Neo4J协同过滤比预期慢

我正在Neo4J图上实现一个推荐系统，我刚开始研究我计划使用的查询，但它的执行速度比我预期的慢得多

统计数据

Neo4J Version: 2.3.1
Nodes: 820K
Relationships: 7.6M

Indexes
ON :LIKES(created_at)     ONLINE  
ON :Product(id)           ONLINE  
ON :Product(created_at)   ONLINE  
ON :User(id)              ONLINE  
ON :User(date_joined)     ONLINE

No constraints

我已经对查询优化做了很多研究，但就我所见，我没有在查询结构中犯任何常见/常见的错误（但我不是专家）

这是一个带有测试数据集的开发人员控制台：

查询

MATCH (u1:User {id: {user_id}})-[l1:LIKES]->(p1:Product)
WITH u1, l1, p1
ORDER BY p1.created_at DESC
LIMIT 10

MATCH (p1)<-[:LIKES]-(u2:User)
WHERE NOT u1=u2
WITH u1, l1, p1, u2, COUNT(u2) as rating
ORDER BY rating DESC
LIMIT 50

MATCH (u2)-[l2:LIKES]->(recommendation:Product)
WHERE NOT (p1)=(recommendation)
WITH recommendation, COUNT(recommendation) as weight
RETURN recommendation.id as id
ORDER BY weight DESC
LIMIT {limit}

查询配置文件输出（对照我们的生产数据集副本）

+-------------------+----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|运算符|估计行|行|数据库命中数|标识符|其他|
+-------------------+----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+生产结果| 7 | 100 | 0 | id | id|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+投影| 7 | 100 | 0 | anon[382]，id，建议，权重| anon[382]|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+Top | 7 | 100 | 0 | anon[382]，推荐，权重|文字（100）；重量|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+投影| 7 | 129342 | 129342 | anon[382]，建议，权重|建议id；重量|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|加总| 7 | 129342 | 0 |推荐，权重|推荐|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+过滤器| 44 | 442432 | 471953 | l1、l2、p1、评级、建议、u1、u2和（非（p1=建议）、建议：产品）|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+扩展（全部）| 44 | 472039 | 472089 | l1、l2、p1、评级、推荐、u1、u2 |（u2）-[l2:LIKES]->（推荐）|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+Top | 10 | 50 | 0 | l1、p1、评级、u1、u2 |文字（50）；评级|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+10 | 527 | 0 | l1，p1，评级，u1，u2 | u1，l1，p1，u2|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+过滤器| 92 | 563 | 563 | anon[82]，anon[119]，l1，p1，u1，u2 | Ands（非（u1==u2），u2:用户）|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+扩展（全部）| 92 | 574 | 584 | anon[82]，anon[119]，l1，p1，u1，u2 |（p1）（p1）|
| |                 +----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+
|+NodeIndexSeek | 1 | 1 | 2 | u1 |：用户（id）|
+-------------------+----------------+--------+---------+--------------------------------------------+---------------------------------------------------------+

我看过一些案例研究，其中人们正在使用Neo4j进行实时协同过滤，因此我认为一定有可能在这种数据集上进行这种查询。我是否不切实际？我们在AmazonEC2计算优化节点（c4.large）上运行这个程序，所以我认为它的性能相当好

我在这里挠头，非常感谢任何意见

干杯， David.

[旁白：重新打开开发人员控制台时，不会重新创建索引，因此必须手动重新创建索引。]

我不知道这对您来说是否足够好，但您可以通过简单地不指定查询中大多数节点（

p1

、

u2

和

建议

）的标签来消除分析结果中约44%的DB命中：

MATCH (u1:User {id: {user_id}})-[l1:LIKES]->(p1)
WITH u1, l1, p1
ORDER BY p1.created_at DESC
LIMIT 10

MATCH (p1)<-[:LIKES]-(u2)
WHERE NOT u1=u2
WITH u1, l1, p1, u2, COUNT(u2) as rating
ORDER BY rating DESC
LIMIT 50

MATCH (u2)-[l2:LIKES]->(recommendation)
WHERE NOT (p1)=(recommendation)
WITH recommendation, COUNT(recommendation) as weight
RETURN recommendation.id as id
ORDER BY weight DESC
LIMIT {limit}

MATCH（u1:User{id:{User\u id}}）-[l1:LIKES]>（p1）
用u1，l1，p1
由p1.U在DESC创建的订单
限制10
匹配（p1）（推荐）
如果不是（p1）=（建议）
对于推荐，将（推荐）计算为
MATCH (u1:User {id: {user_id}})-[l1:LIKES]->(p1)
WITH u1, l1, p1
ORDER BY p1.created_at DESC
LIMIT 10

MATCH (p1)<-[:LIKES]-(u2)
WHERE NOT u1=u2
WITH u1, l1, p1, u2, COUNT(u2) as rating
ORDER BY rating DESC
LIMIT 50

MATCH (u2)-[l2:LIKES]->(recommendation)
WHERE NOT (p1)=(recommendation)
WITH recommendation, COUNT(recommendation) as weight
RETURN recommendation.id as id
ORDER BY weight DESC
LIMIT {limit}