Oracle database reports outliers detection

Project description
There were plenty of the alerts in the database and all of them were ignored by the development team due to an overwhelming quantity. The task was to find which of these alerts require attention and are related specifically to the server crash.
We solved this problem not through the alerts themselves, but through assessing the state of the running database instance analyzing the number of requests, disk IOPS for abnormal conditions. Database html-report parsing using python built-in libraries Data preparation and analytics using pandas, scipy, matplotlib and altair Independent anomalies research using classic anomaly-detection techniques like One-class SVM, Isolation forest and Local Outlier Factor via scikit-learn library Time-series-based approach for anomaly detection using statsmodel, fbprophet AWS-Elasticsearch log scraping via Kibana and Elasticsearch python-API.
Technologies
python, elasticsearch, kibana; numpy, pandas, scipy, matplotlib, altair, statsmodels, fbprophet; Git, GitHub