Expert Systems with Applications Submission Guide

Source: ESWA on ScienceDirect, ESWA insights, Clarivate JCR 2024, accessed May 2026.

This is the ESWA-specific editorial detail authors most often miss:

ESWA emphasizes substantive real-world application over method novelty alone. The journal's public scope describes expert and intelligent systems applied in industry, government, and universities, including finance, engineering, marketing, medicine, energy, information retrieval, risk assessment, production management, and many other applied settings. That breadth is not permission to submit generic AI work. It is a signal that the application context must be real.

The journal's editorial position favors:

1. AI/ML applied to a specific real-world domain

1. Application-domain validation (not just benchmark performance)

1. Practical implications for the application domain

The strategic implication: pure-methods papers without strong application focus fit IEEE TPAMI, JMLR, Pattern Recognition, or methods-specialist venues better. ESWA's bar is application substance, not methodological generality.

ESWA has a few details authors should not treat as generic Elsevier boilerplate.

These are small operational details, but together they show the same editorial preference: ESWA wants applied intelligent-systems work that is clear, comparable, and practically meaningful.

Three operational signals govern editorial assessment:

1. Application substance. ESWA requires substantive real-world application, not just methodological tour-de-force.

2. Methodological rigor. AI/ML methodology must be appropriate and rigorous; benchmark comparisons should be fair.

3. Reproducibility. Code/data sharing is increasingly expected for high-impact applied-AI work.

Recent ESWA issues span:

• Healthcare AI (medical imaging, clinical decision, drug discovery)

• Financial AI (fraud detection, credit risk, algorithmic trading)

• Cybersecurity AI (intrusion detection, malware analysis)

• Manufacturing AI (quality control, predictive maintenance, scheduling)

• Transportation AI (autonomous driving, traffic, logistics)

• E-commerce AI (recommendation, pricing, customer analytics)

• Energy AI (grid management, renewable forecasting)

• LLM and generative AI applications

For specific recent papers and paper-level DOIs, use the live ESWA article list on ScienceDirect. Do not rely on old sample DOIs or cached article lists when preparing a cover letter. Actual ScienceDirect examples from an older ESWA issue include 10.1016/S0957-4174(22)02374-0, 10.1016/j.eswa.2022.119072, and 10.1016/j.eswa.2022.119052; use them only as DOI-format examples, not as evidence about current editorial preference.

The pattern we saw in recent ESWA papers used for this guide is that the strongest submissions do more than report a model score. They connect the system to a real decision setting, explain why the baseline set is current, and make the application consequence clear enough for a non-specialist editor to understand why ESWA is the right journal.

ScienceDirect publishes ESWA's median times openly: 5 days to first decision, 62 days post-review, 147 days submission-to-acceptance, 7 days acceptance-to-publication. Treat as planning ranges, not promises.

• Day 0: Editorial Manager upload. The Editorial Manager submission portal portal accepts the package, runs integrity checks, and routes to a handling editor matching the application domain.

• Days 1 to 5: First editor read. The handling editor evaluates application substance, methodological rigor, and scope fit. The 5-day first-decision median lands here; most desk rejections happen at this stage.

• Days 5 to 20: Reviewer invitations. ESWA typically invites two to three reviewers with expertise in the relevant application domain plus AI/ML methodology.

• Days 20 to 62: Peer review. Reviewer reports return on roughly a 6-week cadence; the 62-day-post-review median reflects this band.

• Days 62 to 147: Revisions and acceptance. Major revision is the most common outcome for papers that pass peer review. The 147-day submission-to-acceptance median assumes one revision round.

• Days 147 to 154: Online publication. Elsevier production typically pushes papers online within 7 days of acceptance.

Treat those as planning signals, not guarantees. A paper with unclear application substance can still be screened early, while a technically strong manuscript can slow down if reviewer expertise is hard to match.

Across applied-AI manuscripts targeting Expert Systems with Applications, three recurring decision risks matter most across submissions that ESWA editors filter out within the 5-day first-decision window. Use the three checks below before you open Editorial Manager upload slot.

