Chuyển tới nội dung chính

Neo4jService - Graph Database Client

TypeScript client for Neo4j graph database providing entity relationship querying, graph traversal, and semantic network analysis for traffic and environmental data.

Overview​

graph TB
    subgraph Neo4jService
        DRIVER[Neo4j Driver]
        SESSION[Session Manager]
        QUERY[Cypher Executor]
    end
    
    subgraph Neo4j["Neo4j Database"]
        ACCIDENTS[(Accidents)]
        CAMERAS[(Cameras)]
        ROADS[(Road Segments)]
        WEATHER[(Weather)]
    end
    
    subgraph Relationships
        NEAR[NEAR_TO]
        CAUSED[CAUSED_BY]
        AFFECTS[AFFECTS]
        LOCATED[LOCATED_ON]
    end
    
    DRIVER --> SESSION
    SESSION --> QUERY
    QUERY --> ACCIDENTS
    QUERY --> CAMERAS
    
    ACCIDENTS --- NEAR
    ACCIDENTS --- CAUSED
    CAMERAS --- LOCATED
    WEATHER --- AFFECTS

Features​

FeatureDescription
Cypher QueriesParameterized query execution with binding
Relationship TraversalMulti-hop queries for connected entities
Accident AnalysisCorrelation and clustering of incidents
Geo-Spatial QueriesDistance-based relationship discovery
Connection PoolingTimeout management and recovery
Transaction SupportRead/write transaction isolation

Class Definition​

import neo4j, { Driver, Session } from 'neo4j-driver';

export class Neo4jService {
  private driver: Driver;

  constructor();
  
  // Accident Queries
  getAccidents(): Promise<Accident[]>;
  getAccidentById(id: string): Promise<Accident | null>;
  getAccidentsByArea(lat: number, lon: number, radiusKm?: number): Promise<Accident[]>;
  getAccidentRelationships(accidentId: string): Promise<Relationship[]>;
  
  // Connection Management
  close(): Promise<void>;
}

interface Relationship {
  relationship: string;
  relatedType: string;
  related: Record<string, any>;
}

Configuration​

// Environment Variables
const config = {
  NEO4J_URL: 'bolt://localhost:7687',
  NEO4J_USER: 'neo4j',
  NEO4J_PASSWORD: 'test12345'
};

// Driver Options
const driverConfig = {
  connectionTimeout: 10000,         // 10s connection timeout
  maxConnectionLifetime: 3600000    // 1 hour max connection lifetime
};

Usage Examples​

Basic Accident Queries​

import { Neo4jService } from './services/neo4jService';

const neo4j = new Neo4jService();

// Get all accidents (latest 100)
const accidents = await neo4j.getAccidents();
console.log(`Found ${accidents.length} accidents`);

// Get single accident by ID
const accident = await neo4j.getAccidentById('urn:ngsi-ld:RoadAccident:001');
if (accident) {
  console.log(`Accident at ${accident.location.address}`);
  console.log(`Severity: ${accident.severity}`);
}

Geo-Spatial Queries​

// Find accidents within 5km of a location
const nearbyAccidents = await neo4j.getAccidentsByArea(
  10.8231,  // latitude
  106.6297, // longitude
  5         // radius in km
);

console.log(`Found ${nearbyAccidents.length} accidents nearby`);

Relationship Traversal​

// Get all relationships for an accident
const relationships = await neo4j.getAccidentRelationships('urn:ngsi-ld:RoadAccident:001');

relationships.forEach(rel => {
  console.log(`${rel.relationship} -> ${rel.relatedType}`);
  console.log(rel.related);
});

Cypher Query Examples​

Get Accidents with Ordering​

MATCH (a:Accident)
RETURN a
ORDER BY a.timestamp DESC
LIMIT 100

Geo-Spatial Distance Query​

MATCH (a:Accident)
WITH a, 
     point({latitude: toFloat(a.latitude), longitude: toFloat(a.longitude)}) AS accidentPoint,
     point({latitude: $latitude, longitude: $longitude}) AS centerPoint
WHERE distance(accidentPoint, centerPoint) <= $radiusMeters
RETURN a
ORDER BY a.timestamp DESC

