完成返回模型详情列表
This commit is contained in:
parent
228581bc68
commit
ec5174ca76
@ -1,4 +1,4 @@
|
||||
feature_engineering_methods_parameters:
|
||||
feature_engineering_methods:
|
||||
|
||||
LabelEncoder:
|
||||
description: "将分类标签编码为整数。"
|
||||
|
||||
@ -1,4 +1,4 @@
|
||||
from data_process.method_reader_date_feature import MethodReader
|
||||
from function.method_reader_date_feature import MethodReader
|
||||
|
||||
# 创建方法读取器实例
|
||||
reader = MethodReader()
|
||||
|
||||
@ -1,4 +1,4 @@
|
||||
from data_process.method_reader_date_process import MethodReader
|
||||
from function.method_reader_date_process import MethodReader
|
||||
|
||||
# 创建方法读取器实例
|
||||
reader = MethodReader()
|
||||
|
||||
14
example_method_reader_model.py
Normal file
14
example_method_reader_model.py
Normal file
@ -0,0 +1,14 @@
|
||||
from function.method_reader_model import MethodReader
|
||||
|
||||
# 创建方法读取器实例
|
||||
reader = MethodReader()
|
||||
|
||||
# 获取所有预处理方法
|
||||
methods = reader.get_preprocessing_methods()
|
||||
print("预处理方法列表:")
|
||||
print(methods)
|
||||
|
||||
# 获取特定方法的详细信息
|
||||
method_details = reader.get_method_details('SVC')
|
||||
print("\nSVC方法详情:")
|
||||
print(method_details)
|
||||
BIN
function/__pycache__/method_reader_date_feature.cpython-39.pyc
Normal file
BIN
function/__pycache__/method_reader_date_feature.cpython-39.pyc
Normal file
Binary file not shown.
BIN
function/__pycache__/method_reader_date_process.cpython-39.pyc
Normal file
BIN
function/__pycache__/method_reader_date_process.cpython-39.pyc
Normal file
Binary file not shown.
BIN
function/__pycache__/method_reader_model.cpython-39.pyc
Normal file
BIN
function/__pycache__/method_reader_model.cpython-39.pyc
Normal file
Binary file not shown.
@ -53,12 +53,12 @@ class MethodReader:
|
||||
|
||||
|
||||
# 获取特征工程方法
|
||||
outlier_methods = list(self.method_config.get('feature_engineering_methods', {}).keys())
|
||||
if outlier_methods:
|
||||
feature_engineering_methods = list(self.method_config.get('feature_engineering_methods', {}).keys())
|
||||
if feature_engineering_methods:
|
||||
methods.append({
|
||||
"name": "outlier_detector",
|
||||
"description": "异常值检测",
|
||||
"method": outlier_methods
|
||||
"name": "feature_engineering_methods",
|
||||
"description": "特征工程方法",
|
||||
"method": feature_engineering_methods
|
||||
})
|
||||
|
||||
return {
|
||||
@ -88,7 +88,7 @@ class MethodReader:
|
||||
|
||||
# 查找方法参数信息
|
||||
parameter_info = None
|
||||
for category in ['feature_engineering_methods_parameters']:
|
||||
for category in ['feature_engineering_methods']:
|
||||
if method_name in self.parameter_config.get(category, {}):
|
||||
parameter_info = self.parameter_config[category][method_name]
|
||||
break
|
||||
135
function/method_reader_model.py
Normal file
135
function/method_reader_model.py
Normal file
@ -0,0 +1,135 @@
|
||||
import yaml
|
||||
from typing import Dict, List
|
||||
import os
|
||||
import logging
|
||||
from pathlib import Path
|
||||
|
||||
class MethodReader:
|
||||
"""方法配置读取器"""
|
||||
|
||||
def __init__(self):
|
||||
"""初始化方法读取器"""
|
||||
self.logger = logging.getLogger(__name__)
|
||||
self.method_config = self._load_method_config()
|
||||
self.parameter_config = self._load_parameter_config()
|
||||
|
||||
def _load_method_config(self) -> Dict:
|
||||
"""加载方法配置文件"""
|
||||
try:
|
||||
config_path = Path('model/method.yaml')
|
||||
if not config_path.exists():
|
||||
raise FileNotFoundError(f"Method config file not found at {config_path}")
|
||||
|
||||
with open(config_path, 'r', encoding='utf-8') as f:
|
||||
config = yaml.safe_load(f)
|
||||
|
||||
self.logger.info("Successfully loaded method config")
|
||||
return config
|
||||
|
||||
except Exception as e:
|
||||
self.logger.error(f"Error loading method config: {str(e)}")
|
||||
raise
|
||||
|
||||
def _load_parameter_config(self) -> Dict:
|
||||
"""加载参数配置文件"""
|
||||
try:
|
||||
config_path = Path('model/parameter.yaml')
|
||||
if not config_path.exists():
|
||||
raise FileNotFoundError(f"Parameter config file not found at {config_path}")
|
||||
|
||||
with open(config_path, 'r', encoding='utf-8') as f:
