466 lines
15 KiB
TypeScript
466 lines
15 KiB
TypeScript
import SubAreaTitle from '@/components/SubAreaTitle';
|
||
import {
|
||
AutoMLEnsembleClass,
|
||
AutoMLResamplingStrategy,
|
||
AutoMLTaskType,
|
||
autoMLEnsembleClassOptions,
|
||
autoMLResamplingStrategyOptions,
|
||
autoMLTaskTypeOptions,
|
||
} from '@/enums';
|
||
import { Col, Form, InputNumber, Radio, Row, Select, Switch } from 'antd';
|
||
|
||
// 分类算法
|
||
const classificationAlgorithms = [
|
||
'adaboost',
|
||
'bernoulli_nb',
|
||
'decision_tree',
|
||
'extra_trees',
|
||
'gaussian_nb',
|
||
'gradient_boosting',
|
||
'k_nearest_neighbors',
|
||
'lda',
|
||
'liblinear_svc',
|
||
'libsvm_svc',
|
||
'mlp',
|
||
'multinomial_nb',
|
||
'passive_aggressive',
|
||
'qda',
|
||
'random_forest',
|
||
'sgd',
|
||
'LightGBMClassification',
|
||
'XGBoostClassification',
|
||
'StackingClassification',
|
||
].map((name) => ({ label: name, value: name }));
|
||
|
||
// 回归算法
|
||
const regressorAlgorithms = [
|
||
'adaboost',
|
||
'ard_regression',
|
||
'decision_tree',
|
||
'extra_trees',
|
||
'gaussian_process',
|
||
'gradient_boosting',
|
||
'k_nearest_neighbors',
|
||
'liblinear_svr',
|
||
'libsvm_svr',
|
||
'mlp',
|
||
'random_forest',
|
||
'sgd',
|
||
'LightGBMRegression',
|
||
'XGBoostRegression',
|
||
].map((name) => ({ label: name, value: name }));
|
||
|
||
// 特征预处理算法
|
||
const featureAlgorithms = [
|
||
'densifier',
|
||
'extra_trees_preproc_for_classification',
|
||
'extra_trees_preproc_for_regression',
|
||
'fast_ica',
|
||
'feature_agglomeration',
|
||
'kernel_pca',
|
||
'kitchen_sinks',
|
||
'liblinear_svc_preprocessor',
|
||
'no_preprocessing',
|
||
'nystroem_sampler',
|
||
'pca',
|
||
'polynomial',
|
||
'random_trees_embedding',
|
||
'select_percentile_classification',
|
||
'select_percentile_regression',
|
||
'select_rates_classification',
|
||
'select_rates_regression',
|
||
'truncatedSVD',
|
||
].map((name) => ({ label: name, value: name }));
|
||
|
||
// 分类指标
|
||
export const classificationMetrics = [
|
||
'accuracy',
|
||
'balanced_accuracy',
|
||
'roc_auc',
|
||
'average_precision',
|
||
'log_loss',
|
||
'precision_macro',
|
||
'precision_micro',
|
||
'precision_samples',
|
||
'precision_weighted',
|
||
'recall_macro',
|
||
'recall_micro',
|
||
'recall_samples',
|
||
'recall_weighted',
|
||
'f1_macro',
|
||
'f1_micro',
|
||
'f1_samples',
|
||
'f1_weighted',
|
||
].map((name) => ({ label: name, value: name }));
|
||
|
||
// 回归指标
|
||
export const regressionMetrics = [
|
||
'mean_absolute_error',
|
||
'mean_squared_error',
|
||
'root_mean_squared_error',
|
||
'mean_squared_log_error',
|
||
'median_absolute_error',
|
||
'r2',
|
||
].map((name) => ({ label: name, value: name }));
|
||
|
||
function ExecuteConfig() {
|
||
const form = Form.useFormInstance();
|
||
const task_type = Form.useWatch('task_type', form);
|
||
const include_classifier = Form.useWatch('include_classifier', form);
|
||
const exclude_classifier = Form.useWatch('exclude_classifier', form);
|
||
const include_regressor = Form.useWatch('include_regressor', form);
|
||
const exclude_regressor = Form.useWatch('exclude_regressor', form);
|
||
const include_feature_preprocessor = Form.useWatch('include_feature_preprocessor', form);
|
||
const exclude_feature_preprocessor = Form.useWatch('exclude_feature_preprocessor', form);
|
||
|
||
return (
|
||
<>
|
||
<SubAreaTitle
|
