Anomaly Detection is the identification of rare items, events or observations which raise suspicions as these differ significantly from the majority class. Some examples of anomalies are transactional fraud in consumer financing, rare medical conditions such as malignancy in tumor or specific behaviour traits of employees or customers which are rare. Fraud Detection is a specific type of anomaly detection, which has the potential of causing losses to the company.
By nature, anomalies are difficult to detect since they can be rarest of the rare with an occurance in less than 5% in entire dataset. At Alletec, we have extensive experience in working on Anomaly Detection using Machine Learning techniques. There are three main approaches used to detect anomalies:
At Alletec, we’ve applied these approaches using a variety of techniques such as Density-based techniques (K-nearest neighbour, Isolation Forest), Subspace Correlation based and tensor-based outlier detection, One-class SVM, RNN, Bayesian Networks, Hidden Markov models, Cluster analysis-based outlier detection, association rules based techniques, Fuzzy Logic based techniques or Ensemble techniques.
Consumer Electronics companies attract customers with No-Cost EMI Promotional offers. Dealers and buyers often exploit such schemes to get undue advantage by selling them in open market in cash. We’ve helped several enterprise companies by detecting anomalies in their business processes using Machine Learning.