Machine Learning: Trend Lines

The Machine Learning: Trend Lines indicator applies K-Nearest Neighbors (KNN) logic and Rational Quadratic kernels to traditional trend line calculations to identify non-standard support and resistance zones. It identifies potential price formations like channels, wedges, and pennants by calculating anchors based on machine learning models rather than simple price extremes.

Usage

The Usage section describes how the script can be used to identify market structures. The indicator automatically plots upward and downward trend lines based on processed price data. These lines act as dynamic support and resistance levels.

• Support and Resistance: While green lines often indicate upward trajectories and red lines indicate downward ones, their role as support or resistance depends on the price position relative to the line. A line may shift roles if the price crosses through it.
• Formations: When two trend lines appear parallel, they define a trading channel. Converging lines may indicate wedges or pennants, signaling potential momentum changes or volatility breakouts.
• Visibility: Trend lines are drawn from their calculated start points. If the start point is not visible on the current chart view (e.g., if it originated far in the past), the line may not display. Users should zoom out to ensure historical anchors are captured.
• Customization: The indicator is highly sensitive to input lengths. Changing the "Fast" or "Slow" ML data lengths will alter the angle and frequency of the lines, requiring users to fine-tune settings based on the specific asset and timeframe.

Details

This script replaces traditional high/low pivots with machine learning sources. It uses a Rational Quadratic kernel to smooth price data, creating a more "rational" base for trend identification.

When KNN sorting is enabled, the script calculates the distance between fast and slow ML data points. It then clusters these distances to find the K-nearest neighbors, averaging these values to determine the most significant pivot points. By using an exponential data average for these neighbors, the script can produce longer-lasting trend lines that reflect historical significance rather than just immediate price action.

Settings

• Use Machine Learning Sources: When enabled, the script uses Rational Quadratics to determine price pivots. If disabled, it defaults to traditional High and Low prices.
• Use KNN Distance Sorting: Toggles the K-Nearest Neighbor logic for data sorting. Disabling this reverts the trend line logic to a more traditional calculation.
• Use Exponential Data Average: Uses a custom exponential average of the KNN results instead of a simple average, often resulting in more stable, long-term trend lines.
• Machine Learning Length: Determines the lookback period for the machine learning memory. High values process more data but may lead to over-smoothing.
• K-Nearest Neighbour (KNN) Length: Sets the number of neighbors used in distance clustering.
• Fast ML Data Length: Controls the short-term processing speed for ML data; values between 3 and 7 are generally recommended.
• Slow ML Data Length: Controls the long-term processing speed for ML data; values between 20 and 50 are generally recommended.

FAQ

How do I adjust the sensitivity of the trend lines? You can adjust the "Fast ML Data Length" and "Slow ML Data Length" settings. Smaller values will create more reactive, frequent lines, while larger values will produce fewer, more significant structural lines.

Why are some trend lines not appearing on my chart? Trend lines require their origin point to be within the visible range or memory of the chart. If a trend line was initiated a long time ago, try zooming out or increasing the bars on the chart.

How can I access the Machine Learning: Trend Lines indicator? You can get access on the LuxAlgo Library for charting platforms like TradingView, MetaTrader (MT4/MT5), and NinjaTrader for free.