Fin Irjmets1681152119

e-ISSN: 2582-5208
International Research Journal of Modernization in Engineering Technology and Science

( Peer-Reviewed, Open Access, Fully Refereed International Journal )
Volume:05/Issue:04/April-2023 Impact Factor- 7.868 www.irjmets.com
MENTAL ILLNESS AND TRAUMA PREDICTION BASED ON MACHINE

LEARNING
Reddi Prasadu*1, Sri Vidhya Karapa*2, Alekya P*3, Paramesh Undabatla*4,
Khasim Shaik*5, Rishika Karri*6
*1Assistant professor, Department of Information Technology, Anil Neerukonda Institute of Technology and
Sciences, Sangivasala, Visakhapatnam, Andhra Pradesh, India.
*2,3,4,5,6Department of Information Technology, Anil Neerukonda Institute of Technology and Sciences,
Sangivasala, Visakhapatnam, Andhra Pradesh, India.
ABSTRACT
In modern society, stress is more prominent. According to the World Health Organization, stress is a mental
illness with physical consequences. The effects of stress and depression may affect anybody, anywhere. Stress
is a part of life for everyone. Stress is a major indicator of how well your mind is functioning. The effects of
stress permeate all aspects of a person's being, from their feelings to their ideas to their actions. For user stress
detection, we advocated for a system based on Decision Trees (DT), Ridge Regression (RR), Random Forests
(RF), and K-Nearest Neighbors (KNN). Several factors' effects on consumers' stress levels were investigated in
this study. The dataset includes information from 2,958 persons who took part in surveys targeting a wide
range of ages. The data collection was derived from a well-crafted survey. We utilized the DT, KNN, and RF
classifiers to do the sorting. The accuracy metric was used for performance evaluation. Compared to RF, DT has
been found to provide higher accuracy.
Keywords: Survey Questions, Machine Learning, Stress Detection, K-Nearest Neighbor, Decision Tree, Random
Forest
I. INTRODUCTION
Stress is one of the major problems in many countries today due to the modern lifestyle. A small amount of
stress from a job or school can be assumed. Yet, sustained high-level pressure can be persistent and lead to a
variety of chronic conditions. Continuous stress can result in real physical problems, such as anxiety, insomnia,
muscle pain, hypertension, and a weakened immune system. The well-being of people is being compromised by
mental health issues. They are thought of as the primary factor in determining how well an individual is
progressing as their mental condition declines. Nowadays, customers can be pressured due to social
cooperation in unofficial organizations. The rapid escalation of mental health problems or stress has turned
into an incredible test for human welfare and the very fabric of life. It should be noted that the terms pressure
and unease are frequently used in opposition to one another. The main distinction is that the former is not
always linked to external signals or specific threats. Yet, stress is a fast reaction to regular demands and is
commonly believed to be more malleable than worry. In both stressful and anxious situations, the brain fails to
differentiate between the two, leading to the same bodily manifestations. These include a rapid heartbeat,
sweaty palms, and an upset stomach.
Figure 1: Impact of Stress on the human brain [18]
www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

