Scribd
Upload
14 views14 pages

Paradox

Uploaded by

Vipul Ghodake
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
14 views14 pages

Paradox

Uploaded by

Vipul Ghodake
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/ 14

Paradox

Impact-Metrics
Indian Institute of Technology (Banaras Hindu University)

Tsunami Risk Prediction and Risk Zone Classification

Analyzing Earthquake and Tsunami Data (1995-2023)

Team Name :- Paradox


Members :-
LINKS:-
Hariom Sharma
Dashboard Link
Saurabh Tripathi
Data Set
Parth Nuwal
GitHub
Vipul Ghodake 1
Paradox Project Overview
Objective: Assess high-risk earthquake zones, predict
tsunami risks, and optimize disaster preparedness.

Dataset: Global earthquake data (1995–2023)

Goals:
• Identify high-risk zones.
• Build predictive models for tsunami occurrences.
• Visualize key insights for stakeholders.
Project deliverables: A unique identifier added to the
dataset for tracking individual earthquake events.

- Predictive model
- Classification of regions into Risk zones(high, medium, low)
- Dashboard for analysis

2
Paradox
Dataset Features

• Metadata: Magnitude, Depth, Date/Time, Location. Technology Stack

• Impact Metrics: CDI, MMI, Tsunami Indicator, Alert Level.

• Geographical Data: Continent/Country, Latitude/Longitude.

• Additional Metrics: Significance Score, Station Distance, Gap.

• Data Set Link (Cleaned data):- https://


docs.google.com/spreadsheets/d/1BXelepsoGZHqAjTq_4ug72Lj
mvA7_m9JFRhYGlVP8rQ/edit?usp=sharing

3
Paradox Methodology
Data Cleaning: Steps: Flow Chart
• Downloading data and preprocessing in Google
Colab.
• From date-time data set, we removed year data
• Removed unnecessary data from data set
• One Hot encoding categorical features.
• Standardizing mixed units into a single
consistent format.
• Handling outliers to improve data integrity.

Classification:
• Applied DBSCAN to cluster data points on spacial
density.
• Focused on spatial patterns to enhance
understanding of regional risk zones

4
Paradox
Risk Zone Classification

Method:
Waited score classification based on weight score of Median magnitude Frequency
over 5 years (freq_5years)
Major earthquake frequency (freq_major_5years)
Ratio of occurrences (ratio_5years) Depth.

Results:
• High Risk: Frequent, high-magnitude, low depth, high ratio of occurence events.
• Medium Risk: Moderate seismic activity, medium parameter.
• Low Risk: Infrequent or low-impact zones lower parameter except for depth.

5
Predictive Modeling
Models Used:
• XGBoost: Optimized gradient-boosting Sequential ANN: 3-layer feedforward neural
algorithm for high-performance tasks. network for binary classification.
Key Features: Architecture:
Input seismic Parameters output whether a • Input Layer: Takes features like magnitude and
tsunami will occur or not depth.
• Zero precision is 0.96 and class one recall is • Hidden Layers: 3 layers with ReLU activation for
0.92. non-linear patterns.
• Accuracy of our model 92.8%. • Output Layer: Predicts tsunami occurrence

XGBoost Model Architecture of ann

6
Predictive Modeling
XGBoost:

7
Predictive Modeling
Ann Model :

8
Paradox Exploratory Data Analysis (EDA) & Results

EDA Techniques:
• Univariate and bivariate analysis to identify Results:-
patterns. • Established strong correlations between
• Correlational analysis (correlation matrix). seismic parameters and tsunami occurrences.
• Classification of regions into High, Medium, and • Identified high-risk zones through effective
Low risk based on key features. classification techniques.

Findings: Median magnitude Frequency over 5 years


• Majority of events had 0 km < depths < 50 km and (freq_5years)
magnitudes between 6.5–7.0. Major earthquake frequency (freq_major_5years)
• Parameters like depth time, dmin, cdi, gap, sig, Ratio of occurrences (ratio_5years) Depth.
magnitude, and depth demonstrated significant
positive correlation.

9
Paradox
Visualization Insights
High-Medium-Low Risk Distribution Cluster

10
Paradox
Visualization Insights

11
Paradox
Visualization Insights
Magnitude

12
Paradox
Visualization Insights

13
Resources
Dashboard Link :-
https://
public.tableau.com/app/profile/parth.nuwal4883/viz/EarthquakeAnalysis_17354827858960/Dashboard1?publ
ish=yes

Data Set :-
https://
docs.google.com/spreadsheets/d/1BXelepsoGZHqAjTq_4ug72LjmvA7_m9JFRhYGlVP8rQ/edit?usp=sharin
g

Git-Hub:-
https://github.com/Parthnuwal7/Tsunami-risk-prediction-and-Risk-zone-classification/tree/main

14

You might also like