【101】NLP 自然语言处理14种分类算法】

  

内容目录

一、数据集介绍二、解压文件明确需求三、批量读取和合并文本数据集四、中文文本分词五、停止词使用六、编码器处理文本标签七、常规算法模型1、k近邻算法2、决策树3、多层感知器4、伯努力贝叶斯5、高斯贝叶斯6、多项式贝叶斯7、逻辑回归8、支持向量机八、集成算法模型1、随机森林算法2、自适应增强算法3、lightgbm算法4、xgboost算法九、深度学习1、前馈神经网络2、LSTM 神经网络十、算法之间性能比较

一、数据集介绍

文档共有4中类型：女性、体育、文学、校园 

训练集放到train文件夹里，测试集放到test文件夹里。

停用词放到stop文件夹里。 

训练集：3305         [女性，体育，文学，校园]

测试集           200          [女性，体育，文学，校园]

import os
import shutil
import zipfile
import jieba
import time
import warnings
import xgboost
import lightgbm
import numpy as np
import pandas as pd
from keras import models
from keras import layers
from keras.utils.np_utils import to_categorical
from keras.preprocessing.text import Tokenizer
from sklearn import svm
from sklearn import metrics
from sklearn.neural_network import MLPClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.neighbors import KNeighborsClassifier
from sklearn.naive_bayes import BernoulliNB
from sklearn.naive_bayes import GaussianNB
from sklearn.naive_bayes import MultinomialNB
from sklearn.linear_model import LogisticRegression
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import AdaBoostClassifier
from sklearn.preprocessing import LabelEncoder
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.feature_extraction.text import TfidfVectorizer 

warnings.filterwarnings('ignore')

二、解压文件明确需求

《获取数据》—《数据分析和处理》—《特征工程与选择》—《算法模型》—《性能评估/参数调优》

目标需求1：批量读取文档文件.txt
目标需求2：文本向量化使用词袋法或者TFIDF
目标需求3：比较各个算法的性能优劣，包括模型消耗用时、模型准确率

三、批量读取和合并文本数据集

path = 'D:\\A\\AI-master\\py-data\\text_classification-master\\text classification'
def read_text(path, text_list):
    '''
    path: 必选参数，文件夹路径
    text_list: 必选参数，文件夹 path 下的所有 .txt 文件名列表
    return: 返回值
        features 文本(特征)数据，以列表形式返回;
        labels 分类标签，以列表形式返回
    '''

    features, labels = [], []
    for text in text_list:
        if text.split('.')[-1] == 'txt':
            try:
                with open(path + text, encoding='gbk') as fp:
                    features.append(fp.read())  # 特征
                    labels.append(path.split('\\')[-2])  # 标签
            except Exception as erro:
                print('\n>>>发现错误, 正在输出错误信息...\n', erro)

    return features, labels


def merge_text(train_or_test, label_name):
    '''
    train_or_test: 必选参数，train 训练数据集 or test 测试数据集
    label_name: 必选参数，分类标签的名字
    return: 返回值
        merge_features 合并好的所有特征数据，以列表形式返回;
        merge_labels   合并好的所有分类标签数据，以列表形式返回
    '''

    print('\n>>>文本读取和合并程序已经启动, 请稍候...')

    merge_features, merge_labels = [], []  # 函数全局变量
    for name in label_name:
        path = 'D:\\A\\AI-master\\py-data\\text_classification-master\\text classification\\' + train_or_test + '\\' + name + '\\'
        text_list = os.listdir(path)
        features, labels = read_text(path=path, text_list=text_list)  # 调用函数
        merge_features += features  # 特征
        merge_labels += labels  # 标签

    # 可以自定义添加一些想要知道的信息
    print('\n>>>你正在处理的数据类型是...\n', train_or_test)
    print('\n>>>[', train_or_test, ']数据具体情况如下...')
    print('样本数量\t', len(merge_features), '\t类别名称\t', set(merge_labels))
    print('\n>>>文本读取和合并工作已经处理完毕...\n')

    return merge_features, merge_labels


 #获取训练集

train_or_test = 'train'

label_name = ['女性', '体育', '校园', '文学']
X_train, y_train = merge_text(train_or_test, label_name)


# 获取测试集

train_or_test = 'test'

label_name = ['女性', '体育', '校园', '文学']
X_test, y_test = merge_text(train_or_test, label_name)

print('X_train, y_train',len(X_train), len(y_train))
print('X_test, y_test',len(X_test), len(y_test))

X_train, y_train 3305 3305
X_test, y_test 200 200

四、中文文本分词

# 训练集

X_train_word = [jieba.cut(words) for words in X_train]
X_train_cut = [' '.join(word) for word in X_train_word]

X_train_cut[:5]

['明天 三里屯 见 瑞丽 服饰 美容 ：2011 瑞丽 造型 大赏 派对 瑞丽 专属 模特 黄美熙 康猴 猴 康乐 帕丽扎 提 也 会 参加 哦 瑞丽 专属 模特 转发 ( 53 ) 评论 ( 5 ) 12 月 8 日 17 : 10 来自 新浪 微博',
 '转发 微博 水草 温 ：北京 早安 中国 宝迪沃 减肥 训练营 2012 年 寒假 班 开始 报名 早安 中国 吧 : 51 （ 分享 自 百度 贴 吧 ） 您 还 在 为 肥胖 为难 吗 冬季 减肥 训练 帮助 您 快速 减掉 多余 脂肪 咨询 顾问 温 老师 1071108677 原文 转发 原文 评论',
 '多谢 支持 ， 鱼子 精华 中 含有 太平洋 深海 凝胶 能 给 秀发 源源 滋养 ， 鱼子 精华 发膜 含有 羟基 积雪草 苷 ， 三重 神经酰胺 能 补充 秀发 流失 的 胶质 ， 可以 有 焕然 新生 的 感觉 。鲍艺芳 ：我 在 倍 丽莎 美发 沙龙 做 的 奢华 鱼子酱 护理 超级 好用 以后 就 用 它 了 我 的 最 爱 巴黎 卡诗 原文 转发 原文 评论',
 '注射 除皱 针 时会 疼 吗 ？其他 产品 注射 会 产生 难以忍受 的 疼痛 ， 除皱 针中 含有 微量 的 利多卡因 以 减轻 注射 过程 中 的 疼痛 ， 注射 过程 只有 非常 轻微 的 疼痛感 。另外 ， 除皱 针 的 胶原蛋白 具有 帮助 凝血 作用 ， 极少 产生 淤血 现象 。一般 注射 后 就 可以 出门 。其他 产品 容易 产生 淤斑 ， 而且 时间 较长 。让 人 难以 接受 。',
 '注射 除皱 针 时会 疼 吗 ？其他 产品 注射 会 产生 难以忍受 的 疼痛 ， 除皱 针中 含有 微量 的 利多卡因 以 减轻 注射 过程 中 的 疼痛 ， 注射 过程 只有 非常 轻微 的 疼痛感 。另外 ， 除皱 针 的 胶原蛋白 具有 帮助 凝血 作用 ， 极少 产生 淤血 现象 。一般 注射 后 就 可以 出门 。其他 产品 容易 产生 淤斑 ， 而且 时间 较长 。让 人 难以 接受 。']

# 训练集

# 测试集

X_test_word = [jieba.cut(words) for words in X_test]
X_test_cut  = [' '.join(word) for word in X_test_word]

X_test_cut[:5]

