机器学习监督学习实战五：六种算法对声呐回波信号进行分类

  本项目基于UCI的声呐目标识别数据集（Sonar, Mines vs. Rocks），通过10种机器学习算法比较，发现集成学习方法表现最优。研究首先对60个声呐能量特征进行可视化分析（分布直方图、相关性矩阵），对比了基础算法（如KNN、SVM）和集成算法（如随机森林、梯度提升）的10折交叉验证准确率。结果显示标准化后模型性能提升，其中额外树（ExtraTrees）表现最佳（准确率88.6%），并通过网格搜索优化超参数。最后利用SHAP值解释模型，揭示关键特征（如Feature 10）对预测的贡献度，为声呐目标分类提供了可解释的解决方案。完整代码已开源在个人GitHubhttps://github.com/KLWU07/Machine-learning-Project-practice。

一、数据集

  onar.all-data 是 UCI 机器学习仓库中的经典二分类数据集，用于声呐信号处理与目标识别任务，区分声呐回波来自岩石（Rock）还是金属圆柱体（如地雷，Mine）。 原始数据通过主动声呐发射宽带声波，接收目标反射信号并进行数字化处理后得到。共 208 个样本，其中：岩石（R）97 个样本金属圆柱体（M）111 个样本；两类样本数量接近平衡，适合二分类算法测试。
  数据集的 60 个特征均为连续型数值范围在 0.0 到 1.0 之间。每个数字代表特定频段内在一定时间内积分的能量。，本质上是声呐信号在不同处理阶段的能量响应值，反映目标对声波的反射特性。每个特征可能对应声呐信号在某个特定频率点的振幅值（通过傅里叶变换得到的频谱分量），部分特征可能对应信号在不同时间窗口的能量分布（短时傅里叶变换的结果），更多数据集信息来自于https://archive.ics.uci.edu/dataset/151，可在里面下载。

二、可视化

1.60个特征的分布情况（直方图和密度图）

2.特征之间的相关性（矩阵图）

三、算法和评估

逻辑回归、线性判别分析、K近邻、决策树、朴素贝叶斯、支持向量机

1.10折交叉验证准确率（箱线图）

2.数据标准化和10折交叉验证准确率（箱线图）

数据标准化后准确率有提升。

四，改进算法和集合算法

• ** 网格搜索GridSearchCV寻找超参数，从上面挑出最好的：K近邻KNN和支持向量机SVR**
• 集成学习方法：AdaBoost、梯度提升、随机森林、额外树

1.数据标准化、10折交叉验证和网格搜索准确率（结果）

KNN最优：0.8360294117647058 使用{'n_neighbors': 1}  

SVR最优：0.85 使用{'C': 1.7, 'kernel': 'rbf'}

2.集成算法标准化和10折交叉验证（箱线图）

3.梯度提升、数据标准化、10折交叉验证和网格搜索

额外树ETR-ExtraTreesClassifier()

最优：0.8860294117647058 使用{'n_estimators': 400}

4.最终模型（分类报告）

  额外树ETR-ExtraTreesClassifier()，使用10折交叉验证。混淆矩阵、ROC曲线

=== 交叉验证评估 ===
Accuracy: 0.874 (±0.080)
Precision: 0.846 (±0.140)
Recall: 0.910 (±0.101)
F1-score: 0.873 (±0.110)
ROC-AUC: 0.971 (±0.031)
=== 分类报告 ===
              precision    recall  f1-score   support

        Rock       0.90      0.82      0.86        97
        Mine       0.86      0.92      0.89       111

    accuracy                           0.88       208
   macro avg       0.88      0.87      0.87       208
weighted avg       0.88      0.88      0.87       208

5. 解释模型计算特征重要性(SHAP 值的摘要图)

展示Mine类特征贡献度可视化（‘M’: 1）。

• 解释展现关系
  • 单个特征对模型预测的影响方向（正相关/负相关）和影响程度（SHAP值绝对值大小）。
  • 特征取值与SHAP值的分布关系（。
  • 横轴为SHAP值，纵轴为特征名称，点的颜色表示特征取值（如颜色越深代表特征值越大）。
  • 比如：特征x10的SHAP值多为正值，且随特征值增大而增大，表明x10与预测结果呈正相关；特征x2的SHAP值多为负值，表明其与预测结果呈负相关。很多都有正负相关。
    

