脉冲星是弱射电源，需要灵敏度高的射电望远镜，FAST应运而生。但FAST搜寻流程中消色散计算DM使用穷举方法；候选体筛选过程通过滤波器过滤信号造成大量漏选，且需要人工筛选非常耗时。因此，如何改进流程得到快速准确的预测模型成为一个要重点研究的问题。同时，由于脉冲星搜寻规格水平迅速提升导致得到的脉冲星数据量以及候选数量成指数水平增长，原有方法的不利影响越来越大，急需新的快速数据处理方法来提高搜寻速度。本项目计划改进脉冲星的搜索流程从而得到快速准确的预测模型，同时研究新的人工智能算法解决候选体筛选技术的不足。着重研究改进智能快速算法并应用到候选体筛选过程中；将非线性DM计算问题转为线性问题解决穷举耗时问题；将Spark的并行优化技术和集成学习方法应用到搜寻流程并行化过程中，为加速科学发现提供强有力的支撑技术。

The pulsar is a weak radio source, which requires high sensitivity radio telescopes, and FAST emerges as the times require. But in the FAST search process, the DM of the dispersion calculation is exhaustive. The candidate screening process filters out the signal through the filter, resulting in a large number of missed entries, and it needs artificial screening, which is a waste of time. Therefore, how to improve the process to get a fast and accurate prediction model has become a key problem to be studied. Meanwhile, due to the rapid improvement of pulsar search specification level, the number and number of pulsar data increase exponentially. The adverse effects of the original method are also increasing. New data processing methods are urgently needed to improve the search speed. This project improves the pulsar search process from the perspective of big data processing, and gets a fast and accurate prediction model. At the same time, we study the new AI algorithm to solve the shortcomings of candidate selection technology. It focuses on improving the fast algorithm in AI technology and applying it to the process of candidate screening. It transforms the nonlinear DM computing problem into a linear problem and solves the exhaustive time consuming problem, and applies the optimization technology of Spark and ensemble learning method to the process to provide strong support technology for speeding up scientific discovery.