svm支持向量机实现mnist数据集的训练和测试(准确率可达98%)

coding=utf-8

“”"
author:lei
function: 对mnist数据集进行模型训练，预测
“”"

from sklearn.svm import SVC
from sklearn.decomposition import PCA
from sklearn.model_selection import train_test_split
import pandas as pd
import joblib
import matplotlib.pyplot as plt
import random

class Mnist(object):
def init(self, train_path, test_path):
self.train_path = train_path
self.test_path = test_path

def data_reduction(self, data):
    pca = PCA(0.85)
    data = pca.fit_transform(data)
    joblib.dump(pca, "./model/pca.pkl")

    return data

def to_plot(self, num):
    # 3、查看具体图像
    plt.imshow(num)
    plt.axis("off")  # 不显示轴
    plt.show()

# 对第n张测试集图片进行预测
def test(self, n):
    pca = joblib.load("./model/pca.pkl")
    data = pd.read_csv(self.test_path)

    data = data.iloc[n, :].values

    # 画图显示图形
    self.to_plot(data.reshape(28, 28))

    # 将特征值进行降维 将数据转换为二维数据
    x = [data / 255]

    # 将数据进行pca降维
    x_pca = pca.transform(x)

    # 将数据传入模型进行预测
    svm = joblib.load("./model/mnist_svc.pkl")
    pre = svm.predict(x_pca)
    print(pre[0])

def train(self):
    data = pd.read_csv(self.train_path)
    x = data.iloc[:, 1:].values
    y = data.iloc[:, 0].values

    # 特征值归一化
    x = x / 255

    # 对图片进行pca降维
    pca_x = self.data_reduction(x)

    x_train, x_test, y_train, y_test = train_test_split(pca_x, y, test_size=0.2)

    svc = SVC()
    svc.fit(x_train, y_train)
    # 输出准确率 0.9804761904761905
    print(svc.score(x_test, y_test))

    # 对模型进行保存
    joblib.dump(svc, "./model/mnist_svc.pkl")

if name == ‘main’:
mn = Mnist("./data/train.csv", “./data/test.csv”)
# mn.train()
mn.test(random.randint(1, 1000))