||
title="执行配置"
|
||
image={require('@/assets/img/model-deployment.png')}
|
||
style={{ marginTop: '20px', marginBottom: '24px' }}
|
||
></SubAreaTitle>
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="任务类型"
|
||
name="task_type"
|
||
rules={[{ required: true, message: '请选择任务类型' }]}
|
||
>
|
||
<Radio.Group
|
||
options={autoMLTaskTypeOptions}
|
||
onChange={() => form.resetFields(['metrics'])}
|
||
></Radio.Group>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="特征预处理算法"
|
||
name="include_feature_preprocessor"
|
||
tooltip="如果不选，则使用所有可能的特征预处理算法。否则，将只使用包含的特征预处理算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="请选择特征预处理算法"
|
||
options={featureAlgorithms}
|
||
disabled={exclude_feature_preprocessor?.length > 0}
|
||
mode="multiple"
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="排除特征预处理算法"
|
||
name="exclude_feature_preprocessor"
|
||
tooltip="如果不选，则使用所有可能的特征预处理算法。否则，将排除包含的特征预处理算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="排除特征预处理算法"
|
||
options={featureAlgorithms}
|
||
disabled={include_feature_preprocessor?.length > 0}
|
||
mode="multiple"
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Form.Item dependencies={['task_type']} noStyle>
|
||
{({ getFieldValue }) => {
|
||
return getFieldValue('task_type') === AutoMLTaskType.Classification ? (
|
||
<>
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="分类算法"
|
||
name="include_classifier"
|
||
tooltip="如果不选，则使用所有可能的分类算法。否则，将只使用包含的算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="请选择分类算法"
|
||
options={classificationAlgorithms}
|
||
mode="multiple"
|
||
disabled={exclude_classifier?.length > 0}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="排除分类算法"
|
||
name="exclude_classifier"
|
||
tooltip="如果不选，则使用所有可能的分类算法。否则，将排除包含的算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="排除分类算法"
|
||
options={classificationAlgorithms}
|
||
mode="multiple"
|
||
disabled={include_classifier?.length > 0}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
</>
|
||
) : (
|
||
<>
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="回归算法"
|
||
name="include_regressor"
|
||
tooltip="如果不选，则使用所有可能的回归算法。否则，将只使用包含的算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="请选择回归算法"
|
||
options={regressorAlgorithms}
|
||
mode="multiple"
|
||
disabled={exclude_regressor?.length > 0}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="排除的回归算法"
|
||
name="exclude_regressor"
|
||
tooltip="如果不选，则使用所有可能的回归算法。否则，将排除包含的算法"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="排除回归算法"
|
||
options={regressorAlgorithms}
|
||
mode="multiple"
|
||
disabled={include_regressor?.length > 0}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
</>
|
||
);
|
||
}}
|
||
</Form.Item>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="集成方式"
|
||
name="ensemble_class"
|
||
tooltip="仅使用单个最佳模型还是集成模型"
|
||
>
|
||
<Radio.Group options={autoMLEnsembleClassOptions}></Radio.Group>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Form.Item dependencies={['ensemble_class']} noStyle>
|
||
{({ getFieldValue }) => {
|
||
return getFieldValue('ensemble_class') === AutoMLEnsembleClass.Default ? (
|
||
<>
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="集成模型数量"
|
||
name="ensemble_size"
|
||
tooltip="集成模型数量，如果设置为0，则没有集成。默认50"
|
||
>
|
||
<InputNumber placeholder="请输入集成模型数量" min={0} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="集成最佳模型数量"
|
||
