43v3r8 / knowledge /graph.py
43v3r Tech
initial
fdeb336
"""
Knowledge Graph for crypto market relationships and insights
Uses NetworkX for graph operations
"""
import networkx as nx
import pickle
import logging
from typing import Dict, List, Tuple, Optional
from datetime import datetime
from config import Config
logging.basicConfig(level=logging.INFO)
logger = logging.getLogger(__name__)
class MarketKnowledgeGraph:
"""
Knowledge graph to store and query relationships between:
- Cryptocurrencies
- Market events
- Trading patterns
- Correlations
- News/sentiment
"""
def __init__(self, graph_path: Optional[str] = None):
self.graph_path = graph_path or Config.KNOWLEDGE_GRAPH_PATH
self.graph = nx.MultiDiGraph()
self.load_graph()
def load_graph(self):
"""Load graph from disk"""
try:
with open(self.graph_path, 'rb') as f:
self.graph = pickle.load(f)
logger.info(f"Loaded knowledge graph with {self.graph.number_of_nodes()} nodes")
except FileNotFoundError:
logger.info("Creating new knowledge graph")
self._initialize_base_graph()
def save_graph(self):
"""Save graph to disk"""
try:
with open(self.graph_path, 'wb') as f:
pickle.dump(self.graph, f)
logger.info("Knowledge graph saved")
except Exception as e:
logger.error(f"Error saving graph: {e}")
def _initialize_base_graph(self):
"""Initialize graph with base crypto knowledge"""
# Major cryptocurrencies
cryptos = {
'BTC': {'name': 'Bitcoin', 'category': 'Currency', 'layer': 'L1'},
'ETH': {'name': 'Ethereum', 'category': 'Smart Contract', 'layer': 'L1'},
'SOL': {'name': 'Solana', 'category': 'Smart Contract', 'layer': 'L1'},
'BNB': {'name': 'Binance Coin', 'category': 'Exchange', 'layer': 'L1'},
'XRP': {'name': 'Ripple', 'category': 'Payment', 'layer': 'L1'},
'ADA': {'name': 'Cardano', 'category': 'Smart Contract', 'layer': 'L1'},
'AVAX': {'name': 'Avalanche', 'category': 'Smart Contract', 'layer': 'L1'},
'MATIC': {'name': 'Polygon', 'category': 'Scaling', 'layer': 'L2'},
'ARB': {'name': 'Arbitrum', 'category': 'Scaling', 'layer': 'L2'},
}
for symbol, attrs in cryptos.items():
self.add_crypto_node(symbol, attrs)
# Known correlations
correlations = [
('BTC', 'ETH', 0.85, 'high_correlation'),
('BTC', 'SOL', 0.75, 'high_correlation'),
('BTC', 'BNB', 0.70, 'medium_correlation'),
('ETH', 'MATIC', 0.80, 'high_correlation'),
('ETH', 'ARB', 0.75, 'ecosystem_related'),
]
for crypto1, crypto2, weight, rel_type in correlations:
self.add_relationship(crypto1, crypto2, rel_type, weight=weight)
# Sectors and relationships
sectors = ['DeFi', 'NFT', 'Gaming', 'L1', 'L2', 'DEX', 'Lending']
for sector in sectors:
self.add_sector_node(sector)
# Connect cryptos to sectors
sector_connections = [
('ETH', 'DeFi', 'powers'),
('SOL', 'DeFi', 'powers'),
('ETH', 'NFT', 'powers'),
('MATIC', 'L2', 'is_type'),
('ARB', 'L2', 'is_type'),
]
for crypto, sector, rel_type in sector_connections:
self.add_relationship(crypto, sector, rel_type)
self.save_graph()
def add_crypto_node(self, symbol: str, attributes: Dict):
"""Add cryptocurrency node"""
self.graph.add_node(
symbol,
node_type='cryptocurrency',
**attributes,
created_at=datetime.now().isoformat()
)
def add_sector_node(self, sector: str):
"""Add sector/category node"""
self.graph.add_node(
sector,
node_type='sector',
created_at=datetime.now().isoformat()
)
def add_event_node(self, event_id: str, event_data: Dict):
"""Add market event node"""
self.graph.add_node(
event_id,
node_type='event',
**event_data,
created_at=datetime.now().isoformat()
)
def add_pattern_node(self, pattern_id: str, pattern_data: Dict):
