ML Lorentzian Classification PRO is a technical analysis indicator that applies a machine learning approach using an approximate k-nearest-neighbours classifier with Lorentzian Distance to measure similarity between market states. This distance metric reduces the impact of outliers and volatility shocks, maintaining comparability across different market regimes. The model scores each bar against chronologically spaced neighbors and confirms predictions via Nadaraya-Watson kernel regression before issuing BUY/SELL signals.

Key features include:
- Adjustable neighbors count (1–100) with a minimum 4-bar spacing to avoid oversampling.
- Up to five configurable feature inputs selectable from normalized RSI, WaveTrend, CCI, or ADX, each with independent length and smoothing parameters.
- Multiple filters including volatility (ATR-based), regime (Kalman-style adaptive slope), ADX threshold, EMA/SMA trend filters, and kernel direction filter.
- Kernel regression using rational quadratic and Gaussian kernels with adjustable parameters and enhanced smoothing options.
- Visual BUY/SELL alerts with ATR-scaled positioning, customizable labels, optional exit markers, and sound, popup, and email alerts triggered on bar close or intrabar.
- A collapsible on-chart panel displaying live prediction values, confidence, kernel and trend states, filter results, and trade statistics (win rate, trade count, win/loss ratio).
- Bar coloring by prediction strength with adjustable gradient or solid modes and visual range settings.

The indicator evaluates forming bars in real time, with values settling at bar close. It supports detailed feature and filter tuning but does not perform backtesting or account for trading costs. Recommended for use on 5-minute to 4-hour timeframes with default settings including eight neighbors and all five features enabled.