['【 L 丰胸 — — 安全 的 丰胸 手术 】 华美 L 美胸 非常 安全 ， 5 大 理由 ：① 丰胸 手术 在世界上 已有 超过 100 年 的 历史 ; ② 假体 硅胶 有 40 多年 的 安全 使用 史 ; ③ 每年 都 有 超过 600 万 女性 接受 丰胸 手术 ; ④ 不 损伤 乳腺 组织 ， 不 影响 哺乳 ; ⑤ 华美 美莱 率先 发布 中国 首部 《 整形 美容 安全 消费 白皮书 》 ， 维护 社会 大众 整形 美容 消费 利益 。\t \t \t \t $ LOTOzf $',
 '【 穿 高跟鞋 不磨 脚 小 方法 】 1 . 用 热 毛巾 在 磨 脚 的 部位 捂 几分钟 , 再拿块 干 的 软 毛巾 垫 着 , 用 锤子 把 鞋子 磨脚 的 地方 敲 平整 就 可以 啦 ；2 . 把 报纸 捏 成团 沾点 水 , 不要 太湿 , 但 要 整团 都 沾 到 水 , 再 拿 张干 的 报纸 裹住 湿 的 报纸 , 塞 在 挤 脚 部位 , 用 塑料袋 密封 一夜 ；3 . 穿鞋 之前 , 拿 香皂 ( 蜡烛 亦可 ) 在 磨脚 位置 薄薄的 涂上一层 。\t \t \t \t $ LOTOzf $',
 '要 在 有限 的 空间 里 实现 最大 的 价值 ， 布 上网 线 、 音频线 、 高清 线 、 投影 线 、 电源线 、 闭 路线 等 一切 能够 想到 的 线 ；在 32 ㎡ 里 放 一套 沙发 和 预留 5 - 6 个人 的 办公 环境 ；5 ㎡ 里 做 一个 小 卧室 ；在 6 ㎡ 里装 一个 袖珍 厨房 和 一个 厕所 ；还要 在 2 米 * 5 米 的 阳台 上 做 一个 小 花园 ；除 阳台 外套 内 只有 43 ㎡ 的 房子 要 怎么 做 才能 合理 利用 ？$ LOTOzf $',
 '【 女人 命好 的 几点 特征 】   1 . 圆脸 ( 对 人际关系 及 财运 都 有加 分 ) ；2 . 下巴 丰满 （ 可能 会 拥有 两栋 以上 的 房产 ） ；3 . 臀大 ( 代表 有 财运 ) ；4 . 腿 不能 细 （ 腿 长脚 瘦 ， 奔走 不停 ， 辛苦 之相 也 ） ；5 . 小腹 有 脂肪 ( 状是 一种 福寿之 相 ) 。$ LOTOzf $',
 '中央 经济 工作 会议 12 月 12 日至 14 日 在 北京 举行 ， 很 引人瞩目 的 一个 新 的 提法 是 ， 要 提高 中等 收入者 比重 ， 但是 股市 却 把 “ 中等 收入者 ” 一批 又 一批 地 变成 了 “ 低收入 人群 ” ！$ LOTOzf $']

五、停止词使用

1 停止词是为了过滤掉已经确认对训练模型没有实际价值的分词
2 停止词作用：提高模型的精确度和稳定性

# 加载停止词语料

stoplist = [word.strip() for word in open(
'D:\\A\\AI-master\\py-data\\text_classification-master\\
text classification\\stop\\stopword.txt', encoding='utf-8').readlines()]

stoplist[:10]

['\ufeff,', '?', '、', '。', '“', '”', '《', '》', '！', '，']

六、编码器处理文本标签

le = LabelEncoder()

y_train_le = le.fit_transform(y_train)
y_test_le  = le.fit_transform(y_test)

y_train_le, y_test_le

(array([1, 1, 1, ..., 2, 2, 2], dtype=int64),
 array([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
        1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
        0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 3,
        3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 3, 2, 2, 2, 2, 2, 2, 2,
        2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
        2, 2], dtype=int64))

文本数据转换成数据值数据矩阵

count = CountVectorizer(stop_words=stoplist)

'''注意：
这里要先 count.fit() 训练所有训练和测试集，保证特征数一致，
这样在算法建模时才不会报错
'''

count.fit(list(X_train_cut) + list(X_test_cut))
X_train_count = count.transform(X_train_cut)
X_test_count  = count.transform(X_test_cut)

X_train_count = X_train_count.toarray()
X_test_count  = X_test_count.toarray()

print(X_train_count.shape, X_test_count.shape)
X_train_count, X_test_count

(3305, 23732) (200, 23732)


(array([[0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        ...,
        [1, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0]], dtype=int64), array([[0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        ...,
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0],
        [0, 0, 0, ..., 0, 0, 0]], dtype=int64))

七、常规算法模型

封装一个函数，提高复用率，使用时只需调用函数即可。

# 用于存储所有算法的名字，准确率和所消耗的时间
estimator_list, score_list, time_list = [], [], []
def get_text_classification(estimator, X, y, X_test, y_test):
    '''
    estimator: 分类器，必选参数
            X: 特征训练数据，必选参数
            y: 标签训练数据，必选参数
       X_test: 特征测试数据，必选参数
        y_tes: 标签测试数据，必选参数
       return: 返回值
           y_pred_model: 预测值
             classifier: 分类器名字
                  score: 准确率
                      t: 消耗的时间
                  matrix: 混淆矩阵
                  report: 分类评价函数

    '''
    start = time.time()

    print('\n>>>算法正在启动，请稍候...')
    model = estimator

    print('\n>>>算法正在进行训练，请稍候...')
    model.fit(X, y)
    print(model)

    print('\n>>>算法正在进行预测，请稍候...')
    y_pred_model = model.predict(X_test)
    print(y_pred_model)

    print('\n>>>算法正在进行性能评估，请稍候...')
    score = metrics.accuracy_score(y_test, y_pred_model)
    matrix = metrics.confusion_matrix(y_test, y_pred_model)
    report = metrics.classification_report(y_test, y_pred_model)

    print('>>>准确率\n', score)
    print('\n>>>混淆矩阵\n', matrix)
    print('\n>>>召回率\n', report)
    print('>>>算法程序已经结束...')

    end = time.time()
    t = end - start
    print('\n>>>算法消耗时间为：', t, '秒\n')
    classifier = str(model).split('(')[0]

    return y_pred_model, classifier, score, round(t, 2), matrix, report

1、k近邻算法

knc = KNeighborsClassifier()

result = get_text_classification(knc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',                     metric_params=None, n_jobs=None, n_neighbors=5, p=2,                     weights='uniform')算法正在进行预测，请稍候...[1 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 1 1 1 0 1 0 1 2 0 1 0 0 1 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 2 2 2 0 0 0 0 2 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 2 2 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 0 2 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 2 2 0 0 2 0 3 0 2 2 2 2 2 2 0 2 0 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 0 2 2 0 0 0 0 0 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.665混淆矩阵 [[99  0 16  0] [26 10  2  0] [ 8  0 23  0] [ 6  0  9  1]]召回率               precision    recall  f1-score   support           0       0.71      0.86      0.78       115           1       1.00      0.26      0.42        38           2       0.46      0.74      0.57        31           3       1.00      0.06      0.12        16    accuracy                           0.67       200   macro avg       0.79      0.48      0.47       200weighted avg       0.75      0.67      0.62       200算法程序已经结束...算法消耗时间为： 28.78205680847168 秒
(None, None, None)

