Novasign - Hybrid Modeling Toolbox
Unleash the power of advanced machine learning with our Hybrid Model Toolbox. This standalone application, effortlessly installable on your PC, revolutionizes how you create, train, and evaluate hybrid models.
Data Import
Easily import datasets directly from your PC with a single click. Our toolbox is compatible with widely-used formats like Excel and CSV, capable of handling large datasets, including Excel files with multiple tabs for streamlined data integration.
Data Grid Exploration
Advanced Data Plotting
-
Run Splitting
When dataset contains multiple runs, splitting is provided out of the box, being able to visualize your runs as independent time series without pre-processing them.
-
Dynamic Selection
Select variable of interest with the simple selectors. Plot variables against time axis or other variables to explore visual correlations
-
Multiplot
Plot multiple times in the same active figure, overlaying plots within one.
-
Flexible Export Options
Export generated figures to PDF, SVG, PNG, C#, Text...
-
Customizable View
Setup figure axes, titles, legends, margins...
After creation, dive into fine-tuning your model. Adjust configurations, equations, and datasets as needed.
Our Toolbox’s flexibility allows for comprehensive retraining and optimization of your models
-
Integrative
Models that predict outcomes over time by integrating successive predictions, ideal for scenarios where continuous monitoring and forecasting are essential
-
Autoregressive
Models that use their previous outputs as inputs for future predictions, enhancing their forecasting accuracy, especially in time-dependent scenarios.
-
Mass Balances
Models capable of automatically adjusting for changes in mass or concentration due to additions or removals in a system, ensuring accurate mass conservation
-
Custom ANN
Flexible neural networks where you can tailor inputs, hidden layers, transfer functions, and even individual neuron connections, offering a high degree of control and customization
-
Bootstrapping
Models trained on various distributions of training, validation, and testing data sets, increasing robustness and reducing overfitting
-
Mechanistic
Models that combine the predictability of parametric models with the adaptability of non-parametric neural networks, suitable for applications needing both mechanistic understanding and predictive power