Machine Learning Pivot Points (KNN)

The Machine Learning Pivot Points (KNN) indicator utilizes a K-Nearest Neighbors (KNN) algorithm and linear regression slopes to predict potential market reversals by comparing current price movement to a localized database of historical pivot signatures.

Usage

The Usage section describes how the script can be used to identify potential market turns before they are fully confirmed. The indicator identifies historical pivot highs and lows and extracts their "slope signature"—the mathematical curve of price leading into the turn.

Signal Interpretation

• Approaching Pivot High: Triggered when the current rolling slope shows high mathematical similarity to historical peak signatures. This is visualized with red glowing polyline geometry and a bearish bias in the dashboard.
• Approaching Pivot Low: Triggered when the current rolling slope matches the signature of historical troughs. This is visualized with green glowing polyline geometry and a bullish bias in the dashboard.

Dashboard Analysis

The script includes two primary tables for real-time analysis:

• Directional Helper: Displays the live KNN bias and a confidence metric. The confidence percentage represents cluster similarity; a value above 90% indicates the current price curve is nearly identical to the most significant historical reversals in the current lookback.
• Backtest Table: Provides a performance summary for the specific ticker, including the success rate of signals and the average price move following a prediction.

Details

The core of the system is the Historical Slope Function. Unlike static machine learning models, this script "trains" itself dynamically on the active chart. When a pivot (defined by the Bars Left/Right setting) is confirmed, the script calculates the linear regression slope of the preceding 20 bars and stores it in a dynamic array.

As new price data arrives, the KNN algorithm calculates the Euclidean distance between the current rolling slope and the stored historical slopes. By classifying the current movement based on its proximity to either the "High" or "Low" slope clusters, the indicator can flag a potential reversal while the price is still in motion. To optimize performance and visual clarity, the script uses v6 Polyline Geometry for recent signals and reverts to standard shapes for deep historical data.

Settings

Pivots

• Bars Left/Right: Defines the number of bars required on either side to confirm a historical pivot point for the memory bank.

Machine Learning Parameters

• KNN Clusters (K): Determines how many of the most similar past events are averaged for the prediction. Higher values make the model stricter and signals less frequent.
• Pivot Window: Sets the lookback length for the linear regression slope calculation. It is recommended to set this to at least double the "Bars Left/Right" value to capture a complete curve signature.

Helper Tables

• Show Directional Helper (Top Right): Toggles the visibility of the real-time bias and confidence table.
• Show backtest table: Toggles the visibility of the performance metrics table in the bottom right.

FAQ

How do I access Machine Learning Pivot Points (KNN)?

You can get access on the LuxAlgo Library for charting platforms like TradingView, MetaTrader (MT4/MT5), and NinjaTrader for free.

Why do some signals appear as triangles instead of glowing lines?

Because of platform limits on polyline geometry, the script uses glowing polylines for the most recent signals and reverts to standard triangles for older historical data to maintain performance and visibility across the chart.

What does the "Confidence" metric represent?

Confidence measures how much closer the current price slope is to one cluster (e.g., historical lows) compared to the other (e.g., historical highs). A higher percentage indicates a stronger mathematical match to one specific type of historical reversal.