/exp-anomaly-detector-AIOps

Using DNN for univariate time series anomaly detection over AIOps Competition dataset

Primary LanguageJupyter Notebook

exp-anomaly-detector-AIOps

Using DNN for univariate time series anomaly detection over AIOps Competition dataset

Please log in into http://iops.ai/competition_detail/?competition_id=5&flag=1 for downloading the input files: (unzip KPI异常检测决赛数据集.zip)

  • phase2_train.csv
  • phase2_ground_truth.hdf

Please check the "code" folder for details. The ipython notebook should be self-explanatory.

Please initialize your virtual environment with python 3.6 and then the following command in the terminal: $: pip install --upgrade pip $: pip install numpy tables pandas sklearn tensorflow keras matplotlib

Hope you enjoy it.

Some take away from the experiments:

  • The most critical factor that determines the result is the identification of the vital features
  • The second critical factor is the scale of the features, different scale methods lead to very distinct results
  • The tuning of the parameters is not as critical as expected, e.g., epoches and batch size in neural network training, and thresholds selection.