支持向量机matlab代码程序_自然语言处理--支持向量机分类（理论+代码）

一、支持向量机

1.基本思想

支持向量机的最终目的是在特征空间中寻找到一个尽可能将两个数据集合分开的超级平面，之所以名字前面加了前缀“超级”，是因为我们的数据特征空间很有可能是高维度空间，而且我们希望这个超级平面能够尽可能大的将两个数据分开。

2.优点

（1）可用于线性、非线性分类（核函数），也可以用于回归

（2）低泛化错误

（3）推导过程清晰，容易解释

（4）计算复杂度较低

3.缺点

（1）对参数和核函数的选择比较敏感

（2）原始的svm只适合处理二分类问题

4.理论推导

二、效果

正确率 = 84.91%

三、代码

from gensim.models.word2vec import Word2Vec
import numpy as np
import pandas as pd
import jieba
from sklearn.externals import joblib
from sklearn.svm import SVC
path = 'D:/SvmModel1.0/'

## 加载文件，导入数据,分词
def loadfile():
    neg=pd.read_excel(path+'data/neg.xls',header=None,index=None)
    pos=pd.read_excel(path+'data/pos.xls',header=None,index=None)
    y = np.concatenate((np.ones(len(pos)), np.zeros(len(neg))))#use 1 for positive sentiment, 0 for negative
    x_train0, x_test0, y_train, y_test = train_test_split(np.concatenate((pos[0], neg[0])), y, test_size=0.2)

    
    x_trainframe = pd.DataFrame(x_train0)
    x_testframe = pd.DataFrame(x_test0)
    cw = lambda x: list(jieba.cut(x))
    x_trainframe['words'] = x_trainframe[0].apply(cw)
    x_testframe['words'] = x_testframe[0].apply(cw)
    x_trainframe.to_excel(path+'data/x_trainframe.xlsx')
    x_testframe.to_excel(path+'data/x_testframe.xlsx')
    
    x_train = np.array(x_trainframe['words'])
    x_test = np.array(x_testframe['words'])
    return x_train, x_test, y_train, y_test, x_trainframe, x_testframe
 

## 对每个句子的所有词向量取均值
def buildWordVector(text, size, imdb_w2v):
    size = 300
    vec = np.zeros(size).reshape((1, size))
    count = 0.
    for word in text:
        try:
            vec += imdb_w2v[word].reshape((1, size))
            count += 1.
        except KeyError:
            continue
    if count != 0:
        vec /= count
    return vec


## 计算词向量
def get_train_vecs(x_train,x_test):
    n_dim = 300
    
    #Initialize model and build vocab
    imdb_w2v = Word2Vec(size=n_dim, min_count=10)
    imdb_w2v.build_vocab(x_train)
    
    #Train the model over train_reviews (this may take several minutes)
    imdb_w2v.train(sentences=x_train, total_examples=imdb_w2v.corpus_count, epochs=20)
    train_vecs = np.concatenate([buildWordVector(z, n_dim, imdb_w2v) for z in x_train])    
    np.save(path+'train_vecs.npy',train_vecs)
    print(train_vecs.shape)
    
    #Train word2vec on test tweets
    imdb_w2v.train(sentences=x_test, total_examples=imdb_w2v.corpus_count, epochs=20)
    imdb_w2v.save(path+'w2v_model.pkl')
    
    #Build test tweet vectors then scale
    test_vecs = np.concatenate([buildWordVector(z, n_dim,imdb_w2v) for z in x_test])
    
    #test_vecs = scale(test_vecs)
    np.save(path+'test_vecs.npy',test_vecs)
    print(test_vecs.shape)
    
    return train_vecs, test_vecs


## 训练svm模型
def svm_train(train_vecs, y_train, test_vecs, y_test):
    clf = SVC(kernel='rbf', verbose=True)
    clf.fit(train_vecs, y_train)
    joblib.dump(clf, path+'model.pkl')
    score = clf.score(test_vecs,y_test)
    print(score)
    return score


## 对单个句子进行情感判断    
if __name__=='__main__':  
    x_train, x_test, y_train, y_test, x_trainframe, x_testframe = loadfile()
    train_vecs, test_vecs = get_train_vecs(x_train,x_test)
    score = svm_train(train_vecs, y_train, test_vecs, y_test)
    
    #得到待预测单个句子的词向量    
    def get_predict_vecs(words):
        n_dim = 300
        imdb_w2v = Word2Vec.load(path+'w2v_model.pkl')
        train_vecs = buildWordVector(words, n_dim, imdb_w2v)
        return train_vecs
    
    text = pd.read_excel(path+'data/x_testframe.xlsx',header=0, index_col=0)
    strings = text[0]
    
    labels = []
    for string in strings:
        words = jieba.lcut(string)
        words_vecs=get_predict_vecs(words)
        clf = joblib.load(path+'model.pkl')
         
        result=clf.predict(words_vecs)
        labels.append(result[0])
        
        if int(result[0])==1:
            print(string,'Yes')
        else:
            print(string,'No')
    text['标记--真实'] = y_test
    text['标记--预测'] = labels
    text.to_excel(path+'text.xlsx')