Research Degree Committee

Presentation
On
A Generation Perspective :
AI Innovation in Mental
health care via Questionnaires
and Social Media
School of Doctoral Research And Innovation,
GLS University,
Ahmedabad, Gujarat
Presented By: Supervisor Name:
Name:Sanket Shah Dr.Tripti Dodiya
Roll No - PH231069
Table of contents
01 02 03
Introduction Study of Mental Motivation
Disorders

04 05 06
Literature Survey Insights Objectives
Reviews

07 08 09
Research Methodology Conclusion References
01
Introduction
What is Artificial Intelligence?

● AI, or Artificial Intelligence,

refers to the development of
computer systems that can
perform tasks that typically
require human intelligence.

Key Component of AI
Application of Artificial Intelligence
02
Study of Mental
Disorder
What is the Status of Mental Healthcare in India?

mindvoyage.in/depression-in-india-latest-statistics/ https://scroll.in/article/1050928/high-mental-healthcare-
https://miro.medium.com/v2/resize:fit:1400/format:webp/
costs-are-pushing-indian-families-into-poverty
1*cVxbOvRx9Lk3q1elYnBe2w.jpeg
 AI and Mental Health Care

Enhanced Personalized Virtual Therapists

Diagnosis Treatment and Chatbots
and Screening

Continuous Challenges Research

Monitoring and and Ethical and
Support Considerations Development
AI and Mental Health Care (Cont..)
● Mental health issues like depression and
anxiety are on the rise globally, especially
among the youth, possibly due to career
pressures.
● Detecting these problems early is crucial for
effective treatment.
● Traditionally, symptoms are identified
through clinical consultations or
questionnaires, but social media data offers a
new avenue for detection.
● This research aims to investigate the role
of AI in mental health care through social
media, focusing on generational
differences.
● With the use of AI to detect mental
disorders by analyzing both questionnaire https://www.amazingathome.com/rethinking-social-networks/

responses and social media activity.

03
Motivation
How Mental Health's Impact on Celebrities Lives
How Mental Health's Impact on Students Lives
On-Site Study at Civil Hospital, A’bad
Sample of Major Depressive Disorder Symptom in Patient