[1064]
e-ISSN: 2582-5208
• Women are more likely than men to report showing physical and emotional signs of stress due to longer
workdays. However, the majority of people are unsure whether or not to seek proper therapy for it. An
American organization claims that 80% of employees experience work-related stress and that 42% of
employees and almost half of those who say they require counselling on stress management agree. The
Health and Safety Executive (HSE) reports that 54% of all workers have a chronic illness, and 44% of cases
involving occupational stress, misfortune, or discomfort as well as disability of all vocations have been lost
in 2018–19. The importance of stress identification is due to this.
Project Scope and Direction:
Finding the precise stress requires many machine learning methods to be combined. To gauge user anxiety,
decision trees are employed in machine learning algorithms [20]. K-Near Neighbors (KNN), Random Forest
(RF) [6]- [8]-[17], and the Decision Tree [17] are only a few of the machine learning methods used to foretell
anxiety levels. The majority of votes will determine the final classification.
II. RELATED WORK
Stress is a physiological response to mental excitement or actual challenges. Little amounts of stress could be
desirable, beneficial, and surprisingly effective. Unnecessary pressure, however, can result in serious harm and
raise the risk of illness.The unquestionable levels of mental instability and pain can be dangerously exacerbated
by high blood pressure, asthma, failure, ulcers, and stress. Several studies on stress identification are available;
they are described as follows.
1. Implementation of Stress Detection using SignalProcessing-
Noppadon Jatupaiboon and colleagues [1]. It was suggested to use a continuous EEG signal to organize
happy and unhappy sentiments motivated by old-style music and images. A worldly pair of channels
provides a desired outcome over the other territory does, and high recurrence groups provide a
preferred outcome over low recurrence groups do, when considering each pair of channels and
distinctive recurrence groups. They help to continuously improve the emotion grouping system by
applying insignificant EEG diverts. Creators use just one set of channels to implement a continuous
framework for bliss recognition based on the findings.Nermine Munla and others .[4] Describe the Auto
Driver data set (DRIVEDB). A few grouping techniques that included SVM-RBF, KNN, and RBF classifier
were investigated. Stress site was predicted using the SVM-RBF classifier with an accuracy of 83%.Adnan
Ghaderi and others [6] displayed the user's level of stress based on biological information from sensors
like heart rate, breathing rate, hand and foot movement, etc. Several features were extracted at various
intervals of time. KNN and SVM techniques were utilised for classification in order to determine the
precise stress level. Yuchae Jung and other .[7] Using multimodal sensors, assess mental tension and
anxiety as well as track physical alterations in various body areas. EEG, ECG, SPO2, BP, and RR sensors,
among others, were used to measure the level of mental anxiety. The SVM classifier was used to
categorise anxiety levels. Ronald Crescentini and others .[8] In the recently published study, the impacts
of an immersive virtual programme environment used for eight weeks of mindfulness-based meditation
were examined.Mario Salai et.al.[9] Presents the ideas and after effects of two investigations focused at
pressure location with a minimal effort of sensor like chest belt. In the gadget approval study (n=5), they
thought about pulse information and different highlights from the belt to those deliberate by a best
quality level gadget to evaluate the dependability of the sensor.Alexandros Zenonos et.al.[10] Investigate
the chance of utilizing such gadgets for temperament acknowledgment, zeroing in on workplaces. They
suggest a brand-new mind-set acknowledgment framework that can automatically identify five force
levels for eight various types of mind-sets. Provide also a mobile application called "Healthy Office" that
aims to promote systematic self-detailing and supply models with real-world data. Yoshiki Nakashima
and others .[11] Stress is categorized into three states: calm, focused, and also exhibit choice level
combination.In [14], Analyze the stress level of working people using the various sensors and classify the
stress level of individual based on the SVM, KNN algorithms. For this research extract the SWELL-KW
dataset and achieve the maximum accuracy. In [15] Create an enormous scope dataset of clients with self-
announced sorrow messages. Use different normal language handling (NLP) devices and methods to
uncover the etymological examples and the conclusions communicatedby these tweets. Finally, apply AI

