為什么 knn 總是預測相同的數字？我該如何解決這個問題？資料集在這里。
代碼：
import numpy as np
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import os
import scipy.io
from sklearn.neighbors import KNeighborsClassifier
from sklearn import metrics
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from torch.utils.data import Dataset, DataLoader
from sklearn import preprocessing
import torch
import numpy as np
from sklearn.model_selection import KFold
from sklearn.neighbors import KNeighborsClassifier
from sklearn import metrics
def load_mat_data(path):
mat = scipy.io.loadmat(DATA_PATH)
x,y = mat['data'], mat['class']
x = x.astype('float32')
# stadardize values
standardizer = preprocessing.StandardScaler()
x = standardizer.fit_transform(x)
return x, standardizer, y
def numpyToTensor(x):
x_train = torch.from_numpy(x)
return x_train
class DataBuilder(Dataset):
def __init__(self, path):
self.x, self.standardizer, self.y = load_mat_data(DATA_PATH)
self.x = numpyToTensor(self.x)
self.len=self.x.shape[0]
self.y = numpyToTensor(self.y)
def __getitem__(self,index):
return (self.x[index], self.y[index])
def __len__(self):
return self.len
datasets = ['/home/katerina/Desktop/datasets/GSE75110.mat']
for DATA_PATH in datasets:
print(DATA_PATH)
data_set=DataBuilder(DATA_PATH)
pred_rpknn = [0] * len(data_set.y)
kf = KFold(n_splits=10, shuffle = True, random_state=7)
for train_index, test_index in kf.split(data_set.x):
#Create KNN Classifier
knn = KNeighborsClassifier(n_neighbors=5)
#print("TRAIN:", train_index, "TEST:", test_index)
x_train, x_test = data_set.x[train_index], data_set.x[test_index]
y_train, y_test = data_set.y[train_index], data_set.y[test_index]
#Train the model using the training sets
y1_train = y_train.ravel()
knn.fit(x_train, y1_train)
#Predict the response for test dataset
y_pred = knn.predict(x_test)
#print(y_pred)
# Model Accuracy, how often is the classifier correct?
print("Accuracy:",metrics.accuracy_score(y_test, y_pred))
c = 0
for idx in test_index:
pred_rpknn[idx] = y_pred[c]
c =1
print("Accuracy:",metrics.accuracy_score(data_set.y, pred_rpknn))
print(pred_rpknn, data_set.y.reshape(1,-1))
輸出：
/home/katerina/Desktop/datasets/GSE75110.mat
Accuracy: 0.2857142857142857
Accuracy: 0.38095238095238093
Accuracy: 0.14285714285714285
Accuracy: 0.4
Accuracy: 0.3
Accuracy: 0.25
Accuracy: 0.3
Accuracy: 0.6
Accuracy: 0.25
Accuracy: 0.45
Accuracy: 0.33497536945812806
[3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3]
我正在嘗試將 knn 與 k 折疊結合起來，以便使用 10 折疊來測驗整個資料集。問題是 knn 總是為每個折疊預測 3 的陣列。我想預測的類是這些：
tensor([[1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 2, 2, 2,2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3,3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3]]
uj5u.com熱心網友回復：
TL;DR
它與StandardScaler，將其更改為簡單的規范化。
例如
from sklearn import preprocessing
...
x = preprocessing.normalize(x)
解釋：
使用它時的標準標量將執行以下操作：
The standard score of a sample `x` is calculated as: z = (x - u) / s where `u` is the mean of the training samples or zero if `with_mean=False`, and `s` is the standard deviation of the training samples or one if `with_std=False`.
當您確實希望此功能幫助 KNN 確定哪個向量更接近時。
在歸一化中，對每個向量分別進行歸一化，因此它不會影響甚至幫助 KNN 區分向量
使用 KNNStandardScaler實際上會損害您的預測。最好在其他形式的資料中使用它。
import scipy.io
from torch.utils.data import Dataset
from sklearn import preprocessing
import torch
import numpy as np
from sklearn.model_selection import KFold
from sklearn.neighbors import KNeighborsClassifier
from sklearn import metrics
def load_mat_data(path):
mat = scipy.io.loadmat(DATA_PATH)
x, y = mat['data'], mat['class']
x = x.astype('float32')
# stadardize values
x = preprocessing.normalize(x)
return x, y
def numpyToTensor(x):
x_train = torch.from_numpy(x)
return x_train
class DataBuilder(Dataset):
def __init__(self, path):
self.x, self.y = load_mat_data(DATA_PATH)
self.x = numpyToTensor(self.x)
self.len=self.x.shape[0]
self.y = numpyToTensor(self.y)
def __getitem__(self,index):
return (self.x[index], self.y[index])
def __len__(self):
return self.len
datasets = ['/home/katerina/Desktop/datasets/GSE75110.mat']
for DATA_PATH in datasets:
print(DATA_PATH)
data_set=DataBuilder(DATA_PATH)
pred_rpknn = [0] * len(data_set.y)
kf = KFold(n_splits=10, shuffle = True, random_state=7)
for train_index, test_index in kf.split(data_set.x):
#Create KNN Classifier
knn = KNeighborsClassifier(n_neighbors=5)
#print("TRAIN:", train_index, "TEST:", test_index)
x_train, x_test = data_set.x[train_index], data_set.x[test_index]
y_train, y_test = data_set.y[train_index], data_set.y[test_index]
#Train the model using the training sets
y1_train = y_train.view(-1)
knn.fit(x_train, y1_train)
#Predict the response for test dataset
y_pred = knn.predict(x_test)
#print(y_pred)
# Model Accuracy, how often is the classifier correct?
print("Accuracy in loop:", metrics.accuracy_score(y_test, y_pred))
c = 0
for idx in test_index:
pred_rpknn[idx] = y_pred[c]
c =1
print("Accuracy:",metrics.accuracy_score(data_set.y, pred_rpknn))
print(pred_rpknn, data_set.y.reshape(1,-1))
Accuracy in loop: 1.0
Accuracy in loop: 0.8571428571428571
Accuracy in loop: 0.8571428571428571
Accuracy in loop: 1.0
Accuracy in loop: 0.9
Accuracy in loop: 0.9
Accuracy in loop: 0.95
Accuracy in loop: 1.0
Accuracy in loop: 0.9
Accuracy in loop: 1.0
Accuracy: 0.9359605911330049
轉載請註明出處，本文鏈接：https://www.uj5u.com/qianduan/323197.html
標籤：python machine-learning scikit-learn knn