6.特征重要性排序

各特征的平均SHAP值绝对值排名，直观展示特征重要性排序。横轴为平均SHAP值绝对值，纵轴为特征名称，特征按重要性降序排列。特征Feature 10的平均SHAP值最大，对模型预测的贡献度最高。

7.单个样本的SHAP决策图（第一个样本）

五、完整代码

# 导入类库
import numpy as np
import pandas as pd
from matplotlib import pyplot
from pandas.plotting import scatter_matrix
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.model_selection import KFold
from sklearn.model_selection import cross_val_score
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import classification_report
from sklearn.metrics import confusion_matrix
from sklearn.metrics import accuracy_score
from sklearn.pipeline import Pipeline
from sklearn.linear_model import LogisticRegression
from sklearn.neighbors import KNeighborsClassifier
from sklearn.tree import DecisionTreeClassifier
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis
from sklearn.naive_bayes import GaussianNB
from sklearn.svm import SVC
from sklearn.ensemble import AdaBoostClassifier
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.ensemble import RandomForestClassifier
from sklearn.ensemble import ExtraTreesClassifier
import shap

# 导入数据
filename = 'sonar.all-data.csv'
dataset = pd.read_csv(filename, header=None)

# 将目标变量转换为数值类型
dataset[60] = dataset[60].map({'R': 0, 'M': 1})

# 数据维度
print(dataset.shape)

# 查看数据类型
print(dataset.dtypes)

# 查看最初的20条记录
pd.set_option('display.width', 100)
print(dataset.head(20))

# 描述性统计信息
pd.set_option('display.precision', 3)
print(dataset.describe())

# 数据的分类分布
print(dataset.groupby(60).size())

# 直方图
dataset.hist(sharex=False, sharey=False, xlabelsize=1, ylabelsize=1)
pyplot.show()

# 密度图
dataset.plot(kind='density', subplots=True, layout=(8, 8), sharex=False, legend=False, fontsize=1)
pyplot.show()

# 关系矩阵图
fig = pyplot.figure()
ax = fig.add_subplot(111)
cax = ax.matshow(dataset.corr(), vmin=-1, vmax=1, interpolation='none')
fig.colorbar(cax)
pyplot.show()

# 分离评估数据集
array = dataset.values
X = array[:, 0:60].astype(float)
Y = array[:, 60]
validation_size = 0.2
seed = 7
X_train, X_validation, Y_train, Y_validation = train_test_split(X, Y, test_size=validation_size, random_state=seed)

# 评估算法的基准
num_folds = 10
seed = 7
scoring = 'accuracy'

# 评估算法 - 原始数据
models = {}
models['LR'] = LogisticRegression()
models['LDA'] = LinearDiscriminantAnalysis()
models['KNN'] = KNeighborsClassifier()
models['CART'] = DecisionTreeClassifier()
models['NB'] = GaussianNB()
models['SVM'] = SVC()
results = []
for key in models:
    kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
    cv_results = cross_val_score(models[key], X_train, Y_train, cv=kfold, scoring=scoring)
    results.append(cv_results)
    print('%s : %f (%f)' % (key, cv_results.mean(), cv_results.std()))

# 评估算法 - 箱线图
fig = pyplot.figure()
fig.suptitle('Algorithm Comparison')
ax = fig.add_subplot(111)
pyplot.boxplot(results)
ax.set_xticklabels(models.keys())
pyplot.show()

# 评估算法 - 正态化数据
pipelines = {}
pipelines['ScalerLR'] = Pipeline([('Scaler', StandardScaler()), ('LR', LogisticRegression())])
pipelines['ScalerLDA'] = Pipeline([('Scaler', StandardScaler()), ('LDA', LinearDiscriminantAnalysis())])
pipelines['ScalerKNN'] = Pipeline([('Scaler', StandardScaler()), ('KNN', KNeighborsClassifier())])
pipelines['ScalerCART'] = Pipeline([('Scaler', StandardScaler()), ('CART', DecisionTreeClassifier())])
pipelines['ScalerNB'] = Pipeline([('Scaler', StandardScaler()), ('NB', GaussianNB())])
pipelines['ScalerSVM'] = Pipeline([('Scaler', StandardScaler()), ('SVM', SVC())])
results = []
for key in pipelines:
    kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
    cv_results = cross_val_score(pipelines[key], X_train, Y_train, cv=kfold, scoring=scoring)
    results.append(cv_results)
    print('%s : %f (%f)' % (key, cv_results.mean(), cv_results.std()))

