【机器学习案列】使用随机森林（RF）进行白葡萄酒质量预测

🧑 博主简介：曾任某智慧城市类企业算法总监，目前在美国市场的物流公司从事高级算法工程师一职，深耕人工智能领域，精通python数据挖掘、可视化、机器学习等，发表过AI相关的专利并多次在AI类比赛中获奖。CSDN人工智能领域的优质创作者，提供AI相关的技术咨询、项目开发和个性化解决方案等服务，如有需要请站内私信或者联系任意文章底部的的VX名片（ID：xf982831907）
💬 博主粉丝群介绍：① 群内初中生、高中生、本科生、研究生、博士生遍布，可互相学习，交流困惑。② 热榜top10的常客也在群里，也有数不清的万粉大佬，可以交流写作技巧，上榜经验，涨粉秘籍。③ 群内也有职场精英，大厂大佬，可交流技术、面试、找工作的经验。④ 进群免费赠送写作秘籍一份，助你由写作小白晋升为创作大佬。⑤ 进群赠送CSDN评论防封脚本，送真活跃粉丝，助你提升文章热度。有兴趣的加文末联系方式，备注自己的CSDN昵称，拉你进群，互相学习共同进步。

【机器学习案列】使用随机森林（RF）进行白葡萄酒质量预测

- 引言
- 数据集介绍
- 环境准备
- 数据预处理
- - [1. 导入相应的分析库和数据加载](#1. 导入相应的分析库和数据加载)
  - [2. 数据探索](#2. 数据探索)
  - [3. 异常值处理](#3. 异常值处理)
  - - [3.1 缺失值处理](#3.1 缺失值处理)
    - [3.2 异常值处理](#3.2 异常值处理)
- 模型训练
- - [1. 数据标签的0/1化](#1. 数据标签的0/1化)
  - [2. 数据的归一化](#2. 数据的归一化)
  - [3. 划分数据集](#3. 划分数据集)
  - [4. 训练模型](#4. 训练模型)
  - [5. 模型评估](#5. 模型评估)
  - [6. 特征重要性](#6. 特征重要性)
- 结论

引言

在葡萄酒产业中，质量评估是一个复杂的过程，涉及到多个化学和感官因素。随着机器学习技术的发展，我们可以使用这些技术来预测葡萄酒的质量。在这篇文章中，我们将使用Python中的RandomForestClassifier来预测白葡萄酒的质量。我们将通过分析数据集中的多个化学成分来训练一个模型，该模型能够预测葡萄酒的质量评分。

数据集介绍

我们的数据集包含了多个与葡萄酒质量相关的化学成分，这些成分包括：

非挥发性酸（fixed acidity）
挥发性酸（volatile acidity）
柠檬酸（citric acid）
残糖（residual sugar）
氯化物（chlorides）
游离二氧化硫（free sulfur dioxide）
总二氧化硫（total sulfur dioxide）
密度（density）
酸碱度（pH）
硫酸盐（sulphates）
酒精（alcohol）
葡萄酒质量（quality，0-10）

环境准备

在开始之前，确保你的Python环境中安装了以下库：

pandas
numpy
scikit-learn
matplotlib
seaborn

你可以通过以下命令安装这些库：

bash 复制代码

pip install pandas numpy scikit-learn matplotlib seaborn

数据预处理

1. 导入相应的分析库和数据加载

首先，我们需要加载数据集。

python 复制代码

import warnings # For warning handling

# Third-party imports
import pandas as pd # For data processing, CSV file I/O
import numpy as np # For numerical operations and mathematical functions
import matplotlib.pyplot as plt # For data visualization
import seaborn as sns # For statistical graphics
import plotly.express as px # For interactive plotting
from sklearn.model_selection import train_test_split # For data splitting for machine learning
from sklearn.preprocessing import MinMaxScaler, StandardScaler # For feature standardization
from sklearn.metrics import accuracy_score # For model evaluation
from termcolor import colored # For colored text printing
from sklearn.ensemble import RandomForestClassifier # For random forest classifier model

