男声女声分类识别SVM

男声女声分类识别SVM

内容为对测试集中声音数据进行性别预测，测试文件夹中有951行数据

数据类别一共两类：female、male

推荐使用SVM分类，也可使用其他如决策树等方法分类

在jupyterlab中将识别结果写入文件 /home/ilab/submission 

每行输出id号（第一列数据）+'t'+ female或者是male（不要改变测试集顺序）

例如：

6 female

7 male

8 male

9 female

代码：评分91.3

import pandas as pd
import numpy as np
from sklearn import svm
from sklearn import model_selection
import matplotlib.pyplot as plt
import matplotlib as mpl

train_data = pd.read_csv('/ilab/datasets/local/voice/train.csv')#导入数据集
test = pd.read_csv('/ilab/datasets/local/voice/test.csv')

train = train_data.values.tolist() #转化成列表
test1 = test.values.tolist()
ss1 = {'male':0, 'female':1}  #设置标签
ss2 = {}
a = -1
yy= []
for i in train: #一共二十列变量，逐个分析
    a = a + 1
    i[21] = ss1[i[21]]
    yy.append(i[21])
    i.pop(21)
train = np.mat(train)
train2 = train[:, 1:3]
test = np.mat(test)
test1 = test[:,1:3]

yy = np.array(yy)
print(train)

def classifier():
    clf = svm.SVC(C=0.5,                            # 误差惩罚系数，默认1
                  kernel='linear',                  # 线性核
                  decision_function_shape='ovr')    # 决策函数
    return clf
clf = classifier()
print(yy)

def train1(clf, x_train, y_train):
    clf.fit(x_train,            # 训练集特征向量
            y_train.ravel())    # 训练集目标值



# 训练SVM 模型
train2 = train[:, 1:19]
test1 = test[:, 1:19]
print(test1)
train1(clf, train2, yy)

ha = clf.predict(test1)

# 读取test.csv文件，输出submission

import pandas as pd

# 将id读取
test = pd.read_csv('/ilab/datasets/local/voice/test.csv')
test_sub = test.iloc[:, [0]]
# 训练出来性别
gender = []
for i in range(len(test_sub)):
    if ha[i] == 1:
        gender.append('female')
    else:
        gender.append('male')
print(gender)
df_gender = pd.Dataframe(gender)
# 将id列和预测出来的性别列拼接并输出submission文件
df=pd.concat([test_sub,df_gender],axis=1)
df.to_csv('/home/ilab/submission', sep='t', header=None, index=False)