Repeated-day priority did not raise both mean pair and time coverage at a 20% budget with three-day histories in two workplace records

We tested whether priority for pairs seen on more past days improves coverage of contacts on the next recorded workday. We used two public workplace records from 2013 and 2015. Each record had ten observed workdays. Seven forecasts used three past days and a budget of 20% of past observed pairs. We compared total contact time with a rule that first ranks pairs by contact-day count and then by total time. Mean pair coverage changed from 19.67% to 19.71% in 2013 and from 19.02% to 22.18% in 2015. Mean contact-time coverage fell from 40.65% to 32.48% and from 35.98% to 35.86%, respectively. Thus, repeated-day priority did not increase both means at this budget. In 2015, pair coverage within departments rose from 27.30% to 32.60%, while coverage between departments fell from 3.23% to 2.43%. Other budgets gave different time-coverage changes. Both records come from one workplace setting. These results are descriptive and do not establish a general workplace effect. The package contains original CC0 source files, provenance, code, daily pair counts, main forecast scores, all tested results, and figures. A separate calculation checked all main method-day results.

temporal networks · face-to-face contacts · workplace · contact prediction · recurrence · coverage · department mixing · open data