[1065]
e-ISSN: 2582-5208
(ML) strategies to assemble conduct expectation devices utilizing the explained corpus. In [17], analyze the
stress level of college Students. Examine the various parameters of 206 students like exam pressure, and
recruitment stress. The dataset ware extracted from Jaypee Institute of Information Technology. For the
classification LR, NB, RF and SVM classifier was used. In [18], presented the design of stress detection
system to acquire the various human body signal based on the machine learning. For the simulation and
implementation used the MATLAB software. Physionet drivedb dataset was used for training and testing.
The result ware showed in terms of stress level classification accuracy using SVM algorithm. In [19],
presented the model for automatically recognizes the stress level of user in daily life based on the
XGBoost machine learning technique. In this research wearable stress and affect dataset was used for
stress analysis. For the feature selection electrodermal activity was used. The various evaluation
parameters ware examined such as accuracy, f1-score. In [20], designs were introduced to estimate user
stress levels with social media data. Facebook post was used as data. CNN was used to remove Facebook
posts. KNN and Transductive SVM were used to categorize users' posts and to estimate user stress levels
on a positive and negative basis.
2. Implementation of Stress Detection using Image& Video Processing-
Tong Chen et.al.[3] Introduced an HSI-based technique for the location of mental pressure. The solid
material-separating capacity of HSI was used in this investigation to separate and measure the measure of
blood chromophores (Hb and HbO2) utilizing the Beer-Lambert Law. The way where HSI signals are
acquired (caught picture) portrays this strategy as a without contact pressure location method.Andra
Adams et.al.[5] Investigate the impact of head movement from outward appearance in the view of
complex feelings in videos. They show that head movements convey enthusiastic data that is integral as
opposed to repetitive to the feeling content in outward appearances. At last, they show that passionate
expressivity in head movement isn't restricted to gestures and shakes and that extra motions, (for
example, head slants, raises and general measure of movement) could be helpful to computerized
acknowledgment frameworks.G. Giannakakis et.al.[13] Builds up a system for the location and
investigation of stress/tension enthusiastic states through video-recorded facial signals. An exhaustive
test convention was set up to incite orderly changeability in full of feeling states (unbiased, loose and
focused/anxious) through an assortment of outside and inner stressors. The examination was centered
basically around non-deliberate and semi- intentional facial signals to gauge the feeling portrayal all the
more impartially. Highlights under scrutiny included eye related occasions, mouth movement, head
movement boundaries and pulse assessed through camera-based photoplethysmography. A component
choice system was utilized to choose the strongest highlights followed by characterization
plansseparatingbetween pressure/uneasiness and unbiased states regarding a casual
state in each experimental phase. In addition, a ranking transformation was proposed utilizing self-
reports in order tomouth activity, head movements and camera-based heart activity achieve good
accuracy and are suitable as discriminative indicators of stress and anxiety.
Table 1: Summary table of stress identification
Authors Dataset Used Classifier Classification Accuracy
Adnan Ghaderi et.al.[6] PHYSIONET KNN, SVM Low, Medium, High 100 sec & 200
sec-98.41% 300
sec- 90%
Nermine Munla et.al. [8] DRIVEDB SVM-RBF, KNN, RBF Normal, Stressed Highest
accuracy SVM-
RBF-83%
Sriramprakash.S et.al.[14] SWELL-KW SVM Stressed, Normal 92.75%
Ravinder Ahuja et.al. [17] 206 students of JIIT Random forest,Highly

Naïve stressed, Stressed,Highest accuracy
Noida Bayes, SVM, KNN Normal SVM- 85.71%

[1066]
e-ISSN: 2582-5208
Md Fahim Rizwan et.al. [18] PHYSIONET SVM Stressed, Non- 98.6%
Stressed
Cheng-Ping Hsieh WESAD XGBoost Stressed, Non-Stressed 92.38%
S. M. Chaware et.al. [20] Information TVSM Positive, Negative 84.2%
extraction
fromFacebook
attributes
III. STRESS DETECTION APPROACH USING ML
Stress dataset, pre-processing, training, and testing data, stress detection model, and expected result are only
some of the major components of the overall stress detection system architecture, which is represented in
Figure 2.
Stress Dataset: Data is gathered throughout this time using various online stress scale questionnaires, etc.
Based on a standardized user questionnaire, the data is gathered. This dataset creates attributes and class
labels that incorporate stress estimation. The dataset's tuples each contain integer values for each attribute. A
total of 2958 male and female users participated in this study, and there were 9 questions on the questionnaire
for stress analysis. The participants are all between the ages of 18 and 50.
Table 1: Number of Participant
Male 1059
Female 1899
Pre-processing: At this phase, the input data undergoes a series of transformations designed to remove any
details that are irrelevant to the stress detection process. The user's name, residence, age, education, type of
employment, and workplace are all factors considered when conducting a survey. There are null values in
several fields. The dataset has undergone data pre-processing to exclude any extraneous data.
Splitting data: Once the data has been cleaned and organized, it may be split into a training set (consisting of
70% of the data) and a testing set (consisting of 30% of the data) for further analysis.
Stress Detection Model: To identify and categorize user mental stress in this phase, employ the K-Nearest
Neighbor, Decision Tree, and random forest machine learning algorithms. It's a stress level predictor.
DT, KNN and RF classifiers are trained using user stress level classification, which is data gathered from survey
participants. It is possible to classify stress level as a value ranging between 0-9.
Predicted Result: At this step, the model's performance in terms of Accuracy is evaluated.
Algorithm: Learning the Classifiers Input: Dataset D
Output: Trained the classifiers
Step 1: Data Exploration
Step 2: Data Cleaning
Step 3: Data Visualization
Step 4: Data Training
Step 5: Model Creation
Step 6: Performance Evaluation
for the dataset being used
Stress Level: The user's primary goal is to determine the system's stress index. After considerable data had
been gathered, the stress index value was calculated using physical and cognitive data. The user's stress index
value is used to determine the threshold value. The user's EEG data varies depending on their behaviour. Also,
the individual's threshold value must be calculated in order to determine their personal stress level.
Proposed Framework-To implement the mental illness and trauma prediction we first need to import python
libraries (pandas, numpy, seaborn, matplotlib, sklearn)

