Tanagra (machine learning)
| Company type | Free software |
|---|---|
| Industry | Machine learning, Data mining, Statistical software |
| Founded | June 2003 |
| Founder | Ricco Rakotomalala |
| Headquarters | Lyon , France |
Area served | Worldwide (mainly French-speaking universities) |
| Products | Machine learning and data mining software package |
| Website | eric |
Tanagra is a no-cost machine learning software package created by Ricco Rakotomalala at Lumière University Lyon 2, France, mainly intended for academic research purposes. It encompasses essential data mining functionalities such as data visualization, descriptive statistics, instance and feature selection, feature construction, regression analysis, factor analysis, clustering, classification, and association rule mining.
As an educational initiative, Tanagra enjoys widespread use within French-speaking universities and is also employed in practical research projects and comparative studies of software tools.
History
The inception of Tanagra dates back to June 2003, with its initial version launched in December of the same year. It was developed as the successor to Sipina, a free data mining application centered on supervised learning and interactive decision tree creation; Sipina continues to be maintained and accessible. Tanagra operates under an open-source license, granting researchers access to its source code and the ability to contribute new algorithms.
The objective behind this project is to deliver an intuitive data mining tool for academics and students that complies with modern development practices—especially regarding user interface design—and supports analysis of both actual and simulated datasets.
From 2006 onward, Ricco Rakotomalala has considerably enhanced the documentation by releasing numerous tutorials on a dedicated platform. These materials explain statistical and machine learning techniques using Tanagra applied to real-world examples and include comparisons with other free data mining software to emphasize differences in result presentation.
Description
Users of Tanagra can construct data mining workflows visually through diagrams where each node corresponds to a specific statistical or machine learning method, and the links denote the flow of data. Unlike many workflow-oriented platforms, Tanagra employs a tree structure for processing steps, streamlining the design process. Outputs are generated in HTML format, which allows for straightforward export and viewing in web browsers or spreadsheet applications.
The software integrates a balance of statistical approaches (including both parametric and nonparametric tests), multivariate techniques (such as factor analysis, correspondence analysis, cluster analysis, and regression), alongside machine learning algorithms (including neural networks, support vector machines, decision trees, and random forests), ensuring flexibility across diverse data mining tasks.