PCA

用于数据自编码，非线性压缩，降维的一种方法；与PCA不同，pca为线性降维方式(Data auto encoding, nonlinear compression)

说明：  隐马尔科夫实现，包含forward-hmm, Viterbi-hmm, Baum-Welch-hmm(Hidden Markov implementation, including forward-hmm, Viterbi-hmm, Baum-Welch-hmm)

用EM算法求解高斯混合模型并可视化，数据是男女生的身高分布，前提是初始化男女生身高各自的均值和方差和比例，然后由EM算法求解，男女生身高的均值方差，以拟合数据。(The EM algorithm is used to solve the Gauss mixture model and visualize. The data is the height distribution of male and female. The premise is to initialize the mean, variance and proportion of the male and female height, then the mean variance of the height of male and female is solved by the EM algorithm, so as to fit the data.)

遗传算法优化小波神经网络的源程序 1 构造的非线性函数 位于nninit_test m 2 直接用WNN逼近非线性 Wnn_test()

隐马尔科夫实现，包含forward-hmm, Viterbi-hmm, Baum-Welch-hmm(Hidden Markov implementation, including forward-hmm, Viterbi-hmm, Baum-Welch-hmm)

Numerical Recipes Software 提供的算法子程序集,一共有上千种算法,部分算法可以在书店买到书()

利用python调用新浪api接口实现的新浪微博数据的挖掘，能够获取指定的经纬度地点的微博动态信息，还能够将其写入MySQL数据库中。

从数据库获取车辆在一段时间内的所有行驶记录的相关数据，确定所需数据为GPS经纬度坐标和驾驶时长等，QB模型采用MDF的思想,其基本思想为:通过平均直接翻转距离函数定义两条轨迹之间的距离，两条轨迹需要具有相同的经纬度点数，具有相同点数的轨迹最大的优点是对轨迹距离成对计算，且相同轨迹之间具有更高的分辨率，对于轨迹聚类的结果有一定的优化。(Retrieved from the database cars all over a period of time, record the related data, determine the required data for the GPS latitude and longitude coordinates, and the driving time, QB model by adopting the idea of MDF, its basic idea is: flip directly by the average distance function definition of the distance between two trajectories, two tracks will have the same latitude and longitude points, and has the biggest advantages of the same points of trajectory track distance calculation in pairs, and has higher resolution, between the same trajectory for trajectory clustering results have certain optimization.)

多重分形去趋势波动分析法，用于不同时间序列的重分形交叉相关性分析。(Multifractal detrended fluctuation analysis)

将数据集转换为拉普拉斯矩阵，然后利用基于图论的谱聚类进行聚类。拉普拉斯矩阵采用高斯核函数，全连接方法计算。谱聚类擅长处理高维数据或非凸数据集。(The data set is transformed into Laplacian matrix, and then clustered by spectral clustering based on graph theory. The Laplacian matrix is calculated by using the Gauss kernel function and the full connection method. Spectral clustering is good at dealing with high-dimensional or non-convex data sets.)