Rejected from Expert Systems with Applications? The 6 Best Journals to Submit Next

Treat the next submission as a ladder keyed to the reason for rejection, not a generic step down in prestige.

Tier 1, scope or application-substance desk rejection. The science is sound, but ESWA decided the work read as method-first rather than application-first. Move sideways to a same-tier Elsevier sibling: Knowledge-Based Systems if the contribution is knowledge-driven or decision-support, Applied Soft Computing if it is fuzzy, evolutionary, or hybrid soft computing, or Engineering Applications of Artificial Intelligence if the application lives in an engineering or industrial system. These are first-choice next targets because the bar is comparable and reviewer reports can travel.

Tier 2, the work is really a learning-method contribution. If the paper is genuinely about a neural architecture or learning algorithm and the "application" was a thin wrapper, Neurocomputing or Pattern Recognition is the honest venue. Do not try to re-dress a methods paper as an application paper for another applied venue; the next editor will see the same seam.

Tier 3, you need a decision fast or the contribution is incremental. IEEE Access publishes technically sound work on a roughly 4-to-6-week submission-to-publication timeline at a much higher acceptance rate. It does not gate on novelty the way ESWA does, so it is the right step down when the paper is correct and useful but not surprising.

The Elsevier Article Transfer Service. Because ESWA is an Elsevier journal, a desk rejection can come with a transfer offer to Knowledge-Based Systems, Applied Soft Computing, Engineering Applications of Artificial Intelligence, Pattern Recognition, or Neurocomputing. The transfer carries your files and any reviewer reports without a fresh upload. Accept it when the suggested venue actually fits your contribution; a transfer is a routing suggestion, not an acceptance, and the receiving editor still screens the paper.

The table below maps the rejection reason to the next target so you can route on cause rather than prestige.

In our pre-submission review work with Expert Systems with Applications submissions, the rejections we see most often are not about broken science. They are about a mismatch between what the manuscript claims and what the visible package proves, and ESWA's 5-day desk screen catches that mismatch fast. Three named patterns account for most of the returns, and each one is checkable against your own draft before you resubmit.

Application invisibility: the abstract claims a real setting, but Figure 1 is a model-architecture diagram. Across our ESWA pre-submission reviews, the single most common desk-rejection trigger is an abstract that opens with a generic "we apply deep learning to X" while Figure 1 is a CNN block diagram, transformer encoder, or attention schematic instead of the application context.

ESWA editors read the abstract-and-Figure-1 pair to decide whether the application is real before the methodology takes over. When Figure 1 is an architecture diagram with no application-domain figure ahead of it, the paper reads as a methods submission that wandered into an application journal, and it gets returned toward Pattern Recognition, Neurocomputing, or a methods venue.

The fix is concrete: make Figure 1 the decision setting, data sample, or operational pipeline, and demote the architecture diagram to Figure 2 or supplementary material.

A benchmark and baseline comparison anchored on stale general-AI numbers. In our review work, ESWA-targeted manuscripts frequently build the comparison table from 2020 to 2022 baselines lifted from the method's original publication trajectory, then skip the recent application-domain methods that have appeared in Knowledge-Based Systems, Applied Soft Computing, Engineering Applications of Artificial Intelligence, or Pattern Recognition in the last 18 months.

ESWA reviewers are application-domain specialists who track their own literature, so a table missing current sibling-venue baselines looks strategically incomplete and invites a reject-or-major-revision. The testable fix before resubmission: add at least two baselines from sibling applied-AI venues published recently, and state in the methods why each is the right comparator for your application claim.

If an ablation is missing for the component you are claiming credit for, add it now; reviewers at every journal on this list will ask for it.

Reproducibility stated as "code available on request." The third pattern we see across ESWA submissions is a data and code availability statement that promises access "on request" or "from the corresponding author" instead of naming a public repository with a permanent identifier. ESWA's editorial culture increasingly favors reproducible applied-AI work, because performance claims are only checkable when the code, trained weights, and evaluation splits are inspectable.

Manuscripts that describe their splits in prose but provide nothing downloadable draw extra scrutiny on whether the reported statistical results hold, and that scrutiny often costs a full revision round. Before resubmitting, deposit code, evaluation splits, and model settings to a public repository (GitHub, Zenodo, or OSF) with a DOI or version tag, and reference that identifier in the data availability statement.

Statistical significance asserted without testing. A quieter pattern worth a self-check: applied-AI papers that report a higher mean accuracy or F1 than the baseline but never run a significance test across seeds or folds. A 0.3-point gain with no variance reported and no paired test is not a result a careful reviewer will accept, at ESWA or its siblings. Report the spread across runs and a significance test before you resubmit.

Do not just blast the same PDF down the ladder. ESWA's desk screen is fast, but the sibling journals are not less rigorous; KBS, Applied Soft Computing, and EAAI sit at a slightly higher impact-factor tier and use the same pool of application-domain reviewers. A paper that was returned for thin application substance or a weak baseline set will be returned again unless the manuscript actually changes.

Separate the two cases honestly. A desk rejection on scope needs a new cover letter and a re-pointed framing, not new experiments; you can move within days. A post-review rejection that named a missing ablation, an unconvincing baseline comparison, or unverifiable reproducibility needs real work first, because the same reviewers read for the next journal.

And if the rejection said the application was a thin wrapper on a methods contribution, the move-on answer may be to stop fighting it and submit to a methods venue instead of forcing it into another applied journal.