Failure pattern: Generic "we applied X to Y" abstract paired with a method-architecture Figure 1

Across ESWA-targeted manuscripts, we consistently see authors open the abstract with a generic application framing ("we propose a deep learning approach for medical image classification") then make Figure 1 a model-architecture diagram (CNN block diagram, transformer encoder, attention heatmap schematic) rather than a figure that grounds the application context.

ESWA editors specifically screen the abstract-plus-Figure-1 pair for whether the application context is established before the methodology takes over. Manuscripts where Figure 1 is an architecture diagram before any application-domain figure consistently get returned with the suggestion that the work fits a methods venue (TPAMI, Pattern Recognition, JMLR) rather than ESWA.

The fix is to make Figure 1 the application context (data sample, decision setting, operational pipeline, before-and-after comparison) and push the architecture diagram to Figure 2 or supplementary material.

Check whether your ESWA abstract and Figure 1 make the application setting visible →

Failure pattern: Benchmark comparison table omits current 2025-2026 baselines from sister AI venues

We frequently see ESWA manuscripts include comparison tables that cite 2020-2022 baselines from the methodology's original publication trajectory (often TPAMI / CVPR / NeurIPS papers) but skip the recent 2025-2026 application-domain methods that have appeared in Knowledge-Based Systems, Engineering Applications of Artificial Intelligence, Applied Soft Computing, or Pattern Recognition.

ESWA reviewers are application-domain specialists who actively track the recent literature in their domain; comparison tables anchored on older general-AI baselines look strategically incomplete to them. The fix is to add at least 2 baselines from sister ESWA-tier venues published in the last 18 months, with the methods section explicitly noting why each baseline is the right comparator for the application-domain claim being made.

Check whether your ESWA comparison table uses current application-domain baselines →

Failure pattern: Reproducibility package references "code available on request" rather than a public repository

The third recurring pattern in ESWA-targeted manuscripts is data and code availability statements that say "code available on request" or "data available from the corresponding author upon reasonable request" rather than naming a specific public repository (GitHub / Zenodo / OSF) with a permanent identifier (DOI / version tag / commit hash).

ESWA's editorial culture increasingly favors reproducibility-track submissions because applied-AI claims are testable only when the code and evaluation splits are inspectable.

Manuscripts where the methods section describes evaluation splits in prose but does not provide the splits as a downloadable supplementary file face additional reviewer scrutiny on the validity of the reported performance numbers, often extending the revision cycle by one full round.

The fix is to deposit code + evaluation splits + trained model weights to a public repository before submission and reference the permanent identifier in the data availability statement.

Check whether your ESWA reproducibility package is specific enough for review →

Check whether your ESWA manuscript is submission-ready →

This guide tells you what ESWA editors look for. The review tells you whether your paper clears the application-substance, sister-journal-routing, and reproducibility check before the Editorial Manager upload. Paid reports include a 60-day money-back guarantee, and Manusights does not train models on submitted manuscripts; we do not train on your paper.

Submission caps: ESWA operates under Elsevier's free-format submission model with no rigid word ceiling, but accepted Research Articles typically run 8,000 to 12,000 words across 25 to 35 pages with 6 to 12 figures and complete supplemental code archives; abstracts cap at roughly 250 words per Elsevier convention. The Editorial Manager portal at Editorial Manager submission portal enforces format completeness at upload.

• Best machine learning journals

• Pre-submission review for machine learning papers

• Pre-submission review for artificial intelligence papers

• International Journal of Information Management Submission Guide

• IEEE Transactions on Pattern Analysis and Machine Intelligence Submission Guide

• If the paper is really a wireless-channel or wireless-system contribution with AI as the method, compare IEEE Transactions on Wireless Communications.

• If the paper is policy-first energy analysis with AI as the tool, compare Energy Policy.

Last verified: May 2026 against ESWA editorial pages.

If your manuscript is already in the portal, use the Expert Systems with Applications Under Review status guide to interpret the current status before sending a follow-up email.