Property Inspection
===================

The "Property EDA" (*Property Exploration Data Analysis*) panel enables users to visualize six histograms for each property. This property can be selected from the dropdown menu at the top of the panel. The histograms are as follows:

 - (1-2) The histogram for all data excluding missing values, for both positive and negative points.
 - (3-4) The histogram for positive and negative points within the training dataset.
 - (5-6) The histogram for positive and negative points within the testing dataset.

.. raw:: html
    :file: ./images/property_inspection/property_inspection.html

*"Property EDA" panel allows users to visualize histograms for each property*

Users can toggle the visibility of these histograms by clicking on the legend of the figure. It is important to note that histograms are scaled, as the data are often significantly imbalanced. These visualizations assist users in making several critical decisions regarding each property, such as:

 - Does this property contribute meaningfully to the prediction, demonstrating good statistical representativity? Indeed, a property that does not present a meaningful statistical distribution might not be beneficial for the model.
 - Do the training and testing datasets exhibit similar distributions? Discrepancies could be attributed to local geological contexts and may introduce bias into the algorithm.
 - Is the property effective at distinguishing positive values? Overly discriminative properties may be derived from the labels themselves, potentially leading to trivial results (for example, the mapping of proximal alteration, which is a target, not a feature to be used for targeting).