"""Add trading pattern node"""
self.graph.add_node(
pattern_id,
node_type='pattern',
**pattern_data,
created_at=datetime.now().isoformat()
)
def add_relationship(self, source: str, target: str, rel_type: str, **attributes):
"""Add relationship between nodes"""
self.graph.add_edge(
source,
target,
relationship=rel_type,
**attributes,
created_at=datetime.now().isoformat()
)
def update_correlation(self, symbol1: str, symbol2: str, correlation: float):
"""Update or add correlation between two cryptos"""
if correlation > 0.7:
rel_type = 'high_correlation'
elif correlation > 0.4:
rel_type = 'medium_correlation'
elif correlation < -0.7:
rel_type = 'negative_correlation'
else:
rel_type = 'low_correlation'
self.add_relationship(symbol1, symbol2, rel_type, weight=correlation)
def get_related_cryptos(self, symbol: str, max_distance: int = 2) -> List[Tuple[str, float]]:
"""
Get related cryptocurrencies within max_distance
Returns:
List of (crypto, relevance_score) tuples
"""
if symbol not in self.graph:
return []
related = []
# Direct connections
for neighbor in self.graph.neighbors(symbol):
node = self.graph.nodes[neighbor]
if node.get('node_type') == 'cryptocurrency':
edges = self.graph[symbol][neighbor]
if edges:
weight = list(edges.values())[0].get('weight', 0.5)
related.append((neighbor, weight))
# Sort by relevance
related.sort(key=lambda x: x[1], reverse=True)
return related
def get_sector_cryptos(self, sector: str) -> List[str]:
"""Get all cryptos in a sector"""
if sector not in self.graph:
return []
cryptos = []
for node in self.graph.predecessors(sector):
if self.graph.nodes[node].get('node_type') == 'cryptocurrency':
cryptos.append(node)
return cryptos
def find_arbitrage_opportunities(self) -> List[Dict]:
"""
Find potential arbitrage opportunities based on correlations
Returns:
List of opportunity dicts
"""
opportunities = []
# Find cryptos with high correlation but different performance
# This is a simplified version - real arbitrage is more complex
crypto_nodes = [n for n, d in self.graph.nodes(data=True)
if d.get('node_type') == 'cryptocurrency']
for crypto in crypto_nodes:
related = self.get_related_cryptos(crypto, max_distance=1)
for related_crypto, correlation in related:
if correlation > 0.7: # High correlation
opportunities.append({
'crypto1': crypto,
'crypto2': related_crypto,
'correlation': correlation,
'opportunity_type': 'correlation_arbitrage'
})
return opportunities[:5] # Top 5
def get_market_narrative(self, symbol: str) -> Dict:
"""
Get market narrative for a cryptocurrency
Returns:
Dict with narrative information
"""
if symbol not in self.graph:
return {}
node_data = self.graph.nodes[symbol]
related = self.get_related_cryptos(symbol)
# Find connected sectors
sectors = []
for neighbor in self.graph.neighbors(symbol):
node = self.graph.nodes[neighbor]
if node.get('node_type') == 'sector':
sectors.append(neighbor)
return {
'symbol': symbol,
'name': node_data.get('name', symbol),
'category': node_data.get('category', 'Unknown'),
'layer': node_data.get('layer', 'Unknown'),
'sectors': sectors,
'related_cryptos': [r[0] for r in related[:5]],
'correlation_strength': sum(r[1] for r in related[:5]) / len(related) if related else 0
}
def add_market_event(self, event_type: str, affected_cryptos: List[str], description: str, impact: str):
"""
Add market event and connect to affected cryptos
Args:
event_type: Type of event (e.g., 'regulation', 'hack', 'upgrade')
affected_cryptos: List of crypto symbols
description: Event description
