Behind a successful ML project is the ability to do the computations then interpret outcomes and visualize results.
Interpret & Visualize
Model interpretation and visualization are built-in functions of the Toolkit library. To make it easy to explore results and understand what is going inside the machine, Flek includes a variety of components that enable the data scientist to:
-
Plot probabilities using various diagrams: bar, scatter, pie, heatmap, etc.
-
Generate a meaningful English statements for each Nugget or Probability stored in the probabilistic model.
-
Produce traceable predictions that associate to each score the rational and the rule used during classification.
-
Carry out tests using various prediction error or accuracy metrics
Knowing how complex it is to work with probabilities these 4 techniques are indispensable when trying to: (i) discover relationships between variables, (ii) solve classification problems or (iii) explore probabilistic results.
By providing a holistic approach to interpretation and visualization, Flek opens a new door for data scientists and programmers alike to tackle complex machine learning problems from all sides.