Discussing the article: "Volumetric neural network analysis as a key to future trends"

[MetaQuotes]

Check out the new article: Volumetric neural network analysis as a key to future trends.

The article explores the possibility of improving price forecasting based on trading volume analysis by integrating technical analysis principles with LSTM neural network architecture. Particular attention is paid to the detection and interpretation of anomalous volumes, the use of clustering and the creation of features based on volumes and their definition in the context of machine learning.

Our first graph is the most common chart of Sber prices with the obtained model signals. We also supplement the signals by highlighting those candles where there are abnormal volumes. This helps us understand the moments when the system reads the market perfectly, like an open book.

The second graph is the predicted return. Here we can clearly see that before powerful movements of the chosen asset quotes, a very powerful series of forecasts often begins. So this suggests the idea of considering creating a system just on this particular observation. Of course, the number of transactions will fall, but we are not chasing quantity, we are striving for quality, aren’t we?

The third chart is the cumulative return with drawdowns highlighted.

Author: Yevgeniy Koshtenko

[Evgeniy Chernish]

"Once the signals are generated, they are applied to prices with a 24 period forward shift. This allows to take into account the lag between making a trading decision and its realisation."

So we enter the market 24 bars after the neural network generates a signal ?

Or we enter on the next bar and hold a position on this signal for 24 bars ?