时间序列分析是概率统计学中应用性较强的一个分支，它是运用概率统计的理论和方法来分析随机数据序列（或称动态数据序列），并对其建立数学模型，进行参数估计，对模型定阶，以及进一步应用于预报、预测、自适应控制、最佳滤波等诸多方面。时间序列分析方法与随机过程理论有所区别，前者是先对实测数据建立数学模型，并在此基础上进一步分析随机数据的统计特性；后者是在对实测数据统计所得的先验概率知识基础上来分析其统计特性。由于人们所能获得的实测数据总是有限的，而理论上的先验概率要求在无限多的样本数据基础上统计才能获得，因此实际上能够获得的先验概率只能是在一定置信度条件下的近似，亦即尽量接近真实的概率（密度）分布，这是随机过程理论和方法在实际应用时的困难。时间序列分析方法可以克服这个困难，它是在有限个样本数据总量的情况下建立起相当精确的数学模型，从而获得具有一定精度（用模型误差方差来表示）的统计特性，与真实结果非常接近，因此在实际应用时比较方便，可操作性较好。

Time Series Analysis is a branch of Probablity and Statistics focus on practice such as forecast, predict, adaptive control etc. It adopts the theory of Probablity and Statistics to analysis random sequence data (dynamical sequence) through some certain mathematical models, to estimate the model's parameters, and the order of the model, etc.There are some difference between time series analysis and stochastic processes. For time series analysis, it tries to set up a mathematical model related to observed data, and then to analysis system's statistical properties. For stochastic processes, it tries to evaluate the prior probalities or joint probability distribution to describing the process. The prior probability evaluated on limited observed data is also a random variable, this makes it hard for practice. On contrast, time series tries to set up a precious mathematical model with the observed data. With such a model, we can extrapolate the process for pactice.