EUVIMEDEUVIMED entdecken
Crossref

An Analytical Framework for Reconstructing Missing Time-Series Data Using Interpretable AI Methods

Sneh Patel, Dr. Darshanaben Pandya

SPU - Journal of Science, Technology and Management Research · 01.07.2026 · 18-25

DOI: 10.63766/spujstmr.26.000069 ↗

Kurz gefasst

Fachartikel in der Zeitschrift SPU - Journal of Science, Technology and Management Research, online erschienen am 01.07.2026. Verfasst von Sneh Patel und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Large AI systems and foundation models rely on continuous and consistent numeric data. In real life, missing values are common in the time-series data because of gaps in reporting, measurement failures, or archival errors. If these are not han dled well, they have the potential to alter the distribution of the data, bias learning, and hurt model performance. In this paper, we presented a simple, intuitive, deterministic interpolation approach to fill gaps in long-term numeric time series. Unlike the imputation approach, which has latent, black-box-type imputation, this uses a step-by-step mathematical process; as such, trac ing and verifying statistically are easily achieved. We tested the method on a historical Standard Yield (SY) dataset from 1950 2009, to which we artificially introduce missing values and then reconstruct. We utilize a comprehensive set of statistical tests for the descriptive statistics, variance and covariance comparisons, and one-way ANOVA to demonstrate that in average value, the dispersion, and the time-varying process, the original data is closely matched with the recovered series. This demonstrates that classic, interpretable reconstruction methods are still relevant in a modern AI pipeline, during the preprocessing stage of foundation models, which demands data quality, stability, and explain ability.

Zitieren

APA

Patel, S., & Pandya, D. D. (2026). An Analytical Framework for Reconstructing Missing Time-Series Data Using Interpretable AI Methods. SPU - Journal of Science, Technology and Management Research, 18-25. https://doi.org/10.63766/spujstmr.26.000069

Vancouver

Patel S, Pandya DD. An Analytical Framework for Reconstructing Missing Time-Series Data Using Interpretable AI Methods. SPU - Journal of Science, Technology and Management Research. 2026:18-25. doi:10.63766/spujstmr.26.000069

BibTeX
@article{Patel2026,
  author = {Patel, Sneh and Pandya, Dr. Darshanaben},
  title = {An Analytical Framework for Reconstructing Missing Time-Series Data Using Interpretable AI Methods},
  journal = {SPU - Journal of Science, Technology and Management Research},
  year = {2026},
  pages = {18--25},
  doi = {10.63766/spujstmr.26.000069}
}
Kostenlos starten →