zhaozhengcoder
diff --git a/‎序列预测/PCA去趋势化/__pycache__/pca.cpython-35.pyc
1.32 KB b/‎序列预测/PCA去趋势化/__pycache__/pca.cpython-35.pyc
1.32 KB
diff --git a/‎序列预测/PCA去趋势化/dump_arr.txt
90.2 KB b/‎序列预测/PCA去趋势化/dump_arr.txt
90.2 KB
diff --git a/‎序列预测/PCA去趋势化/dump_main_x.txt
90.2 KB b/‎序列预测/PCA去趋势化/dump_main_x.txt
90.2 KB
diff --git a/‎序列预测/PCA去趋势化/dump_rest_x.txt
90.2 KB b/‎序列预测/PCA去趋势化/dump_rest_x.txt
90.2 KB
diff --git a/‎序列预测/PCA去趋势化/pca.py
Lines changed: 33 additions & 0 deletions b/‎序列预测/PCA去趋势化/pca.py
Lines changed: 33 additions & 0 deletions
diff --git a/‎序列预测/PCA去趋势化/pre_porcess.py
Lines changed: 20 additions & 0 deletions b/‎序列预测/PCA去趋势化/pre_porcess.py
Lines changed: 20 additions & 0 deletions
diff --git a/‎序列预测/PCA去趋势化/show.py
Lines changed: 32 additions & 0 deletions b/‎序列预测/PCA去趋势化/show.py
Lines changed: 32 additions & 0 deletions
@@ -0,0 +1,33 @@
+import numpy as np
+
+class PCA:
+    def __init__(self, x, retain):
+        # X=m*n m为样本个数，n为样本的维度
+        # X必须均值为0,只有在均值为0时，协方差sigma=xT*x=[n,m]*[m,n]=[n,n]
+        self.x = x
+        m, n = x.shape
+        sigma = np.matmul(self.x.T, self.x) / m  # sigma是协方差矩阵，T是转置,sigma=xT*x=[n,m]*[m,n]=[n,n]
+        self.u, self.s, _ = np.linalg.svd(sigma)  # u的每一列都是特征向量，s是特征值与u一一对应
+        # here iteration is over rows but the columns are the eigenvectors of sigma
+        # u_sum = np.cumsum(self.s)
+        self.retain_num = retain
+        # for i in range(n):
+        #     if u_sum[i] / u_sum[-1] >= retain:
+        #         self.retain_num = i
+        #         break
+        self.main_vector = self.u[:, 0:self.retain_num + 1]
+        self.rest_vector = self.u[:, self.retain_num + 1:]
+        main_x_rot = np.matmul(self.x, self.main_vector)
+        rest_x_rot = np.matmul(self.x, self.rest_vector)
+        self.main_x = np.matmul(main_x_rot, self.main_vector.T)
+        self.rest_x = np.matmul(rest_x_rot, self.rest_vector.T)
+
+    def reduce(self, data):
+        main_data_rot = np.matmul(data, self.main_vector)
+        rest_data_rot = np.matmul(data, self.rest_vector)
+        data_main = np.matmul(main_data_rot, self.main_vector.T)
+        data_rest = np.matmul(rest_data_rot, self.rest_vector.T)
+        return data_main, data_rest
+
+    def reconstruct(self, rest_x):
+        return self.main_x + rest_x
@@ -4,6 +4,8 @@
 import numpy as np 
 import matplotlib.pyplot as plt 
 import pickle
+import pca
+import pickle
 
 def read_excel(filepath):
     df = pd.read_excel(filepath,skip_footer=1)
@@ -108,6 +110,24 @@ def pre_process():
     arr = transfer(dfs)
     return arr
 
+
+def mypickle(filepath,data):
+    abs_path='C:/Users/wwwa8/Documents/GitHub/Machine-Learning/序列预测/PCA去趋势化/'
+    f=open(abs_path+filepath,'wb')
+    pickle.dump(data,f)
+    f.close()
+
 if __name__ =="__main__":
     arr = pre_process()
     print ("shape : ",arr.shape)
+    pca_obj=pca.PCA(arr,2)
+    #print (pca_obj.main_x.shape)
+    #print (pca_obj.rest_x.shape)
+
+    mypickle('dump_arr.txt',arr)
+    mypickle('dump_main_x.txt',pca_obj.main_x)
+    mypickle('dump_rest_x.txt',pca_obj.rest_x)
+
+    
+
+
@@ -0,0 +1,32 @@
+import pandas as pd 
+from datetime import datetime 
+import os 
+import numpy as np 
+import matplotlib.pyplot as plt 
+import pickle
+import pca
+import pickle
+
+def myload(filename):
+    abs_path='C:/Users/wwwa8/Documents/GitHub/Machine-Learning/序列预测/PCA去趋势化/'
+    f = open(abs_path+filename,'rb')
+    data =pickle.load(f)
+    f.close()
+    return data
+
+
+def show(main_x,rest_x):
+    #for item in main_x:
+    #    plt.plot(item)
+    for item in rest_x:
+        plt.plot(item)
+    plt.show()
+    
+
+if __name__=="__main__":
+    arr = myload("dump_arr.txt")
+    main_x = myload("dump_main_x.txt")
+    rest_x = myload("dump_rest_x.txt")
+    show(arr,arr)
+
+