2、决策树

dtc = DecisionTreeClassifier()

result = get_text_classification(dtc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在进行训练，请稍候...
DecisionTreeClassifier(class_weight=None, criterion='gini', max_depth=None,
                       max_features=None, max_leaf_nodes=None,
                       min_impurity_decrease=0.0, min_impurity_split=None,
                       min_samples_leaf=1, min_samples_split=2,
                       min_weight_fraction_leaf=0.0, presort=False,
                       random_state=None, splitter='best')
算法正在进行预测，请稍候...
[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 2 1 1 0
 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 2 2 0 0 0 1 0 2 0 0 2 0
 1 0 0 0 0 1 0 1 1 1 0 0 0 0 0 0 3 3 0 0 1 1 0 0 0 0 0 0 0 0 1 0 0 0 0 2 0
 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 1 1 1 0 1 1 1 1
 1 0 0 0 0 1 3 1 3 2 3 3 3 3 3 0 3 2 2 2 3 2 1 2 1 2 2 2 2 2 1 2 2 2 2 2 3
 3 2 2 0 1 2 3 3 0 2 1 0 2 2 2]
算法正在进行性能评估，请稍候...
准确率
 0.735
混淆矩阵
 [[84 24  5  2]
 [ 1 35  1  1]
 [ 3  5 19  4]
 [ 1  2  4  9]]
召回率
               precision    recall  f1-score   support
           0       0.94      0.73      0.82       115
           1       0.53      0.92      0.67        38
           2       0.66      0.61      0.63        31
           3       0.56      0.56      0.56        16
    accuracy                           0.73       200
   macro avg       0.67      0.71      0.67       200
weighted avg       0.79      0.73      0.74       200
算法程序已经结束...
算法消耗时间为： 20.55004119873047 秒
Out[35]: (None, None, None)

3、多层感知器

mlpc = MLPClassifier()

result = get_text_classification(mlpc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...MLPClassifier(activation='relu', alpha=0.0001, batch_size='auto', beta_1=0.9,              beta_2=0.999, early_stopping=False, epsilon=1e-08,              hidden_layer_sizes=(100,), learning_rate='constant',              learning_rate_init=0.001, max_iter=200, momentum=0.9,              n_iter_no_change=10, nesterovs_momentum=True, power_t=0.5,              random_state=None, shuffle=True, solver='adam', tol=0.0001,              validation_fraction=0.1, verbose=False, warm_start=False)算法正在进行预测，请稍候...[1 1 2 1 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 1 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 1 0 0 0 0 0 0 3 2 3 2 3 1 3 3 3 0 3 2 2 2 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 0 2 1 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.89混淆矩阵 [[105   3   6   1] [  0  36   1   1] [  1   1  29   0] [  2   1   5   8]]召回率               precision    recall  f1-score   support           0       0.97      0.91      0.94       115           1       0.88      0.95      0.91        38           2       0.71      0.94      0.81        31           3       0.80      0.50      0.62        16    accuracy                           0.89       200   macro avg       0.84      0.82      0.82       200weighted avg       0.90      0.89      0.89       200算法程序已经结束...算法消耗时间为： 234.1950967311859 秒
(None, None, None)

4、伯努力贝叶斯

bnb = BernoulliNB()

result = get_text_classification(bnb, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...BernoulliNB(alpha=1.0, binarize=0.0, class_prior=None, fit_prior=True)算法正在进行预测，请稍候...[1 1 0 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 1 1 1 1 1 1 0 1 0 1 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 2 0 0 0 2 0 0 0 0 0 0 2 2 0 0 0 2 2 2 2 0 0 2 0 0 2 2 0]算法正在进行性能评估，请稍候...准确率 0.785混淆矩阵 [[113   0   2   0] [  6  32   0   0] [ 19   0  12   0] [ 15   0   1   0]]召回率               precision    recall  f1-score   support           0       0.74      0.98      0.84       115           1       1.00      0.84      0.91        38           2       0.80      0.39      0.52        31           3       0.00      0.00      0.00        16    accuracy                           0.79       200   macro avg       0.63      0.55      0.57       200weighted avg       0.74      0.79      0.74       200算法程序已经结束...算法消耗时间为： 1.1219978332519531 秒
(None, None, None)

5、高斯贝叶斯

gnb = GaussianNB()

result = get_text_classification(gnb, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...GaussianNB(priors=None, var_smoothing=1e-09)算法正在进行预测，请稍候...[1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 2 1 2 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 0 0 0 1 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 1 2 3 3 1 3 3 3 2 3 3 3 3 3 2 3 2 2 2 2 0 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.89混淆矩阵 [[105   1   8   1] [  2  34   2   0] [  1   1  28   1] [  1   2   2  11]]召回率               precision    recall  f1-score   support           0       0.96      0.91      0.94       115           1       0.89      0.89      0.89        38           2       0.70      0.90      0.79        31           3       0.85      0.69      0.76        16    accuracy                           0.89       200   macro avg       0.85      0.85      0.84       200weighted avg       0.90      0.89      0.89       200算法程序已经结束...算法消耗时间为： 2.2403738498687744 秒
 (None, None, None)

6、多项式贝叶斯

mnb = MultinomialNB()

result = get_text_classification(mnb, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...MultinomialNB(alpha=1.0, class_prior=None, fit_prior=True)算法正在进行预测，请稍候...[1 1 2 1 3 3 1 1 1 1 1 2 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 0 2 3 3 3 0 0 0 0 3 3 2 3 0 3 2 3 3 3 3 3 0 0 0 3 2 2 2 2 2 2 2 2 2 2 0 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.905混淆矩阵 [[108   1   2   4] [  0  33   3   2] [  1   0  30   0] [  4   0   2  10]]召回率               precision    recall  f1-score   support           0       0.96      0.94      0.95       115           1       0.97      0.87      0.92        38           2       0.81      0.97      0.88        31           3       0.62      0.62      0.62        16    accuracy                           0.91       200   macro avg       0.84      0.85      0.84       200weighted avg       0.91      0.91      0.91       200算法程序已经结束...算法消耗时间为： 0.5864312648773193 秒
 (None, None, None)

7、逻辑回归

lgr = LogisticRegression()

result = get_text_classification(lgr, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,                   intercept_scaling=1, l1_ratio=None, max_iter=100,                   multi_class='warn', n_jobs=None, penalty='l2',                   random_state=None, solver='warn', tol=0.0001, verbose=0,                   warm_start=False)算法正在进行预测，请稍候...[1 1 2 1 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 1 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 1 0 2 0 0 0 1 2 0 0 0 0 0 3 2 3 2 3 3 3 3 3 0 3 0 0 0 3 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 1 2 2 2 2 2 2 1 0 2 2 2]算法正在进行性能评估，请稍候...准确率 0.89混淆矩阵 [[105   3   7   0] [  0  36   1   1] [  1   2  28   0] [  5   0   2   9]]召回率               precision    recall  f1-score   support           0       0.95      0.91      0.93       115           1       0.88      0.95      0.91        38           2       0.74      0.90      0.81        31           3       0.90      0.56      0.69        16    accuracy                           0.89       200   macro avg       0.87      0.83      0.84       200weighted avg       0.90      0.89      0.89       200算法程序已经结束...算法消耗时间为： 0.7747480869293213 秒
(None, None, None)

8、支持向量机

svc = svm.SVC()

result = get_text_classification(svc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...SVC(C=1.0, cache_size=200, class_weight=None, coef0=0.0,    decision_function_shape='ovr', degree=3, gamma='auto_deprecated',    kernel='rbf', max_iter=-1, probability=False, random_state=None,    shrinking=True, tol=0.001, verbose=False)算法正在进行预测，请稍候...[0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]算法正在进行性能评估，请稍候...准确率 0.575混淆矩阵 [[115   0   0   0] [ 38   0   0   0] [ 31   0   0   0] [ 16   0   0   0]]召回率               precision    recall  f1-score   support           0       0.57      1.00      0.73       115           1       0.00      0.00      0.00        38           2       0.00      0.00      0.00        31           3       0.00      0.00      0.00        16    accuracy                           0.57       200   macro avg       0.14      0.25      0.18       200weighted avg       0.33      0.57      0.42       200算法程序已经结束...算法消耗时间为： 401.34070205688477 秒
 (None, None, None)