|
||||
config = yaml.safe_load(f)
|
||||
|
||||
self.logger.info("Successfully loaded parameter config")
|
||||
return config
|
||||
except Exception as e:
|
||||
self.logger.error(f"Error loading parameter config: {str(e)}")
|
||||
raise
|
||||
|
||||
def get_preprocessing_methods(self) -> Dict:
|
||||
"""获取预处理方法列表"""
|
||||
try:
|
||||
methods = []
|
||||
|
||||
# 分类方法
|
||||
classification_algorithms = list(self.method_config.get('classification_algorithms', {}).keys())
|
||||
if classification_algorithms:
|
||||
methods.append({
|
||||
"name": "classification_algorithms",
|
||||
"description": "分类方法",
|
||||
"method": classification_algorithms
|
||||
})
|
||||
|
||||
# 回归方法
|
||||
regression_algorithms = list(self.method_config.get('regression_algorithms', {}).keys())
|
||||
if regression_algorithms:
|
||||
methods.append({
|
||||
"name": "regression_algorithms",
|
||||
"description": "回归方法",
|
||||
"method": regression_algorithms
|
||||
})
|
||||
|
||||
# 聚类方法
|
||||
clustering_algorithms = list(self.method_config.get('clustering_algorithms', {}).keys())
|
||||
if clustering_algorithms:
|
||||
methods.append({
|
||||
"name": "clustering_algorithms",
|
||||
"description": "聚类方法",
|
||||
"method": clustering_algorithms
|
||||
})
|
||||
|
||||
return {
|
||||
"status": "success",
|
||||
"methods": methods
|
||||
}
|
||||
|
||||
except Exception as e:
|
||||
self.logger.error(f"Error getting preprocessing methods: {str(e)}")
|
||||
return {
|
||||
"status": "error",
|
||||
"error": str(e)
|
||||
}
|
||||
|
||||
def get_method_details(self, method_name: str) -> Dict:
|
||||
"""获取指定方法的详细信息"""
|
||||
try:
|
||||
# 在各个方法类别中查找方法原理和优缺点
|
||||
method_info = None
|
||||
for category in ['classification_algorithms', 'regression_algorithms', 'clustering_algorithms']:
|
||||
if method_name in self.method_config.get(category, {}):
|
||||
method_info = self.method_config[category][method_name]
|
||||
break
|
||||
|
||||
if method_info is None:
|
||||
raise ValueError(f"Method {method_name} not found in method config")
|
||||
|
||||
# 查找方法参数信息
|
||||
parameter_info = None
|
||||
for category in ['classification_algorithms', 'regression_algorithms', 'clustering_algorithms']:
|
||||
if method_name in self.parameter_config.get(category, {}):
|
||||
parameter_info = self.parameter_config[category][method_name]
|
||||
break
|
||||
|
||||
if parameter_info is None:
|
||||
raise ValueError(f"Method {method_name} not found in parameter config")
|
||||
|
||||
# 组合返回信息
|
||||
return {
|
||||
"status": "success",
|
||||
"method": {
|
||||
"name": method_name,
|
||||
"description": parameter_info.get('description', ''),
|
||||
"principle": method_info.get('principle', ''),
|
||||
"advantages": method_info.get('advantages', []),
|
||||
"disadvantages": method_info.get('disadvantages', []),
|
||||
"applicable_scenarios": method_info.get('applicable_scenarios', []),
|
||||
"parameters": parameter_info.get('parameters', [])
|
||||
}
|
||||
}
|
||||
|
||||
except Exception as e:
|
||||
self.logger.error(f"Error getting method details: {str(e)}")
|
||||
return {
|
||||
"status": "error",
|
||||
"error": str(e)
|
||||
}
|
||||
373
model/method.yaml
Normal file
373
model/method.yaml
Normal file
@ -0,0 +1,373 @@
|
||||
classification_algorithms:
|
||||
LogisticRegression:
|
||||
principle: "逻辑回归是一种线性分类模型，使用 Sigmoid 函数将线性回归的输出映射到 [0,1] 之间，适用于二分类问题。"
|
||||
advantages:
|
||||
- "计算效率高，适用于大规模数据。"
|
||||
- "可解释性强，系数可用于判断特征的重要性。"
|
||||
- "对线性可分数据表现良好。"
|
||||
disadvantages:
|
||||
- "对非线性数据的处理能力较弱。"
|
||||
- "容易受到异常值的影响。"
|
||||
applicable_scenarios:
|
||||
- "医学诊断，如是否患有某种疾病。"
|
||||
- "信用风险评估，如贷款违约预测。"
|
||||
- "市场营销中的客户分类。"
|
||||
|
||||
SVC:
|
||||
principle: "支持向量机（SVM）通过寻找最大化类别间隔的超平面进行分类，支持线性和非线性分类（通过核函数）。"
|
||||
advantages:
|
||||
- "适用于高维数据，尤其是样本较少的情况。"
|
||||
- "通过核函数可以处理非线性分类问题。"
|
||||
disadvantages:
|
||||
- "对大规模数据的计算复杂度较高。"
|
||||
- "对参数和核函数的选择较敏感。"
|
||||
applicable_scenarios:
|
||||
- "文本分类，如垃圾邮件检测。"
|
||||
- "图像识别，如手写数字分类。"
|
||||
|
||||
SVDD:
|
||||
principle: "支持向量数据描述（SVDD）是一种基于支持向量机的单类分类方法，通过寻找最小球面来包围正常数据点，从而检测异常值。"
|
||||
advantages:
