NTIDE framework

NTIDE is a research implementation of nonlocal trend-inhibition delay dynamics and physics-informed stability indicators. This framework-only distribution is designed for user-supplied data and deliberately contains no empirical dataset, crisis labels, historical processed artifacts or paper result tables.

Installation

python -m pip install ntide

For a local checkout:

python -m pip install -e .

Minimal use

import pandas as pd
from ntide import CrisisDetector

data = pd.read_csv("market_data.csv", index_col="date", parse_dates=True)
detector = CrisisDetector(identification_window=160)
components = detector.fit(data).predict_components(data)
print(components.tail())

The input must have a unique, increasing DatetimeIndex and a positive numeric volume column. Warm-up outputs are NaN until enough past observations are available.

Included modules

• ntide.core: nonlocal kernels, NTIDE field solver and delay stability margin.
• ntide.data: causal activity-field preprocessing for user-provided volume.
• ntide.models: online parameter identification and component scores.
• ntide.evaluation: generic AUROC/AUPRC, RGA and SAFE utilities.

Read SCIENTIFIC_SCOPE.md before making empirical or trading claims. The default hazard is a research index, not a calibrated crisis probability.

Article

Zhang, H. (2026). Early Warning of Financial Crises Through Critical Field Dynamics: A Nonlocal Trend-Inhibition Delay Equation Framework. Physica A, 682, 131161. https://doi.org/10.1016/j.physa.2025.131161

The article's lightweight computational reproduction is distributed separately in ntide-physica-a-reproduction.