Scribd
Upload
25 views

Manu

Uploaded by

andrewsdj50
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
25 views

Manu

Uploaded by

andrewsdj50
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 26

A SEMINAR ON

BLOCKCHAIN-BASED EVENT DETECTION AND


TRUST VERIFICATION USING NATURAL
LANGUAGE PROCESSING AND MACHINE
LEARNING
BASED ON
IEEE ACCESS ON REAL-TIME EVENT DETECTION
ON SOCIAL DATA STREAMS

Volume 10,January
2022
By ZEINAB SHAHBAZI AND YUNG-CHEOL
BYUN
01 Introduction

02 Related Works

03 Proposed Event Detection System based on


Crisis Management and Social Media

CONTENTS 04
Analytics
Predictive Analysis for Event Detection

05 Performance Evaluation

06 Conclusion

07 References
INTRODUCTION
• Social media has increasingly become a vital source for reporting
disasters, including natural and human-made events.
• With the rapid growth of user-generated content, platforms like Twitter
offer a vast amount of real-time information. This presents both
opportunities and challenges for early event detection and crisis
management.
• Tools such as Ushahidi and new analytical methods are being explored
to enhance the accuracy and effectiveness of disaster response using
social media data.
RELATED WORKS
1. DISASTER MANAGEMENT CYCLE
Disaster events are divided mainly
into:
• Preparation and Mitigation: Action taken before a
• Response and Recovery: Action taken after a Disaster.
Disaster

Social media serves as a critical tool throughout these stages,


especially in the preparation and response phases.
ML analyzes social media data to identify and extract critical
information related to disasters.
2. SOCIAL MEDIA ACTIONABLE INFORMATION ON DISASTERS
• Social media is vital for gathering actionable disaster
information, but the data volume requires careful filtering to
ensure accuracy.

• Techniques like NLP and NER extract meaningful content for


reports and disaster response.

• One of the challenges is the lack of related documents to


extract the necessary information for creating response reports.
3. BLOCKCHAIN CONSENSUS MECHANISM IN EVENT
DETECTION
• Provides secure and decentralized records for event detection.
• Avoids authentication issues through trust and consensus
mechanisms.

Consensus Mechanisms:

• Proof of Work (PoW): Based on computing power and hash


value evaluation.
• Proof of Stake (PoS): Validators are chosen based on their stake
in the network.
• Proof of Authority (PoA): Trusted nodes manage block creation in
permissioned blockchains.
PROPOSED EVENT DETECTION SYSTEM BASED
ON CRISIS MANAGEMENT AND SOCIAL MEDIA
ANALYTICS
Cloud-Based Crisis Management System
• Objective: Enhance disaster management using social media
analytics within a cloud-based environment.

The architecture has four main components:


• Event Identification
• Automatic Reasoning
• Incident Monitoring
• Blockchain
Architecture:

• Event Identification: Real-time data extraction from social


networks.
• Automatic Reasoning: Intelligent techniques to extract and
process information.
• Incident Monitoring: Knowledge-based emergency processing
with sensory interfaces.
• Blockchain: Ensures security, transparency, and trust through a
proof-of-authority system.
Fig 1: Social media analysis architecture for event detection and management of crisis
EVENT DETECTION USING NATURAL
Representation LANGUAGE
and identification of
eventsboth automatic and manual triggering.
• Event detection involves
• Data Crawling Parameters: Location-based settings (e.g., window size,
predefined area) and social media configurations guide the crawler.
• The crawler collects content related to specific search terms.
• Multi-language detection: translation using Google and Microsoft APIs.
• The processed data is stored in a knowledge-based special keywords
database.
• The system analyzes the crawled information, applies semantic
analysis, and evaluates the number of events appearing in shared
These equations represent the number of events that appear in shared
content.
w represents the terms that appear in document t.
a represent the unobserved variable class.
Event Detection Using Blockchain
• Blockchain uses public keys for secure user identification,
making the system tamper-proof.
• In government services, these keys can verify identities, much
like social media or email verification.
• Smart Contracts: Automatically run on blockchain when
conditions are met, enabling secure, decentralized interactions.
• Event Detection & Aggregation: Separates event detection from
aggregation, with reports based on user input, securely
processed and stored on the blockchain.
• Data Security: Only essential user reports are stored, protecting
data and using it solely for event detection.
Fig 2: Relationship between persistent storage and smart contract in event detection
Event Detection Using Deep Learning

• Enhance disaster detection by analyzing social media posts with deep


learning.
• Classify Posts: Categorize posts into relevant humanitarian categories.
• Improve Responses: Use classified data to generate targeted
responses
• Several deep learning models, including CNNs, GRUs, and LSTMs, have
been used to classify crucial content during critical periods.