impact: 'positive', 'negative', or 'neutral'
"""
event_id = f"event_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
self.add_event_node(event_id, {
'type': event_type,
'description': description,
'impact': impact,
'timestamp': datetime.now().isoformat()
})
for crypto in affected_cryptos:
if crypto in self.graph:
self.add_relationship(event_id, crypto, f'affects_{impact}')
self.save_graph()
def record_pattern_occurrence(self, symbol: str, pattern_name: str, timeframe: str, metadata: Dict):
"""
Record occurrence of a trading pattern
Args:
symbol: Crypto symbol
pattern_name: Pattern name (e.g., 'golden_cross')
timeframe: Timeframe where pattern occurred
metadata: Additional pattern data
"""
pattern_id = f"pattern_{symbol}_{pattern_name}_{datetime.now().strftime('%Y%m%d_%H%M%S')}"
self.add_pattern_node(pattern_id, {
'name': pattern_name,
'symbol': symbol,
'timeframe': timeframe,
**metadata
})
if symbol in self.graph:
self.add_relationship(pattern_id, symbol, 'occurred_on')
self.save_graph()
def get_pattern_history(self, symbol: str, pattern_name: Optional[str] = None) -> List[Dict]:
"""
Get historical pattern occurrences for a symbol
Args:
symbol: Crypto symbol
pattern_name: Optional pattern name filter
Returns:
List of pattern occurrence dicts
"""
patterns = []
pattern_nodes = [n for n, d in self.graph.nodes(data=True)
if d.get('node_type') == 'pattern' and
d.get('symbol') == symbol]
if pattern_name:
pattern_nodes = [n for n in pattern_nodes
if self.graph.nodes[n].get('name') == pattern_name]
for node in pattern_nodes:
patterns.append(self.graph.nodes[node])
return sorted(patterns, key=lambda x: x.get('created_at', ''), reverse=True)
def get_graph_statistics(self) -> Dict:
"""Get knowledge graph statistics"""
crypto_nodes = sum(1 for _, d in self.graph.nodes(data=True)
if d.get('node_type') == 'cryptocurrency')
sector_nodes = sum(1 for _, d in self.graph.nodes(data=True)
if d.get('node_type') == 'sector')
event_nodes = sum(1 for _, d in self.graph.nodes(data=True)
if d.get('node_type') == 'event')
pattern_nodes = sum(1 for _, d in self.graph.nodes(data=True)
if d.get('node_type') == 'pattern')
return {
'total_nodes': self.graph.number_of_nodes(),
'total_edges': self.graph.number_of_edges(),
'cryptocurrencies': crypto_nodes,
'sectors': sector_nodes,
'events': event_nodes,
'patterns': pattern_nodes,
'density': nx.density(self.graph)
}
def export_subgraph(self, center_node: str, radius: int = 2) -> Dict:
"""
Export subgraph centered on a node
Args:
center_node: Central node
radius: Number of hops to include
Returns:
Dict with nodes and edges for visualization
"""
if center_node not in self.graph:
return {'nodes': [], 'edges': []}
# Get all nodes within radius
nodes = {center_node}
current_layer = {center_node}
for _ in range(radius):
next_layer = set()
for node in current_layer:
next_layer.update(self.graph.neighbors(node))
next_layer.update(self.graph.predecessors(node))
nodes.update(next_layer)
current_layer = next_layer
# Build subgraph
subgraph = self.graph.subgraph(nodes)
# Format for visualization
nodes_list = []
for node in subgraph.nodes():
node_data = subgraph.nodes[node]
nodes_list.append({
'id': node,
'label': node_data.get('name', node),
'type': node_data.get('node_type', 'unknown'),
**node_data
})
edges_list = []
for source, target, data in subgraph.edges(data=True):
edges_list.append({
'source': source,
'target': target,
'relationship': data.get('relationship', 'related'),
'weight': data.get('weight', 1.0)
})
return {
'nodes': nodes_list,
'edges': edges_list
}