八、集成算法模型

1、随机森林算法

rfc = RandomForestClassifier()

result = get_text_classification(rfc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',                       max_depth=None, max_features='auto', max_leaf_nodes=None,                       min_impurity_decrease=0.0, min_impurity_split=None,                       min_samples_leaf=1, min_samples_split=2,                       min_weight_fraction_leaf=0.0, n_estimators=10,                       n_jobs=None, oob_score=False, random_state=None,                       verbose=0, warm_start=False)算法正在进行预测，请稍候...[1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 2 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 1 2 1 1 1 2 1 0 0 0 0 0 1 2 2 3 2 3 3 2 3 3 0 3 2 2 2 3 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 0 2 2 1 2 2 3 0 0 2 0 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.82混淆矩阵 [[97  6 12  0] [ 1 36  1  0] [ 4  2 24  1] [ 1  1  7  7]]召回率               precision    recall  f1-score   support           0       0.94      0.84      0.89       115           1       0.80      0.95      0.87        38           2       0.55      0.77      0.64        31           3       0.88      0.44      0.58        16    accuracy                           0.82       200   macro avg       0.79      0.75      0.75       200weighted avg       0.85      0.82      0.82       200算法程序已经结束...算法消耗时间为： 4.107015609741211 秒
 (None, None, None)

2、自适应增强算法

abc = AdaBoostClassifier()

result = get_text_classification(abc, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',                       max_depth=None, max_features='auto', max_leaf_nodes=None,                       min_impurity_decrease=0.0, min_impurity_split=None,                       min_samples_leaf=1, min_samples_split=2,                       min_weight_fraction_leaf=0.0, n_estimators=10,                       n_jobs=None, oob_score=False, random_state=None,                       verbose=0, warm_start=False)算法正在进行预测，请稍候...[1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 2 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 1 2 1 1 1 2 1 0 0 0 0 0 1 2 2 3 2 3 3 2 3 3 0 3 2 2 2 3 2 1 2 2 2 2 2 2 2 2 2 2 2 2 2 2 0 2 2 1 2 2 3 0 0 2 0 2 2 2 2]算法正在进行性能评估，请稍候...准确率 0.82混淆矩阵 [[97  6 12  0] [ 1 36  1  0] [ 4  2 24  1] [ 1  1  7  7]]召回率               precision    recall  f1-score   support           0       0.94      0.84      0.89       115           1       0.80      0.95      0.87        38           2       0.55      0.77      0.64        31           3       0.88      0.44      0.58        16    accuracy                           0.82       200   macro avg       0.79      0.75      0.75       200weighted avg       0.85      0.82      0.82       200算法程序已经结束...算法消耗时间为： 4.107015609741211 秒Out[43]: (None, None, None)abc = AdaBoostClassifier()result = get_text_classification(abc, X_train_count, y_train_le, X_test_count, y_test_le)estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])算法正在启动，请稍候...算法正在进行训练，请稍候...AdaBoostClassifier(algorithm='SAMME.R', base_estimator=None, learning_rate=1.0,                   n_estimators=50, random_state=None)算法正在进行预测，请稍候...[1 1 1 1 0 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 1 0 0 1 0 1 1 0 1 1 0 0 0 1 1 0 0 0 1 1 1 0 1 0 1 0 1 1 1 1 1 0 1 0 0 0 1 1 0 1 1 1 0 1 0 1 1 1 0 1 1 1 0 1 1 1 0 1 0 0 0 0 0 0 0 1 0 0 0 1 1 1 0 0 0 1 1 0 1 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 1 3 1 3 1 3 3 3 3 3 1 3 1 1 1 3 2 1 2 1 2 2 0 2 2 1 1 1 1 1 1 3 3 1 1 1 1 2 3 0 1 2 1 1 1 1 2]算法正在进行性能评估，请稍候...准确率 0.57混淆矩阵 [[59 56  0  0] [ 1 37  0  0] [ 2 17  9  3] [ 0  7  0  9]]召回率               precision    recall  f1-score   support           0       0.95      0.51      0.67       115           1       0.32      0.97      0.48        38           2       1.00      0.29      0.45        31           3       0.75      0.56      0.64        16    accuracy                           0.57       200   macro avg       0.75      0.58      0.56       200weighted avg       0.82      0.57      0.60       200算法程序已经结束...算法消耗时间为： 94.61701512336731 秒
(None, None, None)

3、lightgbm算法

gbm = lightgbm.LGBMClassifier()

result = get_text_classification(gbm, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...LGBMClassifier(boosting_type='gbdt', class_weight=None, colsample_bytree=1.0,               importance_type='split', learning_rate=0.1, max_depth=-1,               min_child_samples=20, min_child_weight=0.001, min_split_gain=0.0,               n_estimators=100, n_jobs=-1, num_leaves=31, objective=None,               random_state=None, reg_alpha=0.0, reg_lambda=0.0, silent=True,               subsample=1.0, subsample_for_bin=200000, subsample_freq=0)算法正在进行预测，请稍候...[1 2 1 1 3 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 3 1 2 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 2 2 0 0 0 2 0 0 0 0 2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 2 3 1 2 0 1 2 2 1 2 0 0 0 0 1 2 2 3 2 3 3 2 3 3 0 3 2 2 2 3 2 2 2 2 2 2 0 2 2 2 3 3 2 2 2 2 2 2 2 1 2 2 3 2 0 2 1 0 2 1 2]算法正在进行性能评估，请稍候...准确率 0.785混淆矩阵 [[94  6 13  2] [ 0 34  2  2] [ 3  3 22  3] [ 1  1  7  7]]召回率               precision    recall  f1-score   support           0       0.96      0.82      0.88       115           1       0.77      0.89      0.83        38           2       0.50      0.71      0.59        31           3       0.50      0.44      0.47        16    accuracy                           0.79       200   macro avg       0.68      0.71      0.69       200weighted avg       0.82      0.79      0.79       200算法程序已经结束...算法消耗时间为： 1.754338264465332 秒
(None, None, None)

4、xgboost算法

xgb = xgboost.XGBClassifier()

result = get_text_classification(xgb, X_train_count, y_train_le, X_test_count, y_test_le)
estimator_list.append(result[1]), score_list.append(result[2]), time_list.append(result[3])

算法正在启动，请稍候...算法正在进行训练，请稍候...XGBClassifier(base_score=0.5, booster='gbtree', colsample_bylevel=1,              colsample_bynode=1, colsample_bytree=1, gamma=0,              learning_rate=0.1, max_delta_step=0, max_depth=3,              min_child_weight=1, missing=None, n_estimators=100, n_jobs=1,              nthread=None, objective='multi:softprob', random_state=0,              reg_alpha=0, reg_lambda=1, scale_pos_weight=1, seed=None,              silent=None, subsample=1, verbosity=1)算法正在进行预测，请稍候...[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 1 0 1 1 0 0 0 1 0 0 0 0 0 1 1 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 1 0 0 1 1 0 0 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 1 0 1 0 1 0 0 1 0 1 0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 1 0 0 0 0 1 3 1 3 0 3 3 3 3 3 0 3 1 1 1 3 2 2 2 1 2 2 2 2 2 1 1 2 2 2 2 3 1 1 2 1 1 2 3 2 0 2 1 1 1 1 2]算法正在进行性能评估，请稍候...准确率 0.735混淆矩阵 [[83 32  0  0] [ 0 38  0  0] [ 1 11 17  2] [ 2  5  0  9]]召回率               precision    recall  f1-score   support           0       0.97      0.72      0.83       115           1       0.44      1.00      0.61        38           2       1.00      0.55      0.71        31           3       0.82      0.56      0.67        16    accuracy                           0.73       200   macro avg       0.81      0.71      0.70       200weighted avg       0.86      0.73      0.75       200算法程序已经结束...算法消耗时间为： 974.3143627643585 秒Out[46]: (None, None, None)