name="ensemble_nbest"
|
||
tooltip="仅集成最佳的N个模型"
|
||
>
|
||
<InputNumber placeholder="请输入集成最佳模型数量" min={1} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
</>
|
||
) : null;
|
||
}}
|
||
</Form.Item>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="最大数量"
|
||
name="max_models_on_disc"
|
||
tooltip="定义在磁盘中保存的模型的最大数量。额外的模型数量将被永久删除，它设置了一个集成可以使用多少个模型的上限。必须是大于等于1的整数，默认50"
|
||
>
|
||
<InputNumber placeholder="请输入最大数量" min={1} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="内存限制（MB）"
|
||
name="memory_limit"
|
||
tooltip="机器学习算法的内存限制（MB）。如果自动机器学习试图分配超过memory_limit MB，它将停止拟合机器学习算法。默认3072"
|
||
>
|
||
<InputNumber placeholder="请输入内存限制" min={0} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="单次时间限制（秒）"
|
||
name="per_run_time_limit"
|
||
tooltip="单次调用机器学习模型的时间限制（以秒为单位）。如果机器学习算法运行超过时间限制，将终止模型拟合，默认600"
|
||
>
|
||
<InputNumber placeholder="请输入时间限制" min={0} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="搜索时间限制（秒）"
|
||
name="time_left_for_this_task"
|
||
tooltip="搜索合适模型的时间限制（以秒为单位）。通过增加这个值，自动机器学习有更高的机会找到更好的模型。默认3600。"
|
||
>
|
||
<InputNumber placeholder="请输入搜索时间限制" min={0} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="测试集比率"
|
||
name="test_size"
|
||
tooltip="将数据划分为训练数据和测试数据，测试数据集所占比例，0到1之间"
|
||
>
|
||
<InputNumber placeholder="请输入测试集比率" min={0} max={1} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item label="计算指标" name="scoring_functions" tooltip="需要计算并打印的指标">
|
||
<Select
|
||
allowClear
|
||
placeholder="请选择计算指标"
|
||
options={
|
||
task_type === AutoMLTaskType.Classification
|
||
? classificationMetrics
|
||
: regressionMetrics
|
||
}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item label="随机种子" name="seed" tooltip="随机种子，将决定输出文件名">
|
||
<InputNumber placeholder="请输入随机种子" min={0} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<SubAreaTitle
|
||
title="重采样策略"
|
||
image={require('@/assets/img/resample-icon.png')}
|
||
style={{ marginTop: '20px', marginBottom: '24px' }}
|
||
></SubAreaTitle>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="重采样策略"
|
||
name="resampling_strategy"
|
||
tooltip="重采样策略，分为holdout和crossValid。holdout指定训练数据划分为训练集和验证集的比例。crossValid为交叉验证。"
|
||
>
|
||
<Select
|
||
allowClear
|
||
placeholder="请选择重采样策略"
|
||
options={autoMLResamplingStrategyOptions}
|
||
showSearch
|
||
/>
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Form.Item dependencies={['resampling_strategy']} noStyle>
|
||
{({ getFieldValue }) => {
|
||
return getFieldValue('resampling_strategy') === AutoMLResamplingStrategy.CrossValid ? (
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="交叉验证折数"
|
||
name="folds"
|
||
rules={[
|
||
{
|
||
required: true,
|
||
message: '请输入交叉验证折数',
|
||
},
|
||
]}
|
||
>
|
||
<InputNumber placeholder="请输入交叉验证折数" min={1} precision={0} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
) : null;
|
||
}}
|
||
</Form.Item>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="是否打乱"
|
||
name="shuffle"
|
||
tooltip="拆分数据前是否打乱顺序"
|
||
valuePropName="checked"
|
||
>
|
||
<Switch />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
|
||
<Row gutter={8}>
|
||
<Col span={10}>
|
||
<Form.Item
|
||
label="训练集比率"
|
||
name="train_size"
|
||
tooltip="重采样划分训练集和验证集，训练集的比率，0到1之间"
|
||
>
|
||
<InputNumber placeholder="请输入训练集比率" min={0} max={1} />
|
||
</Form.Item>
|
||
</Col>
|
||
</Row>
|
||
</>
|
||
);
|
||
}
|
||
|
||
export default ExecuteConfig;
|