Skip to content

Latest commit

 

History

History
48 lines (35 loc) · 1.38 KB

demo.md

File metadata and controls

48 lines (35 loc) · 1.38 KB
Raw

Demo

Dependency

The loglizer toolkit is implemented with Python and requires a number of dependency requirements installed.

  • python 3.6
  • scipy
  • numpy
  • scikit-learn=0.20.3
  • pandas

We recommend users to use Anaconda, which is a popular Python data science platform with many common packages pre-installed. The virtual enviorment can be set up via conda:

$ conda create -n py36 -c anaconda python=3.6
$ conda activate py36

For ease of reproducing our benchmarking results, we have also built a docker image for the running evironment. If you have docker installed, you can easily pull and run a docker container as follows:

$ mkdir loglizer
$ git clone https://github.com/logpai/loglizer.git loglizer/
$ docker run --name loglizer -v loglizer:/loglizer -it logpai/anaconda:py3.6 bash
$ cd /loglizer/demo

Run loglizer

You can try the demo scripts of loglizer on HDFS_100k.log_structured.csv as follows:

# Clone the project from Github
$ git clone https://github.com/logpai/loglizer.git

# Run PCA demo
$ cd loglizer/demo/
$ python PCA_demo.py

# Or run InvariantsMiner demo
$ python InvariantsMiner_demo.py

# If you want to apply loglizer to your own log data, and even have no label data, 
# you can follow the following script to run an unsupervised anomaly detection model. 
$ python PCA_demo_without_labels.py
$ python InvariantsMiner_demo_without_labels.py