脱发因素影响MLP预测

1、数据分析

Id

Genetics

Hormonal Changes

Medical Conditions

Medications & Treatments

Nutritional Deficiencies

Stress

Age

Poor Hair Care Habits

Environmental Factors

Smoking

Weight Loss

Hair Loss

数据处理

1、缺省值舍去

# 数据清洗
import pandas as pd
import numpy as np

def clean_data():
    HA = pd.read_csv("./Predict Hair Fall.csv")

    for i in range(len(HA)):
        for j in range(len(HA.columns)):
            if HA.iloc[i, j] == 'No Data':
                HA.iloc[i, j] = np.nan

    if 'Unnamed: 0' in HA.columns:
        HA.drop(columns=['Unnamed: 0'], inplace=True)


    clean_URL = './Cleaned Predict Hair Fall.csv'
    HA.to_csv(clean_URL, index=False)

    HA.info()
    print(HA)

    return clean_URL

cleaned_file = clean_data()
HB = pd.read_csv(cleaned_file)
HB.info()

删除多余的id，在训练中不起作用，并进行映射

# 定义一个字典来映射 'yes' 到 1 和 'No' 到 0
mapping = {'Yes': 1, 'No': 0}
# 使用 map 函数来应用映射到 DataFrame 的某一列（假设列名为 'Column_Name'）
HB['Genetics'] = HB['Genetics'].map(mapping)
HB['Weight Loss '] = HB['Weight Loss '].map(mapping)
HB['Hormonal Changes'] = HB['Hormonal Changes'].map(mapping)
HB['Environmental Factors'] = HB['Environmental Factors'].map(mapping)
HB['Poor Hair Care Habits '] = HB['Poor Hair Care Habits '].map(mapping)
HB['Smoking'] = HB['Smoking'].map(mapping)
HB.head()

将hair——loss移到最后一列做y

# # 最后，将该列插回到最后一列的位置
# df = pd.concat([df, column_to_move], axis=1)
# HB.head()
import pandas as pd
# 假设 df 是你的 DataFrame，'Column_to_move' 是你想要移动的列名

# 获取除了 'Column_to_move' 以外的所有列名
columns = HB.columns.drop('Hair Loss')

# 创建一个新的列名列表，将 'Column_to_move' 放在最后
new_columns = list(columns) + ['Hair Loss']
# 使用 reindex 方法按照新的列名列表重新排序
HB =HB.reindex(columns=new_columns)
HB.head(5)

计算相关性热土

划分数据集

from sklearn.model_selection import train_test_split
# 划分数据集
X = HB.iloc[:, :-1]
y = HB.iloc[:, -1]

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
print(X_train.shape)
print(X_test.shape)
print(y_train.shape)
print(y_test.shape)
print(HB.columns[-1])

构建模型

# 构建2分类器模型
from sklearn.neural_network import MLPClassifier
mlp = MLPClassifier(hidden_layer_sizes=(50, 75, 100), activation='tanh', solver='lbfgs', learning_rate='invscaling',max_iter=500)
mlp.fit(X_train, y_train)

进行随即搜索优化

   网格搜索法寻找最优的超参数


from sklearn.model_selection import GridSearchCV
from sklearn.neural_network import MLPClassifier

# 定义参数网格
params = {
    'activation': ['relu', 'tanh', 'logistic', 'identity'],
    'hidden_layer_sizes': [(100,), (50, 100,), (50, 75, 100,)],
    'solver': ['adam', 'sgd', 'lbfgs'],
    'learning_rate': ['constant', 'adaptive', 'invscaling']
}

# 初始化GridSearchCV，注意这里传入的是MLPClassifier类，而不是实例
mlp_classif_grid = GridSearchCV(MLPClassifier(), param_grid=params, n_jobs=-1, cv=5, verbose=5)

# 拟合模型
mlp_classif_grid.fit(X_train, y_train)

# 输出结果
print('Train Accuracy : %.3f' % mlp_classif_grid.best_estimator_.score(X_train, y_train))
print('Test Accuracy : %.3f' % mlp_classif_grid.best_estimator_.score(X_test, y_test))
print('Best Accuracy Through Grid Search : %.3f' % mlp_classif_grid.best_score_)
print('Best Parameters : ', mlp_classif_grid.best_params_)

Fitting 5 folds for each of 108 candidates, totalling 540 fits
Train Accuracy : 1.000
Test Accuracy : 0.420
Best Accuracy Through Grid Search : 0.560
Best Parameters :  {'activation': 'relu', 'hidden_layer_sizes': (100,), 'learning_rate': 'adaptive', 'solver': 'lbfgs'}

考虑对模型中数据删除

1.年龄：未对年龄编码，年龄分布差异较大

HC = HB.drop(['Age'], axis=1)
HC.head(5)

同上构建模型优化

性能提升

Fitting 5 folds for each of 10 candidates, totalling 50 fits
Train Accuracy : 1.000
Test Accuracy : 0.500
Best Accuracy Through Randomized Search : 0.556
Best Parameters :  {'solver': 'lbfgs', 'learning_rate': 'invscaling', 'hidden_layer_sizes': (50, 150), 'activation': 'tanh'}