Abstract:When analyzing the data sampled at discrete times, one encounters successive discontinuities in the trajectory of the sampled time series, even if the underlying path is continuous. On the other hand, the distinction between discontinuities caused by finite sampling of continuous stochastic process and real discontinuities in the sample path is one of the main problems. Clues like these led us to the question: Is it possible to provide a model that treats any random variation in the data set as a jump event, regardless of whether the given time series is classified as diffusion or jump-diffusion processes? To address this question, we wrote a new stochastic dynamical equation, which includes a drift term and a combination of Poisson jump processes with different distributed sizes. In this article, we first introduce this equation in its simplest form including a drift term and a jump process, and show that such a jump-drift equation is able to describe the discrete time evolution of a diffusion process. Afterwards, we extend the modeling by considering more jump processes in the equation, which can be used to model complex systems with various distributed amplitudes. At each step, we also show that all the unknown functions and parameters required for modeling can be obtained non-parametrically from the measured time series.
What problem does this paper attempt to address?
The main problem that this paper attempts to solve is to propose a new time - series modeling method, namely the "jump - jump" modeling method. The core idea of this method is to attribute any random change in the time series to a jump event, regardless of whether the time series is classified as a diffusion process or a jump - diffusion process. Specifically, the paper aims to:
1. **Distinguish between continuous processes and jump - diffusion processes**: When analyzing data sampled at discrete times, even if the underlying path is continuous, continuous discontinuities in the sample time - series trajectory will be encountered. Therefore, distinguishing whether these discontinuities are caused by the finite sampling of a continuous random process or by real discontinuities in the sample path is a major problem.
2. **Introduce a new dynamic stochastic equation**: To address the above problems, the author proposes a new dynamic stochastic equation, called the "jump - jump" equation. This equation includes a deterministic drift term and a combination of multiple Poisson jump processes with different distribution sizes. In this way, the equation can describe random changes of various distribution amplitudes in complex systems.
3. **Non - parametric estimation method**: The paper also provides a method for non - parametrically estimating all unknown functions and parameters from the measured time - series data. This enables the model to directly obtain the required parameters from the data without pre - assuming the specific form of the parameters.
### Specific content
1. **Background introduction**:
- The paper first reviews the importance of stochastic process analysis, especially its applications in fields such as physics, engineering, economics, and medicine.
- The traditional Langevin equation is used to describe continuous diffusion processes, but many actual processes contain jump discontinuities and need to be extended to jump - diffusion equations.
2. **Jump - diffusion equation**:
- The jump - diffusion equation includes a diffusion term and a random jump term and can describe random processes containing discontinuities.
- By estimating the Kramers - Moyal (KM) coefficients, the jump parameters and the diffusion coefficient can be determined from the measured data.
3. **Jump - jump modeling**:
- The paper proposes a new dynamic stochastic equation, called the "jump - jump" equation, whose general form is as follows:
\[
dx(t) = D^{(1)}(x,t)dt+\sum_{i = 1}^N\xi_idJ_i(t)
\]
where \(D^{(1)}(x,t)\) represents the deterministic drift term, \(J_i(t)\) is a Poisson jump process, and \(\xi_i\) is the size of the jump.
4. **Non - parametric estimation method**:
- The paper provides a method for non - parametrically estimating all unknown functions and parameters from the measured time - series data. The specific steps include:
- Estimating the first - order KM coefficient \(M^{(1)}(x,t)\) to determine the drift function \(D^{(1)}(x,t)\).
- Estimating the jump amplitude \(\sigma^2_{\xi}(x,t)\) and the jump rate \(\lambda(x,t)\) through the high - order KM coefficients \(M^{(n)}(x,t)\).
5. **Verification and application**:
- The validity of the model is verified by generating synthetic time series and using the proposed jump - jump equation for reconstruction.
- The results show that the time series of the original diffusion process can be well reconstructed using the jump - jump equation, and the estimated parameters are very close to the preset values.
In summary, the paper proposes a new time - series modeling method that can handle complex systems containing jump discontinuities and provides a method for non - parametrically estimating model parameters from data. This provides a new tool for understanding and predicting various actual processes.