Rule Generation from Several Types of Table Data Sets and Its Application: Decision-Making with Transparency and an Improved Execution Environment

Abstract

This paper copes with rule generation from table data sets and applies the obtained rules to decision support. Here, two types of table data sets are considered. One type of them is specified as a Deterministic Information System (DIS). The other type is specified as a Non-deterministic Information System (NIS) for dealing with incomplete information. Two rule generation algorithms are refined and newly implemented in Python. Every obtained rule is applied as evidence of decision-making. Therefore, the reasoning process preserves its transparency, which will be an essential characteristic for Explainable AI. The decision support environment is strengthened due to some described improvements and is also brushed up in Python. Some running videos of Python are available on the web page. This framework applies to almost any table data sets, and we can generate rules from them. This framework based on discrete data will complement statistical data analysis based on numerical data.