九、深度学习

1、前馈神经网络

 Keras 也是一个高级封装的接口，简单介绍如下。

1） 算法流程

创建神经网络——添加神经层——编译神经网络——训练神经网络——预测——性能评估——保存模型

 

2） 添加神经层

至少要有两层神经层，第一层必须是输入神经层，最后一层必须是输出层；

输入神经层主要设置输入的维度，而最后一层主要是设置激活函数的类型来指明是分类还是回归问题

 

3 ）编译神经网络

分类问题的 metrics，一般以 accuracy 准确率来衡量

回归问题的 metrics, 一般以 mae 平均绝对误差来衡量

start = time.time()
# --------------------------------
# np.random.seed(0)     # 设置随机数种子
feature_num = X_train_count.shape[1]     # 设置所希望的特征数量

# ---------------------------------
# 独热编码目标向量来创建目标矩阵
y_train_cate = to_categorical(y_train_le)
y_test_cate = to_categorical(y_test_le)
print(y_train_cate)

[[0. 1. 0. 0.]
 [0. 1. 0. 0.]
 [0. 1. 0. 0.]
 ...
 [0. 0. 1. 0.]
 [0. 0. 1. 0.]
 [0. 0. 1. 0.]]

# 1 创建神经网络
network = models.Sequential() 

# ----------------------------------------------------
# 2 添加神经连接层
# 第一层必须有并且一定是 [输入层], 必选
network.add(layers.Dense(     # 添加带有 relu 激活函数的全连接层
                         units=128, 
                         activation='relu', 
                         input_shape=(feature_num, )
                         ))

# 介于第一层和最后一层之间的称为 [隐藏层]，可选
network.add(layers.Dense(     # 添加带有 relu 激活函数的全连接层
                         units=128, 
                         activation='relu'
                         ))
network.add(layers.Dropout(0.8))
# 最后一层必须有并且一定是 [输出层], 必选                         
network.add(layers.Dense(     # 添加带有 softmax 激活函数的全连接层
                         units=4,
                         activation='sigmoid'
                         ))

# -----------------------------------------------------
# 3 编译神经网络
network.compile(loss='categorical_crossentropy',  # 分类交叉熵损失函数    
                optimizer='rmsprop',  
                metrics=['accuracy']              # 准确率度量
                )

# -----------------------------------------------------
# 4 开始训练神经网络
history = network.fit(X_train_count,     # 训练集特征
            y_train_cate,        # 训练集标签
            epochs=20,          # 迭代次数
            batch_size=300,    # 每个批量的观测数  可做优化
            validation_data=(X_test_count, y_test_cate)  # 验证测试集数据
            )
network.summary()