[1067]
e-ISSN: 2582-5208
For Stress Level Prediction we will use 3 algorithms:
1. Decision Tree Regressor
2.Random Forest Regressor
3. K-Nearest Neighbour
We will first examine the data we have and look for any null or missing values using the aforementioned
approaches. If discovered, the data will be cleaned before being shown for greater comprehension. The next
step will be data training, which involves dividing the data into training and testing data. Next, we'll train our
model using training data, and after it's trained, we'll make predictions. After making a forecast, we will check
these algorithms' accuracy and error rates to see how well they performed.
Dataset:
Data Cleaning: Since Question 7 from the dataset has null values, we will replace them with the mode.
The above is done for Q8 as well.

Data Visualization: Representing the data in the form of graphs.
In the same way, graphs are plotted for all the attributes in the dataset.
Data Training:

[1068]
e-ISSN: 2582-5208
Algorithms: Models used are:
• Decision Tree Regressor
• Random Forest Regressor
• K- Nearest Neighbour
Decision Tree-A decision tree is a supervised learning approach useful for tackling classification and
regression problems. It's a classifier structured like a tree, where the trunk and branches indicate the dataset's
attributes, and the leaves represent the classifier's ultimate verdict.
Random Forest-The random forest is an ensemble of decision trees used to make a more accurate
categorization. By employing bagging and feature randomization to generate each tree, it aims to create a forest
of trees whose collective prediction is more accurate than that of any individual tree.
As its name suggests, Random Forest is a classifier that utilizes many decision trees on various subsets of the
given dataset and then averages the results to improve the dataset's predicted accuracy. The random forest
does not rely on just one set of decision trees, but rather leverages the predictions many to arrive at a final
verdict.
KNN Algorithm-One of the simplest Machine Learning algorithms, K-Nearest Neighbor makes use of the
Supervised Learning approach. The K-Nearest Neighbors (K-NN) method classifies new data or cases based on
their assumed similarities to previously classified data or instances.

[1069]
e-ISSN: 2582-5208
IV. RESULTS

[1070]
e-ISSN: 2582-5208
The stress level values are predicted by the model for various algorithms. (DT, RF, RR, KNN)
We have developed a graphical user interface (GUI) that has a login/register portal. All the data is stored in the
SQLYog Database. After successful authentication, the user answers the survey questions. Based on this the
model is applied and predicts the stress level of the individual. If the stress level is greater than the threshold
then it indicates that the person has high stress. And therefore, stress management techniques popped up.
V. PERFORMANCE EVALUATION
Figure 3: Comparative analysis of both classifiers in terms of accuracy

After applying DT, RF, KNN, and Ridge regression algorithms, the accuracy, and root mean square are as
follows:
VI. CONCLUSION
In everyday life, mental stress is a significant concern. It is found many methods are employed to detect stress.
Also, there are a number of datasets that can be used to detect stress. Several classifiers are utilized, and the
accuracy of each is displayed in the summary table. This makes it feasible to recognize stress. According to the
current methodology, stress can be anticipated through facial reading, interviews, and other interactions in
which individuals analyze one another. In this study, a machine learning approach was used to create a system
for predicting user mental stress. Using DT and RF classifiers, the acquired data was classified and a mental
stress prediction was made. DT provides greater accuracy than the RF algorithm.

