Prediction Model of Strong Electromagnetic Effect Phenomenon Based on Complement Naive Bayes

Strong electromagnetic effects imperil the safety of shipboard electronic systems and components. To ensure their safety, it is necessary to predict strong electromagnetic effect phenomena. However, the acquisition and simulation of strong electromagnetic effect data are troublesome, resulting in the problem of small sample sizes and sample imbalance when predicting phenomena involving strong electromagnetic effects. To resolve the aforementioned issues, this paper proposes a prediction model based on the CNB classifier. The model selects the features, concatenates the discrete features prior to data processing, and then uses the term frequency–inverse document frequency (TF-IDF) algorithm to encode the feature word weights. The vectors are finally fed to a CNB classifier calibrated with an isotonic probability for prediction. In experiments with traditional plain Bayes, the Isotonic probabilistic calibrated CNB classifier achieves an accuracy score of 0.94 on the strong EM effect dataset. This result demonstrates the model's superior performance. The model avoids the problems of small sample size influencing model performance and the tendency of the classifier to favor large categories while ignoring small ones, safeguarding against the prediction of the strong EM effect phenomenon.