Train on 3305 samples, validate on 200 samples
Epoch 1/20
 300/3305 [=>............................] - ETA: 3s - loss: 1.3948 - accuracy: 0.2167
 600/3305 [====>.........................] - ETA: 2s - loss: 1.3805 - accuracy: 0.3167
 900/3305 [=======>......................] - ETA: 1s - loss: 1.3690 - accuracy: 0.3667
1200/3305 [=========>....................] - ETA: 1s - loss: 1.3561 - accuracy: 0.3950
1500/3305 [============>.................] - ETA: 1s - loss: 1.3446 - accuracy: 0.4100
1800/3305 [===============>..............] - ETA: 0s - loss: 1.3301 - accuracy: 0.4283
2100/3305 [==================>...........] - ETA: 0s - loss: 1.3139 - accuracy: 0.4429
2400/3305 [====================>.........] - ETA: 0s - loss: 1.2963 - accuracy: 0.4563
2700/3305 [=======================>......] - ETA: 0s - loss: 1.2767 - accuracy: 0.4707
3000/3305 [==========================>...] - ETA: 0s - loss: 1.2578 - accuracy: 0.4810
3300/3305 [============================>.] - ETA: 0s - loss: 1.2416 - accuracy: 0.4912
3305/3305 [==============================] - 2s 637us/step - loss: 1.2411 - accuracy: 0.4914 - val_loss: 0.9422 - val_accuracy: 0.7000
Epoch 2/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.9911 - accuracy: 0.6033
 600/3305 [====>.........................] - ETA: 1s - loss: 0.9473 - accuracy: 0.6433
 900/3305 [=======>......................] - ETA: 1s - loss: 0.9485 - accuracy: 0.6411
1200/3305 [=========>....................] - ETA: 1s - loss: 0.9281 - accuracy: 0.6567
1500/3305 [============>.................] - ETA: 0s - loss: 0.9095 - accuracy: 0.6733
1800/3305 [===============>..............] - ETA: 0s - loss: 0.8946 - accuracy: 0.6850
2100/3305 [==================>...........] - ETA: 0s - loss: 0.8755 - accuracy: 0.7005
2400/3305 [====================>.........] - ETA: 0s - loss: 0.8618 - accuracy: 0.7067
2700/3305 [=======================>......] - ETA: 0s - loss: 0.8496 - accuracy: 0.7126
3000/3305 [==========================>...] - ETA: 0s - loss: 0.8383 - accuracy: 0.7157
3300/3305 [============================>.] - ETA: 0s - loss: 0.8260 - accuracy: 0.7233
3305/3305 [==============================] - 2s 481us/step - loss: 0.8256 - accuracy: 0.7238 - val_loss: 0.6232 - val_accuracy: 0.8400
Epoch 3/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.6145 - accuracy: 0.8500
 600/3305 [====>.........................] - ETA: 1s - loss: 0.5991 - accuracy: 0.8550
 900/3305 [=======>......................] - ETA: 0s - loss: 0.5806 - accuracy: 0.8622
1200/3305 [=========>....................] - ETA: 0s - loss: 0.5707 - accuracy: 0.8667
1500/3305 [============>.................] - ETA: 0s - loss: 0.5585 - accuracy: 0.8693
1800/3305 [===============>..............] - ETA: 0s - loss: 0.5467 - accuracy: 0.8739
2100/3305 [==================>...........] - ETA: 0s - loss: 0.5445 - accuracy: 0.8705
2400/3305 [====================>.........] - ETA: 0s - loss: 0.5359 - accuracy: 0.8700
2700/3305 [=======================>......] - ETA: 0s - loss: 0.5235 - accuracy: 0.8778
3000/3305 [==========================>...] - ETA: 0s - loss: 0.5224 - accuracy: 0.8763
3300/3305 [============================>.] - ETA: 0s - loss: 0.5158 - accuracy: 0.8806
3305/3305 [==============================] - 1s 447us/step - loss: 0.5162 - accuracy: 0.8805 - val_loss: 0.4481 - val_accuracy: 0.8800
Epoch 4/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.3936 - accuracy: 0.9167
 600/3305 [====>.........................] - ETA: 1s - loss: 0.3881 - accuracy: 0.9133
 900/3305 [=======>......................] - ETA: 0s - loss: 0.3891 - accuracy: 0.9089
1200/3305 [=========>....................] - ETA: 0s - loss: 0.3737 - accuracy: 0.9158
1500/3305 [============>.................] - ETA: 0s - loss: 0.3714 - accuracy: 0.9147
1800/3305 [===============>..............] - ETA: 0s - loss: 0.3725 - accuracy: 0.9139
2100/3305 [==================>...........] - ETA: 0s - loss: 0.3613 - accuracy: 0.9200
2400/3305 [====================>.........] - ETA: 0s - loss: 0.3559 - accuracy: 0.9192
2700/3305 [=======================>......] - ETA: 0s - loss: 0.3526 - accuracy: 0.9207
3000/3305 [==========================>...] - ETA: 0s - loss: 0.3429 - accuracy: 0.9247
3300/3305 [============================>.] - ETA: 0s - loss: 0.3367 - accuracy: 0.9264
3305/3305 [==============================] - 1s 447us/step - loss: 0.3369 - accuracy: 0.9259 - val_loss: 0.3506 - val_accuracy: 0.9000
Epoch 5/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.2625 - accuracy: 0.9567
 600/3305 [====>.........................] - ETA: 1s - loss: 0.2679 - accuracy: 0.9483
 900/3305 [=======>......................] - ETA: 0s - loss: 0.2576 - accuracy: 0.9567
1200/3305 [=========>....................] - ETA: 0s - loss: 0.2515 - accuracy: 0.9558
1500/3305 [============>.................] - ETA: 0s - loss: 0.2498 - accuracy: 0.9567
1800/3305 [===============>..............] - ETA: 0s - loss: 0.2446 - accuracy: 0.9572
2100/3305 [==================>...........] - ETA: 0s - loss: 0.2459 - accuracy: 0.9519
2400/3305 [====================>.........] - ETA: 0s - loss: 0.2445 - accuracy: 0.9521
2700/3305 [=======================>......] - ETA: 0s - loss: 0.2407 - accuracy: 0.9537
3000/3305 [==========================>...] - ETA: 0s - loss: 0.2361 - accuracy: 0.9553
3300/3305 [============================>.] - ETA: 0s - loss: 0.2350 - accuracy: 0.9536
3305/3305 [==============================] - 2s 479us/step - loss: 0.2347 - accuracy: 0.9537 - val_loss: 0.2831 - val_accuracy: 0.9050
Epoch 6/20
 300/3305 [=>............................] - ETA: 2s - loss: 0.1655 - accuracy: 0.9700
 600/3305 [====>.........................] - ETA: 2s - loss: 0.1610 - accuracy: 0.9733
 900/3305 [=======>......................] - ETA: 1s - loss: 0.1643 - accuracy: 0.9767
1200/3305 [=========>....................] - ETA: 1s - loss: 0.1644 - accuracy: 0.9758
1500/3305 [============>.................] - ETA: 1s - loss: 0.1683 - accuracy: 0.9740
1800/3305 [===============>..............] - ETA: 0s - loss: 0.1602 - accuracy: 0.9767
2100/3305 [==================>...........] - ETA: 0s - loss: 0.1545 - accuracy: 0.9781
2400/3305 [====================>.........] - ETA: 0s - loss: 0.1519 - accuracy: 0.9787
2700/3305 [=======================>......] - ETA: 0s - loss: 0.1523 - accuracy: 0.9774
3000/3305 [==========================>...] - ETA: 0s - loss: 0.1515 - accuracy: 0.9777
3300/3305 [============================>.] - ETA: 0s - loss: 0.1510 - accuracy: 0.9773
3305/3305 [==============================] - 2s 524us/step - loss: 0.1511 - accuracy: 0.9773 - val_loss: 0.2449 - val_accuracy: 0.9050
Epoch 7/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.1142 - accuracy: 0.9933
 600/3305 [====>.........................] - ETA: 1s - loss: 0.1093 - accuracy: 0.9850
 900/3305 [=======>......................] - ETA: 0s - loss: 0.1182 - accuracy: 0.9767
1200/3305 [=========>....................] - ETA: 0s - loss: 0.1182 - accuracy: 0.9775
1500/3305 [============>.................] - ETA: 0s - loss: 0.1195 - accuracy: 0.9800
1800/3305 [===============>..............] - ETA: 0s - loss: 0.1158 - accuracy: 0.9817
2100/3305 [==================>...........] - ETA: 0s - loss: 0.1146 - accuracy: 0.9819
2400/3305 [====================>.........] - ETA: 0s - loss: 0.1124 - accuracy: 0.9833
2700/3305 [=======================>......] - ETA: 0s - loss: 0.1104 - accuracy: 0.9841
3000/3305 [==========================>...] - ETA: 0s - loss: 0.1078 - accuracy: 0.9847
3300/3305 [============================>.] - ETA: 0s - loss: 0.1078 - accuracy: 0.9842
3305/3305 [==============================] - 1s 447us/step - loss: 0.1076 - accuracy: 0.9843 - val_loss: 0.2377 - val_accuracy: 0.9000
Epoch 8/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0846 - accuracy: 0.9867
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0838 - accuracy: 0.9867
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0844 - accuracy: 0.9867
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0877 - accuracy: 0.9842
1500/3305 [============>.................] - ETA: 0s - loss: 0.0853 - accuracy: 0.9853
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0852 - accuracy: 0.9856
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0825 - accuracy: 0.9862
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0821 - accuracy: 0.9867
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0807 - accuracy: 0.9874
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0826 - accuracy: 0.9863
3300/3305 [============================>.] - ETA: 0s - loss: 0.0817 - accuracy: 0.9861
3305/3305 [==============================] - 1s 449us/step - loss: 0.0816 - accuracy: 0.9861 - val_loss: 0.2296 - val_accuracy: 0.8850
Epoch 9/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0553 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0574 - accuracy: 0.9950
 900/3305 [=======>......................] - ETA: 0s - loss: 0.0586 - accuracy: 0.9944
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0690 - accuracy: 0.9908
1500/3305 [============>.................] - ETA: 0s - loss: 0.0693 - accuracy: 0.9920
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0673 - accuracy: 0.9922
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0657 - accuracy: 0.9924
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0686 - accuracy: 0.9896
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0685 - accuracy: 0.9893
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0670 - accuracy: 0.9897
3300/3305 [============================>.] - ETA: 0s - loss: 0.0660 - accuracy: 0.9897
3305/3305 [==============================] - 1s 448us/step - loss: 0.0659 - accuracy: 0.9897 - val_loss: 0.2290 - val_accuracy: 0.8950
Epoch 10/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0477 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0562 - accuracy: 0.9917
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0500 - accuracy: 0.9933
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0517 - accuracy: 0.9908
1500/3305 [============>.................] - ETA: 0s - loss: 0.0490 - accuracy: 0.9927
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0499 - accuracy: 0.9928
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0493 - accuracy: 0.9929
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0488 - accuracy: 0.9925
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0474 - accuracy: 0.9933
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0459 - accuracy: 0.9937
3300/3305 [============================>.] - ETA: 0s - loss: 0.0444 - accuracy: 0.9942
3305/3305 [==============================] - 1s 450us/step - loss: 0.0443 - accuracy: 0.9943 - val_loss: 0.2183 - val_accuracy: 0.8900
Epoch 11/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0316 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0464 - accuracy: 0.9917
