A HYBRID MACHINE LEARNING-BASED PREDICTION APPROACH: THE ACCOUNTANT BEHAVIOR

Abstract

The integration of technology in accounting roles raises questions about the adaptability and skills of accountants in utilizing these tools effectively. Understanding how accountants' behavior is influenced by technology is crucial for their professional development and the accounting industry's future. This study focused on the development of a predictive model, leveraging both Naive Bayes and K-Nearest Neighbors (KNN) models. The research methodology involved the use of Pandas DataFrame to establish a robust framework for the dataset, incorporating both established and innovative features as input variables. These datasets were then utilized as the training data for the predictive model, with the primary objective of extracting valuable insights for decision-making and forecasting accountant behavior. The key findings of the study shed light on the performance of the different models employed. The Naïve Bayes model emerged as a standout performer, achieving an accuracy rate of 63% and an exceptional recall rate of 97%. This underscores its effectiveness in predicting accountant behavior, especially in identifying positive instances. On the other hand, the K-Nearest Neighbors model displayed a balanced trade-off between precision and recall, achieving an accuracy rate of 52% and an F1 score of 64%. This suggests that the model provides a reasonable compromise between accurately identifying positive cases and overall performance. Furthermore, the hybrid KNN-NB model, which amalgamates elements from both approaches, also achieved an accuracy rate of 52%. This finding indicates that the hybrid model has the potential to harness the strengths of both algorithms, offering a versatile approach to predicting accountant behavior.