Neo4j 从零到实战：社交网络图谱

1. 背景与概念

1.1 图数据库 vs 关系型数据库

在社交网络中，查询"朋友的朋友"在 SQL 中需要多次 JOIN（O(n²) 复杂度），而图数据库通过指针跳转做到 O(1)。

SQL:   SELECT * FROM friends f1
       JOIN friends f2 ON f1.target = f2.source
       WHERE f1.source = '张三'

Cypher: MATCH (张三)-[:FRIEND]->()-[:FRIEND]->(fof)
        RETURN fof

1.2 Cypher 核心语法

语法	含义
(n:Label {key: value})	匹配/创建节点
-[r:TYPE]->	有向关系
-[*1..3]->	1 到 3 跳变长路径
shortestPath()	最短路径算法
collect()	聚合为数组
size()	列表/字符串长度

2. 分步实战：构建校园社交图谱

场景

模拟大学社交网络：学生、社团、课程，实现好友推荐、社团发现。

步骤一：数据建模

// 创建约束（高效查找）
CREATE CONSTRAINT FOR (p:Person) REQUIRE p.name IS UNIQUE;
CREATE CONSTRAINT FOR (g:Group) REQUIRE g.name IS UNIQUE;
CREATE CONSTRAINT FOR (c:Course) REQUIRE c.name IS UNIQUE;

步骤二：批量导入数据

// 学生
UNWIND [
    {name: 'Alice', major: 'CS', year: 3},
    {name: 'Bob', major: 'CS', year: 3},
    {name: 'Carol', major: 'Math', year: 2},
    {name: 'Dave', major: 'CS', year: 4},
    {name: 'Eve', major: 'Physics', year: 3},
    {name: 'Frank', major: 'Math', year: 2}
] AS person
CREATE (:Person {name: person.name, major: person.major, year: person.year});

// 社团
CREATE (:Group {name: '编程社', type: '学术'});
CREATE (:Group {name: '篮球队', type: '体育'});
CREATE (:Group {name: '数学建模', type: '学术'});

// 课程
CREATE (:Course {name: '数据库原理'});
CREATE (:Course {name: '算法导论'});
CREATE (:Course {name: '高等代数'});

// 好友关系（双向用双向 CREATE）
CREATE
    (:Person {name: 'Alice'})-[:FRIEND]->(:Person {name: 'Bob'}),
    (:Person {name: 'Alice'})-[:FRIEND]->(:Person {name: 'Carol'}),
    (:Person {name: 'Bob'})-[:FRIEND]->(:Person {name: 'Dave'}),
    (:Person {name: 'Carol'})-[:FRIEND]->(:Person {name: 'Eve'}),
    (:Person {name: 'Dave'})-[:FRIEND]->(:Person {name: 'Frank'}),
    (:Person {name: 'Eve'})-[:FRIEND]->(:Person {name: 'Alice'});

// 社团关系
CREATE
    (:Person {name: 'Alice'})-[:MEMBER_OF]->(:Group {name: '编程社'}),
    (:Person {name: 'Bob'})-[:MEMBER_OF]->(:Group {name: '编程社'}),
    (:Person {name: 'Dave'})-[:MEMBER_OF]->(:Group {name: '编程社'}),
    (:Person {name: 'Carol'})-[:MEMBER_OF]->(:Group {name: '数学建模'}),
    (:Person {name: 'Frank'})-[:MEMBER_OF]->(:Group {name: '数学建模'});

// 选课
CREATE
    (:Person {name: 'Alice'})-[:ENROLLED_IN]->(:Course {name: '数据库原理'}),
    (:Person {name: 'Bob'})-[:ENROLLED_IN]->(:Course {name: '数据库原理'}),
    (:Person {name: 'Carol'})-[:ENROLLED_IN]->(:Course {name: '高等代数'});

步骤三：实现推荐查询

// 1. 二度好友推荐（你朋友的朋友，但你还不是朋友）
MATCH (me:Person {name: 'Alice'})-[:FRIEND*2..2]-(recommendation:Person)
WHERE NOT (me)-[:FRIEND]-(recommendation) AND me <> recommendation
RETURN DISTINCT recommendation.name AS 推荐好友;

// 2. 基于社团的多重好友推荐
MATCH (me:Person {name: 'Alice'})-[:MEMBER_OF]->(g:Group)<-[:MEMBER_OF]-(other:Person)
WHERE NOT (me)-[:FRIEND]-(other) AND me <> other
WITH other, COUNT(g) AS common_groups
RETURN other.name, common_groups
ORDER BY common_groups DESC;

// 3. 课程共修者
MATCH (me:Person {name: 'Alice'})-[:ENROLLED_IN]->(c:Course)<-[:ENROLLED_IN]-(peer:Person)
WHERE NOT (me)-[:FRIEND]-(peer)
RETURN peer.name AS 同课同学, c.name AS 课程;

// 4. 图算法：PageRank（需安装 GDS 插件）
// CALL gds.pageRank.stream('social-graph') YIELD nodeId, score
// RETURN gds.util.asNode(nodeId).name AS name, score
// ORDER BY score DESC;

3. 思考题

1. 如果好友关系改为单向（如微博关注），上述推荐查询需要如何修改？
2. 在 [:FRIEND*2..2] 变长路径中，如果社交网络有 100 万节点，如何避免查询爆炸？
3. 设计一个"可能认识的人"功能，综合考虑共同朋友数、共同社团数、同课程数，给出综合评分。