seasonal adjustment

This study addresses the lack of systematic evaluation of stationarity-inducing transformations across diverse non-stationary time series. The authors construct synthetic datasets encompassing trend, seasonality, and heteroskedasticity, complemented by real-world airport passenger flow data, and conduct 3,528 controlled experiments evaluating 14 transformation methods across seven forecasting models and three prediction horizons. Innovatively, stationarity is assessed via consensus from ten statistical tests, and mediation analysis elucidates underlying mechanisms. Results challenge the common assumption that transformations universally improve forecasts: matched transformations enhance accuracy in only 18% of cases; log or Box–Cox transformations are effective for heteroskedastic data (60–65% of cases); and differencing consistently degrades performance on linear-trend series.

Evaluating interpretability in time series classification (TSC) remains challenging due to the lack of principled metrics for assessing simplification-based explanations. Method: We propose a novel complexity–fidelity dual-dimensional metric—formally defining and quantifying the trade-off between simplification fidelity and interpretability—and conduct systematic benchmarking across mainstream TSC models (ROCKET, TSF, InceptionTime) and heterogeneous UCR/UEA datasets using piecewise linear and symbolic simplification techniques. Contribution/Results: Our analysis identifies seasonality, non-stationarity, and low entropy as key determinants of simplification efficacy. Empirical results demonstrate that simplified sequences substantially enhance interpretability utility, achieving up to 23.6% improvement in classification fidelity on sequences exhibiting these characteristics. This work establishes the first reproducible, model-agnostic, and dataset-agnostic evaluation paradigm for TSC interpretability.

This study addresses the challenge of degraded data quality—specifically outliers and missing values—in long-horizon time series forecasting, which critically undermines model robustness. We establish a unified evaluation framework to systematically benchmark mainstream models—including LSTM, Prophet, XGBoost, and Random Forest—under three realistic data conditions: complete, noisy (outlier-contaminated), and incomplete (missing-value) sequences, with ARIMA as the baseline. Methodologically, we employ sliding-window modeling, multi-step rolling prediction, and adaptive imputation for preprocessing. Our key contributions include: (i) a novel, interpretable algorithm selection guideline grounded in data characteristics and forecasting requirements; and (ii) empirical findings demonstrating that XGBoost reduces average MAE by 23% under noise, while Prophet exhibits superior stability for long-term trend forecasting. The results provide reproducible, principled guidance for industrial-scale time series modeling.

This paper addresses the limited predictive accuracy of multi-model ensembling in complex seasonal time series forecasting. We propose a meta-learning-based dynamic weighted ensemble framework. Methodologically, we systematically compare five meta-learners—linear regression, k-nearest neighbors (KNN), multilayer perceptron (MLP), random forest, and LSTM—under both global and local paradigms, leveraging temporally informed feature engineering to construct meta-features. Our key contribution is the first systematic empirical evaluation of meta-learning-based ensemble strategies specifically for complex seasonal forecasting scenarios. Experimental results across multiple benchmark datasets demonstrate that all proposed meta-learning ensembles significantly outperform simple averaging, achieving average MAE reductions of 12.7%–23.4%. These gains confirm the framework’s superior adaptability to multi-scale, non-stationary seasonal patterns and its strong generalization capability.

This paper addresses the challenge of simultaneously achieving statistical fidelity and controllable privacy in synthetic data generation. We propose a statistical mapping method based on empirical conditional distributions, which avoids assumptions about underlying data distributions and systematically models conditional dependencies among features to construct an invertible statistical mapping. Privacy is explicitly controlled via tunable parameters that govern information leakage. Our key contribution is the first systematic application of empirical conditional distributions to correlation-preserving synthetic data generation, significantly improving fidelity of higher-order statistics—particularly Pearson correlation matrices. Experiments on synthetic benchmarks, artificial examples, and real-world household energy consumption data from Madeira Island demonstrate that generated data closely preserve the original correlation structure (mean absolute correlation error < 0.05) while providing a well-defined privacy–utility trade-off mechanism.

This study addresses the challenges of modeling and analyzing complex time series arising in astrophysics, meteorology, finance, and other domains by systematically integrating classical statistical methods—such as ARIMA, exponential smoothing, and state-space models—with modern machine learning techniques, including tree-based ensembles, hidden Markov models, Gaussian processes, and deep learning architectures like RNNs, CNNs, and Transformers. By distilling cross-disciplinary modeling principles, the work establishes a unified framework that combines theoretical rigor with practical guidance, offering researchers a comprehensive and extensible toolkit for time series analysis. This approach significantly enhances the capacity to handle temporal data across diverse scientific and applied contexts.

This study addresses the limitations of rule-based model selection in time series forecasting, which often fails due to variations in data-generating mechanisms. The authors propose a descriptor framework grounded in measurable characteristics—such as trend strength, seasonality, noise level, and temporal dependence—and systematically evaluate the efficacy of static selection rules across diverse real-world datasets. Their analysis reveals, for the first time, that static descriptors are insufficient for reliably predicting model performance: model selection proves highly context-dependent and unstable. Notably, under noisy conditions or mixed generative mechanisms, recommended models frequently diverge substantially from the true best-performing ones, and model rankings exhibit pronounced sensitivity to both data characteristics and forecast horizons.

This work addresses a key ambiguity in traditional hierarchical forecasting: whether performance gains from reconciliation stem from aggregation constraints or the fusion of information across levels. To disentangle these effects, the authors propose IComb (Information-fusion-based Combination), a novel reconciliation method that explicitly models and leverages the differences in information sets underlying base forecasts at each hierarchy level. By incorporating these information disparities within a penalized regression framework—readily implementable with existing software—IComb enhances forecast accuracy while maintaining coherence. Extensive experiments on both simulated and real-world datasets demonstrate that IComb significantly outperforms conventional reconciliation approaches, thereby confirming the critical role of cross-level information fusion in hierarchical forecasting.

Existing time series pretraining methods struggle to generalize effectively across multiple datasets due to discrepancies in input length and channel dimensions. This work proposes ADAPT, a novel pretraining paradigm that enables unified modeling across 162 time series classification datasets by adaptively aligning the physical attributes of time series data. Integrating self-supervised learning with a hybrid batch training strategy, ADAPT overcomes the generalization limitations inherent in conventional many-to-one pretraining approaches. The method achieves state-of-the-art performance on multiple benchmarks, establishing a foundational framework for developing general-purpose foundation models for time series analysis.