【Python Code】Use the nonlinear support vector machine to classify the identified model parameters

This article is recommended for such people !

Person who wants to write program of nonlinear support vector machine
A person studying machine learning to solve a classification problem
A person performing model parameter identification for research

Hi, This is a research archive article, but please see if you want to see this.
So, what am I doing ?

Modeling the EEG signal
Identify model parameters by solving the inverse problem
Classification by SVM to determine concentration or relax state

I’m doing something like this.
If you want to know more detail, please go to this page.

Modeling of EEG (Experimental Parameters Identification and Statistical Processing of Model Parameters)
https://kenyu-life.com/2018/11/03/modeling_of_eeg/
This is my first time to write blog in English. Today, …

Actually, I am writing a dissertation, so I can not write specific data on my blog.

But, python program code is managed in my blog.
So, you can use it for free.

I’m showing some unused data.
Red marks are concentrated state and blue marks are relax state.

Classification is possible for data that can be linearly separated.

The left side shows the data used to draw classification lines while the right side shows the data to test weather the line works well.
The accuracy rate is about 87 % (It is given above).

I think that experimental data itself is not very good, so it is necessary to recover experimental data again.

Programming of nonlinear support vector machine

I will write a program for nonlinear support vector machine!

# -*- coding: utf-8 -*-
import numpy as np
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
from sklearn.preprocessing import PolynomialFeatures
from sklearn.svm import SVC
 
#--plt., Collective setting-----------#
plt.rcParams['font.family'] = 'Times New Roman' #全体のフォントを設定
plt.rcParams['font.size'] = 10 #フォントサイズを設定
plt.rcParams['axes.linewidth'] = 1.5 #軸の太さを設定。目盛りは変わらない
#------------------------------------------#
#----Name for analysis data------#
subject = "b"
task = "100"
#100,game,word
#-----------------------------------------#
#----------Analysis data reading---------------#
inputTarget = 'rectan_3s/target.txt'
inputData_a = 'rectan_3s/para4Freq3_%s_%s_3s.txt' %(subject,task)
 
target = np.loadtxt(inputTarget, dtype='float')
dataA = np.loadtxt(inputData_a, delimiter='\t', dtype='float')
#-----------------------------------------------------#
 
X = np.vstack((dataA[:180, 1:3]))
#rabelを追加 0=relax,1=stress
y = np.hstack((target[:180]))
y = y.astype(np.int64)
 
#解析データの分布を標準化しておく
scaler = StandardScaler()
X = scaler.fit_transform(X)
 
X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state=0)
print(X_test.shape[0])
X_train_number = X_test.shape[0]
 
#----------Function for graph drawing---------------#
def plot_decision_function(model):
    _x0 = np.linspace(-5.0, 5.0, 100)
    _x1 = np.linspace(-5.0, 5.0, 100)
    x0, x1 = np.meshgrid(_x0, _x1)
    X = np.c_[x0.ravel(), x1.ravel()]
    y_pred = model.predict(X).reshape(x0.shape)
    y_decision = model.decision_function(X).reshape(x0.shape)
    plt.contourf(x0, x1, y_pred, cmap=plt.cm.brg, alpha=0.2)
    plt.contourf(x0, x1, y_decision,levels=10,alpha=0.3)
 
def plot_dataset(X,y):
    plt.plot(X[:, 0][y==0], X[:, 1][y==0], "bo", ms=10)
    plt.plot(X[:, 0][y==1], X[:, 1][y==1], "r^", ms=10)
    plt.xlabel("A")
    plt.ylabel("B")
#-----------------------------------------------------------#
#--SVM-Kernel method (using pipeline) ---------#
kernel_svm = Pipeline([
                       ('scaler',StandardScaler()),
                       ('svm', SVC(kernel='poly', degree=3, coef0=1))
                       ])
 
print(kernel_svm.fit(X_train,y_train))
print(X.shape)
kernel_svm.fit(X_train,y_train)
#テストデータを使って，予測した分類がそれぞれ正しいのかを見る
print(kernel_svm.predict(X_test) == y_test)
 
R = kernel_svm.predict(X_test)
total = len(X_test)
success = sum(R == y_test)
print('被験者%sのタスク%sの正解率=' %(subject,task))
print(100.0*success/total)
#----------------------------------------------------------#
 
#--Draw or save a graph (1 by 2)------#
plt.subplot(121)
plot_decision_function(kernel_svm)
plot_dataset(X_train,y_train)
plt.title('Train, Sub:%s,Task:%s'%(subject,task))
 
plt.subplot(122)
#plt.figure(figsize=(12,8))
plot_decision_function(kernel_svm)
plot_dataset(X_test,y_test)
#plt.show()
plt.title('Test, Accuracy:{:.1f}, N:{:.1f}'.format(100.0*success/total, X_train_number) )
plt.savefig('para4Freq3_%s_%s_3s.png' %(subject, task), dpi=1200)
 
