Mosgo
Mosgo: Mosgo is a web platform for the benchmarking of classification algorithms for security, developed in Java, AngularJS, D3 and Weka.
Challenges
- End-User Objective
The main objective of the Mosgo platform is to aid researchers and engineers in their choice of the most efficient algorithms for the anomaly detection. Mosgo provides intuitive tools to compare and contrast the performance the most common classification algorithms according to several different metrics.
- Scientific objective
Anomaly detection is a critical feature in many fields such as fraud detection, intrusion detection systems, an even used in the medical field. Generally these rely on statistical approaches derived from the perspective of artificial intelligence or bio-inspired computing. A major difficulty comes from the plethora number of anomaly detections algorithms . Hence it becomes very difficult to choose the most pertinent one based on one’s given task. In the field of Machine Learning, there are thousands of existing algorithms, with hundreds of new ones being published every year. The scientific objective of Mosgo project is to evaluate a set of classification algorithms used to detect anomalies, with the ultimate challenge of determining the most appropriate approach. In addition, the platform offers the possibility integrating business applications to the platform, due to its integrated components oriented architecture. Thus, it becomes possible to preemptively experiment with different algorithms before directly integrating into applications.
Features
- Addition of personal data set files.
- Implementation of major classification algorithms (AIRS, decision trees, naive Bayesian classification, perceptron, SVM, decision tables)
- Benchmarks realization
- Visualization of results in different metrics (time, false positive-negative, ROC curve...)
- Cross Comparison of results from different experiments
- Server logging
- Distributed execution of algorithms
Implementation
- Architecture
- Service-oriented architecture (SOA) - RESTful
- Technologies
- Developed in Java - Based on Weka Machine Learning library - Document-oriented database (CouchDB) - Frontend developed with AngularJS - Tomcat webserver