Stylianos Kampakis

In this series of liveProjects, you’ll learn different methods for detecting anomalies and outliers with Python machine learning techniques. Anomaly detection is a vital tool for tasks like spotting medical problems, and even detecting seismic events like earthquakes. You’ll explore both supervised and unsupervised learning methods for anomaly detection to master this valuable ML task.

In this liveProject, you’ll utilize undersampling techniques to balance out a seismic activity dataset. To balance this dataset, you will utilize the ClusterCentroids, NearMiss and CondensedNearestNeighbor algorithms to downsample the majority class. Then, the performance is compared using random forest, logistic regression and Naive Bayes binary classification algorithms.

In this liveProject, you’ll go hands-on with supervised learning methods for anomaly detection. You’ll explore an imbalanced dataset of seismic activity. To balance this dataset you will utilize the SMOTE and ADASYN oversampling algorithms to both generate synthetic examples of the minority class and then compare performance using random forest, logistic regression and Naive Bayes binary classification algorithms.

In this liveProject, you’ll explore a dataset with more variables and use scikit-learn and the PyOD library to build an unsupervised machine learning model for detecting cardiac arrhythmias. You’ll develop an algorithm which will detect arrhythmias from device data like EEG, using the Locally Selective Combination in Parallel Outlier Ensembles (LSCP) algorithm. A LSCP model accepts input as various other algorithms, and can be used to set up detectors with different settings.

In this liveProject, you’ll use scikit-learn and the PyOD library to build an unsupervised machine learning model for detecting hyperthyroidism. PyOD is a Python toolkit for detecting outlying objects in multivariate data. You’ll compare performance between four different anomaly detection methods on a specialized thyroid dataset: PCA, Clustering-Based Local Outlier Factor (CBLOF), Histogram-Based Outlier Score (HBOS), and KNN algorithms.

In this liveProject, you’ll explore the basics of anomaly detection by analyzing a medical dataset using unsupervised learning. You’ll create a model that can determine whether patients referred to a clinic have abnormal thyroid function. To accomplish this, you’ll download and prepare your dataset, and then utilize scikit-learn to compare different anomaly detection algorithms to find the most effective. You are going to use Isolation Forest, the Local Outlier Factor (LOF), One-Class SVM and Robust Covariance.