Exploring Complex Decision Trees: Unveiling Data Patterns and Optimal Predictive Power

Abstract

This research investigates the development and analysis of decision tree models in the context of classification tasks. Decision tree models were developed without employing pruning or pre-pruning techniques and were tested on relevant datasets. The research findings demonstrate that complex models without pruning achieved the highest level of accuracy in classifying data. This study was inspired by the potential issue of students facing the risk of not completing their studies (dropout), which could lead to a decline in the college's accreditation rating. Therefore, this model was devised to assist in identifying factors that could influence this outcome as a preventative measure. Additionally, we successfully generated clear visualizations of the decision trees, enhancing the understanding of the model's decision-making process. This research provides insights into the adaptability of decision tree models within this specific case and showcases their potential for enhancing decision-making across various contexts. These findings encourage further discussions on the benefits of pruning methods within this specific context and the broader application potential of decision tree models.