From Comments to Insight: Predictive Classification of Organizational Cultural Entropy Using SBERT, K-Means, and Logistic Regression

Abstrak

This study aims to develop a machine learning-based predictive model based on clustered data to identify cultural entropy in organizations through the analysis of open-ended comments on employee perception surveys of superiors. energy used for unproductive activities in a work environment. Entropy shows the level of conflict, friction and frustration in the environment. With a text mining approach, answers to open-ended questions in the cultural entropy survey were processed with Sentence-BERT and clustered using the K-Means algorithm into two categories, namely cultural entropy and non-cultural entropy. The dataset that already has labels from the clustering results is used to develop a classification model. The algorithms used are Random Forest, Logistic Regression, and Support Vector Machine (SVM), which are evaluated through accuracy, precision, recall, and F1-score metrics and a confusion matrix. The results show that Logistic Regression provides the best performance with an accuracy of 0.985, a precision of 1.00, and an F1-score of 0.978 without any classification errors. These findings indicate that the clustering approach followed by machine learning-based predictive is effective in identifying organizational cultural entropy. This can be used to design appropriate interventions and as an early detection system for cultural entropy in human resource management