Skip to content

Latest commit

 

History

History
102 lines (80 loc) · 5.3 KB

README.md

File metadata and controls

102 lines (80 loc) · 5.3 KB

[Project Logo]

Anomaly Detection System

  • anomaly detection system to detect outliers in server nodes using Gaussian model
  • implemented on two sets of data
    1. dataset with two features: Latency and Throughput
    2. higher dimensional dataset with a set of 11 features

Dependencies

  • jupyter with python 3.10.6 64-bit kernel
  • python libraries: pandas, matplotlib, numpy, pylab

Dataset

  1. 2-dimensional dataset
  • data_train.csv: training set for the 2-D data
  • data_val.csv: validation set including X_val and y_val; separated after loading dataset
  1. 11-dimensional dataset
  • data2_X.csv: training set with 11 features
  • data2_Xval.csv: validation set with 11 features
  • data2_yval.csv: validation labels to assess the model trained on the higher dimensional dataset

Performance Measures

  • performance of this unsupervised Gaussian model is determined using the F1-score for every iteration till it saturates to its optimum
  • the model is trained over the training set to learn the parameters mu and sigmasq denoting the dataset's mean and variance
  • validation set is fed to the model using the learnt mu and sigmasq to learn the new parameter epsilon by maximizing the F1 score
  • using the learnt parameter epsilon outliers are detected in the dataset

Performance Goals

  1. 2-dimensional dataset
  • optimum epsilon expected: 8.99e-05
  • best F1 score: 0.875000
  1. 11-dimensional dataset
  • optimum epsilon expected: 1.38e-18
  • best F1 score: 0.615385

DIVERSION 2k23

This project is part of Diversion 2k23

DIVERSION 2k23 🎉

GitHub issues GitHub PRs GitHub forks

GitHub contributors License Badge

🗣 DIVERSION 2k23 encourages participation in the open source community, which grows bigger every year. Complete the challenge and win amazing rewards.

📢 Register here for DIVERSION 2k23 and start contributing to the projects.

Features to be developed during the event.

[Add Feature list]

DIVERSION 2k23 Contribution Rules

  • Pull requests can be submitted to any opted-in repository on GitHub or GitLab.
  • The pull request must contain commits you made yourself.
  • If a maintainer reports your pull request as spam, it will not be counted toward your participation in Hacktoberfest.
  • If a maintainer reports behavior that’s not in line with the project’s code of conduct, you will be ineligible to participate.
  • To get a shirt, you must make four approved pull requests (PRs) on opted-in projects between October 1-31 in any time zone.
  • This year, the first 40,000 participants can earn a T-shirt.

Contribute

There are DIVERSION 2k23 contributor guidelines as well as individual project guidelines.

Do Read - Code of Conduct

Jump to - DIVERSION 2k23 guidelines (updated)

Why not start your open source journey by merging some codes here. It will be very great if you review the code and find some bugs and issues.

Steps for a successful commit ⚡

  • Choose any of the Ways to contribute for contribution.
  • Create a PR and commit by creating a new branch (:warning: Avoid commiting to the main)
  • Be patience it might take time for the maintainers to review your PR.
  • Congratulations on a successful merge.
  • You can add your name to contributors list.

Ways to Contribute

  • Find a bug by running the project locally
  • Solve the issues which are already set up
  • Sugget some change or create your own isssue
  • Create documentation like site structure, how to write a blog,etc
  • Indent/Beautify/Modify codes.

List of Contributors

Thank You! Keep Contributing :octocat: