zeru-assignment

Wallet Credit Scoring System

This system assigns credit scores (0-1000) to cryptocurrency wallets based on their transaction behavior in the Aave V2 protocol. Higher scores indicate reliable usage, while lower scores suggest risky or exploitative behavior.

Key Files

1. zeru.py - Main processing script

   • Input: input_transactions.json
   • Outputs:
     • output_scores.json (credit scores)
     • wallet_features_scores.csv (features + scores)

2. plot.py - Visualization script

   • Input: wallet_features_scores.csv
   • Outputs:
     • score_distribution.png
     • risk_comparison.png

To Run

1. Setup Environment

git clone https://github.com/saai07/zeru-assignment.git
pip install -r requirements.txt 

2. Add Input File

• Download the dataset and place the input_transactions.json in the root directory

3. Run Credit Scoring

python zeru.py

4. Run Analysis

python plot.py

5. Output Files

output_scores.json: Final wallet credit scores

wallet_features_scores.csv: Features + Scores

score_distribution.png, feature_comparison.png: Visuals for reporting

Methodology

Algorithm: Isolation Forest

• Type: Unsupervised anomaly detection
• Why chosen:
  • Effective for identifying unusual patterns in unlabeled data
  • Handles high-dimensional data well
  • Computationally efficient
• Mechanism:
  • Creates decision trees to isolate anomalies
  • Scores based on path length to isolate observations
  • Higher scores = more normal behavior

Key Features Engineered

Feature	Description	Risk Significance
repay_ratio	Repayments / Borrows	Measures debt responsibility
liquidation_ratio	Liquidations / Borrows	Default risk indicator
utilization	Net Borrows / Net Deposits	Leverage risk
tx_frequency	Transactions per day	Bot-like behavior detection
wallet_age	Days since first transaction	Long-term commitment
time_variance	Variance between transaction times	Pattern consistency

System Architecture

graph TD
    A[Raw Transaction Data] --> B[Feature Engineering]
    B --> C[Isolation Forest Model]
    C --> D[Credit Scores 0-1000]
    D --> E[JSON Output]
    D --> F[CSV Features]
    F --> G[Behavior Analysis]
    G --> H[Risk Comparison Plots]

Loading