OPEN-SOURCE SCRIPT

ML Kernel Regression + EMA Trend

1 257
// This work is based on concepts by Uptrick
// Modified and updated by marjyhen1921

[ML] Nadaraya-Watson Kernel Regression (Multi-EMA Filter Edition)

Overview
This indicator features a Non-Parametric Machine Learning estimator based on the Nadaraya-Watson (NW) Kernel Regression. Unlike traditional moving averages that rely on fixed time-windows, this model uses a Gaussian Kernel to assign weights to historical data points, effectively estimating the underlying price trend with minimal lag.

Credits & Inspiration
Special thanks to Uptrick for the original logic and the machine learning concepts that inspired this implementation. This version has been refined for Pine Script V6 with added multi-filter flexibility.

What’s New:

Triple EMA Filter System: To help traders avoid "whipsaws" in sideways markets, I have integrated independent toggles for EMA 9, EMA 20, and EMA 50.

Default "No Filter" Mode: The script starts with all filters disabled, allowing for pure ML signal analysis out of the box.

Dynamic Volatility Bands: The bands are calculated based on the standard deviation of residuals (price error relative to the kernel), providing a "confidence interval" for price movement.

How to Use:

Mean Reversion: Trade the "B" (Buy) and "S" (Sell) labels when price breaks out of the upper/lower bands and the kernel flips color.

Trend Alignment: In the settings, enable the EMA filter of your choice. A Buy signal will only trigger if the price is above your selected EMA; a Sell signal will only trigger if it is below.

Customization: Use the "Visuals" section to show or hide the EMA lines and change the lookback windows to fit your timeframe.

คำจำกัดสิทธิ์ความรับผิดชอบ

ข้อมูลและบทความไม่ได้มีวัตถุประสงค์เพื่อก่อให้เกิดกิจกรรมทางการเงิน, การลงทุน, การซื้อขาย, ข้อเสนอแนะ หรือคำแนะนำประเภทอื่น ๆ ที่ให้หรือรับรองโดย TradingView อ่านเพิ่มเติมใน ข้อกำหนดการใช้งาน