|
||||
- "适用于异常检测和单类分类问题。"
|
||||
- "能够有效应对非线性分布数据。"
|
||||
disadvantages:
|
||||
- "对参数选择较敏感，训练时间较长。"
|
||||
applicable_scenarios:
|
||||
- "异常检测，如信用卡欺诈检测。"
|
||||
- "工业设备故障检测。"
|
||||
|
||||
DecisionTreeClassifier:
|
||||
principle: "决策树基于特征的划分规则构建一棵树，通过树结构进行分类。"
|
||||
advantages:
|
||||
- "可解释性强，易于理解和可视化。"
|
||||
- "对数据的分布和尺度不敏感，无需归一化。"
|
||||
disadvantages:
|
||||
- "容易过拟合，泛化能力较弱。"
|
||||
- "对噪声数据较敏感。"
|
||||
applicable_scenarios:
|
||||
- "客户细分，如电商用户分类。"
|
||||
- "医学诊断，如肿瘤良恶性分类。"
|
||||
|
||||
RandomForestClassifier:
|
||||
principle: "随机森林是一种集成学习方法，通过构建多个决策树并投票或平均进行分类，提高模型的稳定性和准确性。"
|
||||
advantages:
|
||||
- "较强的泛化能力，能有效防止过拟合。"
|
||||
- "可用于高维数据，支持特征重要性评估。"
|
||||
disadvantages:
|
||||
- "训练时间较长，预测速度相对较慢。"
|
||||
applicable_scenarios:
|
||||
- "信用风险评估。"
|
||||
- "医疗数据分析。"
|
||||
|
||||
XGBClassifier:
|
||||
principle: "XGBoost（Extreme Gradient Boosting）是一种基于梯度提升树（GBDT）的改进算法，具有更强的正则化和并行处理能力。"
|
||||
advantages:
|
||||
- "计算效率高，支持并行计算。"
|
||||
- "具有内置的缺失值处理能力。"
|
||||
disadvantages:
|
||||
- "参数较多，调优较复杂。"
|
||||
applicable_scenarios:
|
||||
- "金融风险预测。"
|
||||
- "搜索引擎排序。"
|
||||
|
||||
AdaBoostClassifier:
|
||||
principle: "AdaBoost 是一种提升方法，它通过组合多个弱分类器来提高整体分类精度，赋予错误分类样本更高的权重。"
|
||||
advantages:
|
||||
- "能有效提升弱分类器的性能。"
|
||||
- "适用于处理非均衡数据。"
|
||||
disadvantages:
|
||||
- "对噪声数据敏感。"
|
||||
applicable_scenarios:
|
||||
- "人脸检测。"
|
||||
- "信用评分模型。"
|
||||
|
||||
CatBoostClassifier:
|
||||
principle: "CatBoost 是一种专为类别特征优化的梯度提升决策树方法，适用于处理高维稀疏数据。"
|
||||
advantages:
|
||||
- "对类别型特征处理能力强。"
|
||||
- "训练速度快，支持 GPU 加速。"
|
||||
disadvantages:
|
||||
- "需要较长的训练时间。"
|
||||
applicable_scenarios:
|
||||
- "推荐系统。"
|
||||
- "搜索排序。"
|
||||
|
||||
LGBMClassifier:
|
||||
principle: "LightGBM 是一种基于直方图优化的梯度提升树方法，优化了计算效率，适用于大规模数据。"
|
||||
advantages:
|
||||
- "计算速度快，适用于大规模数据集。"
|
||||
- "对高维特征数据处理效果良好。"
|
||||
disadvantages:
|
||||
- "对小数据集容易过拟合。"
|
||||
applicable_scenarios:
|
||||
- "广告点击率预测。"
|
||||
- "金融风控。"
|
||||
|
||||
GaussianNB:
|
||||
principle: "高斯朴素贝叶斯基于贝叶斯定理，假设特征服从高斯分布进行分类。"
|
||||
advantages:
|
||||
- "计算效率高，适用于大规模数据。"
|
||||
- "对小数据集具有良好效果。"
|
||||
disadvantages:
|
||||
- "对特征的独立性假设较强，可能不适用于某些数据。"
|
||||
applicable_scenarios:
|
||||
- "文本分类。"
|
||||
- "医疗诊断。"
|
||||
|
||||
KNeighborsClassifier:
|
||||
principle: "K 近邻（KNN）是一种基于距离度量的分类算法，通过找到最近的 K 个邻居来进行分类。"
|
||||
advantages:
|
||||
- "易于理解，实现简单。"
|
||||
- "对异常值不敏感。"
|
||||
disadvantages:
|
||||
- "计算复杂度高，预测速度较慢。"
|
||||
applicable_scenarios:
|
||||
- "推荐系统。"
|
||||
- "生物信息学分类。"
|
||||
|
||||
MLPClassifier:
|
||||
principle: "多层感知机（MLP）是一种前馈神经网络，通过多个隐藏层和非线性激活函数实现复杂分类任务。"
|
||||
advantages:
|
||||
- "能够学习复杂的非线性关系。"
|
||||
- "适用于高维数据。"
|
||||
disadvantages:
|
||||
- "需要大量数据进行训练，容易过拟合。"
|
||||
applicable_scenarios:
|
||||
- "图像分类。"
|
||||
- "语音识别。"
|
||||
|
||||
GradientBoostingClassifier:
|
||||
principle: "梯度提升决策树（GBDT）是一种集成学习方法，通过迭代训练多个决策树，使模型不断优化误差。"
|
||||
advantages:
|
||||
- "较强的泛化能力，适用于大规模数据集。"
|
||||
- "支持特征重要性分析。"
|
||||
disadvantages:
|
||||
- "训练时间较长，调优复杂。"
|
||||
applicable_scenarios:
|
||||
- "搜索引擎排名。"
|
||||
- "金融信用评分。"
|
||||
|
||||
DNN:
|
||||
principle: "深度神经网络（DNN）是一种多层神经网络，通过多层非线性变换提取数据的复杂特征。"
|
||||
advantages:
|
||||
- "适用于复杂非线性问题。"
|
||||
- "能够处理大规模数据。"
|
||||
disadvantages:
|
||||
- "计算开销大，对硬件要求高。"
|
||||
applicable_scenarios:
|
||||
- "自动驾驶。"
|
||||
- "自然语言处理。"
|
||||
|
||||
regression_algorithms:
|
||||
LinearRegression:
|
||||
principle: "线性回归通过最小化数据点与回归线之间的误差平方和，来拟合一条最佳的直线。"
|
||||
advantages:
|
||||
- "实现简单，易于理解"
|
||||
- "计算效率高，适用于小规模数据集"
|
||||
disadvantages:
|
||||
- "对异常值敏感"
|
||||
- "只能建模线性关系，无法处理非线性数据"
|
||||
applicable_scenarios: "适用于线性关系明确，且数据量适中的问题。"
|
||||
|
||||
PolynomialRegression:
|
||||
principle: "多项式回归是线性回归的一种扩展，利用多项式拟合数据，处理非线性关系。"
|
||||
advantages:
|
||||
- "能够拟合非线性关系"
|
||||
- "灵活性强，能够处理复杂数据模式"
|
||||
disadvantages:
|
||||
- "容易过拟合，特别是在多项式度数较高时"
|
||||
- "计算复杂度较高"
|
||||
applicable_scenarios: "适用于非线性数据拟合，但需避免过拟合。"
|
||||
|
||||
Ridge:
|
||||
principle: "岭回归在最小化误差的同时，引入L2正则化，约束模型的复杂度。"
|
||||
advantages:
|
||||
- "有效防止过拟合"
|
||||
- "适用于特征多的情况"
|
||||
disadvantages:
|
||||
- "对于特征相关性较强的数据效果较差"
|
||||
- "结果解释性较差"
|
||||