Relationship Discovery​

MATCH (a:Accident {id: $id})-[r]-(related)
RETURN type(r) AS relationship, labels(related) AS relatedType, related
MATCH (a:Accident {id: $accidentId})-[:NEAR_TO]->(c:Camera)
RETURN c
ORDER BY c.distance ASC
LIMIT 5

Weather-Accident Correlation​

MATCH (a:Accident)-[:OCCURRED_DURING]->(w:Weather)
WHERE w.precipitation > 0.5
RETURN a, w
ORDER BY a.timestamp DESC

Data Model​

erDiagram
    Accident ||--o{ Camera : NEAR_TO
    Accident ||--o{ Weather : OCCURRED_DURING
    Accident ||--o{ RoadSegment : LOCATED_ON
    Camera ||--o{ RoadSegment : MONITORS
    Weather ||--o{ AirQuality : AFFECTS
    
    Accident {
        string id PK
        float latitude
        float longitude
        string address
        string type
        string severity
        string description
        datetime timestamp
        boolean resolved
        int casualties
    }
    
    Camera {
        string id PK
        float latitude
        float longitude
        string name
        string imageUrl
    }
    
    Weather {
        string id PK
        float temperature
        float humidity
        float precipitation
        datetime observedAt
    }

Error Handling​

async getAccidentById(id: string): Promise<Accident | null> {
  const session: Session = this.driver.session();
  try {
    const result = await session.run(
      'MATCH (a:Accident {id: $id}) RETURN a',
      { id }
    );

    if (result.records.length === 0) {
      return null;
    }

    return this.mapAccident(result.records[0].get('a').properties);
  } catch (error) {
    logger.error(`Error fetching accident ${id}:`, error);
    return null;
  } finally {
    await session.close();  // Always close session
  }
}

Session Management​

sequenceDiagram
    participant Service
    participant Driver
    participant Session
    participant Neo4j
    
    Service->>Driver: Get session
    Driver->>Session: Create session
    Service->>Session: Run query
    Session->>Neo4j: Execute Cypher
    Neo4j-->>Session: Results
    Session-->>Service: Records
    Service->>Session: Close session

Important: Always close sessions after use to prevent connection leaks.

Use Cases​

1. Accident Correlation Analysis​

Find accidents that occurred under similar conditions:

// Find correlated accidents (same road, similar time)
const query = `
  MATCH (a1:Accident {id: $id})-[:LOCATED_ON]->(r:RoadSegment)<-[:LOCATED_ON]-(a2:Accident)
  WHERE a1 <> a2 
    AND abs(duration.between(a1.timestamp, a2.timestamp).hours) < 24
  RETURN a2
`;

2. Camera Coverage Analysis​

Find cameras that could have recorded an accident:

const query = `
  MATCH (a:Accident {id: $id})
  WITH a, point({latitude: a.latitude, longitude: a.longitude}) AS accidentPoint
  MATCH (c:Camera)
  WITH c, a, point({latitude: c.latitude, longitude: c.longitude}) AS cameraPoint
  WHERE distance(accidentPoint, cameraPoint) <= 500  // 500m radius
  RETURN c
  ORDER BY distance(accidentPoint, cameraPoint) ASC
`;

3. Weather Impact Analysis​

Analyze how weather affects accident frequency:

const query = `
  MATCH (a:Accident)-[:OCCURRED_DURING]->(w:Weather)
  WITH w.precipitation > 0.5 AS rainy, count(a) AS accidentCount
  RETURN rainy, accidentCount
`;

Performance Considerations​

AspectRecommendation
Session LifecycleCreate per-query, close immediately
IndexesCreate on frequently queried properties (id, timestamp)
Query OptimizationUse LIMIT, avoid unbounded traversals
Connection PoolingReuse driver instance, not sessions
CREATE INDEX accident_id FOR (a:Accident) ON (a.id);
CREATE INDEX accident_timestamp FOR (a:Accident) ON (a.timestamp);
CREATE INDEX camera_id FOR (c:Camera) ON (c.id);
CREATE INDEX road_segment_id FOR (r:RoadSegment) ON (r.id);

References​