Knowledge-Based Systems Submission Guide: How to Submit to KBS (Elsevier)

Knowledge-Based Systems publishes original research articles, review articles, and short communications. The format you choose drives the expectations that apply, but every type is held to the same scope and reproducibility bar.

Original research articles are the main category. There is no rigid page cap, so length is governed by completeness and clarity rather than a hard limit, which means an over-long paper is judged on whether every section earns its space. A typical KBS research article runs in the 6,000 to 9,000 word range once methods, experiments, and ablations are included, though the journal does not enforce a fixed word count and judges on substance.

Review articles are expected to be genuine syntheses that organize a knowledge-based or reasoning subfield, not annotated bibliographies. A survey that lists methods without a taxonomy, a comparison framework, and an argued research agenda reads as incomplete for this journal.

Short communications carry condensed, time-sensitive results and are held to the same novelty and reproducibility standard in a shorter form.

For declarations, prepare the full required-artifacts set before you open the portal:

• a complete CRediT author contributions statement for every author
• a conflicts of interest declaration (Elsevier's declaration of competing interests)
• a data availability statement describing how readers can access the data
• a funding statement naming grants and sponsors
• suggested reviewers with full contact details, plus any opposed reviewers
• supplementary material (extended results, proofs, or appendices) packaged separately
• ORCID iDs for the corresponding author

Knowledge-Based Systems does not require a cover letter on initial submission, but a short cover letter that states the knowledge contribution in one or two sentences helps the handling editor place the work. Manuscripts whose English is unclear can be returned for rewrite before review, so the language bar is enforced at triage, not deferred. The journal also expects reproducibility support: code, data, or a clear statement of how a reader can access them.

Figures should be supplied as separate high-resolution files, each typically under 10 MB, rather than embedded only in the manuscript PDF.

Before the declarations and reproducibility package are locked, a Knowledge-Based Systems reproducibility and ablation check can confirm whether your ablation isolates the knowledge component and whether the data and code availability statements clear the editorial bar.

Knowledge-Based Systems assigns submissions to a handling editor who manages them through Editorial Manager. Community-reported data puts the first review round at roughly 2.8 months, with about 2.2 reports per round and around 1.8 review rounds before a final decision. Treat the stages below as planning ranges, not commitments.

• Day 0: Submission and PDF build. The Editorial Manager portal ingests your files and builds a merged PDF. You approve it, confirm the CRediT statement, data availability statement, and competing-interests declaration, and submit.

• Days 1 to 10: Editorial screening. Editorial staff and the handling editor check scope fit, methodological novelty, completeness of declarations, and language quality.

The fastest returns happen in this window: generic-model papers with no knowledge contribution and out-of-scope work rarely reach external review.

• Days 10 to 35: Editor scope and novelty decision. The handling editor decides whether to send the manuscript out or return it. Community data puts immediate rejection at around 35 days.

Papers that are an existing architecture applied to a new dataset, or that lack an ablation, are commonly returned at this stage.

• Days 35 to 90: Peer review. Reviewers are invited and reports return, typically about two reports per round, on a multi-week cadence. A first round near 2.8 months is common, though subfield and reviewer load shift this.

• Weeks 12 to 28: Decision and revision. Reject, major revision, minor revision, or accept.

A revised manuscript must be accompanied by a point-by-point response letter. Most papers that pass review go through at least one major-revision round, and the community-reported average is about 1.8 rounds.

• Months 7 to 9: Final decision and production. Total handling time for accepted manuscripts runs to roughly eight months from submission, with faster outcomes for clean, in-scope papers and slower ones for multi-round revisions.

In our pre-submission review work with Knowledge-Based Systems submissions, four patterns generate the most consistent early returns. None of them are about the model being technically wrong. They are about contribution type and evidence packaging that this journal screens for before peer review begins.

In our review of knowledge-driven AI manuscripts, each of these is a named rejection pattern you can check your own draft against, and each reflects an editorial triage pattern specific to how Knowledge-Based Systems handling editors read submissions. The journal's scope-first screen raises the stakes on every one of these, because a strong leaderboard number does not survive a contribution-type mismatch. Editors consistently screen for these before sending a manuscript out for review.

Knowledge-Based Systems author policies and the Elsevier Guide for Authors define the mechanics below; the patterns describe how manuscripts coming through pre-submission review for this journal most often fall short of them. SciRev community data on this journal, where authors report a first review round near 2.8 months and a high immediate-rejection rate, is consistent with what we see: most attrition happens at the editor scope-and-novelty screen, before reviewers ever weigh in, and these four patterns are why.

