Methodology

ECERankings counts publications in a fixed list of selective electrical and computer engineering journals and conferences, credits each paper to the institutions its authors listed on it, and combines the per-area totals with a geometric mean. It follows CSRankings in spirit, but uses journals as well as conferences and takes data from OpenAlex instead of DBLP.

What is ranked · How credit is assigned · The score · Years · Data caveats · Venue list · Source and data

What is ranked

• Institutions, not departments. Any author at an institution who publishes in a listed venue counts, whether they sit in ECE, computer science, physics, medicine, or a research institute. The ranking does not know which department an author belongs to.
• Not a faculty roster. There is no list of faculty behind these numbers, and the site makes no claim about who is employed where today. Papers by students, postdocs, and visitors count the same as papers by faculty.
• Education institutions. Credit uses affiliations classified as education by the provider or a curated mapping. A paper co-written with an industry lab gives the university only its authors' share. Classification errors and documented exceptions remain possible; see data coverage.
• Publications only. No citations, funding, awards, or reputation surveys are used.

How credit is assigned

Institution at time of publication. Each paper is credited to the institution an author listed on that paper. A researcher who moves does not take earlier papers along: work done at one university stays with it. This differs from CSRankings, which credits a faculty member's whole career to their current institution. The two answer different questions; this site measures output produced under an institution's name at the time it happened.

Adjusted count. Every paper is worth 1, split evenly among all of its N authors, so each author carries 1/N. An institution's adjusted count is the sum of its authors' shares. Shares belonging to authors with no university affiliation are dropped, not redistributed. An author who lists two universities on one paper has their 1/N split evenly between them.

Papers counts papers with at least one credited author at the institution, without the 1/N split. Counts are summed across venues and years; a paper recorded in more than one qualifying venue can appear more than once. This column is shown for context and does not enter the score.

The score

For each selected area i, let s_i be the institution's adjusted count in that area over the chosen years. With n selected areas, the score is the geometric mean of (1 + s_i), minus one:

score = (∏_i=1..n (1 + s_i))^1/n − 1 = exp( mean( log(1 + s_i) ) ) − 1

• Areas with no papers still count. n is the number of areas selected, including areas where the institution has zero output. A zero area contributes log(1) = 0 to the mean and pulls the score down, so breadth across areas is rewarded and dominance of a single area is damped.
• Changing the area selection changes every score, because n changes. Scores are only comparable within one selection of areas, years, and region.
• Ties. Institutions with equal scores share a rank; the listing order within a tie is alphabetical.
• Region filters the list after scoring, using the institution's recorded country, with documented metadata corrections. Institutions with no recorded country appear only under World.

Years

• The year range is inclusive. Journal papers use their publication year; conference papers use the verified proceedings edition year where the venue registry specifies it. Documented provider year-label errors are corrected against the edition and DOI evidence. The selectable range is derived from the loaded data files.
• The default range ends at the latest complete calendar year. The current year is labelled partial: papers are still being published and indexed, so including it lowers counts relative to full years.
• Even complete years can be incomplete for a particular venue. Conference proceedings are sometimes indexed months after the event, and coverage of older years varies by venue (most areas start at 2016). A complete calendar year is not a promise that every venue in it has been fully collected.

Data caveats

• Affiliations come from OpenAlex, which resolves the free-text affiliation on each paper to an institution record. These resolutions are not always right; for example, some papers listing "University of Pennsylvania" are attributed to California University of Pennsylvania. Where OpenAlex has no resolved institution, the project uses its own curated mapping of affiliation strings.
• Venue membership. OpenAlex does not link many IEEE conference papers to their conference. For those, Crossref supplies the list of papers in each proceedings volume and OpenAlex supplies the affiliations, joined by DOI. Papers without a DOI match or without affiliations are missed.
• Non-research items such as editorials, errata, and letters are excluded by type, and a curated list removes columns and other items OpenAlex labels as articles. Some may remain.
• Institution identity. Campuses and sub-units are credited as OpenAlex records them, without rolling them up to a parent. Curated aliases merge verified duplicate records of the same university. A shared name or country alone does not establish that two records are the same institution.
• Machine Learning is off by default. Proceedings coverage and paper-specific affiliations are incomplete for NeurIPS, ICML and ICLR, because source records often do not identify the accepted proceedings or provide a paper-specific affiliation. DOI matches recover some historical affiliations; current author-profile affiliations are withheld. Selecting this area compares an incomplete subset.
• Interdisciplinary Flagship is off by default. It holds general-science journals (Nature, Science, PNAS, …) whose papers are mostly outside ECE.

Venue list

Areas and venues come from the project's registry, data/areas.json. Areas marked off by default can still be selected on the rankings page.

Source and data

All code, the venue registry, and the aggregated data files are in github.com/eding42/ecerankings.org. The rankings page loads one file per area, data/<area>.json, with rows of institution, year, paper count, and adjusted count; the Download CSV button on the rankings page exports the current table with per-area adjusted counts. Corrections to affiliations or venues are welcome as GitHub issues.