There is excellent tool written by Michael Hunger, which can be cloned from https://github.com/jexp/batch-import.

The CSV file has to be converted into the format specified in the readme file. The tool is very flexible in terms of the number of properties and the types of each property. The nodes and relationships can be within the same file or within multiple files. The example file format is present in the sample directory. To run the tool, use the following command:

$ wget https://dl.dropboxusercontent.com/u/14493611/batch_importer_22.zip
$ unzip batch_importer_22.zip
# Download sample nodes.csv and rels.csv from the github repo under sample
$ import.sh test.db nodes.csv rels.csv
$ cp test.db ${NEO4J_ROOT}/data/graph.db

Each parameter in the command has been fully explained in the readme file.

Benchmark figures claimed by the batch importer tool are 2 billion nodes and 20 billion relationships in 11 hours (500K elements/second).

This is claimed over the EC2 high I/O instance.

Custom scripts can be written in any language to import data from CSV files. Custom scripts give you the advantages of checking various erroneous scenarios, leaving out redundant columns, and other flexibilities. For a smaller number of nodes and relationships, custom scripts can be written in any language of your choice.

The exact format of the script will depend on the CSV file. You can write the script as follows:

#Bash Script for importing nodes
NEO4J_ROOT="/var/lib/neo4j"
while read LINE
do
  name=`echo $LINE | awk -F "," '{print $3}'`
  ${NEO4J_ROOT}/bin/neo4j-shell -c mknode --np \"{'name':$name}\" -v
done

Similar scripts can be written for relationships too, as shown here:

#Bash Script for creating relationships
#Format of csv should be startnode,endnode,type,direction
NEO4J_ROOT="/var/lib/neo4j"
IFS=","
while read LINE
do
  echo $LINE
  array=($LINE)
  ${NEO4J_ROOT}/bin/neo4j-shell -c cd -a ${array[0]} mkrel -d ${array[3]} -t ${array[2]} ${array[1]}
done

This task can also be achieved in Python using the py2neo module, as shown in the following script:

#Sample Python code to create nodes from csv file
import csv
from py2neo import neo4j, cypher
from py2neo import node, rel
graph_db = neo4j.Graph("http://localhost:7474/db/data/")
ifile = open('nodes.csv', "rb")
reader = csv.reader(ifile)
rownum = 0
for row in reader:
  nodes = graph_db.create({"name":row[2]})
ifile.close()

A similar Python code can be written for creating relationships, too. The py2neo module can also be used to create a batch request, wherein there's a whole list with parameters as shown in the following code:

records = [(101, "A"), (102, "B"), (103, "C")]
graph_db = neo4j.Graph ("http://localhost:7474/db/data/")
batch = neo4j.WriteBatch(graph_db)
for emp_no, name in records:
  batch.get_or_create_indexed_node("Employees", "emp_no", emp_no,{
  "emp_no": emp_no, "name": name
})
nodes = batch.submit()