 900/3305 [=======>......................] - ETA: 0s - loss: 0.0444 - accuracy: 0.9933
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0436 - accuracy: 0.9917
1500/3305 [============>.................] - ETA: 0s - loss: 0.0448 - accuracy: 0.9913
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0431 - accuracy: 0.9917
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0410 - accuracy: 0.9929
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0405 - accuracy: 0.9929
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0393 - accuracy: 0.9933
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0399 - accuracy: 0.9933
3300/3305 [============================>.] - ETA: 0s - loss: 0.0394 - accuracy: 0.9936
3305/3305 [==============================] - 1s 451us/step - loss: 0.0394 - accuracy: 0.9936 - val_loss: 0.2331 - val_accuracy: 0.9000
Epoch 12/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0263 - accuracy: 1.0000
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0263 - accuracy: 0.9983
 900/3305 [=======>......................] - ETA: 0s - loss: 0.0314 - accuracy: 0.9967
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0312 - accuracy: 0.9967
1500/3305 [============>.................] - ETA: 0s - loss: 0.0286 - accuracy: 0.9973
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0270 - accuracy: 0.9978
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0254 - accuracy: 0.9981
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0251 - accuracy: 0.9983
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0263 - accuracy: 0.9981
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0277 - accuracy: 0.9973
3300/3305 [============================>.] - ETA: 0s - loss: 0.0298 - accuracy: 0.9967
3305/3305 [==============================] - 2s 482us/step - loss: 0.0298 - accuracy: 0.9967 - val_loss: 0.2208 - val_accuracy: 0.8850
Epoch 13/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0254 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0234 - accuracy: 0.9983
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0220 - accuracy: 0.9978
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0228 - accuracy: 0.9958
1500/3305 [============>.................] - ETA: 0s - loss: 0.0279 - accuracy: 0.9947
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0269 - accuracy: 0.9950
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0275 - accuracy: 0.9952
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0273 - accuracy: 0.9954
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0277 - accuracy: 0.9956
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0273 - accuracy: 0.9957
3300/3305 [============================>.] - ETA: 0s - loss: 0.0274 - accuracy: 0.9955
3305/3305 [==============================] - 2s 459us/step - loss: 0.0273 - accuracy: 0.9955 - val_loss: 0.2477 - val_accuracy: 0.8950
Epoch 14/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0187 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0249 - accuracy: 0.9950
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0240 - accuracy: 0.9956
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0222 - accuracy: 0.9967
1500/3305 [============>.................] - ETA: 0s - loss: 0.0210 - accuracy: 0.9967
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0201 - accuracy: 0.9972
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0203 - accuracy: 0.9971
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0201 - accuracy: 0.9975
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0204 - accuracy: 0.9974
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0205 - accuracy: 0.9970
3300/3305 [============================>.] - ETA: 0s - loss: 0.0204 - accuracy: 0.9970
3305/3305 [==============================] - 2s 455us/step - loss: 0.0204 - accuracy: 0.9970 - val_loss: 0.2115 - val_accuracy: 0.8950
Epoch 15/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0182 - accuracy: 1.0000
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0163 - accuracy: 1.0000
 900/3305 [=======>......................] - ETA: 0s - loss: 0.0156 - accuracy: 0.9989
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0144 - accuracy: 0.9992
1500/3305 [============>.................] - ETA: 0s - loss: 0.0140 - accuracy: 0.9987
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0145 - accuracy: 0.9983
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0144 - accuracy: 0.9986
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0148 - accuracy: 0.9983
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0142 - accuracy: 0.9985
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0149 - accuracy: 0.9980
3300/3305 [============================>.] - ETA: 0s - loss: 0.0154 - accuracy: 0.9973
3305/3305 [==============================] - 1s 450us/step - loss: 0.0154 - accuracy: 0.9973 - val_loss: 0.2775 - val_accuracy: 0.8950
Epoch 16/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0110 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0140 - accuracy: 0.9967
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0123 - accuracy: 0.9978
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0127 - accuracy: 0.9983
1500/3305 [============>.................] - ETA: 0s - loss: 0.0137 - accuracy: 0.9973
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0147 - accuracy: 0.9972
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0157 - accuracy: 0.9967
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0157 - accuracy: 0.9967
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0152 - accuracy: 0.9963
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0149 - accuracy: 0.9963
3300/3305 [============================>.] - ETA: 0s - loss: 0.0149 - accuracy: 0.9964
3305/3305 [==============================] - 1s 453us/step - loss: 0.0149 - accuracy: 0.9964 - val_loss: 0.2708 - val_accuracy: 0.9000
Epoch 17/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0140 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0155 - accuracy: 0.9967
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0147 - accuracy: 0.9978
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0130 - accuracy: 0.9983
1500/3305 [============>.................] - ETA: 0s - loss: 0.0132 - accuracy: 0.9980
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0129 - accuracy: 0.9978
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0137 - accuracy: 0.9976
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0132 - accuracy: 0.9979
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0147 - accuracy: 0.9970
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0142 - accuracy: 0.9973
3300/3305 [============================>.] - ETA: 0s - loss: 0.0145 - accuracy: 0.9973
3305/3305 [==============================] - 1s 452us/step - loss: 0.0145 - accuracy: 0.9973 - val_loss: 0.2197 - val_accuracy: 0.9100
Epoch 18/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0116 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0132 - accuracy: 0.9950
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0123 - accuracy: 0.9967
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0129 - accuracy: 0.9975
1500/3305 [============>.................] - ETA: 0s - loss: 0.0137 - accuracy: 0.9973
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0125 - accuracy: 0.9978
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0123 - accuracy: 0.9976
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0150 - accuracy: 0.9967
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0167 - accuracy: 0.9963
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0162 - accuracy: 0.9967
3300/3305 [============================>.] - ETA: 0s - loss: 0.0156 - accuracy: 0.9970
3305/3305 [==============================] - 2s 469us/step - loss: 0.0156 - accuracy: 0.9970 - val_loss: 0.2645 - val_accuracy: 0.9000
Epoch 19/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0057 - accuracy: 1.0000
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0052 - accuracy: 1.0000
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0072 - accuracy: 0.9989
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0078 - accuracy: 0.9992
1500/3305 [============>.................] - ETA: 0s - loss: 0.0081 - accuracy: 0.9993
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0083 - accuracy: 0.9994
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0119 - accuracy: 0.9986
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0136 - accuracy: 0.9979
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0127 - accuracy: 0.9981
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0142 - accuracy: 0.9980
3300/3305 [============================>.] - ETA: 0s - loss: 0.0144 - accuracy: 0.9976
3305/3305 [==============================] - 2s 466us/step - loss: 0.0144 - accuracy: 0.9976 - val_loss: 0.2941 - val_accuracy: 0.9000
Epoch 20/20
 300/3305 [=>............................] - ETA: 1s - loss: 0.0105 - accuracy: 0.9967
 600/3305 [====>.........................] - ETA: 1s - loss: 0.0172 - accuracy: 0.9933
 900/3305 [=======>......................] - ETA: 1s - loss: 0.0156 - accuracy: 0.9933
1200/3305 [=========>....................] - ETA: 0s - loss: 0.0138 - accuracy: 0.9950
1500/3305 [============>.................] - ETA: 0s - loss: 0.0128 - accuracy: 0.9960
1800/3305 [===============>..............] - ETA: 0s - loss: 0.0120 - accuracy: 0.9961
2100/3305 [==================>...........] - ETA: 0s - loss: 0.0131 - accuracy: 0.9962
2400/3305 [====================>.........] - ETA: 0s - loss: 0.0126 - accuracy: 0.9967
2700/3305 [=======================>......] - ETA: 0s - loss: 0.0142 - accuracy: 0.9959
3000/3305 [==========================>...] - ETA: 0s - loss: 0.0132 - accuracy: 0.9963
3300/3305 [============================>.] - ETA: 0s - loss: 0.0125 - accuracy: 0.9967
3305/3305 [==============================] - 2s 456us/step - loss: 0.0125 - accuracy: 0.9967 - val_loss: 0.3149 - val_accuracy: 0.8950
Model: "sequential_2"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
dense_1 (Dense)              (None, 128)               3037824   
_________________________________________________________________
dense_2 (Dense)              (None, 128)               16512     
_________________________________________________________________
dropout_1 (Dropout)          (None, 128)               0         
_________________________________________________________________
dense_3 (Dense)              (None, 4)                 516       
=================================================================
Total params: 3,054,852
Trainable params: 3,054,852
Non-trainable params: 0
_________________________________________________________________

