Cyber insurance has become increasingly important in today's digital landscape, as organizations face the growing threat of cyberattacks and data breaches. Accurate prediction of cyber insurance policy patterns can help insurance companies assess risk, set appropriate premiums, and develop effective coverage strategies. In this study, we propose a methodology that combines TF-IDF (Term Frequency-Inverse Document Frequency) feature extraction and multinomial naive Bayes classifier to predict cyber insurance policy patterns. We leverage the TF-IDF algorithm to represent policy documents as numerical feature vectors, which capture the importance of terms in the documents. The multinomial naive Bayes classifier is then employed to classify the policy patterns based on the extracted features.