@datascibykashi: Outliers = extreme values that don’t “fit” with the rest of the data ⚠️ They can distort averages, scaling, and ML models if not handled properly. One simple way to detect them is using Percentiles. 🔹 How Percentile Method Works? 1️⃣ Choose lower & upper thresholds (commonly 1st and 99th, or 5th and 95th percentiles). 2️⃣ Any value below the lower bound or above the upper bound is considered an outlier. 📌 Example: If Salary < 1st percentile OR > 99th percentile → mark as outlier. ✅ Advantages: • Very simple to implement • Doesn’t assume data distribution (non-parametric) • Works well for skewed data ❌ Disadvantages: • Arbitrary cutoff points • May remove valid extreme values • Not suitable for very small datasets 💡 Key Takeaway: Percentile method = quick & easy way to detect outliers. Best used when you want a fast, distribution-free check 🚀 #️⃣ #OutlierDetection #PercentileMethod #DataPreprocessing #MachineLearning #DataScience #100DaysOfML

Data Scientist | Kashi

Open In TikTok:

Region: PK

Sunday 14 September 2025 12:10:47 GMT