Predicting COVID‑19 Peaks Using Excess Fever Search Index: Analysis of Taiwan, Hong Kong, Japan and Chinese Cities

The author examined Google search trends for the term "fever" in Taiwan, Hong Kong and Japan, comparing pandemic‑period indices with non‑pandemic averages to compute a baseline fever index.

During the pandemic, the excess fever search index was summed to produce a cumulative area (S), representing the proportion of infected population relative to total population.

Charts show that in the three regions the cumulative area reaches 80 when the epidemic peaks, while first‑wave end values are 160 (Hong Kong), 200 (Taiwan) and 250 (Japan).

Using Baidu search data for Chinese cities (Shijiazhuang, Xingtai, Baoding), similar cumulative areas were calculated, yielding values around 70‑76, indicating comparable magnitude between Google and Baidu metrics.

The author set conservative thresholds (100 for peak, 250 for first‑wave end) and derived peak‑time and end‑time estimates for each city based on the growth speed of the cumulative area.

Updates from December 12‑16 refine the model: adding a massive‑calculation correction, adjusting peak values back to 80, introducing an "end progress bar" variable, fixing bugs that merged multiple waves, and incorporating population‑ratio metrics.

These refinements improve the stability of the predictions, showing that many mainland Chinese cities now exhibit peak and post‑peak trends similar to those observed in Hong Kong and Taiwan.

The author emphasizes that the method is simple and approximate, yet it aligns reasonably with real epidemic trends and can help alleviate public anxiety while awaiting official health data.