Dr.Harshil Shah
Consultant Neuro-psychiatrist
(Senior Resident)MBBS, MD,
Neuro-psychiatry at B.J Medical College and Civil
Hospital, Ahmedabad
04
Literature Review of
Existing System
Literature Review of Existing System
Title Tracking Public Depression from Tweets on
COVID-19 and Its Comparison with Pre-pandemic Time,2022,[Md Khayrul Bashar]
Classifier Naive Bayes, Random Forest, Linear Support Vector Machine, Logistic Regression
Dataset These datasets were extract using standard Twitter API.
Dataset 1: This first tweet dataset was collected from Detecting-Depression-in-Tweets,
which is collected
it contains 4391 annotated tweets, with 3488 “depressive” and 903 “non-depressive” tweets.
Dataset 2 & 3: The tweets in these datasets were collected from Japanese people
Total number of tweets in these datasets were 5965 and 6432, respectively.
Measure F1- Score and Accuracy
Feature Bag – of – Words (BOW)
Extraction
Conclusion The Logistic Regression classifier using the bag-of-words (BOW) feature demonstrated
promising performance on the Twitter dataset, achieving an 83.757% weighted F1 score
and 84.977% accuracy, outperforming the NB, LSVM, and RF models
Future Scope • Perform detailed content analysis to determine whether the source of depression is
predominantly: Economic problems, Family problems, Lack of social awareness
• Explore inter-country comparisons to analyze trends in depression across different
countries.
Literature Review of Existing System
Title Improving Early Detection of Suicide Ideation using Oversampling Methods,2023, [Raje
ev
Verma Avinash Chandra Pandey]
Classifier Naive Bayes, Random Forest, Recurrent Neural Network, Ensemble Learning
Dataset This dataset has been taken from Twitter and comprises 1298 samples.
There are three class labels ‘0’, ‘1’, ‘2’ in the dataset, where ‘0’ corresponds to non-
suicidal, ‘1’ to nearly suicidal, and ‘2’ to suicidal.
This dataset comprises 232074 posts collected from two subreddits on the Reddit
platform: "SuicideWatch" and "depression". In this paper, a subset comprises 5767
samples of suicide intent and 5837 non-suicide intent have been used.
Measure Accuracy,Precision,Recall,F1-Score
Feature Extraction CV,TF-IDF,N-gram
Conclusion Random forest perform better in terms of accuracy, precision, recall, and F1 score
across both datasets and various feature extraction methods.(0.78 to 0.87)
Future Scope Model based on Bert and other RNN models can also be used to automatically extract
features and improve the Accuracy
Literature Review of Existing System
Title A Machine Learning based Depression Analysis
and Suicidal Ideation Detection System using
Questionnaires and Twitter,2019,[Swati Jain, Suraj Prakash Narayan, Rupesh Kumar
Dewang, Utkarsh Bhartiya, Nalini Meena and Varun Kumar]
Classifier Logistic Regression,Random forest Classifier,Support vector Machine,Decision Tree
Classifier, XGBoost
Dataset Dataset 1: Real time data is collected from students and parents by making them fill
questionnaires.[619 responses were recorded]
Dataset 2: Twitter and Reddit dataset
Measure Dataset 1: Accuracy(XGBoost-83.87)
Dataset 2:Accuracy(Logistic Regression – 86.45)
Feature Extraction TF-IDF, Weighted Word Count
Conclusion Major factors of depression among the age group of 15-29 which we found during the
course of the project are parental pressure, love, failures, bullying, body shaming,
inferiority complex, exam pressure, peer pressure, physical and sexual abuse etc.
Future Scope Analyzing social media behavior can help develop methods for detecting depression at
scale.
Literature Review of Existing System
Title A peek into one’s emotion through pen: analyzing
individual’s emotion through their thoughts
expressed on various social media platforms using ensemble classifier,2021,[Shalini
Shree, Arshdeep Singh Chudhey]
Classifier Logistic Regression, Naive Bayes classifier, Random Forest Classifiers, Voting
classifier
Dataset Emotions dataset for NLP
There are 20000 samples - 16000 for training, 2000 for validation, and another
2000 for testing.
Measure Accuracy [Voting classifier- 80.15%]
Feature Extraction TF-IDF,Word2Vec,Doc2Vec
Conclusion Research is not only limited to social distress but can help to generate effective email
and social media writings as well.
Future Scope • Various other ensemble classifiers like stacking classifiers can also be used to
check the efficancy
• GUI based Desktop application that can be used by an individual to predict the
emotion involved.
Literature Review of Existing System
Title Depression Prediction from Combined Reddit
and Twitter Data using Machine Learning,2023, Lovin D’Cruz,Vivishwan Dubey,Prof
. Pallavi Thakur]
Classifier Naïve Bayes, Random Forest, Support Vector Machine, Logistic Regression
Dataset Combined dataset of posts from Twitter and Reddit. This dataset included two
columns: one for the user’s post and another indicating whether the user was
depressed (1) or not (0) based on the content of the post.
Measure Accuracy Score,Precision_Score,F1_Score,recall
[For the Random Forest classifier, the precision was 0.75552431 and 0.8030666,
the recall was 0.8206415 and 0.73363975, and the f1 score was 0.78673779 and
0.76678486.The support was 4583 and 4569 for the two classes.]

