面向有监督学习与文本数据的通用分类器

面向有监督学习与文本数据的通用分类器

方法源码

# 将源码命名为 ml.py
__all__ = ['Classifier']

import numpy as np
import matplotlib.pyplot as plt

class Classifier:
    def __init__(self, num_classes: int, max_iter: int=200, lr: float=0.1):
        self.max_iter = max_iter # max iteration
        self.lr = lr # learning rate
        self.num_classes = num_classes # categories
        self.scores = []
    
    def __data_matrix(self, X):
        '''
        Parameters
        ----------
        X : numpy.array
            input matrix

        Returns
        -------
        numpy.array
            augmented matrix

        '''
        ones = np.ones(X.shape[0])
        return np.insert(X, 0, ones, axis=1)
    
    def __softmax(self, horizon):
        '''
        Parameters
        ----------
        horizon : numpy.array
            one line of data-set.

        Returns
        -------
        numpy.array
            softmax result.

        '''
        return np.exp(horizon) / np.sum(np.exp(horizon))
    
    def fit(self, X, y) -> None:
        '''
        Parameters
        ----------
        X : numpy.array
            data-set to be trained
        
        y : numpy.array
            correct labels

        Returns
        -------
        None

        '''
        augmented = self.__data_matrix(X)
        self.weights = np.zeros((augmented.shape[1], self.num_classes), dtype=np.float64)
        for step in range(self.max_iter):
            for index in range(augmented.shape[0]):
                res = self.__softmax(np.dot(augmented[index], self.weights))
                obj = np.eye(self.num_classes)[int(y[index])]
                err = res - obj
                self.weights -= self.lr * np.transpose([augmented[index]]) * err
            score = self.score(X_test, y_test) # working environment store the two values: X_test, y_test
            self.scores.append(score)
            if step % 20 == 0:
                print("Training Error: {0:<}, Testing Score: {1:<}".format(np.linalg.norm(err), score))
                
    def score(self, X, y) -> float:
        '''
        Parameters
        ----------
        X : numpy.array
            data-set to be tested
        
        y : numpy.array
            correct labels

        Returns
        -------
        float
            correct rate

        '''
        X = self.__data_matrix(X)
        corr = 0
        multiply = np.dot(X, self.weights)
        predicted = np.argmax(multiply, axis=1)
        corr += (predicted == y).sum()
        return corr / X.shape[0]
    
    def predict(self, X):
        '''
        Parameters
        ----------
        X : numpy.array
            data-set to be predicted

        Returns
        -------
        numpy.array
             predicted result

        '''
        X = self.__data_matrix(X)
        multiply = np.dot(X, self.weights)
        return np.argmax(multiply, axis=1)

    def plot(self, color: str="slateblue", mark: str='o', style: str='dashed') -> None:
        '''
        Parameters
        ----------
        color : str
            The color of plot line.

        mark : str
            The mark of points.

        style : str
            The styple of plot.

        Returns
        -------
        None

        '''
        axis_x = [num for num in range(1, self.max_iter + 1)]
        # plt.xlabel, plt.ylabel, plt.title
        plt.plot(axis_x, self.scores, c=color, marker=mark, linestyle=style)
        plt.show()
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134

测试示例

1. 环境依赖

import numpy as np
import matplotlib.pyplot as plt
import scipy.io as sio # 用于解析.mat格式数据
1
2
3

2. 加载分类器

from ml import Classifier # ml.py文件存储着分类器源码
1

3. 数据预处理

下载实验所需的数据集：textretrieval.mat
数据集维度是：2866*6603。

def process_data(url) -> tuple:
    data = sio.loadmat(url)
    return data['X'], data['class']
features, labels = process_data('../data/textretrieval.mat') # 数据集路径由环境部署结构确定
1
2
3
4

4. 拆分数据集

def pretreat(features, labels) -> tuple:
    labels_mo = np.argwhere(labels == 1)[:, 1] # 对标签进行处理
    return features[:2000, :], features[2000:, :], labels_mo[:2000], labels_mo[2000:]
X_train, X_test, y_train, y_test = pretreat(features, labels)
1
2
3
4

5. 分类器训练

model = Classifier(num_classes=10, max_iter=1000, lr=0.1)
model.fit(X_train, y_train)
model.score(X_test, y_test)
1
2
3

6. 训练日志

model.plot(color="slateblue", mark='o', style='dashed')
1