applicable_scenarios: "适用于特征多且存在共线性的线性回归问题。"
|
||||
|
||||
Lasso:
|
||||
principle: "Lasso回归通过L1正则化来限制模型的复杂度，可以实现特征选择。"
|
||||
advantages:
|
||||
- "能够进行特征选择，减少不相关特征"
|
||||
- "减少模型的复杂度，避免过拟合"
|
||||
disadvantages:
|
||||
- "可能会导致某些特征完全被剔除，造成信息丢失"
|
||||
- "对于高相关特征可能不稳定"
|
||||
applicable_scenarios: "适用于需要特征选择或特征较多的线性回归问题。"
|
||||
|
||||
ElasticNet:
|
||||
principle: "弹性网络回归结合了Lasso回归的L1正则化和岭回归的L2正则化，能够处理更多情况。"
|
||||
advantages:
|
||||
- "能够处理相关特征，结合L1和L2的优点"
|
||||
- "适用于特征数大于样本数的情形"
|
||||
disadvantages:
|
||||
- "计算复杂度较高"
|
||||
- "可能需要调参来获得最佳效果"
|
||||
applicable_scenarios: "适用于高维数据，且特征间存在线性相关的情况。"
|
||||
|
||||
SVR:
|
||||
principle: "支持向量回归（SVR）通过在高维空间中寻找一个最优超平面来拟合数据，能够处理非线性回归问题。"
|
||||
advantages:
|
||||
- "能够处理非线性数据，效果较好"
|
||||
- "适用于高维数据"
|
||||
disadvantages:
|
||||
- "对超参数敏感，调参困难"
|
||||
- "计算时间较长，尤其在大数据集上"
|
||||
applicable_scenarios: "适用于数据具有非线性关系且数据量适中的问题。"
|
||||
|
||||
DecisionTreeRegressor:
|
||||
principle: "决策树回归通过构建树状结构来拟合数据，节点上的划分基于特征的不同值。"
|
||||
advantages:
|
||||
- "易于理解和可视化"
|
||||
- "可以处理非线性关系"
|
||||
disadvantages:
|
||||
- "容易过拟合"
|
||||
- "对噪声数据敏感"
|
||||
applicable_scenarios: "适用于非线性回归，且对数据的复杂性有较高的容忍度。"
|
||||
|
||||
RandomForestRegressor:
|
||||
principle: "随机森林回归通过集成多棵决策树的结果来提高预测精度，能够处理高维数据。"
|
||||
advantages:
|
||||
- "不容易过拟合，适用于复杂问题"
|
||||
- "能够处理缺失数据"
|
||||
disadvantages:
|
||||
- "计算复杂度较高"
|
||||
- "结果难以解释"
|
||||
applicable_scenarios: "适用于复杂的回归问题，尤其是特征维度较高的场景。"
|
||||
|
||||
XGBRegressor:
|
||||
principle: "XGBoost回归通过梯度提升算法（GBDT）优化损失函数，通过树的组合来预测目标值。"
|
||||
advantages:
|
||||
- "预测精度高，效果好"
|
||||
- "处理大数据能力强"
|
||||
disadvantages:
|
||||
- "需要较长的训练时间"
|
||||
- "对超参数敏感"
|
||||
applicable_scenarios: "适用于大规模数据集且对预测精度要求较高的问题。"
|
||||
|
||||
AdaBoostRegressor:
|
||||
principle: "AdaBoost回归通过多次训练弱学习器并加权组合，改进模型的预测能力。"
|
||||
advantages:
|
||||
- "能够提高弱学习器的预测精度"
|
||||
- "对噪声较为鲁棒"
|
||||
disadvantages:
|
||||
- "容易受到异常值影响"
|
||||
- "对某些问题的性能较差"
|
||||
applicable_scenarios: "适用于弱学习器的组合优化问题，特别是样本不平衡的场景。"
|
||||
|
||||
CatBoostRegressor:
|
||||
principle: "CatBoost回归基于梯度提升决策树（GBDT），特别优化了类别特征的处理。"
|
||||
advantages:
|
||||
- "能够处理类别特征，减少数据预处理"
|
||||
- "高效且精度较高"
|
||||
disadvantages:
|
||||
- "训练时间较长"
|
||||
- "参数调节较为复杂"
|
||||
applicable_scenarios: "适用于包含大量类别特征的数据集，且数据量较大的问题。"
|
||||
|
||||
LGBMRegressor:
|
||||
principle: "LightGBM回归基于梯度提升决策树，使用了直方图优化算法以加速训练过程。"
|
||||
advantages:
|
||||
- "训练速度快，能够处理大规模数据"
|
||||
- "内存占用较少，适合高维数据"
|
||||
disadvantages:
|
||||
- "需要调参以获得最佳效果"
|
||||
- "模型解释性较差"
|
||||
applicable_scenarios: "适用于大规模数据集，尤其是处理海量数据时效果优越。"
|
||||
|
||||
MLPRegressor:
|
||||
principle: "多层感知机回归是基于神经网络的回归模型，通过多个隐层来拟合复杂的非线性关系。"
|
||||
advantages:
|
||||
- "能够处理复杂的非线性关系"
|
||||
- "在大数据和高维数据中表现良好"
|
||||
disadvantages:
|
||||
- "训练时间较长，且对计算资源要求较高"
|
||||
- "容易过拟合，特别是在数据较少时"
|
||||
applicable_scenarios: "适用于需要捕捉复杂非线性关系的回归问题，尤其在数据量大时效果较好。"
|
||||
|
||||
clustering_algorithms:
|
||||
KMeans:
|
||||
principle: "K均值聚类通过最小化样本点到其最近簇中心的距离来将数据分为K个簇。"
|
||||
advantages:
|
||||
- "实现简单，计算效率高"
|
||||
- "适用于大数据集"
|
||||
disadvantages:
|
||||
- "需要预先指定簇的数量K"
|
||||
- "对初始中心点敏感，容易受到噪声影响"
|
||||
applicable_scenarios: "适用于簇形状较为规则且数据量较大的聚类问题。"
|
||||
|
||||
KMeansPlusPlus:
|
||||
principle: "K-Means++是一种改进的初始化方法，通过选择更远离当前簇中心的样本点作为初始中心，提升K均值聚类的效果。"
|
||||
advantages:
|
||||
- "比传统K均值方法更稳定"
|
||||
- "可以减少聚类结果的变异性"
|
||||
disadvantages:
|
||||
- "依然存在需要指定K的问题"
|
||||
- "初始化仍然可能影响最终结果"
|
||||
applicable_scenarios: "适用于K均值聚类方法，并且希望改进初始中心选择的场景。"
|
||||
|
||||
HierarchicalKMeans:
|
||||
principle: "层次化K均值结合了层次聚类和K均值聚类的方法，逐步将样本合并到已有簇中，形成层次化结构。"
|
||||
advantages:
|
||||
- "能够自动确定簇的数量"
|
||||
- "生成的树状图有助于理解数据结构"
|
||||
disadvantages:
|
||||
- "计算量大，尤其是在数据量较大时"
|
||||
- "对噪声和离群点敏感"
|
||||
applicable_scenarios: "适用于不确定簇的数量且数据结构较复杂的情况。"
|
||||
|
||||
FCM:
|
||||
principle: "模糊C均值（FCM）允许每个数据点属于多个簇，基于隶属度来进行聚类。"
|
||||
advantages:
|
||||
- "能够处理数据点属于多个簇的情况"
|
||||
- "适用于软聚类问题"
|
||||
disadvantages:
|
||||
- "对初始簇中心和隶属度设置较为敏感"
|
||||
- "计算量大，尤其是簇数较多时"
|
||||
applicable_scenarios: "适用于数据点可以属于多个簇的情况，如图像分割等问题。"
|
||||
|
||||
AgglomerativeClustering:
|
||||
principle: "层次聚类通过将相似度较高的样本逐步合并，最终形成树状结构（树状图）。"
|
||||
advantages:
|
||||
- "无需预先指定簇的数量"
|
||||
