Advanced but easy to use tool for predictive modeling and data mining.
GMDH Shell is a handy and reliable application designed for data mining and forecasting of multi-parametric datasets. It performs a fully automatic structural and parametric optimization of mathematical models that reveal underlying patterns in your data.GMDH Shell targets such topics as predictive analytics, predictive modeling, knowledge discovery, time series forecasting software, data mining tools, classification, regression, prediction and curve fitting. The software combines well proven machine learning technology and extended capabilities for effective use of multi-core, multiprocessor and cluster computers.GMDH Shell makes processing of data much easier. It is able automatically detect usable data in the file, transform data according to a problem type, drop irrelevant inputs and construct a set of predictive models at the base of optimal complexity detection and self-organization principles. Here are some key features of \"GMDH Shell\":
· Time series forecasting, classification and regression.
· GMDH-type neural networks, linear and non-linear combinatorial GMDH models.
· Feature ranking and Feature selection.
· Curve fitting and Function finding.
· Export formula to Excel.
· Encoding and binary decomposition of text variables.
· Handling of missing values.
· Save and load models and apply them to new data (Scoring).
· Background execution mode via command line.
· Dataset examples and preconfigured problem templates.
· Reading from TXT, CSV, XLS or XLSX files and file sets.
· One-click result recalculation for dynamically updated data files.
· Support for multi-core processors and cluster computers.
· Doesn\'t export to Excel
· Doens\'t save models
· No command-line interface
· 60 sec onds time limit on task processing What\'s New in This Release: [ read full changelog ]
· Solver: Improved quality of Neural-type models. Elite selection of neurons was replaced by bi-criteria selection composed of testing error minimization and population diversity maximization. Repeated processing of any project now often produce slightly different models depending on how parallel tasks ware distributed across CPU cores.
· Solver: Estimation of model coefficients using Robust regression (locked to MAE selection criterion).
· File explorer: Pie chart, 3D Bubbles (OpenGL required), expanded Statistics table.
· Preprocessor: Handling of spikes.
· Preprocessor: Generation of lags with certain step. For example, preprocessor command VarName@0-30:12 means generation of lags with step 12 (only two lags will be generated within the interval 0-30).
· New command line keys for batch-mode execution.
· Improved standard templates.
· Minor changes in user interface.