SVM回归预测 - 技术栈

svm回归预测，数据，见同名同名公众号
复制代码
# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.svm import SVR
from sklearn.preprocessing import StandardScaler

#加载数据集并归一化
def data_load():
    data = pd.read_csv('data.csv') #读取数据
    y = data.y      #标签
    X = data.drop('y', axis=1)  #特征
    row = data.shape[0] #数据集总行数
    # print(row)
    num_train = int(row * 0.9)  # 训练集比例计算训练集数量
    #划分训练集与测试集
    x_train=X.iloc[0:num_train,:].values
    x_test = X.iloc[num_train:, :].values
    y_train = y.iloc[0:num_train].values.reshape(-1,1) #转换为二维
    y_test=y.iloc[num_train:].values.reshape(-1,1)#转换为二维

    # 归一化标准ss_X,ss_Y
    ss_X = StandardScaler().fit(x_train)
    ss_Y = StandardScaler().fit(y_train)

    #归一化
    x_train = ss_X.transform(x_train)
    x_test = ss_X.transform(x_test)
    y_train = ss_Y.transform(y_train)
    y_test = ss_Y.transform(y_test)

    return x_train,y_train,x_test,y_test,ss_Y


if __name__=='__main__':
    # 加载数据
    x_train,y_train,x_test,y_test,ss_Y=data_load()
    # 训练SVR模型
    model = SVR(kernel='linear') #定义模型
    model.fit(x_train,y_train)   #训练模型
    #预测
    pred_data = model.predict(x_test)
    #反归一化
    y_test    = ss_Y.inverse_transform(y_test) #实际值反归一化
    pred_test = ss_Y.inverse_transform(pred_data.reshape(-1,1))#预测值反归一化
    #设置绘图属性
    plt.rcParams['font.family'] = ['sans-serif']
    plt.rcParams['font.sans-serif'] = ['SimHei']
    plt.rcParams['axes.unicode_minus']=False

    # 绘图
    plt.figure()
    plt.plot(y_test, c='k', marker="*",label='实际值')
    plt.plot(pred_test, c='r', marker="o",label='预测值')
    plt.legend()
    plt.xlabel('样本点')
    plt.ylabel('功率')
    plt.title('测试集对比')
    plt.savefig('测试集对比.jpg')
    plt.show()
预测对比图：