Challenges:
• Input Embedding: Fixed and unidirectional pre-trained embeddings
may not suit all disaster categories without tuning.process.
Predictive Analysis for Event
Detection
Enhance the system’s ability to predict and respond to events by
analyzing historical data and refining prediction algorithms.
• Learning Module:
⚬ Purpose: Adjusts the prediction algorithm to improve accuracy.
⚬ Uses historical data to identify patterns and relationships between
input (e.g., social media posts) and outputs (e.g., event
classifications).
• Prediction Algorithm:
⚬ Forecasts outcomes based on new user data and previously
learned patterns.
⚬ Applies the trained model to predict future events or outcomes.
• The system continuously monitors the prediction algorithm's
performance.
• It updates model parameters based on feedback and external
conditions to enhance accuracy.
• Continuous learning ensures the system refines itself with new data
and environmental changes, keeping predictions relevant and
accurate.

Fig.3: Event detection conceptual view for predictive model


learning
Performance
• Evaluation
F1-Score: A measure that combines precision and recall to
evaluate the accuracy of the model. It’s calculated using the
formula:

• It has been observed that F1-Score improves as the number of


posts increases.
• Also Enriching data helps in better topic identification.
Geotagging and Post Categorization
• Levels: Results were analyzed across different levels of
geotagging: PoI (Point of Interest), district, street, and city.
• Observation: PoI level achieved the highest F1-Score.

Fig.4: Different values of distribution of posts in four


levels
Blockchain
Results:
Objective: Evaluate how well the blockchain framework supports event
detection and manages transactions.
• Transaction Phases:
⚬ Local Synchronization: The blockchain first synchronizes
transactions within a specific geographical area.
⚬ Global Synchronization: It then synchronizes these transactions
across a broader global network to ensure that all messages are
delivered accurately.
Fig 5: Success rate based on threshold Fig 6: Success rate of the false event based on
impact attacker percentage based on threshold impact
CONCLUSION
• The system combines blockchain and machine learning to automate
crisis mapping and support humanitarian efforts.
• Utilizes social media and Twitter data to showcase its application.
• It includes event detection, classification, clustering, and trust
verification.
• This approach not only streamlines relief efforts but also helps in
maintaining the integrity of information during crises.
REFERENCE
[1] P. Williams, ‘‘Crisis
S
management,’’ in Contemporary Strategy.
Evanston, IL, USA: Routledge, 2021, pp. 152–171.

[2] L. Ardito, M. Coccia, and A. M. Petruzzelli, ‘‘Technological exaptation


and crisis management: Evidence from COVID-19 outbreaks,’’ RD
Manage., vol. 51, no. 4, pp. 381–392, Sep. 2021.

[3] J. Abbas, D. Wang, Z. Su, and A. Ziapour, ‘‘The role of social media in
the advent of COVID-19 pandemic: Crisis management, mental health
challenges and implications,’’ Risk Manag. Healthcare Policy, vol. 14, p.
1917, May 2021.
[4] S. Wang, Z. Yang, and Y. Chang, ‘‘Bringing order to episodes: Mining
timeline in social media,’’ Neurocomputing, vol. 450, pp. 80–90, Aug.
2021.

[5] S. G. Arapostathis, ‘‘A methodology for automatic acquisition of flood


event management information from social media: The flood in Messinia,
South Greece, 2016,’’ Inf. Syst. Frontiers, vol. 23, pp. 1127–1144, Jan.
2021.

[6] A. Kruspe, J. Kersten, and F. Klan, ‘‘Detection of informative tweets in


crisis events,’’ Natural Hazards Earth Syst. Sci., 2021. [Online]. Available:
https://nhess.copernicus.org/articles/21/1825/2021/
THANK YOU

You might also like