Data Analysis
读取 CSV/JSON 数据集,按目标列做分层抽样的 train/validation/test 切分(默认 70/15/15,随机种子可复现),或分层 k 折交叉验证;输出每折类别分布报告与偏差条、重复行泄漏检查、SMOTE 过采样预览(仅在训练折上做最近邻插值),并可导出各折 CSV 为 ZIP。
用三种语言从你的代码中调用此工具。
# 1) Request a presigned URL → returns { uploadUrl, storageKey }
curl -X POST 'http://127.0.0.1:3003/api/upload/presign/train-test-split-with-stratification' \
-H 'Content-Type: application/json' \
-d '{"filename":"dataFile.ext","contentType":"application/octet-stream","size":12345}'
# 2) PUT the file bytes directly to the presigned uploadUrl
curl -X PUT '<presigned uploadUrl>' \
--data-binary @/path/to/file.ext
# 3) Call the tool, passing the returned storageKey for each file field
curl -X POST 'http://127.0.0.1:3003/zh/api/tools/train-test-split-with-stratification' \
-F 'dataFile=uploads/2026/01/01/your-tool-1700000000000-abc123.ext' \
-F 'datasetText=text,length,label
The movie was wonderful and I loved every minute,42,positive
Absolutely terrible waste of time,31,negative
Pretty good overall would recommend,31,positive
Bad acting and a worse plot,25,negative
I enjoyed it a lot,16,positive
Not great not terrible just fine,30,neutral
Loved the soundtrack and visuals,28,positive
The pacing was boring and slow,28,negative
It was okay nothing special,24,neutral
Fantastic acting by the whole cast,30,positive
Weak ending ruined it for me,24,negative
Fine for a rainy afternoon,24,neutral
Brilliant and moving film,23,positive
Disappointing sequel,19,negative
Mediocre but watchable,21,neutral
Gorgeous cinematography throughout,29,positive
Painfully dull script,19,negative
A masterpiece of the genre,24,positive
Cliched and forgettable,20,negative
Charming and genuinely funny,25,positive
Skip it you will thank me,21,negative
Delightful from start to finish,27,positive
Two hours I will never get back,29,negative
Watchable if nothing else is on,26,neutral' \
-F 'targetColumn=label' \
-F 'splitMode=standard' \
-F 'trainPercent=70' \
-F 'valPercent=15' \
-F 'kFolds=5' \
-F 'randomSeed=42' \
-F 'stratify=true' \
-F 'shuffle=true' \
-F 'smotePreview=false' \
-F 'exportZip=true'以 JSON 形式 POST 提交输入参数。文件类型参数需先单独上传。
POST http://127.0.0.1:3003/zh/api/tools/train-test-split-with-stratification| 参数名 | 类型 | 必填 | 说明 |
|---|---|---|---|
| dataFile | file需先上传 | 否 | — |
| datasetText | textarea | 否 | — |
| targetColumn | text | 是 | — |
| splitMode | select | 否 | — |
| trainPercent | number | 否 | — |
| valPercent | number | 否 | — |
| kFolds | number | 否 | — |
| randomSeed | number | 否 | — |
| stratify | checkbox | 否 | — |
| shuffle | checkbox | 否 | — |
| smotePreview | checkbox | 否 | — |
| exportZip | checkbox | 否 | — |
HTML 结果
{
"result": "<div>Processed HTML content</div>",
"error": "Error message (optional)",
"message": "Notification message (optional)",
"metadata": {
"key": "value"
}
}将此工具加入你的 Model Context Protocol 服务，让 AI 智能体可以列出并调用它。
将以下内容加入你的 MCP 客户端配置：
{
"mcpServers": {
"elysiatools-train-test-split-with-stratification": {
"name": "train-test-split-with-stratification",
"description": "读取 CSV/JSON 数据集,按目标列做分层抽样的 train/validation/test 切分(默认 70/15/15,随机种子可复现),或分层 k 折交叉验证;输出每折类别分布报告与偏差条、重复行泄漏检查、SMOTE 过采样预览(仅在训练折上做最近邻插值),并可导出各折 CSV 为 ZIP。",
"baseUrl": "http://127.0.0.1:3003/mcp/sse?toolId=train-test-split-with-stratification",
"command": "",
"args": [],
"env": {},
"isActive": true,
"type": "sse"
}
}
}连接到 SSE 端点后，列出已开放的工具：
{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/list"
}通过工具 id 调用，参数由其参数表构建：
{
"jsonrpc": "2.0",
"id": 2,
"method": "tools/call",
"params": {
"name": "train-test-split-with-stratification",
"arguments": {
"dataFile": "https://example.com/file.ext",
"datasetText": "text,length,label\nThe movie was wonderful and I loved every minute,42,positive\nAbsolutely terrible waste of time,31,negative\nPretty good overall would recommend,31,positive\nBad acting and a worse plot,25,negative\nI enjoyed it a lot,16,positive\nNot great not terrible just fine,30,neutral\nLoved the soundtrack and visuals,28,positive\nThe pacing was boring and slow,28,negative\nIt was okay nothing special,24,neutral\nFantastic acting by the whole cast,30,positive\nWeak ending ruined it for me,24,negative\nFine for a rainy afternoon,24,neutral\nBrilliant and moving film,23,positive\nDisappointing sequel,19,negative\nMediocre but watchable,21,neutral\nGorgeous cinematography throughout,29,positive\nPainfully dull script,19,negative\nA masterpiece of the genre,24,positive\nCliched and forgettable,20,negative\nCharming and genuinely funny,25,positive\nSkip it you will thank me,21,negative\nDelightful from start to finish,27,positive\nTwo hours I will never get back,29,negative\nWatchable if nothing else is on,26,neutral",
"targetColumn": "label",
"splitMode": "standard",
"trainPercent": 70,
"valPercent": 15,
"kFolds": 5,
"randomSeed": 42,
"stratify": true,
"shuffle": true,
"smotePreview": false,
"exportZip": true
}
}
}有问题或反馈？请联系 [email protected]