# 评估算法 - 箱线图
fig = pyplot.figure()
fig.suptitle('Scaled Algorithm Comparison')
ax = fig.add_subplot(111)
pyplot.boxplot(results)
ax.set_xticklabels(models.keys())
pyplot.show()

# 调参改进算法 - KNN
scaler = StandardScaler().fit(X_train)
rescaledX = scaler.transform(X_train)
param_grid = {'n_neighbors': [1, 3, 5, 7, 9, 11, 13, 15, 17, 19, 21]}
model = KNeighborsClassifier()
kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
grid = GridSearchCV(estimator=model, param_grid=param_grid, scoring=scoring, cv=kfold)
grid_result = grid.fit(X=rescaledX, y=Y_train)

print('最优：%s 使用%s' % (grid_result.best_score_, grid_result.best_params_))
cv_results = zip(grid_result.cv_results_['mean_test_score'],
                 grid_result.cv_results_['std_test_score'],
                 grid_result.cv_results_['params'])
for mean, std, param in cv_results:
    print('%f (%f) with %r' % (mean, std, param))

# 调参改进算法 - SVM
scaler = StandardScaler().fit(X_train)
rescaledX = scaler.transform(X_train).astype(float)
param_grid = {}
param_grid['C'] = [0.1, 0.3, 0.5, 0.7, 0.9, 1.0, 1.3, 1.5, 1.7, 2.0]
param_grid['kernel'] = ['linear', 'poly', 'rbf', 'sigmoid']
model = SVC()
kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
grid = GridSearchCV(estimator=model, param_grid=param_grid, scoring=scoring, cv=kfold)
grid_result = grid.fit(X=rescaledX, y=Y_train)

print('最优：%s 使用%s' % (grid_result.best_score_, grid_result.best_params_))
cv_results = zip(grid_result.cv_results_['mean_test_score'],
                 grid_result.cv_results_['std_test_score'],
                 grid_result.cv_results_['params'])
for mean, std, param in cv_results:
    print('%f (%f) with %r' % (mean, std, param))

# 集成算法
ensembles = {}
ensembles['ScaledAB'] = Pipeline([('Scaler', StandardScaler()), ('AB', AdaBoostClassifier())])
ensembles['ScaledGBM'] = Pipeline([('Scaler', StandardScaler()), ('GBM', GradientBoostingClassifier())])
ensembles['ScaledRF'] = Pipeline([('Scaler', StandardScaler()), ('RFR', RandomForestClassifier())])
ensembles['ScaledET'] = Pipeline([('Scaler', StandardScaler()), ('ETR', ExtraTreesClassifier())])

results = []
for key in ensembles:
    kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
    cv_result = cross_val_score(ensembles[key], X_train, Y_train, cv=kfold, scoring=scoring)
    results.append(cv_result)
    print('%s: %f (%f)' % (key, cv_result.mean(), cv_result.std()))

# 集成算法 - 箱线图
fig = pyplot.figure()
fig.suptitle('Algorithm Comparison')
ax = fig.add_subplot(111)
pyplot.boxplot(results)
ax.set_xticklabels(ensembles.keys())
pyplot.show()

# 集成算法ETR - 调参
scaler = StandardScaler().fit(X_train)
rescaledX = scaler.transform(X_train)
param_grid = {'n_estimators': [10, 50, 100, 200, 300, 400, 500, 600, 700, 800, 900]}
model = ExtraTreesClassifier()
kfold = KFold(n_splits=num_folds, shuffle=True, random_state=seed)
grid = GridSearchCV(estimator=model, param_grid=param_grid, scoring=scoring, cv=kfold)
grid_result = grid.fit(X=rescaledX, y=Y_train)
print('最优：%s 使用%s' % (grid_result.best_score_, grid_result.best_params_))

# 模型最终化
scaler = StandardScaler().fit(X_train)
rescaledX = scaler.transform(X_train)
model = ExtraTreesClassifier()
model.fit(X=rescaledX, y=Y_train)
# 评估模型
rescaled_validationX = scaler.transform(X_validation)
predictions = model.predict(rescaled_validationX)
print(accuracy_score(Y_validation, predictions))
print(confusion_matrix(Y_validation, predictions))
print(classification_report(Y_validation, predictions))