# For warning handling
warnings.filterwarnings('ignore') # For ignoring warnings

加载相应的数据集。

python 复制代码

# load data
try:
    # Relative file path
    filePath = "winequality-white.csv"
    
    # Read the CSV file and save it in "data" variable
    data= pd.read_csv(filePath,sep=';')
    
    # Check loading data
    print(colored("THE DATASET LOADED SUCCESSFULLY...", "green", attrs=['reverse']))

except FileNotFoundError:
    print(colored("ERROR: File not found!", "red", attrs=['reverse']))

except Exception as e:
    print(colored(f"ERROR: {e}", "red", attrs=['reverse']))

2. 数据探索

在进行任何预处理之前，我们应该对数据有一个基本的了解，首先查看数据前几行。

python 复制代码

# 查看数据集的前几行
dataset_rows = data.head(7) #.head() the default value = 5

print(colored('As you can see, the first 7 rows in the dataset:\n', 'green', attrs=['reverse']))

# Iterate over each row in the dataset_rows DataFrame
for index, row in dataset_rows.iterrows():
    # Print the index label of the current row
    print(colored(f"Row {index + 1}:","white",attrs=['reverse']))
    
    # Print the content of the current row
    print(row)
    
    # Print a separator line
    print("--------------------------------------")

查看数据的基本情况，包括shape、特征、总数等等。

python 复制代码

print("The shape =",data.shape)

# Show information about the dataset
num_rows, num_cols = data.shape
num_features = num_cols - 1
num_data = num_rows * num_cols

# Print the information
print(f"Number of Rows: {num_rows}")
print(f"Number of Columns: {num_cols}")
print(f"Number of Features: {num_features}")
print(f"Number of All Data: {num_data}")

# Check and ensure running
print(colored("The task has been completed without any errors....","green", attrs=['reverse']))

python 复制代码

# 查看数据集的信息
print(data.info())

查看数据统计特征。

python 复制代码

data.describe().T.round(2)

查看数据标签的分布情况。

python 复制代码

# Create a count plot using seaborn
sns.catplot(data=data, x='quality', kind='count')

# Add labels and title to the plot
plt.title('Distribution of Wine Quality')
plt.xlabel('Quality')
plt.ylabel('Count')

# Display the plot
plt.show()

这里的数据EDA部分不在做详细的介绍，具体参考【数据可视化案列】白葡萄酒质量数据的EDA可视化分析一文即可。

3. 异常值处理

3.1 缺失值处理

python 复制代码

# Check for missing values
null_counts = data.isnull().sum() 

# Display the number of null values
print(null_counts)

print("_________________________________________________________________")
print(colored(f"Totally, there are {null_counts.sum()} null values in the dataset.","green", attrs=['reverse']))

发现数据集中无缺失值的存在（万恶的资本主义数据集中都没有空值）。

3.2 异常值处理

python 复制代码

# Set the figure size
plt.figure(figsize=(22, 11))

# Add outliers to the plot
sns.stripplot(data=data, color="red", jitter=0.2, size=5)

# Set the axis labels and title
plt.title("Outliers")
plt.xlabel("X-axis label")
plt.ylabel("Y-axis label")

# Show the plot
plt.show()

python 复制代码

# Delete the outliers
# The data before deleting outliers 
print("Before Removing the outliers", data.shape)

# Deleting outliers (Removing the number of observation where the total sulfur dioxide is more than 160)
data = data[data['total sulfur dioxide']<160]

#The data after deleting outliers
print("After Removing the outliers", data.shape)

python 复制代码

# Set the figure size
plt.figure(figsize=(22, 11))

# Add outliers to the plot
sns.stripplot(data=data, color="red", jitter=0.2, size=5)

# Set the axis labels and title
plt.title("Outliers")
plt.xlabel("X-axis label")
plt.ylabel("Y-axis label")

# Show the plot
plt.show()

模型训练

1. 数据标签的0/1化

python 复制代码

# Split the data into features (X) and target variable (Y)
X = data.drop('quality',axis=1)

# Create a new series 'Y' by applying a lambda function to the 'quality' column of the 'data' DataFrame
# The lambda function assigns a value of 1 if the 'quality' value is greater than or equal to 5, otherwise assigns 0
Y = data['quality'].apply(lambda y_value: 1 if y_value >= 5 else 0)