#plt.show()
#------------------------------------------------#

# -*- coding: utf-8 -*-

import numpy as np

import matplotlib.pyplot as plt

from sklearn.model_selection import train_test_split

from sklearn.preprocessing import StandardScaler

from sklearn.pipeline import Pipeline

from sklearn.preprocessing import PolynomialFeatures

from sklearn.svm import SVC

#--plt., Collective setting-----------#

plt.rcParams['font.family'] = 'Times New Roman' #全体のフォントを設定

plt.rcParams['font.size'] = 10 #フォントサイズを設定

plt.rcParams['axes.linewidth'] = 1.5 #軸の太さを設定。目盛りは変わらない

#------------------------------------------#

#----Name for analysis data------#

subject = "b"

task = "100"

#100,game,word

#-----------------------------------------#

#----------Analysis data reading---------------#

inputTarget = 'rectan_3s/target.txt'

inputData_a = 'rectan_3s/para4Freq3_%s_%s_3s.txt' %(subject,task)

target = np.loadtxt(inputTarget, dtype='float')

dataA = np.loadtxt(inputData_a, delimiter='\t', dtype='float')

#-----------------------------------------------------#

X = np.vstack((dataA[:180, 1:3]))

#rabelを追加 0=relax,1=stress

y = np.hstack((target[:180]))

y = y.astype(np.int64)

#解析データの分布を標準化しておく

scaler = StandardScaler()

X = scaler.fit_transform(X)

X_train, X_test, y_train, y_test = train_test_split(X, y, stratify=y, random_state=0)

print(X_test.shape[0])

X_train_number = X_test.shape[0]

#----------Function for graph drawing---------------#

def plot_decision_function(model):

_x0 = np.linspace(-5.0, 5.0, 100)

_x1 = np.linspace(-5.0, 5.0, 100)

x0, x1 = np.meshgrid(_x0, _x1)

X = np.c_[x0.ravel(), x1.ravel()]

y_pred = model.predict(X).reshape(x0.shape)

y_decision = model.decision_function(X).reshape(x0.shape)

plt.contourf(x0, x1, y_pred, cmap=plt.cm.brg, alpha=0.2)

plt.contourf(x0, x1, y_decision,levels=10,alpha=0.3)

def plot_dataset(X,y):

plt.plot(X[:, 0][y==0], X[:, 1][y==0], "bo", ms=10)

plt.plot(X[:, 0][y==1], X[:, 1][y==1], "r^", ms=10)

plt.xlabel("A")

plt.ylabel("B")

#-----------------------------------------------------------#

#--SVM-Kernel method (using pipeline) ---------#

kernel_svm = Pipeline([

('scaler',StandardScaler()),

('svm', SVC(kernel='poly', degree=3, coef0=1))

])

print(kernel_svm.fit(X_train,y_train))

print(X.shape)

kernel_svm.fit(X_train,y_train)

#テストデータを使って，予測した分類がそれぞれ正しいのかを見る

print(kernel_svm.predict(X_test) == y_test)

R = kernel_svm.predict(X_test)

total = len(X_test)

success = sum(R == y_test)

print('被験者%sのタスク%sの正解率=' %(subject,task))

print(100.0*success/total)

#----------------------------------------------------------#

#--Draw or save a graph (1 by 2)------#

plt.subplot(121)

plot_decision_function(kernel_svm)

plot_dataset(X_train,y_train)

plt.title('Train, Sub:%s,Task:%s'%(subject,task))

plt.subplot(122)

#plt.figure(figsize=(12,8))

plot_decision_function(kernel_svm)

plot_dataset(X_test,y_test)

#plt.show()

plt.title('Test, Accuracy:{:.1f}, N:{:.1f}'.format(100.0*success/total, X_train_number) )

plt.savefig('para4Freq3_%s_%s_3s.png' %(subject, task), dpi=1200)

#plt.show()

#------------------------------------------------#

I commended the code, so it might be understandable

SVM parameters

In this code, the kernel method is described by pipeline.
The data of X_train and y_train are used by the following SVM parameters..

Pipeline(memory=None,
     steps=[('scaler', StandardScaler(copy=True, with_mean=True, with_std=True)), ('svm', SVC(C=1.0, cache_size=200, class_weight=None, coef0=1,
  decision_function_shape='ovr', degree=3, gamma='auto_deprecated',
  kernel='poly', max_iter=-1, probability=False, random_state=None,
  shrinking=True, tol=0.001, verbose=False))])

Pipeline(memory=None,