- "能够处理任意形状的簇"
|
||||
disadvantages:
|
||||
- "计算复杂度较高，数据量大时效率低"
|
||||
- "容易受到噪声的影响"
|
||||
applicable_scenarios: "适用于不确定簇的数量且簇的形状较复杂的场景。"
|
||||
|
||||
DBSCAN:
|
||||
principle: "DBSCAN基于密度的聚类方法，通过寻找密集区域来划分簇，对于稀疏区域则标记为噪声点。"
|
||||
advantages:
|
||||
- "能够发现任意形状的簇"
|
||||
- "不需要预先指定簇的数量，能够自动识别噪声"
|
||||
disadvantages:
|
||||
- "对参数设置敏感，尤其是邻域半径和最小样本数"
|
||||
- "不适用于簇大小差异过大的数据集"
|
||||
applicable_scenarios: "适用于具有明显密度差异的数据集，尤其适合处理含有噪声的数据。"
|
||||
|
||||
GaussianMixture:
|
||||
principle: "高斯混合模型（GMM）假设数据是由多个高斯分布的混合体组成，通过EM算法估计每个数据点的隶属概率。"
|
||||
advantages:
|
||||
- "能够处理重叠的簇"
|
||||
- "可以自动估计每个簇的分布"
|
||||
disadvantages:
|
||||
- "计算复杂度高，容易陷入局部最优解"
|
||||
- "需要预先指定簇的数量"
|
||||
applicable_scenarios: "适用于需要拟合混合高斯分布的数据，尤其适合处理连续数据。"
|
||||
|
||||
SpectralClustering:
|
||||
principle: "谱聚类通过构造样本之间的相似度矩阵，并计算其特征值与特征向量来实现聚类。"
|
||||
advantages:
|
||||
- "能够处理复杂的簇结构"
|
||||
- "适用于非凸形状的簇"
|
||||
disadvantages:
|
||||
- "计算量大，尤其在大数据集上效率低"
|
||||
- "需要计算样本间的相似度矩阵，存储和计算成本较高"
|
||||
applicable_scenarios: "适用于样本之间相似度较强，但簇形状复杂或非凸的聚类任务。"
|
||||
|
||||
|
||||
650
model/parameter.yaml
Normal file
650
model/parameter.yaml
Normal file
@ -0,0 +1,650 @@
|
||||
classification_algorithms:
|
||||
LogisticRegression:
|
||||
parameters:
|
||||
- name: "penalty"
|
||||
type: "str"
|
||||
default: "l2"
|
||||
description: "用于正则化的惩罚项，可选 'l1', 'l2', 'elasticnet', 'none'。"
|
||||
- name: "C"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "正则化强度的倒数，较小的值表示更强的正则化。"
|
||||
- name: "solver"
|
||||
type: "str"
|
||||
default: "lbfgs"
|
||||
description: "优化算法，可选 'newton-cg', 'lbfgs', 'liblinear', 'sag', 'saga'。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "最大迭代次数，控制优化收敛速度。"
|
||||
|
||||
SVC:
|
||||
parameters:
|
||||
- name: "C"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "正则化参数，控制决策边界的松弛程度。"
|
||||
- name: "kernel"
|
||||
type: "str"
|
||||
default: "rbf"
|
||||
description: "核函数类型，可选 'linear', 'poly', 'rbf', 'sigmoid', 'precomputed'。"
|
||||
- name: "degree"
|
||||
type: "int"
|
||||
default: "3"
|
||||
description: "多项式核函数的维度，仅当 kernel='poly' 时有效。"
|
||||
- name: "gamma"
|
||||
type: "str"
|
||||
default: "scale"
|
||||
description: "核函数系数，可选 'scale', 'auto' 或浮点数。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "-1"
|
||||
description: "最大迭代次数，-1 表示无限制。"
|
||||
|
||||
DecisionTreeClassifier:
|
||||
parameters:
|
||||
- name: "criterion"
|
||||
type: "str"
|
||||
default: "gini"
|
||||
description: "用于划分的准则，可选 'gini' 或 'entropy'。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "树的最大深度，None 表示不限制。"
|
||||
- name: "min_samples_split"
|
||||
type: "int"
|
||||
default: "2"
|
||||
description: "内部节点再划分所需的最小样本数。"
|
||||
- name: "min_samples_leaf"
|
||||
type: "int"
|
||||
default: "1"
|
||||
description: "叶子节点包含的最小样本数。"
|
||||
|
||||
RandomForestClassifier:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "森林中树的数量。"
|
||||
- name: "criterion"
|
||||
type: "str"
|
||||
default: "gini"
|
||||
description: "用于划分的准则，可选 'gini' 或 'entropy'。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "树的最大深度，None 表示不限制。"
|
||||
- name: "bootstrap"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否进行自助采样。"
|
||||
|
||||
XGBClassifier:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "树的数量。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "学习率，控制每棵树的贡献程度。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "6"
|
||||
description: "每棵树的最大深度。"
|
||||
- name: "gamma"
|
||||
type: "float"
|
||||
default: "0"
|
||||
description: "节点分裂所需的最小损失减少量。"
|
||||
|
||||
AdaBoostClassifier:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "50"
|
||||
description: "弱分类器的数量。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "更新权重时的缩放因子。"
|
||||
- name: "algorithm"
|
||||
type: "str"
|
||||
default: "SAMME.R"
|
||||
description: "提升算法，可选 'SAMME' 或 'SAMME.R'。"
|
||||
|
||||
CatBoostClassifier:
|
||||
parameters:
|
||||
- name: "iterations"
|
||||
type: "int"
|
||||
default: "1000"
|
||||
description: "训练的迭代次数。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.03"
|
||||
description: "学习率，控制更新步长。"
|
||||
- name: "depth"
|
||||
type: "int"
|
||||
default: "6"
|
||||
