TMVA 4.0.1 is a toolkit for multivariate data analysis integrated into the ROOT framework. It provides a wide range of multivariate classification and regression algorithms, allowing users to train, test, and evaluate classifiers using user-friendly interfaces. Version 4 extends TMVA to support multivariate regression of a real-valued target vector, with regression handled through the same interfaces as classification. TMVA features flexible data handling, enabling the combination of MVA methods. A generalized boosting method is the first to benefit from the new framework. TMVA is developed and maintained by CERN, DESY, MPI-Kernphysik Heidelberg, University of Bonn, and University of Victoria, under a BSD license. It includes various methods such as rectangular cut optimization, projective likelihood estimation, multidimensional likelihood estimation, linear and nonlinear discriminant analysis, artificial neural networks, support vector machines, boosted decision trees, and predictive learning via rule ensembles. These methods are implemented in C++/ROOT and include auxiliary tools for parameter fitting and transformations. TMVA operates in a transparent factory mode to ensure unbiased performance comparisons between MVA methods. The Factory class manages user interaction with TMVA analysis steps, pre-processing data, and calculating correlations. The Reader class reads and interprets weight files, enabling the use of trained MVA methods in C++ executables, ROOT macros, or Python scripts. TMVA also generates lightweight C++ response classes for standalone use, which do not depend on TMVA or ROOT. TMVA supports both classification and regression tasks, with detailed descriptions of all methods and their configurations in the documentation. It provides training, testing, and performance evaluation algorithms, as well as visualization scripts. The toolkit includes example jobs for training and application, with macros for classification and regression. Users can run these examples using the TMVA Factory or Reader, and the results are stored in ROOT output files for further analysis. TMVA is compatible with various ROOT versions and can be used as a ROOT script, standalone executable, or Python script via PyROOT. The toolkit includes help resources, tutorials, and references for configuration options and methods. Users can also access the TMVA mailing list for support and feedback. The documentation provides detailed instructions on using TMVA for classification and regression tasks, including data preprocessing, variable definitions, and performance evaluation.TMVA 4.0.1 is a toolkit for multivariate data analysis integrated into the ROOT framework. It provides a wide range of multivariate classification and regression algorithms, allowing users to train, test, and evaluate classifiers using user-friendly interfaces. Version 4 extends TMVA to support multivariate regression of a real-valued target vector, with regression handled through the same interfaces as classification. TMVA features flexible data handling, enabling the combination of MVA methods. A generalized boosting method is the first to benefit from the new framework. TMVA is developed and maintained by CERN, DESY, MPI-Kernphysik Heidelberg, University of Bonn, and University of Victoria, under a BSD license. It includes various methods such as rectangular cut optimization, projective likelihood estimation, multidimensional likelihood estimation, linear and nonlinear discriminant analysis, artificial neural networks, support vector machines, boosted decision trees, and predictive learning via rule ensembles. These methods are implemented in C++/ROOT and include auxiliary tools for parameter fitting and transformations. TMVA operates in a transparent factory mode to ensure unbiased performance comparisons between MVA methods. The Factory class manages user interaction with TMVA analysis steps, pre-processing data, and calculating correlations. The Reader class reads and interprets weight files, enabling the use of trained MVA methods in C++ executables, ROOT macros, or Python scripts. TMVA also generates lightweight C++ response classes for standalone use, which do not depend on TMVA or ROOT. TMVA supports both classification and regression tasks, with detailed descriptions of all methods and their configurations in the documentation. It provides training, testing, and performance evaluation algorithms, as well as visualization scripts. The toolkit includes example jobs for training and application, with macros for classification and regression. Users can run these examples using the TMVA Factory or Reader, and the results are stored in ROOT output files for further analysis. TMVA is compatible with various ROOT versions and can be used as a ROOT script, standalone executable, or Python script via PyROOT. The toolkit includes help resources, tutorials, and references for configuration options and methods. Users can also access the TMVA mailing list for support and feedback. The documentation provides detailed instructions on using TMVA for classification and regression tasks, including data preprocessing, variable definitions, and performance evaluation.