cafeltk / anomaly-detection-using-unsupervised-learning Goto Github PK

This project forked from sdave-connexion/anomaly-detection-using-unsupervised-learning

Outiers are rare but are very crucial. In this project, several methods to detect anomalies using Unsupervised Learning where no labelled dataset is given is presented. This work was done between August 2019- November 2019. This later on served as the base project for the Master Thesis which is available in other repository. Unfortunately, I am not open to share code for this one but for master thesis code is public. Hope it helps. As we are moving towards the Industry 4.0 era where Artificial Intelligence(AI) and the Internet of Things(IoT) are crucial and integral parts of the revolution. In this transition phase from manual to the automation of work using different machines, sensors are a very important component and they play a vital role in the setup. The connectivity and flow of data/ information between sensors and devices leads us to witness rapid growth of time-based data are known as time series. In this project we will be implementing the techniques and applications of machine learning and statistical analysis, getting familiar with pandas, matplotlib, NumPy and various other libraries using Python on available sensor data from industries and extract useful information and make it possible to detect outliers and perform conditional monitoring which in-turn will help in reducing cost, optimizing manual labour capacity, increase productivity, availability, reliability and keep downtime minimum. The main aim of the Research Project is to develop online multivariate analysis tool which fetches the data, impute the missing data, eliminates outliers and non- compliant data, perform unsupervised learning and inform the user in case of abnormality i.e., out of control situations.