You are on page 1of 5

For You - Detection of Possible Vulnerability to

Depression before its Onset

Sanidhiya Sharma1, Sanjana Jain2, Vartika3, Ayushi Gupta4 Disha Mohini Pathak5

COMPUTER SCIENCE DEPARTMENT

1
sanidhiya.19b111007@abes.ac.in
2
sanjana.19b111002@abes.ac.in
3
vartika.19b111058@abes.ac.in
4
ayushi.19b111008@abes.ac.in
5
disha.pathak@abes.ac.in

COMPUTER SCIENCE DEPARTMENT


ABES ENGINEERING COLLEGE ,GHAZIABAD,INDIA

Abstract— Depression is a mental to get a simple, reliable score. This


disorder that can majorly impact a person's model also uses an emotional chatbot
health. It can affect the person's day-to-day and texts from various social media
life and the people associated with them. In platforms that enable the model to
the fast pace world, psychological issues determine the user's current emotional
such as anxiety, depression, and stress have state.
become common among the general public.
These disorders are a serious problem in
society as they can impact the social well- Using text processing, several machine
being of a person or society. It can learning algorithms can recognize a
disconnect the person from society and person's emotional state. Some of these
provoke suicidal thoughts. algorithms are Naive-Bayes, Support
Vector Machine, Logistic Regression, and
Linear Support Vector.
Hence, there is a requirement for an
automated system that can help the person
self-evaluate the mental score determining Keywords—Depression , ground truth ,
the vulnerability of the particular person to anxiety ,disability
the mental disorder. Various researchers
have developed several approaches and
I. INTRODUCTION
tests to compute the score determining the
person's mental health status. People today have become naturally
ambitious and seek every possible
In this paper, we will discuss a model opportunity to grow and now believes
that uses several machine learning himself to be an integral part to excel in
algorithms to determine the possible professional life. WHO observed that
vulnerability to depression, anxiety, or depression is the most common mental
stress for a person. The user can perform disorder, concerning more than 350
internationally acclaimed self-evaluating million individuals worldwide, and the
tests such as Beck Depression Inventory stringency of the problem has prompted
many health researchers to document the unknown.
field. Contrasting anxiety, depression, and
stress are challenging for the machine.
Accurate diagnosis therefore requires a
suitable learning algorithm. Healthy people
have a healthy brains and physical well-
being. Depression and anxiety are the
highest causes of disability worldwide,
with an assessed lifetime prevalence of
20%. Although the majority of individuals
with depression and anxiety are treated in
primary care, up to 50% of individuals are
Fig. 1. Depression types, causes , gender , race in
not identified or treated adequately. Given
last year
the damaging effects and costs associated
with untreated conditions, and the recent
increase in the prevalence of common
psychiatric disorders, proper diagnosis and
timely treatment of depression and anxiety
is an urgent priority. Depression is a II. METHODOLOGY
prolonged health concern that needs to be
The anxiety, depression, and stress score
constantly observed and handled. ML
are computed as occurring on the five
methods used for detecting and predicting
different levels of severity by using
depression depend upon subjective ground
multiple ML algorithms. These algorithms
truth obtained through psychological
can produce accurate results and are widely
questionnaires such as Taylor's Test and
the Beck Depression Inventory test. ML used to predict psychological disorders
models are conditioned to detect these among persons of different sex, age, and
scores and their outcome is compared with social background.
these scores to measure their precision of In this model, we have used the following
prediction. Acquiring ground truth from five machine-learning algorithms to
users with depression or any mental health compute the severity of depression. These
problem frequently over a prolonged
algorithms are support vector machine,
period of time is not endurable as periodic
decision tree classifier, naive Bayes, KNN,
requests to conduct questionnaires will
and random forest.
over time become an extra obligation
especially when the user is experiencing 2.1 Participants
intense symptoms. The model Is deployed on the internet.
Any user with an active internet connection
The current analyses in detecting and can use the model to determine their
deducing depression have depended on mental health status by performing a self-
periodic dimensions of depression status evaluating test and a text-based evaluation
(example: every month or weeks). of the user's emotions.
Additionally, while existing analysis has
evaluated ML techniques for detecting the 2.2 Questionnaire
existence of depressive signs, whether or
not these methods can capture We have used the BDI test as the
modifications in depressive signs is questionnaire for computing the scores that
will determine the level of severity of the divides the data by creating two separate
depression. It is a commonly used self- classes using a straight line. This line is
evaluation to compute the scale of known as the hyperplane. It is the line with
depression worldwide. The advantage of a maximum distance between the two
using BDI is that it is widely valid, classes.
flexible, and short. It consists of 21
questions. Once the user attempts all the Naive Bayes
questions, the model will return the level This algorithm assumes that the input
based on the answer the user. factors are independent of each other. It
works on the premise that features are
2.3 Classification independent. The formula for Naive Bayes
is as follows:
The machine learning algorithms were
implemented using the Python
programming language using pycharm
IDE. This model will determine the mental
health status of a person by using a BDI
test and text analysis. 70% of the datasets
were used to train the model, while the rest
were used as test sets.