Feature Extraction Tokenization, Lemmatization

Conclusion In conclusion, this paper aimed to build a machine learning classifier model that can
predict depression from a user’s post on social media platforms Twitter and Reddit.
Future Scope • Developing a tool for clinicians to use in their practice or integrating it into social
media platforms to provide support for at-risk users.
• Incorporating additional features such as user demographics and psychological
traits could improve the performance of the model.
Literature Review of Existing System
Title Depression Detection from Bangla Facebook Status
using Machine Learning Approach,2019, [Masum Billah,Enamul Hassan]
Classifier Multinomial Naïve Bayes, Logistic Regression, Linear SVC, Logistic Regression, Nu
SVC
Dataset Facebook statuses of individuals were collected by using own developed crawler where
user tokens were used. Our small dataset contains only 1000 status with Depressed
or Not Depressed tag. [50 facebook user]Among them 800 status are used for
training and 200 status is used for testing purpose
Measure F-Measure,Precision,Recall
Feature Extraction Unigram, Unigram and Emoticons, Bigram and emoticons, Unigram and Emoticons
with TF-IDF values, Unigram + emoticons as well as depressive words.

Conclusion The highest 77.96 % accuracy was found for SGDC with unigram + emoticons as
feature.
Future Scope • The work would be to build the model using neural network for better accuracy
• After detecting depressed people on social media, online counseling service would
be offered for them where free online counseling will be provided
Literature Review of Existing System
Title Depression detection using machine and deep learning models to assess the mental health o
f social media users,2022,[
Smita Ghosh1, Sneha Ghosh2, Diptaraj Sen2 and Pramita Das]
Classifier Multinomial Naive Bayes, Support Vector Machines, Decision Tree, K-Nearest
Neighbours, Long Short-Term Memory (LSTM), and the Universal Sentence Encoder
(USE)
Dataset Sentiment140 [Dataset 1]
Sentimental Analysis for Tweets [Dataset 2]
Twitter dataset [Dataset 3]
Measure Accuracy,F1-Score
Conclusion Based on the F1 scores provided in the conclusion, the Universal Sentence Encoder
(USE) algorithm appears to perform the best on both Dataset 1 and Dataset T, achieving
high F1 scores of 0.99 and 0.98, respectively.
Future Scope • Retrieve user data from a social network via an API and create a dummy social network
using OOP for deploying the MARS recommendation system
• Exploring advanced models like Transformers for depression detection, acquiring more
labeled datasets to enhance classifier evaluation in the study.
05
Survey Insights
Survey Insights
● Limited Scope of Social Media Platforms
• The study focuses on analyzing data from Twitter and Reddit. Future research could explore a broader
range of social media platforms to understand how depression manifests and is expressed across different
platforms, each with its unique user demographics and communication styles.[3]
● Incorporation of Multimodal Data
• The paper primarily focuses on text data from social media posts. Including other modalities such as
images, videos, or audio data could provide a more comprehensive understanding of mental health
indicators and enrich the predictive models.[3]
● Integration with Clinical Practices
• While the study aims to inform interventions and support for individuals at risk of mental health
challenges, there could be a gap in exploring how the predictive models can be integrated into clinical
practices or digital mental health interventions for real-time monitoring and support.[3]
● Cross-Linguistic Validation:
• The Study highlights the lack of prior research on depression detection from Bangla text, emphasizing the
need for cross-linguistic validation of existing models. Future studies could focus on validating and
adapting depression detection models developed in other languages to Bangla, considering cultural and
linguistic differences that may impact the detection accuracy.[24]
● Ethical Considerations:
• Given the sensitive nature of mental health data, further research could explore ethical considerations
related to data privacy, consent, and the responsible use of predictive models in mental health
interventions
06
Objectives
Objectives

