Abstract:The generalized Langevin equation (GLE), derived by projection from a general many-body Hamiltonian, exactly describes the dynamics of an arbitrary coarse-grained variable in a complex environment. However, analysis and prediction of real-world data with the GLE is hampered by slow transient or seasonal data components and time-discretization effects. Machine-learning (ML) techniques work but are computer-resource demanding and difficult to interpret. We show that by convolution filtering, time-series data decompose into fast, transient and seasonal components that each obey Hamiltonian dynamics and, thus, can be separately analyzed by projection techniques. We introduce methods to extract all GLE parameters from highly discretized time-series data and to forecast future data including the environmental stochasticity. For daily-resolved weather data, our analysis reveals non-Markovian memory that decays over a few days. Our prediction accuracy is comparable to commercial (<a class="link-external link-http" href="http://weather.com" rel="external noopener nofollow">this http URL</a>) and ML long short-term memory (LSTM) methods at a reduced computational cost by a factor of $10^2-10^3$ compared to LSTM. For financial data, memory is very short-ranged and the dynamics effectively is Markovian, in agreement with the efficient-market hypothesis; consequently, models simpler than the GLE are sufficient. Our GLE framework is an efficient and interpretable method for the analysis and prediction of complex time-series data.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to use the Generalized Langevin Equation (GLE) to analyze and predict discrete - time - series data, especially weather and financial data. Specifically, the paper focuses on the following aspects: 1. **Multiscale Dynamic Coupling**: Solve the multiscale dynamic coupling problem between short - time stochastic effects and long - time transient or seasonal effects. 2. **Non - Gaussian Correlation**: Deal with the non - Gaussian correlations existing in many non - trivial systems. 3. **Time - Discretization Effect**: Overcome the influence of time - discretization on the application of GLE. 4. **Non - equilibrium Effect**: Handle the slowly relaxing non - equilibrium effects, which make many tools in statistical mechanics ineffective. To address these problems, the author proposes a method that combines convolution filtering and projection techniques to decompose time - series data into rapidly changing, transient and seasonal components, and analyze them with Hamiltonian dynamics respectively. This method can not only extract all GLE parameters, but also predict future data, including the influence of environmental randomness. ### Main Contributions - **Efficiency**: Compared with traditional machine - learning methods (such as LSTM), this method significantly reduces the demand for computational resources while maintaining high prediction accuracy. - **Interpretability**: All model parameters are interpretable, which makes this method more advantageous in natural - science research and high - risk social - political decision - making. - **Scope of Application**: This method is applicable to the analysis and prediction of various complex time - series data, including weather and financial data. ### Application Examples - **Weather Data**: By analyzing the daily maximum temperature data in Berlin through GLE, the non - Markovian memory effect is revealed, and the accuracy and efficiency of this method in weather prediction are demonstrated. - **Financial Data**: By analyzing Bitcoin price, S&P 500 index and Yen / US Dollar exchange rate data through GLE, it is found that the memory effect of financial data is very short, which is in line with the efficient - market hypothesis. In conclusion, this paper provides an efficient and interpretable method to analyze and predict complex time - series data by introducing a new GLE framework, especially outstanding in the applications in the weather and financial fields.

Predictability Analysis and Prediction of Discrete Weather and Financial Time-Series Data with a Hamiltonian-Based Filter-Projection Approach

Efficient Bayesian estimation of the generalized Langevin equation from data

Temporal coarse-graining and elimination of slow dynamics with the generalized Langevin equation for time-filtered observables

Predictive Patterns and Market Efficiency: A Deep Learning Approach to Financial Time Series Forecasting

Multi-step prediction of chaotic time-series with intermittent failures based on the generalized nonlinear filtering methods

Temporal Data Meets LLM -- Explainable Financial Time Series Forecasting

Financial Time Series Analysis and Forecasting with HHT Feature Generation and Machine Learning

Prediction of financial time series using LSTM and data denoising methods

Real-time Forecasting of Time Series in Financial Markets Using Sequentially Trained Many-to-one LSTMs

Text2TimeSeries: Enhancing Financial Forecasting through Time Series Prediction Updates with Event-Driven Insights from Large Language Models

Statistical Inference of 1D Persistent Nonlinear Time Series and Application to Predictions

Bayesian forecasting and scalable multivariate volatility analysis using simultaneous graphical dynamic models

Nonlinear Autoregressive Neural Network and Extended Kalman Filters for Prediction of Financial Time Series

Disentangled Interpretable Representation for Efficient Long-term Time Series Forecasting

Deep Video Prediction for Time Series Forecasting

Filtered not Mixed: Stochastic Filtering-Based Online Gating for Mixture of Large Language Models

FiLM: Frequency improved Legendre Memory Model for Long-term Time Series Forecasting

Ensemble of temporal Transformers for financial time series

Data-driven Prediction of General Hamiltonian Dynamics via Learning Exactly-Symplectic Maps

Functional Latent Dynamics for Irregularly Sampled Time Series Forecasting

A simple statistical approach to prediction in open high dimensional chaotic systems