Source code for the Distributed Anomaly Detection System (DADS)

This algorithm is based on the Series2Graph algorithm, that was published in P. Boniol and T. Palpanas, Series2Graph: Graph-based Subsequence Anomaly Detection in Time Series, PVLDB (2020).

Requirements:

• JDK 8 (Java)
• Maven

mvn package

You can skip the tests with -DskipTests. The created fat JAR is located in the target-folder.

The algorithm only accepts binary input files without timestamps and ordered double values. To convert a text file, use the Python script in script/sequence-converter/convert.py.

Input file:

0.333443
0.466437
0.474765

Converting:

python3 script/sequence-converter/convert.py --input input.txt --output output.bin

Requirements:

• JRE 8 (Java)

java -jar <dads-file-name>.jar master --host localhost --port 7788 --min-slaves 0 \
    --sequence <input_file> \
    --sub-sequence-length 50 --intersection-segments 50 \
    --query-length 75 --convolution-size 16 \
    --output ./results.txt --no-statistics