description: "树的深度。"
|
||||
|
||||
LGBMClassifier:
|
||||
parameters:
|
||||
- name: "num_leaves"
|
||||
type: "int"
|
||||
default: "31"
|
||||
description: "树的最大叶子数。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "学习率。"
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "树的数量。"
|
||||
|
||||
GaussianNB:
|
||||
parameters:
|
||||
- name: "var_smoothing"
|
||||
type: "float"
|
||||
default: "1e-9"
|
||||
description: "加在方差上的小数，防止零方差问题。"
|
||||
|
||||
KNeighborsClassifier:
|
||||
parameters:
|
||||
- name: "n_neighbors"
|
||||
type: "int"
|
||||
default: "5"
|
||||
description: "最近邻的数量。"
|
||||
- name: "weights"
|
||||
type: "str"
|
||||
default: "uniform"
|
||||
description: "权重分配策略，可选 'uniform' 或 'distance'。"
|
||||
- name: "algorithm"
|
||||
type: "str"
|
||||
default: "auto"
|
||||
description: "搜索最近邻的算法。"
|
||||
|
||||
MLPClassifier:
|
||||
parameters:
|
||||
- name: "hidden_layer_sizes"
|
||||
type: "tuple"
|
||||
default: "(100,)"
|
||||
description: "隐藏层的神经元数量。"
|
||||
- name: "activation"
|
||||
type: "str"
|
||||
default: "relu"
|
||||
description: "激活函数，可选 'identity', 'logistic', 'tanh', 'relu'。"
|
||||
- name: "solver"
|
||||
type: "str"
|
||||
default: "adam"
|
||||
description: "优化算法，可选 'lbfgs', 'sgd', 'adam'。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "200"
|
||||
description: "最大迭代次数。"
|
||||
|
||||
GradientBoostingClassifier:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "树的数量。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "学习率。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "3"
|
||||
description: "每棵树的最大深度。"
|
||||
|
||||
|
||||
regression_algorithms:
|
||||
LinearRegression:
|
||||
parameters:
|
||||
- name: "fit_intercept"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否计算截距。"
|
||||
- name: "normalize"
|
||||
type: "bool"
|
||||
default: "False"
|
||||
description: "是否对数据进行归一化处理。（已弃用）"
|
||||
- name: "copy_X"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否复制输入数据。"
|
||||
- name: "n_jobs"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "用于计算的并行作业数。"
|
||||
|
||||
PolynomialRegression:
|
||||
parameters:
|
||||
- name: "degree"
|
||||
type: "int"
|
||||
default: "2"
|
||||
description: "多项式的最高次数。"
|
||||
- name: "interaction_only"
|
||||
type: "bool"
|
||||
default: "False"
|
||||
description: "是否仅考虑特征之间的交互项。"
|
||||
- name: "include_bias"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否包含偏置项。"
|
||||
|
||||
Ridge:
|
||||
parameters:
|
||||
- name: "alpha"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "正则化力度。"
|
||||
- name: "fit_intercept"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否计算截距。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "0.001"
|
||||
description: "容忍误差。"
|
||||
- name: "solver"
|
||||
type: "str"
|
||||
default: "auto"
|
||||
description: "求解器选择。"
|
||||
|
||||
Lasso:
|
||||
parameters:
|
||||
- name: "alpha"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "正则化参数。"
|
||||
- name: "fit_intercept"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否计算截距。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "1000"
|
||||
description: "最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "0.0001"
|
||||
description: "收敛容忍度。"
|
||||
- name: "selection"
|
||||
type: "str"
|
||||
default: "cyclic"
|
||||
description: "特征选择方式，可选 'cyclic' 或 'random'。"
|
||||
|
||||
ElasticNet:
|
||||
parameters:
|
||||
- name: "alpha"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "正则化参数。"
|
||||
- name: "l1_ratio"
|
||||
type: "float"
|
||||
default: "0.5"
|
||||
description: "L1正则化比例，控制L1和L2的混合比例。"
|
||||
- name: "fit_intercept"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否计算截距。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "1000"
|
||||
description: "最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "0.0001"
|
||||
description: "收敛容忍度。"
|
||||
|
||||
SVR:
|
||||
parameters:
|
||||
- name: "kernel"
|
||||
type: "str"
|
||||
default: "rbf"
|
||||
description: "核函数类型，可选 'linear', 'poly', 'rbf', 'sigmoid'。"
|
||||
- name: "C"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "惩罚参数。"
|
||||
