How to use Similarity ? GDS (2.6.5) in neo4j desktop (5.15.0)

[takeshi0525982]

Greetings,

I am using the latest GDS (2.6.5) in neo4j desktop (5.15.0). I am not sure how to use Similarity. The code below is data for three different recipes.

Question:

Can I use Similarity to search for recipes similar to Recipe 1 (A->B->C) from various perspectives?

For example, can I search with Similarity based on the presence or absence of process labels?

Each recipe's process has parameters. Is it possible to calculate similarity taking into account the parameter values as well?

Code:
// recipe1
CREATE (fip1:FirstIntermediateProduct {name: "A", created_date: "2023-11-01", projectId: "P1"})
CREATE (sip1:SecondIntermediateProduct {name: "B", created_date: "2023-11-02", projectId: "P1"})
CREATE (fp1:FinalProduct {name: "C", created_date: "2023-11-03", projectId: "P1"})
CREATE (p1:Process {name: "process1", pro1_para1: 100, pro1_para2: 200, pro1_para3: 300, pro1_para4: 400, pro1_para5: 500})
CREATE (p4:Process {name: "process4", pro4_para1: 150, pro4_para2: 250, pro4_para3: 350, pro4_para4: 450, pro4_para5: 550})
CREATE (fip1)-[:mix]->(sip1)-[:add]->(fp1)
CREATE (fip1)-[:process]->(p1)
CREATE (sip1)-[:process]->(p4)

// recipe2
CREATE (fip2:FirstIntermediateProduct {name: "D", created_date: "2023-11-04", projectId: "P2"})
CREATE (sip2:SecondIntermediateProduct {name: "E", created_date: "2023-11-05", projectId: "P2"})
CREATE (fp2:FinalProduct {name: "F", created_date: "2023-11-06", projectId: "P2"})
CREATE (p1_2:Process {name: "process1", pro1_para1: 120, pro1_para2: 220, pro1_para3: 320, pro1_para4: 420, pro1_para5: 520})
CREATE (p2_2:Process {name: "process2", pro2_para1: 180, pro2_para2: 280, pro2_para3: 380, pro2_para4: 480, pro2_para5: 580})
CREATE (p4_2:Process {name: "process4", pro4_para1: 160, pro4_para2: 260, pro4_para3: 360, pro4_para4: 460, pro4_para5: 560})
CREATE (p5_2:Process {name: "process5", pro5_para1: 190, pro5_para2: 290, pro5_para3: 390, pro5_para4: 490, pro5_para5: 590})
CREATE (p6_2:Process {name: "process6", pro6_para1: 140, pro6_para2: 240, pro6_para3: 340, pro6_para4: 440, pro6_para5: 540})
CREATE (fip2)-[:mix]->(sip2)-[:add]->(fp2)
CREATE (fip2)-[:process]->(p1_2)
CREATE (fip2)-[:process]->(p2_2)
CREATE (sip2)-[:process]->(p4_2)
CREATE (sip2)-[:process]->(p5_2)
CREATE (sip2)-[:process]->(p6_2)

// recipe3
CREATE (fip3:FirstIntermediateProduct {name: "G", created_date: "2023-11-07", projectId: "P3"})
CREATE (sip3:SecondIntermediateProduct {name: "H", created_date: "2023-11-08", projectId: "P3"})
CREATE (fp3:FinalProduct {name: "I", created_date: "2023-11-09", projectId: "P3"})
CREATE (p1_3:Process {name: "process1", pro1_para1: 110, pro1_para2: 210, pro1_para3: 310, pro1_para4: 410, pro1_para5: 510})
CREATE (p2_3:Process {name: "process2", pro2_para1: 170, pro2_para2: 270, pro2_para3: 370, pro2_para4: 470, pro2_para5: 570})
CREATE (p3_3:Process {name: "process3", pro3_para1: 130, pro3_para2: 230, pro3_para3: 330, pro3_para4: 430, pro3_para5: 530})
CREATE (p4_3:Process {name: "process4", pro4_para1: 150, pro4_para2: 250, pro4_para3: 350, pro4_para4: 450, pro4_para5: 550})
CREATE (fip3)-[:mix]->(sip3)-[:add]->(fp3)
CREATE (fip3)-[:process]->(p1_3)
CREATE (fip3)-[:process]->(p2_3)
CREATE (fip3)-[:process]->(p3_3)
CREATE (sip3)-[:process]->(p4_3)

[joshcornejo]

I'm just starting with GDS, so my help here is limited at this stage.