● Investigate the role of AI in mental health care through social media, focusing on
generational differences.
● Investigate how AI can analyze social media questionnaires to assess and improve student
health.
● Leverage AI to analyze unwanted comments in both English and Gujarati on social media,
focusing on identifying and mitigating anxiety and depressive content.
● Acquiring real user data to enhance the analysis based on actual user profiles and posts.
07
Research
Methodology
Research Methodology
Research Methodology
1. Literature Review:
• Review of mental health-related papers. Understand the current state of research, identify
challenges, and explore successful methodologies and models.
2. Define Objectives:
• Analyze attitudes and perceptions of AI tools across different generational cohorts.
• Focus on Gen Z, Millennials, Gen X, and Baby Boomers.
• Identify generational differences in engagement and perceptions of AI tools.
3. Data Collection:
• Mixed-method approach: Combines qualitative and quantitative data.
Qualitative Methods
• Content Analysis: Analyze social media posts, comments, and discussions to identify
common themes and sentiments.
Quantitative Methods
• Surveys and Polls: Conduct surveys and polls targeting generational cohorts using
platform-specific features.
Research Methodology
• Data Collection:
• Participant Selection
Generational Cohorts
1. Gen Z (born 1997-2012)
2. Millennials (born 1981-1996)
3. Gen X (born 1965-1980)
4. Baby Boomers (born 1946-1964)
• Inclusion Criteria: Active social media users, diverse demographics, varying experience
with AI tools.
• Sample Size: 100 participants per cohort for surveys, sufficient social media posts for
content analysis.
• Data Collection Sources:
• Platforms: Facebook, Twitter, Instagram, Reddit.
• Tools: APIs, web scraping tools, platform analytics (e.g., Tweepy, Facebook Graph API,
BeautifulSoup).
Research Methodology
4. Data Pre-processing
• Data Cleaning: Remove irrelevant data, duplicates, noise reduction.
• Tokenization: Break text into tokens.
• Normalization: Standardize text (lowercasing, removing punctuation).
5. Feature Extraction
• TF-IDF Vectorization: Quantify the importance of terms.
• Sentiment Scores: Assign sentiment scores using pre-trained models.
6. Classification and Clustering
• Qualitative Data: Apply natural language processing (NLP) techniques and machine learning
classifiers to categorize sentiments and themes in social media posts.
• Quantitative Data: Use statistical classifiers to analyze survey data, identifying significant
differences between generational cohorts.
7. Comparative Analysis
• Statistical Comparison: Compare engagement levels, sentiment scores, themes across cohorts.
• Visualization: Create visualizations (e.g., heat maps, bar charts) to illustrate differences and
trends.
Research Methodology
Other Points
Ethical Considerations
• Informed Consent: Ensure compliance with social media platforms' terms of service and data
privacy regulations.
• Data Privacy: Anonymize data to protect user privacy.
• Bias Mitigation: Address potential biases to ensure fair representation.
Limitation
• Self-Reporting Bias: Acknowledge potential bias in self-reported survey and poll data.
• Social Media Sample: Recognize that findings may not be generalizable due to the focus on
social media users and potential for fake data.
08
Conclusion
Conclusion
● This research highlights the pivotal role of AI in advancing mental health care through
social media, with a focus on understanding generational differences.
● By employing a mixed-method approach that integrates qualitative and quantitative data
from social media platforms, we have gained insights into how different generational
cohorts perceive and interact with AI-driven mental health interventions
10
References
References
1. K. Chadaga et al., “SADXAI: Predicting Social Anxiety Disorder using Multiple Interpretable
