Facebook
From Speedy Octupus, 3 Years ago, written in Python.
Embed
Download Paste or View Raw
Hits: 150
  1. train_x = aug_x
  2. train_y = aug_y
  3.  
  4. train_x = preprocessing.scale(train_x)
  5.  
  6. reducer = KernelPCA(n_components=300, kernel='rbf')
  7. X_embeded = reducer.fit_transform(train_x)
  8. print(X_embeded.shape)
  9.  
  10. cluster_num = 1
  11. clustering = KMeans(n_clusters=cluster_num).fit(X_embeded)
  12. for i in range(cluster_num):
  13.   print(np.count_nonzero(clustering.labels_ == i))
  14.  
  15. estimator_num = 200
  16. models = []
  17. for i in range(cluster_num):
  18.   trX, tsX, trY, tsY = train_test_split(train_x[clustering.labels_ == i], train_y[clustering.labels_ == i], test_size = 0.1, shuffle=True)
  19.   print("Training with", trX.shape[0], "Samples\nTesting with", tsX.shape[0])
  20.  
  21.  
  22.   test_data = tsX
  23.   train_data = trX
  24.  
  25.   rf = RandomForestRegressor(n_estimators = estimator_num)
  26.   #rf = RandomForestRegressor()
  27.   rf.fit(train_data, trY)
  28.  
  29.   predicted_rf = rf.predict(test_data)
  30.   predicted_rf_train = rf.predict(train_data)
  31.  
  32.  
  33.  
  34.   #gb = GradientBoostingRegressor(learning_rate=0.01, n_estimators=estimator_num)
  35.   #gb.fit(train_data.flatten().reshape(-1,1), trY.flatten())
  36.   #predicted_gb = gb.predict(test_data.flatten().reshape(-1, 1))
  37.  
  38.  
  39.  
  40.   models.append(rf)
  41.   #models.append(gb)
  42.  
  43.   actual = tsY.flatten() #melt t1 dataset (ground truth)
  44.   predicted = predicted_rf.flatten()#melt your prediction
  45.   print("RF Result =",mse(predicted,actual))#returns mse result for two melted matrices
  46.   #print("GB Result =",mse(predicted_gb.flatten(),actual))#returns mse result for two melted matrices
  47.   print("RF Train Result", mse(predicted_rf_train.flatten(),trY.flatten()))#returns mse result for two melted matrices