Decision Tree Algorithm Moderately Coupled to PostgreSQL DBMS

Abstract

Using machine learning for data management is an extraordinary opportunity to move towards a leadership model based on information, which drives the organization towards success in each initiative. However, when incorporating these technologies, a company presents problems associated with the economic and administrative costs generated in this process since these are usually quite high, limiting their implementation in MSMEs. This paper proposes to integrate supervised machine learning techniques into PostgreSQL DBMS in a moderately coupled architecture to provide it with the capabilities of discovering knowledge in databases. Classification and regression algorithms were coupled by developing extensions using one of the procedural languages supported by PostgreSQL. Initially, the C4.5 decision tree classification algorithm was implemented using the PL/pgSQL procedural language. The main advantage of this strategy is that it considers the scalability, administration, and data manipulation of the DBMS. Since PostgreSQL is an open-source manager, organizations such as MSMEs will have a free tool that allows them to perform predictive analysis in order to improve their decision-making processes by anticipating future consumer behavior and making rational decisions based on their findings.

Keywords

• classification techniques
• C4.5 algorithm
• middle coupled architecture
• PostgreSQL DBMS