- name: "epsilon"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "epsilon不敏感损失中的参数。"
|
||||
- name: "degree"
|
||||
type: "int"
|
||||
default: "3"
|
||||
description: "多项式核函数的度数，仅当 kernel='poly' 时有效。"
|
||||
- name: "gamma"
|
||||
type: "str"
|
||||
default: "scale"
|
||||
description: "核函数系数。"
|
||||
|
||||
DecisionTreeRegressor:
|
||||
parameters:
|
||||
- name: "criterion"
|
||||
type: "str"
|
||||
default: "squared_error"
|
||||
description: "衡量分裂质量的指标。"
|
||||
- name: "splitter"
|
||||
type: "str"
|
||||
default: "best"
|
||||
description: "划分策略，可选 'best' 或 'random'。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "树的最大深度。"
|
||||
- name: "min_samples_split"
|
||||
type: "int/float"
|
||||
default: "2"
|
||||
description: "内部节点再划分所需的最小样本数。"
|
||||
- name: "min_samples_leaf"
|
||||
type: "int/float"
|
||||
default: "1"
|
||||
description: "叶子节点最少样本数。"
|
||||
|
||||
RandomForestRegressor:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "森林中树的数量。"
|
||||
- name: "criterion"
|
||||
type: "str"
|
||||
default: "squared_error"
|
||||
description: "衡量分裂质量的指标。"
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "树的最大深度。"
|
||||
- name: "min_samples_split"
|
||||
type: "int"
|
||||
default: "2"
|
||||
description: "内部节点再划分所需的最小样本数。"
|
||||
- name: "n_jobs"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "用于计算的并行作业数。"
|
||||
|
||||
XGBRegressor:
|
||||
parameters:
|
||||
- name: "max_depth"
|
||||
type: "int"
|
||||
default: "3"
|
||||
description: "树的最大深度。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "学习率。"
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "树的数量。"
|
||||
- name: "objective"
|
||||
type: "str"
|
||||
default: "reg:squarederror"
|
||||
description: "损失函数。"
|
||||
- name: "subsample"
|
||||
type: "float"
|
||||
default: "1"
|
||||
description: "采样比例。"
|
||||
|
||||
AdaBoostRegressor:
|
||||
parameters:
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "50"
|
||||
description: "基学习器的数量。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "学习率。"
|
||||
- name: "loss"
|
||||
type: "str"
|
||||
default: "linear"
|
||||
description: "损失函数类型，可选 'linear', 'square', 'exponential'。"
|
||||
|
||||
CatBoostRegressor:
|
||||
parameters:
|
||||
- name: "iterations"
|
||||
type: "int"
|
||||
default: "1000"
|
||||
description: "迭代次数。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.03"
|
||||
description: "学习率。"
|
||||
- name: "depth"
|
||||
type: "int"
|
||||
default: "6"
|
||||
description: "树的深度。"
|
||||
- name: "l2_leaf_reg"
|
||||
type: "float"
|
||||
default: "3.0"
|
||||
description: "L2正则化系数。"
|
||||
- name: "loss_function"
|
||||
type: "str"
|
||||
default: "RMSE"
|
||||
description: "损失函数。"
|
||||
|
||||
LGBMRegressor:
|
||||
parameters:
|
||||
- name: "num_leaves"
|
||||
type: "int"
|
||||
default: "31"
|
||||
description: "叶子节点数量。"
|
||||
- name: "learning_rate"
|
||||
type: "float"
|
||||
default: "0.1"
|
||||
description: "学习率。"
|
||||
- name: "n_estimators"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "树的数量。"
|
||||
- name: "objective"
|
||||
type: "str"
|
||||
default: "regression"
|
||||
description: "目标函数。"
|
||||
- name: "subsample"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "采样比例。"
|
||||
|
||||
MLPRegressor:
|
||||
parameters:
|
||||
- name: "hidden_layer_sizes"
|
||||
type: "tuple"
|
||||
default: "(100,)"
|
||||
description: "隐藏层的神经元数量和层数。"
|
||||
- name: "activation"
|
||||
type: "str"
|
||||
default: "relu"
|
||||
description: "激活函数，可选 'identity', 'logistic', 'tanh', 'relu'。"
|
||||
- name: "solver"
|
||||
type: "str"
|
||||
default: "adam"
|
||||
description: "权重优化算法。"
|
||||
- name: "alpha"
|
||||
type: "float"
|
||||
default: "0.0001"
|
||||
description: "L2正则化参数。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "200"
|
||||
description: "最大迭代次数。"
|
||||
|
||||
|
||||
clustering_algorithms:
|
||||
KMeans:
|
||||
parameters:
|
||||
- name: "n_clusters"
|
||||
type: "int"
|
||||
default: "8"
|
||||
description: "簇的数量，K均值聚类的核心参数。"
|
||||
- name: "init"
|
||||
type: "str"
|
||||
default: "k-means++"
|
||||
description: "初始化方法，可选 'k-means++'、'random' 或数组类型。"
|
||||
- name: "n_init"
|
||||