# 5 模型预测

y_pred_keras = network.predict(X_test_count)

# 6 性能评估
print('>>>多分类前馈神经网络性能评估如下...\n')
score = network.evaluate(X_test_count,
                        y_test_cate,
                        batch_size=32)
print('\n>>>评分\n', score)
print()
end = time.time()

estimator_list.append('前馈网络')
score_list.append(score[1])
time_list.append(round(end-start, 2))

多分类前馈神经网络性能评估如下...
 32/200 [===>..........................] - ETA: 0s
128/200 [==================>...........] - ETA: 0s
200/200 [==============================] - 0s 529us/step
评分
 [0.31489723950624465, 0.8949999809265137]

# 损失函数情况
import matplotlib.pyplot as plt

train_loss = history.history["loss"]
valid_loss = history.history["val_loss"]
epochs = [i for i in range(len(train_loss))]
plt.plot(epochs, train_loss,linewidth=3.0)
plt.plot(epochs, valid_loss,linewidth=3.0)
plt.show()

# 准确率情况
train_loss = history.history['accuracy']
valid_loss = history.history['val_accuracy']
epochs = [i for i in range(len(train_loss))]
plt.plot(epochs, train_loss,linewidth=3.0)
plt.plot(epochs, valid_loss,linewidth=3.0)
plt.show()

# 7 保存/加载模型

# 保存
print('\n>>>你正在进行保存模型操作, 请稍候...\n')

network.save('D:\\A\AI-master\\py-data\\text_classification-master\\my_network_model.h5')

print('>>>保存工作已完成...\n')


# 加载和使用
print('>>>你正在加载已经训练好的模型, 请稍候...\n')

my_load_model = models.load_model('D:\\A\\AI-master\\py-data\\text_classification-master\\my_network_model.h5')

print('>>>你正在使用加载的现成模型进行预测, 请稍候...\n')
print('>>>预测部分结果如下...')

my_load_model.predict(X_test_count)[:20]

你正在进行保存模型操作, 请稍候...
保存工作已完成...


你正在加载已经训练好的模型, 请稍候...
W0401 16:51:40.653452  2520 nn_ops.py:4372] Large dropout rate: 0.8 (>0.5). In TensorFlow 2.x, dropout() uses dropout rate instead of keep_prob. Please ensure that this is intended.
你正在使用加载的现成模型进行预测, 请稍候...
预测部分结果如下...


array([[6.5150915e-14, 9.9956208e-01, 7.6179704e-15, 5.9075337e-14],
       [1.3438159e-07, 7.5693017e-01, 6.7286510e-06, 1.0690396e-06],
       [1.1479392e-05, 1.3596797e-03, 5.2056019e-04, 3.6293927e-06],
       [1.7625215e-08, 9.9138886e-01, 1.0188694e-08, 2.5233907e-08],
       [3.5257504e-04, 8.8972389e-04, 9.2386857e-05, 4.8687063e-02],
       [2.7085954e-04, 7.7684097e-02, 8.9289460e-05, 1.2786119e-04],
       [5.2730764e-13, 9.9920541e-01, 1.9166842e-11, 8.2062039e-12],
       [1.2135930e-13, 9.9911600e-01, 6.6255135e-12, 1.3244053e-12],
       [2.1020567e-11, 3.2926968e-01, 2.7211988e-10, 5.5204083e-12],
       [2.6995275e-14, 9.9954861e-01, 8.7633069e-13, 2.9740466e-12],
       [2.0245436e-13, 9.5068341e-01, 2.3346332e-11, 1.7555527e-12],
       [9.6436650e-13, 3.0095114e-05, 6.6907599e-09, 2.0117261e-16],
       [1.5413009e-11, 9.9399120e-01, 1.2802791e-11, 1.5615023e-11],
       [1.5962692e-25, 9.9999988e-01, 1.4920251e-24, 4.1572804e-24],
       [8.9582224e-11, 9.9145550e-01, 2.9633193e-10, 3.9553510e-10],
       [8.1094669e-09, 1.9613329e-01, 1.3441264e-06, 1.8306956e-08],
       [1.6764140e-11, 9.9539572e-01, 1.2220336e-10, 1.5574391e-11],
       [1.9227930e-06, 8.0996698e-01, 4.0275991e-05, 7.4675190e-06],
       [1.1718333e-05, 2.7493445e-02, 3.0977117e-05, 1.2554987e-06],
       [1.0690468e-10, 9.9689734e-01, 3.5567584e-09, 3.6802743e-09]],
      dtype=float32)

2、LSTM 神经网络

# 设置所希望的特征数
feature_num = X_train_count.shape[1]

# 使用单热编码目标向量对标签进行处理

y_train_cate = to_categorical(y_train_le)
y_test_cate = to_categorical(y_test_le)

print(y_train_cate)


[[0. 1. 0. 0.] [0. 1. 0. 0.] [0. 1. 0. 0.] ... [0. 0. 1. 0.] [0. 0. 1. 0.] [0. 0. 1. 0.]]

# 1 创建神经网络
lstm_network = models.Sequential()

# ----------------------------------------------
# 2 添加神经层
lstm_network.add(layers.Embedding(input_dim=feature_num,  # 添加嵌入层
                                  output_dim=4))

lstm_network.add(layers.LSTM(units=128))                 # 添加 128 个单元的 LSTM 神经层

lstm_network.add(layers.Dense(units=4,
                              activation='sigmoid'))     # 添加 sigmoid 分类激活函数的全连接层

# ----------------------------------------------
# 3 编译神经网络
lstm_network.compile(loss='binary_crossentropy',
                     optimizer='Adam',
                     metrics=['accuracy']
                     )

# ----------------------------------------------
# 4 开始训练模型
lstm_network.fit(X_train_count,
                 y_train_cate,
                 epochs=5,
                 batch_size=128,
                 validation_data=(X_test_count, y_test_cate)
                 )

十、算法之间性能比较

df = pd.DataFrame()
df['分类器'] = estimator_list
df['准确率'] = score_list
df['消耗时间/s'] = time_list
df

                  分类器    准确率  消耗时间/s
0     KNeighborsClassifier  0.665   28.50
1   DecisionTreeClassifier  0.735   21.13
2            MLPClassifier  0.860  327.06
3              BernoulliNB  0.785    1.15
4               GaussianNB  0.890    1.97
5            MultinomialNB  0.905    0.54
6       LogisticRegression  0.890    0.74
7                      SVC  0.575  389.03
8   RandomForestClassifier  0.820    3.95
9       AdaBoostClassifier  0.570   96.50
10          LGBMClassifier  0.785    2.01
11           XGBClassifier  0.735  984.70
12                 前馈网络  0.905   31.13

‍

综上 DataFrame 展示，结合消耗时间和准确率来看，可以得出以下结论：

在同一训练集和测试集、分类器默认参数设置（都未进行调参）的情况下：

 

综合效果最好的是:

MultinomialNB 多项式朴素贝叶斯分类算法：

其准确率达到了 90.5% 并且所消耗的的时间才 0.55 s

综合效果最差的是：

SVC 支持向量机

其准确率才 0.575 并且消耗时间高达 380.72s

准确率最低的是：0.570

 

AdaBoostClassifier 自适应增强集成学习算法

消耗时间最高的是：566.59s

 

XGBClassifier 集成学习算法

 

About Me：小婷儿

● 本文作者：小婷儿，专注于python、数据分析、数据挖掘、机器学习相关技术，也注重技术的运用

● 作者博客地址：https://blog.csdn.net/u010986753

● 本系列题目来源于作者的学习笔记，部分整理自网络，若有侵权或不当之处还请谅解

● 版权所有，欢迎分享本文，转载请保留出处

● 微信：tinghai87605025 联系我加微信群

● QQ：87605025

● QQ交流群py_data ：483766429

● 公众号：python宝 或 DB宝

● 提供OCP、OCM和高可用最实用的技能培训

● 题目解答若有不当之处，还望各位朋友批评指正，共同进步

如果您觉得到文章对您有帮助，欢迎赞赏哦！有您的支持，小婷儿一定会越来越好！