脉冲星是一种能够发出周期性脉冲电波的中子星。1967年10月，科学家首次发现该天体并命名为脉冲星。将射电望远镜巡天搜寻到的周期性脉冲信号经过一系列人工信号处理后，得到可能为脉冲星的信号，称为脉冲星候选体。其中很大部分为干扰信号，比如人为产生的电磁波信号和星际介质闪烁产生的高频脉冲，因此需要对脉冲星候选体进行仔细谨慎的筛选，找出其中真正的脉冲星信号。 对超大量级的脉冲星候选体进行筛选将是一项工作量巨大的项目。传统的人工识别筛选方法已经无法保证对候选体数据处理的时效性和准确性。目前，支持向量机已经是机器学习领域中的大热门，在模式识别中处理小样本、非线性和高维度问题时取得良好效果，而且具有很强的推广能力，是筛选脉冲星候选体的有效工具。 本文运用PCA算法和LIBLINEAR工具包给出的SVM分类器在MATLAB上实现对部分脉冲星候选体筛选仿真实验，通过调整SVM分类器参数和分组训练集和预测集找出最优的分类器，实验表明，SVM分类器搭配PCA算法在对脉冲星候选体筛选上具有极高的准确率并且计算时间较短。

Pulsars are neutron stars that emit periodic pulsed waves. In October 1967, scientists first discovered the object and named it pulsar. The periodic pulse signal searched by the radio telescope through a series of artificial signals is processed to obtain a possible pulsar signal, which is called a pulsar candidate. Most of these are interference signals, such as artificially generated electromagnetic signals and high-frequency pulses generated by scintillation in interstellar medium. Therefore, careful selection of pulsar candidates is needed to find the true pulsar signal. The screening of super-mass pulsar candidates will be a huge workload. Traditional manual identification screening methods cannot guarantee the timeliness and accuracy of candidate data processing. At present, the support vector machine has been a hot topic in the field of machine learning. It achieves good results when dealing with small samples, nonlinear and high-dimensional problems in pattern recognition, and it has strong promotion ability. It is effective for screening pulsar candidates. tool. In this paper, a SVM classifier based on PCA algorithm and LIBLINEAR toolkit is applied to MATLAB partial filter simulation experiments. The optimal classifier is found by adjusting the SVM classifier parameters and the rational grouping training set and prediction set. The experiment shows that the SVM classifier with the PCA algorithm has a very high accuracy and short calculation time for the selection of the pulsar candidate.