Tracing prodromal behaviour by analysing data patterns from social media with ensemble machine learning approach
Bibliographic Data
| ID | 4388665 |
|---|---|
| Authors | Deepali Joshi (0000-0002-8832-9294, Department of Technology Savitribai Phule Pune University India, corresponding author), Manasi Patwardhan (0000-0002-2775-3497, TCS Pune India) |
| Year | 2023 |
| Volume | 73 |
| Issue | 247 |
| Pages | 29-50 |
| Publication date | 2023-03-01 |
| Peer Reviewed | Yes |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | International Social Science Journal (JOURNAL) |
| Journal identifiers | ISSN: 0020-8701 • E-ISSN: 1468-2451 |
| Publisher | Wiley (PUBLISHER • GB) |
| DOI | 10.1111/issj.12368 |
| OpenAlex | W4311058032 |
| Language | EN |
| Citations received | 1 |
| References cited | 39 |
This paper presents a novel solution for tracing prodromal behaviour of Twitter users for early detection and prevention of mental illness. A very large number of people are using Twitter to share their daily happenings. This is creating a rich source of data portraying individual behaviour. Machine learning can be used to screen individual users who show prodromal behavioural change over a period and need to consult a psychiatrist. The system comprises of an ensemble model with multiple modules including classification and clustering in addition to lexicon-based approach, and, for the final verdict, a voting system which emphasises a better precision model is proposed. The model has been validated and tested on standard Self-Reported Mental Health Diagnoses dataset and promises to give a better result than the baseline
Big data · Cluster analysis · Data mining · Data science · Ensemble learning · Lexicon · Machine learning · Majority rule · Medical diagnosis · Social media · Tracing · Voting · World Wide Web · Artificial Intelligence · Complex Network Analysis Techniques · Computer Science · Medicine · Mental Health via Writing · Sentiment Analysis and Opinion Mining
| Unique citing works | 1 |
|---|---|
| Citations per year | 0,33 |
| Citation span | 2023 - 2023 (1) |
| Citation velocity | historical |
| Highly cited | No |
| Citation types | Neutral: 1 |