Usage

TODO

• Convert JSON format graph into interactive graph

Start up

Clone the repo

git clone https://github.com/korvva/RuleEngineNeo6j.git \
&& cd RuleEngineNeo6j

And run

podman compose up

Notice that the image used while composing is private, please reach out for permit granting.

Once successfully run up, neo4j is exposed to port 7474 and is accessible from localhost:7474. For other ports exposures please refer to settings in docker-compose.yml

For developing purpose, neo4j is set to run without database authorization. Make changes accordingly in docker-compose.yml

environment:
    #-NEO4J_AUTH=none
    -NEO4J_AUTH='user/password'

with default user password are neo4j neo4j

Run check

• In neo4j browser(localhost:7474), use cypher-shell command below to check if all required custom procedures are included

  CALL dbms.procedures()
  

  The highlighted procedures above must exist.

• Place training data and answers in import folder and check if neo4j successfully load path. For instance

  LOAD CSV WITH HEADERS FROM "file:///answers.csv" AS row
  CREATE (a:Answer {id: toInteger(row.id), text: row.text})

Run

Create Schema ----> Create Tree ----> Query Tree

• Create Schema simply run the following

  CALL com.hcw.schema.generate

• Create tree is a training process and is only needed for first run or with new data, otherwise, the trained model is stored in data/database/graph.db. To create new tree, call custom procedure as followed

  CALL com.hcw.decision_tree.create('credit', \
  '/import/training.csv', '/import/answers.csv', 0.02)
  

  where 'credit' stands for the name of tree while 0.02 is the splitting and merging threshold. It may take quite a while for training, for up to more than 5 minutes. Nothing prints out in neo4j browser but only in terminal that runs up container

• Query Tree (in neo4j browser)

  CALL com.hcw.decision_tree.traverse('TREENAME',{PAYLOAD})

  For example according to training.csv

  CALL com.hcw.decision_tree.traverse('credit', \
  {RevolvingUtilizationOfUnsecuredLines:'0.7661',\
  Age:'45', Late30to59Days:'2', DebtRatio:'0.803', \
  MonthlyIncome:'9120',OpenCreditLinesAndLoans:'13', \
  Late90Days:'0',RealEstateLoans:'6', \
  Late60to89Days:'0', Dependents:'2'});

  Returns graph in neo4j

• Query Tree (through curl) With NEO4J_AUTH=none

  curl -X POST \
  http://localhost:7474/db/data/transaction/commit \
  -H "Content-Type: application/json" \
  -d "{\"statements\":[
  	{\"statement\":
              \"CALL com.hcw.decision_tree.traverse('credit', \
                  {RevolvingUtilizationOfUnsecuredLines:'0.7661',\
                   Age:'45', Late30to59Days:'2', DebtRatio:'0.803', \
                   MonthlyIncome:'9120',OpenCreditLinesAndLoans:'13', \
                   Late90Days:'0',RealEstateLoans:'6', Late60to89Days:'0',\
                   Dependents:'2'});\"}]
      }"

  with NEO4J_AUTH='neo4j/password'

  curl -u neo4j:password \
  -X POST \
  http://localhost:7474/db/data/transaction/commit \
  -H "Content-Type: application/json" \
  -d "{\"statements\":[]}"

  And returns JSON format graph