changes

Clinton/Linear_reg_reg.py

@@ -8,25 +8,28 @@
 import pandas as pd
 import numpy as np
 
-def error(w_hat):
+def error(w_hat, xe, ye):
     tr_error =  np.empty([1,1])
-    for i in range(x.shape[0])  :
-        tr_error += np.power((y[i] - np.dot(w_hat.T,x[i])), 2) 
-    return tr_error/x.shape[0]
+    for i in range(xe.shape[0])  :
+        tr_error += np.power((ye[i] - np.dot(w_hat.T,xe[i])), 2) 
+    return tr_error/xe.shape[0]
 
 df = pd.read_csv(r'Book3.csv', header=None)
 
 df_rest = df.loc[df.index.isin(df.index)]
-df_master=np.empty([1,1])
+df_master=[]
 
 for o in range (10):
     df_master.append(df_rest.sample(n=8, axis =0))
     df_rest = df_rest.loc[~df_rest.index.isin(df_master[o].index)]
 
 tr_error= []
-
-for p in range(len(df_master)-1):
-    df_num = df_master[p].values
+tst_error=[]
+
+
+for p in range(len(df_master)):
+    temp2 = pd.concat(df_master[:p] + df_master[p+1:])
+    df_num = temp2.values
     x = np.empty([df_num.shape[0],df_num.shape[1]-1])
     y = np.empty([df_num.shape[0],1])
 
@@ -50,6 +53,30 @@ def error(w_hat):
 
     w_hat_ridge = np.dot(np.dot(np.linalg.inv(np.dot(x_t,x) + np.multiply(lam,np.identity(int(x.shape[1])))),x_t),y)
 
-    tr_error.append(error(w_hat_ridge))
+    tr_error.append(error(w_hat_ridge, x, y))
+
+    #validation
+    df_numv = df_master[p].values
+    xv = np.empty([df_numv.shape[0],df_numv.shape[1]-1])
+    yv = np.empty([df_numv.shape[0],1])
+
+
+    for qv in range(df_numv.shape[1]-1):
+        x_meanv = np.mean(df_numv[:,[qv]], dtype=np.float64)
+        x_stdv = np.std(df_numv[:,[qv]])
+        # replace each row in given column with normalized data
+        for rv in range(df_numv.shape[0]):
+            xv[rv][qv] = (df_numv[rv][qv]-x_meanv)/x_stdv
+
+
+    x_tv = xv.T
+
+    for sv in range (df_numv.shape[0]):
+        y_meanv = np.mean(df_numv[:,[df_numv.shape[1]-1]], dtype=np.float64)
+        y_stdv = np.std(df_numv[:,[df_numv.shape[1]-1]])
+        yv[sv][0] = (df_numv[sv,[df_numv.shape[1]-1]]-y_meanv)/y_stdv
+
+    lamv =[0.0]
 
-val_error.append
+    tst_error.append(error(w_hat_ridge, xv, yv))
+