Artificial Intelligence Techniques,” SLAS Technology, vol. 29, no. 2, p. 100129, Apr. 2024, doi:
10.1016/j.slast.2024.100129.
2. R. Verma and A. C. Pandey, “Improving Early Detection of Suicide Ideation using Oversampling
Methods,” IC3 2023: 2023 Fifteenth International Conference on Contemporary Computing, Aug.
2023, doi: 10.1145/3607947.3608003.
3. L. D’Cruz, V. Dubey, and P. Thakur, “Depression Prediction from Combined Reddit and Twitter Data
using Machine Learning,” 2023 2nd International Conference for Innovation in Technology
(INOCON), Mar. 2023, doi: 10.1109/inocon57975.2023.10101174.
4. https://data.mendeley.com/datasets/4jycfwhb4g/2
5. https://www.kaggle.com/datasets/shariful07/student-mental-health
6. H. Han, “Fuzzy clustering algorithm for university students’ psychological fitness and performance
detection,” Heliyon, vol. 9, no. 8, p. e18550, Aug. 2023, doi: 10.1016/j.heliyon.2023.e18550.
7. E. S. Mohamed, T. A. Naqishbandi, S. A. C. Bukhari, I. Rauf, V. Sawrikar, and A. Hussain, “A hybrid
mental health prediction model using Support Vector Machine, Multilayer Perceptron, and Random
Forest algorithms,” Healthcare Analytics, vol. 3, p. 100185, Nov. 2023, doi:
10.1016/j.health.2023.100185.
8. https://adaa.org/sites/default/files/GAD-7_Anxiety-updated_0.pdf
References
9. https://www.kaggle.com/datasets/krishbaisoya/tweets-sentiment-analysis
10. https://www.kaggle.com/datasets/gargmanas/sentimental-analysis-for-tweets
11. https://www.kaggle.com/datasets/vaibhavshahds/twitter-dataset
12. S. Ghosh, “Depression Detection using Machine and Deep Learning Models to Assess Mental Health of
Social Media Users”, Nov. 2022, doi:10.5121/csit.2022.122101.
13. P. Mann, E. H. Matsushima, and A. Paes, “Detecting Depression from Social Media Data as a Multiple-
Instance Learning Task,” 2023 2nd International Conference for Innovation in Technology (INOCON),
Oct. 2022, doi: 10.1109/acii55700.2022.9953811.
14. W. P. Aulia and A. Purwarianti, “Early Risks Depression Prediction of Indonesian Twitter Users Through
Indonesian Text Using Transfer Learning and Linguistic Metadata Features Approaches,” 2022 9th
International Conference on Advanced Informatics: Concepts, Theory and Applications (ICAICTA),
Sep. 2022, doi: 10.1109/icaicta56449.2022.9932951
15. R. S. Skaik and D. Inkpen, “Predicting depression in Canada by automatic filling of Beck’s Depression
Inventory Questionnaire,” IEEE Access, vol. 10, pp. 102033–102047, Jan. 2022, doi:
10.1109/access.2022.3208470.
16. https://github.com/swcwang/depression-detection/
17. M. K. Bashar, “Tracking Public Depression from Tweets on COVID-19 and Its Comparison with Pre-
pandemic Time,” 2022 7th International Conference on Intelligent Informatics and Biomedical Science
(ICIIBMS), Nov. 2022, doi: 10.1109/iciibms55689.2022.9971633.
References
18. A. C. De Oliveira, E. J. S. Diniz, S. Teixeira, and A. S. Teles, “How can machine learning identify
suicidal ideation from user’s texts? Towards the explanation of the Boamente system,” Procedia
Computer Science, vol. 206, pp. 141–150, Jan. 2022, doi: 10.1016/j.procs.2022.09.093.
19. https://maic.qld.gov.au/wp-content/uploads/2016/07/DASS-21.pdf
20. https://www.irjet.net/archives/V10/i7/IRJET-V10I742.pdf
21. https://www.kaggle.com/datasets/praveengovi/emotions-dataset-for-nlp
22. P. Mann, A. Paes, and E. H. Matsushima, “See and Read: Detecting Depression Symptoms in Higher
Education Students Using Multimodal Social Media Data,” arXiv (Cornell University), Jan. 2019,
doi: 10.48550/arxiv.1912.01131.
23. S. Jain, S. P. Narayan, R. K. Dewang, U. Bhartiya, N. Meena, and V. Kumar, "A Machine Learning
based Depression Analysis and Suicidal Ideation Detection System using Questionnaires and Twitter,"
in Proc. of the IEEE, 2019,doi:10.1109/SCES46477.2019.8977211.
24. M. Billah and E. Hassan, "Depression Detection from Bangla Facebook Status using Machine
Learning Approach," International Journal of Computer Applications, vol. 178, no. 43, pp. 1-5, Aug.
2019.
Thank You