Blockchain & Big Data Analytics
A data pipeline for exploring blockchain transactions with graph features and XGBoost.
Project screenshots are being prepared.
Overview
This data-science project explores blockchain transaction data from Kaggle. The workflow cleans the raw records, derives behavioural features from transaction relationships, trains an XGBoost model, and presents analysis through an interactive Streamlit application.
Challenge
Blockchain transaction datasets are large and relational. Reviewing raw rows alone makes it difficult to identify behavioural patterns or prepare useful inputs for a classification model.
Solution
The pipeline handles missing values and scaling, converts transaction relationships into graph-based features, trains and evaluates XGBoost, and exposes the results in a Streamlit interface for exploration.
Key features
- Data preparation: Cleans transaction records, handles missing values, and prepares model-ready features.
- Graph feature engineering: Transforms wallet and transaction relationships into measurable network properties.
- XGBoost modelling: Trains and evaluates a classification model using the engineered feature set.
- Interactive analysis: Uses Streamlit to present charts, inputs, and model outputs in a web interface.