Looking at your CYPHER, you have nodes with the same structure (viewed on the image) but they don't share anything.

"The Node Similarity algorithm compares a set of nodes based on the nodes they are connected to. Two nodes are considered similar if they share many of the same neighbours."

Node Similarity - Neo4j Graph Data Science

I would suggest that at least some/all of your 'leaf' nodes (:Process) are common across the graphs.

// recipe1
CREATE (fip1:FirstIntermediateProduct {name: "A", created_date: "2023-11-01", projectId: "P1"})
CREATE (sip1:SecondIntermediateProduct {name: "B", created_date: "2023-11-02", projectId: "P1"})
CREATE (fp1:FinalProduct {name: "C", created_date: "2023-11-03", projectId: "P1"})
CREATE (p1:Process {name: "process1", pro1_para1: 100, pro1_para2: 200, pro1_para3: 300, pro1_para4: 400, pro1_para5: 500})
CREATE (p4:Process {name: "process4", pro4_para1: 150, pro4_para2: 250, pro4_para3: 350, pro4_para4: 450, pro4_para5: 550})
CREATE (fip1)-[:mix]->(sip1)-[:add]->(fp1)
CREATE (fip1)-[:process]->(p1)
CREATE (sip1)-[:process]->(p4)

// recipe2
CREATE (fip2:FirstIntermediateProduct {name: "D", created_date: "2023-11-04", projectId: "P2"})
CREATE (sip2:SecondIntermediateProduct {name: "E", created_date: "2023-11-05", projectId: "P2"})
CREATE (fp2:FinalProduct {name: "F", created_date: "2023-11-06", projectId: "P2"})
CREATE (p2:Process {name: "process2", pro2_para1: 180, pro2_para2: 280, pro2_para3: 380, pro2_para4: 480, pro2_para5: 580})
CREATE (p5:Process {name: "process5", pro5_para1: 190, pro5_para2: 290, pro5_para3: 390, pro5_para4: 490, pro5_para5: 590})
CREATE (p6:Process {name: "process6", pro6_para1: 140, pro6_para2: 240, pro6_para3: 340, pro6_para4: 440, pro6_para5: 540})
CREATE (fip2)-[:mix]->(sip2)-[:add]->(fp2)
CREATE (fip2)-[:process]->(p1)
CREATE (fip2)-[:process]->(p2)
CREATE (sip2)-[:process]->(p4)
CREATE (sip2)-[:process]->(p5)
CREATE (sip2)-[:process]->(p6)

// recipe3
CREATE (fip3:FirstIntermediateProduct {name: "G", created_date: "2023-11-07", projectId: "P3"})
CREATE (sip3:SecondIntermediateProduct {name: "H", created_date: "2023-11-08", projectId: "P3"})
CREATE (fp3:FinalProduct {name: "I", created_date: "2023-11-09", projectId: "P3"})
CREATE (fip3)-[:mix]->(sip3)-[:add]->(fp3)
CREATE (fip3)-[:process]->(p1)
CREATE (fip3)-[:process]->(p2)
CREATE (fip3)-[:process]->(p3)
CREATE (sip3)-[:process]->(p4)

[ameyasoft]

Here is a way:
I added 'finalproductProjectID' property to all the nodes. Also I changed the process para1 property name to 'para1' without the prefix like pro1. 

match (a:SecondIntermediateProduct)-[]-(b:Process)
where a.finalproductProjectID = "P1" = b.finalproductProjectID

match (a1:SecondIntermediateProduct)-[]-(b1:Process)
where a1.finalproductProjectID <>  "P1" 
//= b1.finalproductProjectID
with a, b, a1, b1, (100 - toInteger(apoc.text.jaroWinklerDistance(toString(b.para1), toString(b1.para1)) * 100)) as similarity

with a, b,a1, b1, similarity 
//where similarity = 100
with a.finalproductProjectID as recipe1, b.name as name1, b.para1 as bpara1, similarity, a1.finalproductProjectID as recipe2, b1.name as name2, b1.para1 as b1para1
return recipe1, name1, bpara1, similarity, recipe2, name2, b1para1 order by name1
You can set the minimum value for your Similarity.

Result: