Training and Checking a Neural Trend Indicator Based on Kaufman’s Adaptive Average

Summary

This document outlines an attempt to build a trend indicator using a neural network and Kaufman’s adaptive moving average as a reference. The described initialization phase trains the network on bars numbered 200 through 300 by selecting weight coefficients. It then calculates values across the history from bar 300 toward bar zero, allowing the outputs to be compared with the standard indicator. The proposed diagnostic is whether the network makes fewer errors within the interval used for learning than outside it.

If errors increase sharply on bars outside the training interval, the author treats that as a sign of overfitting and suggests reducing the number of input neurons or changing the model. The note gives a conceptual training and validation check, not a full evaluation: it reports no error measurements, trading results, or details sufficient to reproduce the network. Its focus is indicator construction and model fitting rather than a complete trading strategy.

Key ideas

• The proposed neural trend indicator uses Kaufman’s adaptive moving average as a reference.
• The network is fitted on a historical interval and its outputs are compared with the reference indicator.
• A sharp rise in errors outside the training interval is treated as evidence of overfitting.
• The suggested responses are to reduce input neurons or change the model.
• The note provides no quantified accuracy or trading performance evidence.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.