OPUS-C Modeling Process
STEP 1: MODELING
Creating the Causal Graph
Domain expertise on a particular type of process (business, supply chain, manufacturing), etc. is used to create a Causal Model. This model is then trained using observable data in the real world.
​
-
A Causal Graphical Tool will be provided to capture the Causal Graph
-
Example to the right is a high level causal graph of how a piece of equipment from the electronics industry functions. It may be used to answer questions on how various factors (nodes) can impact the 3 types of output (Accuracy, UPH and Errors).
Causal Discovery
In most cases, the Causal Graph is not fully known. Hence the need for CAUSAL DISCOVERY:
-
MANUAL—Domain expert (user) creates the causal graph and specifies formulae/constraints on the node interactions
-
SEMI AUTO - Domain expert sets certain restrictions on the node interactions (e.g. linear relationship) but does not specify formulae for their interaction. OPUS-C figures out the rest.
-
FULL AUTO—OPUS-C computes the most likely Causal Model based on the data at hand.
STEP 2 : DATA CLEANING & FITTING
OPUS-C comes with an option for data integration, a critical step in the modeling process. This includes the usual data cleaning/scrubbing/filtering/processing activities that run on servers, either on-site or on a cloud. Once the data has been cleaned, it is stored in a Data Warehouse, which then becomes the source for the data fitting step to create the causal model.