from sklearn.model_selection import train_test_split
import numpy as np
import pandas as pd
import jieba as jb
from sklearn.externals import joblib
from sklearn.svm import SVC
from gensim.models.word2vec import Word2Vec

 neg =pd.read_excel("data/neg.xls",header=None,index=None)
 pos =pd.read_excel("data/pos.xls",header=None,index=None)
 # 這是兩類數(shù)據(jù)都是x值
 pos['words'] = pos[0].apply(lambda x:list(jb.cut(x)))
 neg['words'] = neg[0].apply(lambda x:list(jb.cut(x)))
 #需要y值 0 代表neg 1代表是pos
 y = np.concatenate((np.ones(len(pos)),np.zeros(len(neg))))
 X = np.concatenate((pos['words'],neg['words']))

 X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.2,random_state=3)
 #保存數(shù)據(jù)
 np.save("data/y_train.npy",y_train)
 np.save("data/y_test.npy",y_test)

#初始化模型和詞表
 wv = Word2Vec(size=300,min_count=10)
 wv.build_vocab(x_train)
 # 訓(xùn)練并建模
 wv.train(x_train,total_examples=1, epochs=1)
 #獲取train_vecs
 train_vecs = np.concatenate([ build_vector(z,300,wv) for z in x_train])
 #保存處理后的詞向量
 np.save('data/train_vecs.npy',train_vecs)
 #保存模型
 wv.save("data/model3.pkl")
 
 wv.train(x_test,total_examples=1, epochs=1)
 test_vecs = np.concatenate([build_vector(z,300,wv) for z in x_test])
 np.save('data/test_vecs.npy',test_vecs)
 

def build_vector(text,size,wv):
 #創(chuàng)建一個(gè)指定大小的數(shù)據(jù)空間
 vec = np.zeros(size).reshape((1,size))
 #count是統(tǒng)計(jì)有多少詞向量
 count = 0
 #循環(huán)所有的詞向量進(jìn)行求和
 for w in text:
 try:
 vec += wv[w].reshape((1,size))
 count +=1
 except:
 continue
 
 #循環(huán)完成后求均值
 if count!=0:
 vec/=count
 return vec

 #創(chuàng)建SVC模型
 cls = SVC(kernel="rbf",verbose=True)
 #訓(xùn)練模型
 cls.fit(train_vecs,y_train)
 #保存模型
 joblib.dump(cls,"data/svcmodel.pkl")
 #輸出評(píng)分
 print(cls.score(test_vecs,y_test))

from tkinter import *
import numpy as np
import jieba as jb
import joblib
from gensim.models.word2vec import Word2Vec

class core():
 def __init__(self,str):
 self.string=str

 def build_vector(self,text,size,wv):
 #創(chuàng)建一個(gè)指定大小的數(shù)據(jù)空間
 vec = np.zeros(size).reshape((1,size))
 #count是統(tǒng)計(jì)有多少詞向量
 count = 0
 #循環(huán)所有的詞向量進(jìn)行求和
 for w in text:
 try:
 vec += wv[w].reshape((1,size))
 count +=1
 except:
 continue
 #循環(huán)完成后求均值
 if count!=0:
 vec/=count
 return vec
 def get_predict_vecs(self,words):
 # 加載模型
 wv = Word2Vec.load("data/model3.pkl")
 #將新的詞轉(zhuǎn)換為向量
 train_vecs = self.build_vector(words,300,wv)
 return train_vecs
 def svm_predict(self,string):
 # 對(duì)語(yǔ)句進(jìn)行分詞
 words = jb.cut(string)
 # 將分詞結(jié)果轉(zhuǎn)換為詞向量
 word_vecs = self.get_predict_vecs(words)
 #加載模型
 cls = joblib.load("data/svcmodel.pkl")
 #預(yù)測(cè)得到結(jié)果
 result = cls.predict(word_vecs)
 #輸出結(jié)果
 if result[0]==1:
 return "好感"
 else:
 return "反感"
 def main(self):
 s=self.svm_predict(self.string)
 return s

root=Tk()
root.title("情感分析")
sw = root.winfo_screenwidth()
#得到屏幕寬度
sh = root.winfo_screenheight()
#得到屏幕高度
ww = 500
wh = 300
x = (sw-ww) / 2
y = (sh-wh) / 2-50
root.geometry("%dx%d+%d+%d" %(ww,wh,x,y))
# root.iconbitmap('tb.ico')

lb2=Label(root,text="輸入內(nèi)容，按回車鍵分析")
lb2.place(relx=0, rely=0.05)

txt = Text(root,font=("宋體",20))
txt.place(rely=0.7, relheight=0.3,relwidth=1)

inp1 = Text(root, height=15, width=65,font=("宋體",18))
inp1.place(relx=0, rely=0.2, relwidth=1, relheight=0.4)

def run1():
 txt.delete("0.0",END)
 a = inp1.get('0.0',(END))
 p=core(a)
 s=p.main()
 print(s)
 txt.insert(END, s) # 追加顯示運(yùn)算結(jié)果

def button1(event):
 btn1 = Button(root, text='分析', font=("",12),command=run1) #鼠標(biāo)響應(yīng)
 btn1.place(relx=0.35, rely=0.6, relwidth=0.15, relheight=0.1)
 # inp1.bind("<Return>",run2) #鍵盤響應(yīng)

button1(1)
root.mainloop()