An orchestra without a conductor: the musicians take their seats, tune their instruments, and start playing in harmony because each one hears the common rhythm. That's how TreeDOX works: a method that predicts chaotic time series without requiring a human to choose parameters or fine-tune anything.

The secret of TreeDOX lies in explicit memory. Windows are formed from the time series: each next value is predicted from several previous ones. Then Extra Trees Regression is used twice: first to estimate feature importance, then to train on a reduced set. The window size and the number of features are determined automatically via mutual information and statistical tests. No knobs to turn.

On the Hénon map and the Lorenz system, TreeDOX reproduced chaotic attractors with correlation dimension matching to the fourth decimal place. For the Kuramoto–Sivashinsky equation with 64 spatial points, the forecast held for about 12.5 Lyapunov times. On the real Southern Oscillation Index, the method showed lower error and higher mutual information than LSTM and NG-RC. Training time was 0.63 seconds versus 200 seconds for neural networks with Bayesian optimization.

Previously, predicting chaos was an art of tuning: an expert adjusted hyperparameters for each task, spending hours and computational resources. TreeDOX shows that an ensemble of trees with automatic delay selection can compete without a single hand-tuned parameter. Chaos prediction becomes a routine operation, accessible to anyone who can call a ready-made library.

The orchestra played the concert without a conductor, and the hall applauds. Trees that tune themselves predict climate indices, biomedical signals, and financial series. The method can still evolve: GPU parallelization would speed up training by tens of times, and adaptation to non-stationary data would allow the delay window to change over time. But already now, TreeDOX makes chaos prediction as simple as calling a ready-made library.