type: "int"
|
||||
default: "10"
|
||||
description: "初始化次数，K均值算法会运行该次数，选择最优结果。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "300"
|
||||
description: "每次运行的最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "1e-4"
|
||||
description: "算法收敛的容忍误差，当目标函数变化小于此值时，认为算法收敛。"
|
||||
- name: "precompute_distances"
|
||||
type: "bool"
|
||||
default: "True"
|
||||
description: "是否预计算距离矩阵（会增加计算量）。"
|
||||
|
||||
KMeansPlusPlus:
|
||||
parameters:
|
||||
- name: "n_clusters"
|
||||
type: "int"
|
||||
default: "8"
|
||||
description: "簇的数量。"
|
||||
- name: "n_init"
|
||||
type: "int"
|
||||
default: "10"
|
||||
description: "初始化次数。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "300"
|
||||
description: "最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "1e-4"
|
||||
description: "容忍误差，用于判断算法是否收敛。"
|
||||
|
||||
AgglomerativeClustering:
|
||||
parameters:
|
||||
- name: "n_clusters"
|
||||
type: "int"
|
||||
default: "2"
|
||||
description: "簇的数量。"
|
||||
- name: "affinity"
|
||||
type: "str"
|
||||
default: "euclidean"
|
||||
description: "衡量样本间距离的方式，可选 'euclidean'、'manhattan'、'cosine'。"
|
||||
- name: "memory"
|
||||
type: "str/None"
|
||||
default: "None"
|
||||
description: "缓存路径，存储树状结构。"
|
||||
- name: "linkage"
|
||||
type: "str"
|
||||
default: "ward"
|
||||
description: "聚类方式，可选 'ward'（最小化平方误差）、'average'、'complete'。"
|
||||
- name: "compute_full_tree"
|
||||
type: "str"
|
||||
default: "auto"
|
||||
description: "是否计算完全树结构，'auto'为自动，'True'为计算，'False'为不计算。"
|
||||
|
||||
FCM:
|
||||
parameters:
|
||||
- name: "n_clusters"
|
||||
type: "int"
|
||||
default: "2"
|
||||
description: "簇的数量。"
|
||||
- name: "m"
|
||||
type: "float"
|
||||
default: "2.0"
|
||||
description: "模糊指数，控制隶属度的模糊程度，通常设置为2。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "最大迭代次数。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "1e-5"
|
||||
description: "收敛容忍度，当隶属度变化小于此值时，认为算法收敛。"
|
||||
- name: "random_state"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "随机种子，用于初始化隶属度矩阵。"
|
||||
|
||||
DBSCAN:
|
||||
parameters:
|
||||
- name: "eps"
|
||||
type: "float"
|
||||
default: "0.5"
|
||||
description: "邻域的最大距离，决定样本是否在同一簇中。"
|
||||
- name: "min_samples"
|
||||
type: "int"
|
||||
default: "5"
|
||||
description: "形成簇的最小样本数。"
|
||||
- name: "metric"
|
||||
type: "str"
|
||||
default: "euclidean"
|
||||
description: "计算距离的方式，可选 'euclidean', 'manhattan', 'cosine' 等。"
|
||||
- name: "algorithm"
|
||||
type: "str"
|
||||
default: "auto"
|
||||
description: "计算最近邻的方法，可选 'auto', 'ball_tree', 'kd_tree', 'brute'。"
|
||||
- name: "leaf_size"
|
||||
type: "int"
|
||||
default: "30"
|
||||
description: "对于BallTree或KDTree的叶子大小，影响效率。"
|
||||
|
||||
GaussianMixture:
|
||||
parameters:
|
||||
- name: "n_components"
|
||||
type: "int"
|
||||
default: "1"
|
||||
description: "高斯混合分布的组件数，表示簇的数量。"
|
||||
- name: "covariance_type"
|
||||
type: "str"
|
||||
default: "full"
|
||||
description: "协方差类型，可选 'full', 'tied', 'diag', 'spherical'。"
|
||||
- name: "tol"
|
||||
type: "float"
|
||||
default: "1e-3"
|
||||
description: "收敛容忍度，当对数似然变化小于此值时停止算法。"
|
||||
- name: "max_iter"
|
||||
type: "int"
|
||||
default: "100"
|
||||
description: "最大迭代次数。"
|
||||
- name: "random_state"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "随机种子，用于初始化高斯混合模型的参数。"
|
||||
|
||||
SpectralClustering:
|
||||
parameters:
|
||||
- name: "n_clusters"
|
||||
type: "int"
|
||||
default: "8"
|
||||
description: "簇的数量。"
|
||||
- name: "affinity"
|
||||
type: "str"
|
||||
default: "rbf"
|
||||
description: "计算相似度的方式，可选 'rbf'、'nearest_neighbors'、'precomputed'。"
|
||||
- name: "n_neighbors"
|
||||
type: "int"
|
||||
default: "10"
|
||||
description: "对于 'nearest_neighbors' 相似度，使用的邻居数。"
|
||||
- name: "gamma"
|
||||
type: "float"
|
||||
default: "1.0"
|
||||
description: "用于计算径向基函数的gamma参数，影响相似度的计算。"
|
||||
- name: "eigen_solver"
|
||||
type: "str"
|
||||
default: "auto"
|
||||
description: "求解特征值的方式，可选 'arpack', 'lobpcg', 'auto'。"
|
||||
- name: "random_state"
|
||||
type: "int"
|
||||
default: "None"
|
||||
description: "随机种子，用于初始化谱聚类的计算。"
|
||||
|
||||
|
||||
Loading…
Reference in New Issue
Block a user