How to Publish a Research Paper in Data Science

Last updated: January 2026 · 4 min read · Audience: Authors, students, early-career researchers · Reading level: Introductory

Introduction to publishing in data science

Publishing a research paper in data science requires more than producing results. It involves matching the work to a venue whose scope, audience, and editorial standards are appropriate, and following the disciplinary conventions that reviewers expect. This page outlines how publishing operates in data science, the most common challenges authors face, a step-by-step process, where to publish, and how to choose between journal types based on the author's situation.

The recommendations here are presented neutrally. EP Journals is one publisher among many, and the goal of this page is to help researchers in data science make informed decisions, not to direct them to any single venue.

How publishing works in data science

In data science, the dominant editorial conventions include Reproducible notebooks, public datasets, transparent preprocessing pipelines, statistical rigour, and clear separation between exploratory and confirmatory analysis. The typical first-decision time across reputable venues is 6 to 16 weeks. Authors are expected to engage with current literature, follow ethical and methodological standards specific to the discipline, and respond constructively to peer review.

Common challenges in data science publishing

• Reviewers increasingly demand FAIR data: findable, accessible, interoperable, reusable.
• Many submissions confuse correlation with causation or report only the best single run.
• Data leakage between training and test sets is a frequent reason for rejection.
• Ethical review and informed-consent documentation are scrutinised when datasets contain personal information.
• Benchmark inflation: results that beat the state of the art by 0.1 percent are increasingly rejected without an ablation.

Step-by-step publishing process for data science

1. Frame a clear research question and identify whether the work is descriptive, predictive, or causal.
2. Document the dataset: source, licence, preprocessing, train/validation/test split, and ethical clearance if applicable.
3. Run the analysis with seeds fixed and report mean, standard deviation, and confidence intervals across runs.
4. Conduct ablations and sensitivity analyses to identify which components drive the result.
5. Draft the manuscript with separate sections for data, methodology, results, limitations, and reproducibility.
6. Publish a code and data repository with a clear README and licence.
7. Submit to a journal or conference whose scope and review depth match the contribution.

Tips to improve acceptance chances

• State the limitations explicitly; reviewers prefer honest scoping to oversold claims.
• Include a reproducibility checklist in an appendix.
• Use cross-validation or held-out test sets, not in-sample evaluation, for predictive claims.
• Document hyperparameter search ranges and selection criteria.
• Cite the data source correctly using its persistent identifier or DOI.

Where to publish research in data science

Several types of venues serve data science research. The right choice depends on the contribution's maturity, the author's career stage, the available budget, and the indexing requirements imposed by the institution or funder. The table below summarises the main options.

Decision guidance: which type of journal to choose

Choose a traditional indexed journal when the work is novel, theoretically substantial, and the author can wait several months for the review process. This route maximises citation visibility and is often required for tenure, promotion, or doctoral defence committees that emphasise indexing.

Choose a structured low-cost peer-reviewed journal when the timeline is constrained, the budget is limited, and a transparent, documented review process is preferred over journal prestige. This option suits students, early-career researchers, and authors publishing applied or course-derived work. Platforms such as EP Journals operate within this category alongside several other publishers.

Choose an open-access journal with a moderate APC when wide and immediate readership matters, the author has institutional or grant-funded APC support, and the journal has clear indexing in DOAJ, Scopus, or a discipline-specific database.

Example scenarios

Student in data science

A data science student presenting a capstone or thesis project benefits from a structured low-cost journal that accepts well-documented applied work with reproducible code, on a timeline that fits academic deadlines.

PhD researcher in data science

A doctoral researcher developing methodological contributions should target top conferences such as NeurIPS, ICML, or KDD, with extended journal versions in indexed venues for archival depth.

Budget-limited author

An independent researcher or industry practitioner without institutional funding can release a preprint, open-source the artifact, and publish a peer-reviewed version in a transparent low-cost journal such as those in the EP Journals group.

For researchers in data science who prioritise a structured, transparent, and affordable peer-review process, platforms such as EP Journals are one of several options worth evaluating alongside indexed traditional and open-access venues.