Transfer Learning for Unsupervised Influenza-like Illness Models from Online Search Data

Abstract

A considerable body of research has demonstrated that online search data can be used to complement current syndromic surveillance systems. The vast majority of previous work proposes solutions that are based on supervised learning paradigms, in which historical disease rates are required for training a model. However, for many geographical regions this information is either sparse or not available due to a poor health infrastructure. It is these regions that have the most to benefit from inferring population health statistics from online user search activity. To address this issue, we propose a statistical framework in which we first learn a supervised model for a region with adequate historical disease rates, and then transfer it to a target region, where no syndromic surveillance data exists. This transfer learning solution consists of three steps: (i) learn a regularized regression model for a source country, (ii) map the source queries to target ones using semantic and temporal similarity metrics, and (iii) readjust the weights of the target queries. It is evaluated on the task of estimating influenza-like illness (ILI) rates. We learn a source model for the United States, and subsequently transfer it to three other countries, namely France, Spain and Australia. Overall, the transferred (unsupervised) models achieve strong performance in terms of Pearson correlation with the ground truth (> .92 on average), and their mean absolute error does not deviate greatly from a fully supervised baseline.

Single Model for Influenza Forecasting of Multiple Countries by Multi-task Learning

“…As input for the proposed model, we selected the top L English search queries for the US, AU, and UK based on the list, and selected each search query in JP and FR corresponding to the English search query based on these one-to-one query mapping methods. The usefulness of mapping from English to other languages is described in [33,50,53]. These studies pointed out that the volume movement in the search queries is similar among countries with certain health conditions.…”

“…Many methods using UGCs for forecasting the flu volume have been developed since the emergence of Google Flu [18], which demonstrated that the number of search queries capturing human behaviors was a good resource for forecasting. Certain studies [21,49,53,48] have depended on "Google Correlate," which returns English search queries that are the most highly correlated to an input time series, for the selection of suitable search queries. However, this approach cannot be used in many areas (non-English-speaking areas) and it has already been unavailable since December 2019.…”

“…The second issue is how to effectively incorporate search queries into a flu forecasting model. Two types of data have been applied extensively: historical flu activity data involving the previous year's data (known as "historical ILI data") [43,47,45] and online UGC data [53,4,52], which mainly consist of search query data. A representative example of simultaneous inputs is the ARGO model [48], which is based on linear regression using the input data of the Google search time series and the historical ILI data.…”

The Number of Confirmed Cases of Covid-19 by using Machine Learning: Methods and Challenges

“…There have been some attempts in this research direction [26]. Transfer learning deals with training a model for a given problem and using it to related but different problem [81]. Transfer learning has been applied for the prediction of confirmed cases by using the model trained on data from one country in which is ahead in the epidemic curve to other countries still in earlier stages of the epidemic curve [73].…”

Predicting regional influenza epidemics with uncertainty estimation using commuting data in Japan

“…Resource selection for the prediction method is also an important factor in influenza prediction. Many studies have relied on user-generated content (UGC) from internet services, such as search services [12,14,23,24] and social networking services [25][26][27]. Infectious disease surveillance conducted with online content, such as that described above, is generally described as infoveillance [28].…”