A novel model for accurate and fast prediction of cancer incidence

TL;DR

This paper introduces a new model using Google Trends data to accurately predict cancer incidence rates in the U.S. and globally.

Contribution

The novel approach combines cancer incidence data with yearly changes in Google Trends search volume for improved predictions.

Findings

The model achieved less than 6% error in predicting cancer rates across most U.S. states.

Similar accuracy was observed when predicting cancer incidence in 54 countries.

The model successfully provided short-term predictions from 2017 to 2023.

Abstract

Predicting cancer incidence has long been a challenge for clinicians and researchers. Accurate predictions are essential for health planning to ensure adequate resources for diagnosis, treatment, and rehabilitation. Current prediction methods rely on historical data, assuming persistent patterns of cancer incidence.

In this study, the Google Trends tool was used to obtain the relative search volume index (RSVI) for the topic “cancer” each year from 2017 to 2023 in the United States and worldwide. The proposed model incorporated actual cancer incidence rates and yearly changes in RSVI.

The model was applied to predict the rates of new cancer cases in fifty American states over four consecutive years (2017, 2018, 2019, 2020). The selection of years was restricted with data availability. In most states, the percentage error did not exceed 6%. The high degree of similarity between the actual and predicted cancer incidence rates was notable. Similar results were obtained when predicting cancer incidence rates in the countries studied.

The model has successfully provided accurate short-term predictions of cancer incidence rates across all 50 American states and 54 countries since 2017.

The online version contains supplementary material available at 10.1186/s12889-025-22624-4.

Figures