古人云“以史为鉴”，说的是吸取历史的经验教训，对未来的情况做出预判或者改变。生活中，亦是存在相似的利用历史数据对未来变化趋势进行预测分析的时间序列问题。本文就时间序列一类的问题进行研究，探讨如何更好地根据历史统计数据，对未来的变化趋势进行预测分析。本文基于神经网络，以气象观测历史数据作为研究的对象，建立了气温变化时序预测模型。本模型利用大数据相关技术对数据进行特征处理，通过深度神经网络，学习特征数据和标签数据之间复杂的非线性关系，从而实现对气温变化的趋势预测。实验结果表明，相较其他模型，本文的模型能够更好地进行时序预测，同时也证明了神经网络用于气象预测的可行性。

As the ancient saying goes, history acts as a mirror. Drawing lessons from historical and empirical experience can make changes and prejudgment to the futuristic situation. A similar time sequence problem, which is related to trend prediction and trend analysis through historical data, can also happen in our modern life. This article is a research on time sequence related issues, a research on how to analyze and predict the futuristic change trend through historical statistical data. Based on neural network technology, by taking meteorological observation data as a research object, this article establishes an air temperature change time sequence prediction model. This mod-elling carries on characteristic processing on historical data through big data technology and learns complex nonlinear relationship between characteristic data and tag data through deep neural network automatically. Thus, an air temperature change trend prediction can be achieved. The experimental results show that the model mentioned above has more advantages than other model on time sequence prediction, which proves the feasibility of neural networks in meteorological prediction.