Target threat assessment is an important issue in command decision-making system. The naive Bayes, as an effective way to deal with uncertainty information, can be used to evaluate the target threat level. Unfortunately, most existing threat assessment methods base on naive Bayes have the main problem of ignoring the assumption of conditional independence under incomplete data, which does not meet the requirements of target threat assessment in complex battlefield situation. To solve this main problem, we propose a target threat assessment method using weighted naive Bayes under incomplete data. The core parts are each missing sample is replaced with the maximum weight fraction sample obtained from the E-step of expectation maximization algorithm and a weighted conditional log-likelihood function is constructed to learn the attribute weight of weighted naive Bayes. Additionally, an adaptive differential evolution is provided to solve the optimization problem of attribute weighting. The effectiveness of the proposed method has been validated through two types of experiments, including sample set test and specific case test. The results show that the proposed method is superior to the competitors, and can improve the accuracy and belief of target threat assessment.