[1071]
e-ISSN: 2582-5208
VII. REFERENCES
[1] Setha Pan-ngum, and Pasin Israsena Noppadon Jatupaiboon, "Real-Time EEG-Based Happiness Detection
System," in Hindawi Publishing Corporation the Scientific World Journal Volume 2013 , 2013, p. 12.
[2] Omayya Murad Mohammad Malkawi, "Artificial neuro fuzzy logic system for detecting human emotions,"
in Human-centric Computing and Information Sciences, Springer , 2013.
[3] Peter Yuen, Mark Richardson, Guangyuan Liu, and Zhishun She Tong Chen, "Detection of Psychological
StressUsing a Hyperspectral Imaging Technique," in IEEE Transactions on Affective Computing, VOL. 5, NO.
4 , 2014.
[4] Mohamad Khalil, Ahmad Shahin ,Azzam Mourad Nermine Munla, "Driver Stress Level Detection Using HRV
Analysis," in International Conference on Advances in Biomedical Engineering (ICABME), 2015.
[5] M. Mahmoud, T. Baltrusaitis, P. Robinson A. Adams, "Decoupling facial expressions and head motions in
complexemotions," in Affective Computing and Intelligent Interaction (ACII), 2015.
[6] Javad Frounchi ,Alireza Farnam Adnan Ghaderi, "Machine Learning-based Signal Processing Using
Physiological Signals for Stress Detection," in 22nd Iranian Conference on Biomedical Engineering(ICBME
2015), Iranian Research Organization for Science and Technology, Tehran, Iran, 25-27 November 2015,
2015.
[7] Yuchae Jung & Yong Ik Yoon, "Multi-level assessment model for wellness service based on human mental
stress level," in Springer , 2016.
[8] Luca Chittaro ,Viviana Capurso ,Riccardo Sioni Franco Fabbro Cristiano Crescentini, "Psychological and
physiological responses to stressful situations in immersive virtual reality: Differences between users who
practice mindfulness meditation and controls," in Computers in Human Behavior , 2016.
[9] István Vassányi, and István Kósa Mario Salai, "Stress Detection Using Low-Cost Heart Rate Sensors," in
Hindawi Publishing Corporation Journal of Healthcare Engineering , 2016.
[10] Aftab Khan, Georgios Kalogridis, Stefanos Vatsikas Alexandros Zenonos, "Healthy Office: Mood
Recognition at Work Using Smartphones and Wearable Sensors," in The Second IEEE International
Workshop on Sensing Systems and Applications Using Wrist Worn Smart Devices , 2016.
[11] Jonghwa Kim,Simon Flutura Yoshiki Nakashima, "Stress Recognition in Daily Work," in Springer
InternationalPublishing, 2016.
[12] Mark A. Neerincx, Wessel Kraaij Saskia Koldijk, "Detecting work stress in offices by combining
unobtrusive sensors," in IEEE Transactions of Affective Computing, 2016.
[13] M. Pediaditis , D. Manousos , E. Kazantzaki , F. Chiarugi, P.G. Simos , K. Marias , M. Tsiknakis G. Giannakakis,
"Stress and anxiety detection using facial cues from videos," in Biomedical Signal Processing and Control ,
2017, pp. 89- 101.
[14] Prasanna Vadana. D, O. V. Ramana Murthy Sriramprakash.S, "Stress Detection in Working People," in
Procedia Computer Science, Elsevier , 2017.
[15] Wajdi Zaghouani, "A Large-Scale Social Media Corpus for the Detection of Youth Depression," in the 4th
International Conference on Arabic Computational Linguistics , Dubai, United Arab Emirates., November
17-192018.
[16] Gao Huang, Kilian Q. Weinberger &Alice X. Zheng Zhixiang (Eddie) Xu"Gradient BoostedFeature
Selection”," in International Journal of Engineering and Technology, 13 Jan 2019, p. 10.
[17] Alisha Banga, Ravinder Ahuja, "Mental Stress Detection in University Students using Machine Learning
Algorithms,"in Procedia Computer Science, Elsevier ,2019.
[18] Rayed Farhad ,Md Fahim Rizwan, et.al. "Design of a Bio signal Based Stress Detection System Using Machine
Learning Techniques," in International Conference on Robotics, Electrical and Signal Processing
Techniques(ICREST) 364, 2019.
[19] Y. Chen, W. Beh and A. A. Wu C. Hsieh, "Feature Selection Framework for XGBoost Based on Electrodermal
Activity in Stress Detection," in IEEE International.

[1072]

Fin Irjmets1681152119

Uploaded by

Copyright:

Available Formats

Fin Irjmets1681152119

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fin Irjmets1681152119

Uploaded by

Copyright:

Available Formats

e-ISSN: 2582-5208

International Research Journal of Modernization in Engineering Technology and Science

MENTAL ILLNESS AND TRAUMA PREDICTION BASED ON MACHINE

Figure 1: Impact of Stress on the human brain [18]

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

Ravinder Ahuja et.al. [17] 206 students of JIIT Random forest,Highly

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

The above is done for Q8 as well.

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

Figure 3: Comparative analysis of both classifiers in terms of accuracy

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

www.irjmets.com @International Research Journal of Modernization in Engineering, Technology and Science

You might also like