A generic machine-learning model with no knowledge-representation, reasoning, or decision-support contribution. This is the single most common stall we see. The manuscript trains a neural network, a transformer, or an ensemble on a benchmark, reports an accuracy or F1 number, and frames the result as the contribution. There is no knowledge representation, no reasoning mechanism, no decision-support framing, and no expert-system or recommender structure.

The methods section is competent, but a Knowledge-Based Systems handling editor reads it and asks the journal's defining question: where is the knowledge? When the answer is "the model is the knowledge," the paper is a scope mismatch. KBS expects the knowledge, reasoning, or decision-support component to be the protagonist, and a manuscript where the model architecture is the only novelty belongs at a neural-network venue, not here.

This is the fastest desk return at this journal.

Check whether your Knowledge-Based Systems paper has a real knowledge or reasoning contribution →

An applied-X-to-dataset-Y paper with no methodological novelty. The parallel failure is the application paper that takes an existing method and runs it on a new domain or dataset. The introduction frames novelty as "first application of method X to problem Y," the experiments report that it works, and the discussion claims practical value.

But the method itself is unchanged, the knowledge engineering is shallow, and the contribution is an application result rather than a methodological advance. Reviewers at this journal treat "we applied a known approach and it performed well" as insufficient: the bar is a new idea in how knowledge is represented, acquired, reasoned over, or used for decisions.

A manuscript whose only claim is a competitive number on a new dataset, with the method borrowed wholesale, is consistently returned as lacking novelty before or early in review.

Check if your Knowledge-Based Systems method clears the novelty bar →

Results reported without an ablation that isolates the knowledge component. Even when the contribution is genuinely knowledge-driven, a recurring stall is a results section that reports an overall improvement without showing which part of the system caused it. The paper proposes a knowledge-augmented model, reports that it beats baselines, and stops there.

Reviewers in this field treat the ablation as part of the result, not an appendix to it: if the knowledge graph, the reasoning module, or the expert-rule layer is the claimed advance, the experiments must isolate its effect by removing it and measuring the drop. When there is no ablation, no component-wise analysis, and no evidence tying the gain to the knowledge mechanism rather than to scale or tuning, the central claim is unsupported.

A figure that shows "our full system wins" without "and here is the contribution of each knowledge component" reads as incomplete at KBS regardless of how strong the headline number is.

Check whether your Knowledge-Based Systems experiments include a component ablation →

Scope drift toward a pure neural-network or pure-applications venue. Knowledge-Based Systems sits between several adjacent journals, and a recurring desk return is a manuscript whose real home is one of them.

The introduction frames the work as knowledge-based, but the novel contribution is a new network architecture (which belongs at a neural-computation venue), a broad information-processing result (which belongs at a general informatics venue), or a deployed industry system with no methodological advance (which belongs at an applications venue).

Handling editors at KBS know their neighbors well, and they identify quickly when the knowledge framing is a wrapper over a contribution that another journal would evaluate more naturally. A manuscript where the genuine advance would be judged better by a neural-network, broad-informatics, or pure-applications reviewer is consistently identified as a scope mismatch before review.

This guide tells you what Knowledge-Based Systems editors look for; a Manusights review tells you whether YOUR paper passes that screen. A Manusights review checks the knowledge contribution, the methodological novelty, the ablation evidence, and the scope framing against the editorial bar this journal applies before peer review. Paid Manusights reviews include a 60-day money-back guarantee, and we do not train models on submitted manuscripts.

Before submitting, a Knowledge-Based Systems scope and novelty readiness check tests whether your knowledge contribution, ablation evidence, and scope framing clear the editorial bar this journal applies before peer review.

This guide was built from the Knowledge-Based Systems Guide for Authors, Elsevier author policies, the Editorial Manager submission system, and Manusights pre-submission review patterns from knowledge-driven AI manuscripts. We checked the scope topics, the CRediT and data-availability requirements, and the submission portal against the journal's own pages, and we cross-checked review-timing ranges against SciRev community data and the Clarivate JCR 2024 metric. The failure patterns describe what we see most often when knowledge-based AI manuscripts come through pre-submission review for this journal.

Use this page before you upload, when the official instructions cannot answer the real question: whether your knowledge contribution, methodological novelty, ablation evidence, and scope framing are already defensible. Source limitation: Elsevier updates format details, charges, and policies after this review date, so confirm administrative specifics against the journal's official pages before submission. To pressure-test the manuscript itself, run a manuscript readiness check.

• Expert Systems with Applications submission guide
• Information Sciences submission guide
• For the broader cluster, see the artificial intelligence journals overview.

Before you upload, run your manuscript through a Knowledge-Based Systems submission package check to catch the scope, novelty, and ablation issues editors filter for on first read. The check is free to run (/ai-review) and takes a single upload.