Below we have described the principle


behind the ML algorithms used in this
model. Fig2. Naives Bayes Formula

Decision Tree
This algorithm uses the tree as the data K-Nearest Neighbour
structure to make decisions. This is KNN gathers the data and classifies recent
particularly used for making predictions, information points. Then it arranges the
and they are easy to interpret, and the tree gathered information to the closest
is a stable data structure. The algorithm neighbor. It is widely used because of its
enables the user to cover both the simplicity and adequacy.
classification and regression aspects of the After implementing different classifiers,
problem. we computed the F1 score for the model. It
was found that the Random Forest
classifier returned the most accurate result
among all five different machine learning
Random Forest algorithms.
It is an advanced version of the Decision
Tree algorithm. It creates several decision
trees by randomly creating a subset of the
training dataset. Then it accumulates the result
from all th decision trees and determines the final
class of objects.

Support Vector Machine


This algorithm is used for performing
both regression and classification. It
increase body dopamine
hormones .According to his preferences
which will divert his mind and help him
with his anxiety. Users can also consult
with doctors as there will be doctors
information available. There will be a
chatbot available with which a person can
share his own feelings without any fear of
being judged.

Fig. 3. Methodology Class Diagram

IV. PROPOSED WORK


To construct ML models that are capable
of deriving accurate predictions from long- Fig. 4. Application overview [use case diagram].
term data with limited ground truth
acquisition, data needs to be processed and
aggregated without compromising key
behavioral information during different V. FEATURES
time periods that could be useful in
1. Determine the anxiety and
detecting and predicting depression.
depression levels of the user by
Therefore, we extract finely detailed
performing multiple internationally
features to capture behavioral markers in
acclaimed tests which are known as
different time windows with varying
ground truth results.
granularity throughout the day, week, and 2. A ML model made on specially
semester. Although this step results in a researched questions and daily
greater number of features than the number diary entry, a score will be
of samples , the hierarchical and generated depicting different level
incremental modeling component and of anxiety and depression.
stable feature selection in the pipeline are 3. It has an automated chatbot with
both effective in identifying the most which users can share their
crucial features, i.e., features that are thoughts and feelings. This will
selected most often in different validation. provide a user with the chance of
If the user scores below the required score, sharing their thoughts without fear
then he will be suggested some activities to of being judged.
4. It will provide some optional datasets to verify our technique's efficiency
activities that will allow the user to and efficacy with more precision.
divert their mind from a stressed VIII. REFERENCES
situation and provide them with a
cool-off period. [1]https://dl.acm.org/doi/fullHtml/
5. If the score is below a certain level 10.1145/3422821#:~:text=ML
that it will recommend a %20methods%20used%20for
consultation to the user. %20detecting,measure%20their
6. Instead of doing long ground truth %20accuracy%20of%20prediction.
tests ,that take up to twenty minutes
average , a set of specially curated
set of questions would be asked
periodically to get regular analysis
and check the progression in
health. This will be time efficient
and people with severe symptoms
would not attempt such long time
taking tests ,This is the key feature
that is does not exists in the past
systems made.

VI. CONCLUSION
The main objective of this project is to
construct a machine learning model that
will compute the anxiety and depression
score of the user by performing
internationally acclaimed psychological
tests. It will enable the user to determine
the current anxiety level and based on the
severity of the vulnerability to depression
the model will provide certain activities to
the user to divert their minds and provide a
cool-off period. This will provide an easy
method to keep track of one mental health.
It will also provide an option to share your
feelings and thoughts with an automated
chatbot.
VII. FUTURE SCOPE
In future work, we intend to use another
procedure to extract paraphrases from
more additional types of emotional
features. Also, we plan to use more

You might also like