language processing

Introduction
Mental health can be seen as a person’s emotional, psychological, and social well-being. It can
be harmed by various mental health conditions, which negatively influence a person’s
intellectual capacity, emotions, and social relationships. To combat these disorders, early
prediction of disorder is necessary.
Machine learning (ML) is a subfield of artificial intelligence (AI) that
deals with three problems: classification, regression, and clustering. It utilizes datasets and
algorithms to mimic how people learn while progressively improving performing tasks. ML
has been applied to multiple areas of psychological treatments and offers excellent potential
for predicting and treating mental health conditions and analogous health outcomes. Typically,
these algorithms require significant data to learn patterns and perform classification tasks. One
of the most widely applied ML approaches in the prediction of mental illnesses is supervised
learning. Different ML algorithms used are SVM, RANDOM FOREST, KNN.
Deep learning is a subset of machine learning. Deep learning plays vital
role in various fields. In deep learning, algorithms automatically and instantly learn beneficial
attributes from raw data. DL algorithms used here are convolutional neural networks (CNN),
deep belief networks (DBN), auto-encoders (AE), and recurrent neural networks (RNN).
 Natural Language Processing (NLP) is
ability of computer to understand and manipulate human language
in a way that is both meaningful and useful. In this paper NLP is
used to predict mental state of an individual based on social media
data like Facebook, Twitter etc. NLP powers many applications
that use language, such as text translation, voice recognition, text
summarization, and chatbots. Some of the applications voice-
operated GPS systems, digital assistants, speech-to-text software,
and customer service bots. NLP also helps businesses improve
their efficiency, productivity, and performance by simplifying complex tasks that involve
language.

Some of the Datasets used in predicting mental health illness are

Literature Survey
Questionnaire is created to collect data from individuals of all ages, responses and Verbal
interviews with closed questions have been used for a long time to identify and classify mental
disorders among people with mental illness symptoms. From this data ,we can observe that one
can identify different mental health issues like stress, depression, anxiety, and bipolar disorder.
For classifying depression by analysing the response to open-ended
questions, language models like LSTM and GRU-BiLSTM are employed where GRU-
BiLSTM outperforms LSTM. But in most cases, interviews and surveys contain close-ended
questions. Reducing features using appropriate feature selection techniques can also help
increase the accuracy of the classifiers [1]. Ensemble methods outperform single
machine learning models [2]. The performance of machine learning algorithms greatly depends
on the wise choice of questions. Also, the datasets are often large dimensional. However, survey
and interview-based data are very suitable for classifying mental diseases with higher
accuracy [3].
Table 2. Survey and Interviews’ data-based works.

Source Year Disease Age Location Approach Performance Limitations

[1] 2018 BD – China Different ML AUC score Small dataset that

models with 0.913 have restricted
feature the depth,
selection breadth, and
generalizability
of the research
findings.

[2] 2018 Multiple 18– India Clustering accuracy Authors

26 based 90%. suggested
Classification applying a deep
learning model to
increase the
accuracy and to
use more classes
on their utilized
dataset which
now contains
only three classes
including
Mentally
distressed,
Neutral, and
Happy.
Source Year Disease Age Location Approach Performance Limitations

[3] 2019 Depression 50+ UK Ensemble Logloss For specific age

techniques 0.251 group. Focusing
with feature only on the top 50
ranking features for XGB
and LGB models
may overlook
crucial factors in
predicting
depression in
older age.

[4] 2018 Depression – – LSTM f1 score 0.67 Low

performance,
Potential biases
in training data,
the small sample
size of 142
individuals, and
poor model
calibration.

[5] 2022 Depression – China GRU- f1 score 0.85 The authors

BiLSTM released this
dataset based on
162 persons and
applied ML to
this. Their sample
size and accuracy
rate can be
increased.

[6] 2020 AD – Israel ML models. accuracy Small dataset and

71.44% the accuracy is
low as well as the
results are not
validated with an
external dataset.

[7] 2020 Generic 9– Sweeden ML models AUC score Challenging &

Mental 15 0.739 low performing
Health model and the
model is not
Source Year Disease Age Location Approach Performance Limitations

accurate enough
for clinical usage

[8] 2019 AD – India ML models. Accuracy Profession-

82.6% specific data,
lacks justification
for feature
selection
technique and
proper
discussion.

[9] 2023 Generic – – Different ML 88.80% Use of Binary

Mental models accuracy. Classification
Health limits the
application in
multi-class,
complex mental
health issues.
Occupation-
specific data
lacks
generalization.

Conclusion
In this paper we have discussed about Machine learning and deep learning algorithms to predict
various mental disorders. In our study, we have included all the recent works that provide some
original experimental results in this domain. We have compared the works based on the
different algorithms and datasets.
In the future, we would like to consider more articles from different databases and
also provide a comparative analysis among related works. We will gather sufficient data for
our experiment and provide experimental analysis for our proposed integrated system in the
future. However, this paper will provide a solid overview of recent works and a future direction
for the new researchers in this field.
References
[1] C. Feng, H. Gao, X.B. Ling, J. Ji, Y. Ma, shorten bipolarity checklist for the differentiation
of subtypes of bipolar disorder using machine learning, in: Proceedings of the 2018 6th
International Conference on Bioinformatics and Computational Biology, in: ICBCB 2018,
Association for Computing Machinery, New York, NY, USA, 2018, pp. 162–166,
http://dx.doi.org/10.1145/3194480. 3194508
[2] M. Srividya, S. Mohanavalli, N. Bhalaji, Behavioral modeling for mental health using
machine learning algorithms, J. Med. Syst. 42 (5) (2018) 1–12.
[3] H. Yang, P.A. Bath, Automatic prediction of depression in older age, in: Proceedings of the
Third International Conference on Medical and Health Informatics 2019, 2019, pp. 36–44
[4] T. Al Hanai, M.M. Ghassemi, J.R. Glass, Detecting depression with audio/text sequence
modeling of interviews, in: Interspeech, 2018, pp. 1716–1720.
[5] Y. Shen, H. Yang, L. Lin, Automatic depression detection: An emotional audiotextual
corpus and a GRU/BiLSTM-based model, in: ICASSP 2022-2022 IEEE International
Conference on Acoustics, Speech and Signal Processing, ICASSP, IEEE, 2022, pp. 6247–6251
[6] T. Richter, B. Fishbain, A. Markus, G. Richter-Levin, H. Okon-Singer, Using machine
learning-based analysis for behavioral differentiation between anxiety and depression, Sci.
Rep. 10 (1) (2020) 1–12.
[7] A.E. Tate, R.C. McCabe, H. Larsson, S. Lundström, P. Lichtenstein, R. KujaHalkola,
Predicting mental health problems in adolescence using machine learning techniques, PLoS
One 15 (4) (2020) e0230389
[8] A. Sau, I. Bhakta, Screening of anxiety and depression among the seafarers using machine
learning technology, Inform. Med. Unlocked 16 (2019) 100149.
[9] J. Chung, J. Teo, Single classifier vs. ensemble machine learning approaches for mental
health prediction, Brain Inform. 10 (1) (2023) 1.
[10] Twitter suicidal data, 2024, Online at https://www.kaggle.com/datasets/
hosammhmdali/twitter-suicidal-data

[11] F. Yin, J. Du, X. Xu, L. Zhao, Depression detection in speech using transformer and
parallel convolutional neural networks, Electronics 12 (2) (2023) 328.