Step 1: Ingest the Data and Start the Automated Analysis
Our software decomposes the high-dimensional input data into a set of independent data patterns. Each pattern consists of data items that behave similarly in terms of a given target variable and are succinctly defined by just a small set of attributes, making them easy to understand. This analysis typically only takes a few minutes, or less.
Step 2: See the Data Patterns in the Context of Their Relevant Attributes
After the analyst has selected data patterns that appear interesting, the system produces an easy-to-read visual layout of these patterns in the context of their relevant attributes. In the image on the right, the green and red disks are patterns with unusually high or low values in the target attribute (disk radius encodes pattern size), and the triangles nearby are the attributes that define these them.
Step 3: Add the Causal Interactions Among the Data Patterns
Many data analytics packages compute correlations among attributes to indicate relationships. But correlation does not necessarily imply causation. Akai Kaeru’s causal inference engine establishes true causal relations which imply a direction like, for example, smoking causes cancer but not vice versa. Our visualization interface conveys the intricate web of causal relationships in a single intuitive map.
(Optional) Step 4: Use the Identified Patterns and Relations in Downstream Applications
Our software is available as a Jupyter Notebook plug-in and can be easily integrated into an existing data analyst’s workflow. Simply select one (or more) data pattern in the visualization interface and export it as a data frame into your downstream application, such as, training a neural network, decision tree, etc.
More Analytics tools
See How Patterns Change Over Time to Aid in Predictive Analysis
Patterns often change over time. Our immigration visualization interface allows analysts to observe these changes and gain valuable insight for predictive analysis. This interface is further boosted but our temporal causal analysis engine.
Form Hypotheses using our Pattern and Feature Engineering Suite
Feeling adventurous? Use our interactive pattern sculpting interface to view, edit and specify new patterns of interest. Then quickly test the effectiveness of new attributes, and features derived from them by examining their ability to form new patterns and causal relations.
many diverse application areas
A particular strength of our software is that it's very general and widely applicable. Here is a list of use cases where we successfully addressed the needs of the customer:
- Predict unusually high returns or losses in financial trading
- Unravel the interactions of risk factors leading to stroke
- Diagnose bias and fairness in data-driven decision making
- Identify population needs for government health insurance
- Optimize throughput and energy use of large compute clusters
- And several others
