Data Exploration

Before starting model building it is very important to understand the possibilities and limitations in the data. This data exploration can be done visually and using data mining algorithms.

Sarchitect has a wide range of graphical visualizations to enable visual analysis. The viewing options include: 2D/3D Chemical Structure viewer, 2D/3D Scatter Plots, Profile Plot, Heat Maps, Box-whisker Plots, Histograms, Matrix Plots, Bar Charts and Chemical Space views.

All views in Sarchitect are dynamically linked – selections in one view get highlighted all other open views.

Similarly, a range of powerful algorithms enable launching algorithmic inquiry.

For feature selection:

• Statistical Tests & Correlation
• Forward/Backward Selection
• GA Wrappers

For clustering:

• K-Means
• Hierarchical
• SOM
• Random Walk

Principal Component Analysis