tensorflow1.4.0和R语言实现KNN(k-Nearest Neighbors)

说明

• KNN作为惰性学习方法，训练和测试同时进行
• Classifying with k-Nearest Neighbors

tensorflow实现KNN对手写字的分类

https://github.com/aymericdamien/TensorFlow-Examples/blob/master/examples/2_BasicModels/nearest_neighbor.py

import tensorflow as tf
import numpy as np
print(tf.__version__)
print(np.__version__)

使用TensorFlow自带的数据集作为测试

from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("MNIST_data", one_hot=True)

划分训练集和测试集

Xtrain, Ytrain = mnist.train.next_batch(5000)  #从数据集中选取5000个样本作为训练集
Xtest, Ytest = mnist.test.next_batch(200)    #从数据集中选取200个样本作为测试集

创建tensorflow结构

# 输入占位符
xtr = tf.placeholder("float", [None, 784])
xte = tf.placeholder("float", [784])
# 计算L1距离
distance = tf.reduce_sum(tf.abs(tf.add(xtr, tf.negative(xte))), reduction_indices=1)
# 获取最小距离的索引
pred = tf.arg_min(distance, 0) # np.argmin(sess.run(distance, feed_dict={xtr: Xtrain, xte: Xtest[i, :]}))

测试模型

# 初始化变量
init = tf.global_variables_initializer()
#分类精确度
accuracy = 0.
# 运行会话，训练模型
with tf.Session() as sess:
    # 运行初始化
    sess.run(init)
    # 遍历测试数据
    for i in range(len(Xtest)):
        # 获取当前样本的最近邻索引
        nn_index = sess.run(pred, feed_dict={xtr: Xtrain, xte: Xtest[i, :]})   #向占位符传入训练数据
#         print("Test", i, "Prediction:", Ytrain[nn_index],"True Class:", Ytest[i])
#         print(np.argmin(sess.run(distance, feed_dict={xtr: Xtrain, xte: Xtest[i, :]})),nn_index)
        # 最近邻分类标签与真实标签比较
        if i%50==0:
            print("Test", i, "Prediction:", np.argmax(Ytrain[nn_index]),"True Class:", np.argmax(Ytest[i]))# argmax最大值的下标
        # 计算精确度
        if np.argmax(Ytrain[nn_index]) == np.argmax(Ytest[i]):
            accuracy += 1./len(Xtest)
print("Accuracy:", accuracy)

tensorflow实现KNN对文理分科的分类

模拟数据下载

导入数据

import pandas as pd
df = pd.read_csv("data/cjb.csv")

划分数据集

train_set = np.random.choice(len(df), 500, replace=False)
Xtrain = np.array(df.iloc[train_set,3:12])
Ytrain = np.array(df.iloc[train_set,[12]])
Xtest =  np.array(df.iloc[-train_set,3:12])
Ytest =  np.array(df.iloc[-train_set,[12]])

创建tensorflow结构

# 输入占位符
xtr = tf.placeholder("float", [None, 9]) # 不规定有多少sample,每个sample有9个属性
xte = tf.placeholder("float", [9])
# 计算L1距离
distance = tf.reduce_sum(tf.abs(tf.add(xtr, tf.negative(xte))), reduction_indices=1)
# 获取最小距离的索引
pred = tf.arg_min(distance, 0) # np.argmin(sess.run(distance, feed_dict={xtr: Xtrain, xte: Xtest[i, :]}))

测试模型

# 初始化变量
init = tf.global_variables_initializer()
#分类精确度
accuracy = 0.
# 运行会话，训练模型
with tf.Session() as sess:
    # 运行初始化
    sess.run(init)
    # 遍历测试数据
    for i in range(len(Xtest)):
        # 获取当前样本的最近邻索引
        nn_index = sess.run(pred, feed_dict={xtr: Xtrain, xte: Xtest[i, :]})   #向占位符传入训练数据
        if i%50==0:
            print("Test", i, "Prediction:", Ytrain[nn_index],"True Class:", Ytest[i])
#         计算精确度
        if Ytrain[nn_index] == Ytest[i]:
            accuracy += 1./len(Xtest)
print("Accuracy:", accuracy)

Ｒ语言实现KNN对文理分科的分类

导入数据划分测试集和训练集

library("tidyverse")
library('kknn')
cjb <- read.csv("data/cjb.csv")
cjb1 <- cjb %>%
    select(4:13) %>%
    mutate(wlfk = factor(wlfk))

train_set_idx <- sample(nrow(cjb), 0.7 * nrow(cjb))
test_set_idx <- (1:nrow(cjb))[-train_set_idx]

建立模型

imodel <- kknn(wlfk~., train = cjb[train_set_idx,], test = cjb[train_set_idx,])

查看测试集合预测结果

predicted_train <- imodel$fit

模型评估

Metrics::ce(cjb$wlfk[train_set_idx], predicted_train)