# Print the shapes of X and Y to verify the splitting
print("Shape of X:", X.shape)
print("Shape of Y:", Y.shape)

2. 数据的归一化

python 复制代码

# Rescale and normalize the features
'''
# Standardization (Normalization)
standard_scaler = StandardScaler()
X = standard_scaler.fit_transform(X)
'''

# Min-Max Scaling (Rescaling)
min_max_scaler = MinMaxScaler()
X = min_max_scaler.fit_transform(X)

#I will choose one of them in the future part "model selection" based on the highest accuracy

3. 划分数据集

我们将数据集划分为训练集和测试集。

python 复制代码

X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2, random_state=44)

# Print the shapes of the training and testing sets to verify the splitting
print("Shape of X_train:", X_train.shape)
print("Shape of X_test:", X_test.shape)
print("Shape of Y_train:", Y_train.shape)
print("Shape of Y_test:", Y_test.shape)

4. 训练模型

使用RandomForestClassifier训练模型。

python 复制代码

# Initialize lists to store training and testing accuracies
scoreListRF_Train = []
scoreListRF_Test = []

'''
max_dep      ----------> (1, 5),(1, 10) 
rand_state   ----------> (1, 35),(1, 50)
n_est        ----------> (1, 30),(1, 30)
'''

# Iterate over different values of max_depth
for max_dep in range(1, 5):
    # Iterate over different values of random_state
    for rand_state in range(1, 20):
        # Iterate over different values of n_estimators
        for n_est in range(1, 15):
            # Create a Random Forest model with the different values of max_depth, random_state, and n_estimators
            Model = RandomForestClassifier(n_estimators=n_est, random_state=rand_state, max_depth=max_dep)            
            
            # Fit the model on the training data
            Model.fit(X_train, Y_train)
            
            # Calculate and store the training accuracy
            scoreListRF_Train.append(Model.score(X_train, Y_train))
            
            # Calculate and store the testing accuracy
            scoreListRF_Test.append(Model.score(X_test, Y_test))

# Find the maximum accuracy for both training and testing
RF_Accuracy_Train = max(scoreListRF_Train) 
RF_Accuracy_Test = max(scoreListRF_Test)

# Print the best accuracies achieved
print(f"Random Forest best accuracy (Training): {RF_Accuracy_Train*100:.2f}%")
print(f"Random Forest best accuracy (Testing): {RF_Accuracy_Test*100:.2f}%")

# Print a success message indicating that the model has been trained successfully
print(colored("The Random Forest model has been trained successfully","green", attrs=['reverse']))

5. 模型评估

评估模型的性能。

python 复制代码

from sklearn.metrics import accuracy_score, classification_report, confusion_matrix

# 预测测试集
y_pred = Model.predict(X_test)

# 计算准确率
accuracy = accuracy_score(Y_test, y_pred)
print(f'Accuracy: {accuracy:.2f}')

# 打印分类报告
print(classification_report(Y_test, y_pred))

# 打印混淆矩阵
print(confusion_matrix(Y_test, y_pred))

6. 特征重要性

RandomForestClassifier提供了一个方便的特性，即可以查看每个特征的重要性。

python 复制代码

import matplotlib.pyplot as plt
import seaborn as sns

# 获取特征重要性
feature_importances = Model.feature_importances_

# 创建一个DataFrame来存储特征和它们的重要性
feature_importance_df = pd.DataFrame({
    'Feature': data.columns.tolist()[:-1],
    'Importance': feature_importances
})

# 对特征重要性进行排序
feature_importance_df = feature_importance_df.sort_values(by='Importance', ascending=False)

# 绘制特征重要性图
plt.figure(figsize=(10, 8))
sns.barplot(x='Importance', y='Feature', data=feature_importance_df)
plt.title('Feature Importance')
plt.show()

结论

通过使用RandomForestClassifier，我们能够预测白葡萄酒的质量。在这个过程中，我们进行了数据预处理、特征选择、模型训练和评估，并分析了特征的重要性。这只是一个简单的示例，实际应用中可能需要更复杂的数据预处理和模型调优。