CSV Outlier Detector
Find numeric outliers with the interquartile-range method
CSV Outlier Detector — what it does
Select a numeric column and the tool applies the interquartile range method: it finds the middle 50 percent of your values, extends a boundary above and below it by a multiplier you control, and flags everything outside as an outlier. The lower and upper bounds are shown, so you can see exactly what threshold produced each flag.
When this tool helps
- Finding data entry mistakes, such as a price recorded in cents when every other row is in whole units.
- Screening sensor or log readings for spikes before they distort an average.
- Checking an expense or invoice column for unusually large values that deserve manual review.
How to use it
- Choose the CSV file.
- Select the numeric column to examine.
- Set the IQR multiplier — 1.5 is the conventional default.
- Process to see the bounds and the outlier count.
Accuracy, limits, and good practice
- A multiplier of 1.5 marks mild outliers and 3.0 marks extreme ones; raise it when a broad natural spread produces too many flags.
- The IQR method is resistant to extreme values in a way that standard deviation is not, which is why it works on skewed data.
- An outlier is not automatically an error. Genuine large orders and rare events are real data, and deleting them silently is a mistake.
Frequently asked questions
What is the interquartile range?
The distance between the 25th and 75th percentiles. Bounds are set at the quartiles plus or minus the multiplier times that distance.
Why does a small dataset flag so much?
Quartiles are unstable with few rows. Below about twenty values, treat the result as a hint rather than a verdict.
Can I remove the outliers?
Filter them out with the CSV row filter using the bounds shown here, so the removal is explicit and reversible.
Related tools on this site
- CSV to Markdown Table — Convert a CSV file into a Markdown table
- Excel to CSV — Export one worksheet or every worksheet as CSV files
- JSON to Excel — Turn JSON arrays into an XLSX worksheet with optional flattening
Everything above runs inside this page. CSV Outlier Detector needs no account, no upload, and no server round trip — close the tab and nothing is left behind.
CSV 離群值偵測能做什麼
選定一個數值欄位，工具會套用四分位距法：先找出資料中間 50% 的範圍，再依你設定的倍數往上下各延伸出界線，落在界線外的值就標記為離群值。上下界都會顯示，你能清楚看到每個標記是依哪個門檻判定的。
什麼時候用得上
- 找出輸入錯誤，例如其他列都以元為單位、唯獨某列以分為單位的價格。
- 在平均值被拉歪之前，先篩出感測器或日誌讀數中的尖峰。
- 檢查費用或發票欄位中異常龐大的數字，安排人工複核。
使用步驟
- 選擇 CSV 檔案。
- 選定要檢查的數值欄位。
- 設定 IQR 倍數，慣例預設為 1.5。
- 執行後查看上下界與離群值數量。
精確度、限制與實務建議
- 倍數 1.5 標記中度離群，3.0 標記極端離群；資料本身分布很廣而標記過多時，可把倍數調高。
- 四分位距法對極端值的抵抗力遠優於標準差，因此在偏態資料上也管用。
- 離群值不等於錯誤。真實的大額訂單與罕見事件也是資料，直接刪掉是危險的做法。
常見問題
什麼是四分位距？
第 25 與第 75 百分位之間的距離。上下界即為四分位數加減「倍數乘以該距離」。
資料量少時為什麼標記特別多？
列數少時四分位數本來就不穩定。少於約二十筆時，請把結果當作提示而非結論。
可以直接移除離群值嗎？
建議用這裡顯示的上下界，搭配 CSV 資料篩選工具過濾，讓移除動作明確且可回溯。
本站相關工具
- CSV 轉 Markdown 表格 — 將 CSV 檔案轉成 Markdown 表格
- Excel 轉 CSV — 將指定工作表或全部工作表匯出為 CSV
- JSON 轉 Excel — 將 JSON 陣列轉為 XLSX，並可攤平巢狀資料
以上步驟全部在這個頁面內完成，CSV 離群值偵測 不需要註冊、不上傳檔案、也不經過伺服器，關閉分頁後不會留下任何資料。