Accelerating Parametric Inference in Healthcare Data Streams Using Symbolic Histogram Likelihoods
DOI:
https://doi.org/10.19139/soic-2310-5070-4242Keywords:
Symbolic Data Analysis, Maximum Likelihood Estimation, Histogram-Valued Data, Streaming Inference, Computational Statistics, Healthcare AnalyticsAbstract
Maximum likelihood estimation (MLE) is a fundamental tool for parametric inference, but likelihood evaluations become computationally prohibitive for very large samples and for moving-window analyses of streaming data. We develop a symbolic maximum likelihood framework for univariate continuous data in which the raw sample is replaced by a histogram-valued symbol. The resulting objective is a multinomial log-likelihood whose cost per evaluation depends on the number of histogram boundary points $B$, rather than directly on the sample size $n$. Utilising the symbolic data analysis (SDA) framework, we demonstrate the substantial computational efficiency improvements of MLE via histogram-valued symbols where the raw sample is replaced by a compressed symbolic representation. The evaluation cost of the resulting multinomial log-likelihood objective depends on the number of histogram boundary points $B$, rather than directly on the sample size $n$. We show that the symbolic likelihood is exact for the aggregated histogram counts and that, under increasingly fine partitions, it approximates the classical log-likelihood based on the individual observations up to an additive bin-width term that does not depend on the unknown parameter. We present a numerically stable implementation, compare equal-width and equal-probability binning, and describe a sliding-window workflow with constant-time count updates and $\mathcal{O}(B)$ likelihood evaluation per optimisation step. The simulation benchmark covers Exponential, Normal, Gamma, Weibull, and Lognormal models for sample sizes up to $10^6$ and histogram resolutions from $B-1=10$ to $B-1=500$ bins. In the largest-sample setting, the equal-width symbolic estimator achieves speed-ups ranging from 2.64-fold to 90.05-fold across the reported configurations. Accuracy improves steadily with finer partitions: at $n=10^6$ and $B-1=500$, the relative RMSE of the equal-width estimator ranges from 1.000 to 1.042 across the five families, while equal-probability bins are especially helpful for strongly skewed models at coarse resolution. A public-data illustration on the MIMIC-III heart-rate stream shows that the symbolic approach remains practical in window-wise fitting. For the preferred model, the median per-window speed-up is 5-fold, with close agreement between classical and symbolic parameter estimates. These findings position symbolic MLE as a practical likelihood-based compression strategy for large-scale simulation, estimation, and near-real-time monitoring.Downloads
Published
2026-11-16
How to Cite
Jamaluddin, A. H., Anam, S., Mohd Nasir, M. J., Tri Rian Dani, A., & Chamidah, N. (2026). Accelerating Parametric Inference in Healthcare Data Streams Using Symbolic Histogram Likelihoods. Statistics, Optimization & Information Computing. https://doi.org/10.19139/soic-2310-5070-4242
License
Copyright (c) 2026 Ahmad Hakiim Jamaluddin, Syaiful Anam, Muhammad Jaffri Mohd Nasir, Andrea Tri Rian Dani, Nur Chamidah

This work is licensed under a Creative Commons Attribution 4.0 International License.
Authors who publish with this journal agree to the following terms:
- Authors retain copyright and grant the journal right of first publication with the work simultaneously licensed under a Creative Commons Attribution License that allows others to share the work with an acknowledgement of the work's authorship and initial publication in this journal.
- Authors are able to enter into separate, additional contractual arrangements for the non-exclusive distribution of the journal's published version of the work (e.g., post it to an institutional repository or publish it in a book), with an acknowledgement of its initial publication in this journal.
- Authors are permitted and encouraged to post their work online (e.g., in institutional repositories or on their website) prior to and during the submission process, as it can lead to productive exchanges, as well as earlier and greater citation of published work (See The Effect of Open Access).