Seasonal-adjustment Based Feature Selection Method for Large-scale Search Engine Logs

Thien Q. Tran,Jun Sakuma

DOI: https://doi.org/10.1145/3292500.3330766

2020-08-22

Abstract:Search engine logs have a great potential in tracking and predicting outbreaks of infectious disease. More precisely, one can use the search volume of some search terms to predict the infection rate of an infectious disease in nearly real-time. However, conducting accurate and stable prediction of outbreaks using search engine logs is a challenging task due to the following two-way instability characteristics of the search logs. First, the search volume of a search term may change irregularly in the short-term, for example, due to environmental factors such as the amount of media or news. Second, the search volume may also change in the long-term due to the demographic change of the search engine. That is to say, if a model is trained with such search logs with ignoring such characteristic, the resulting prediction would contain serious mispredictions when these changes occur. In this work, we proposed a novel feature selection method to overcome this instability problem. In particular, we employ a seasonal-adjustment method that decomposes each time series into three components: seasonal, trend and irregular component and build prediction models for each component individually. We also carefully design a feature selection method to select proper search terms to predict each component. We conducted comprehensive experiments on ten different kinds of infectious diseases. The experimental results show that the proposed method outperforms all comparative methods in prediction accuracy for seven of ten diseases, in both now-casting and forecasting setting. Also, the proposed method is more successful in selecting search terms that are semantically related to target diseases.

Machine Learning

What problem does this paper attempt to address?

### What problems does this paper attempt to solve?

This paper aims to solve several key problems encountered when using large - scale search engine logs to predict infectious disease outbreaks:

1. **Data Instability**:
- **Short - term Fluctuations**: The search volume of search terms may change irregularly in the short - term due to environmental factors (such as media and news reports).
- **Long - term Changes**: The search volume of search terms may also gradually change in the long - term due to changes in the search engine user base. If these changes are ignored, the prediction model may have serious mispredictions.

2. **Computational Complexity**:
- Search engine logs contain a large number of search terms (for example, Yahoo! Japan has an average of about 30 million different search terms per day), resulting in a very large dimension of explanatory variables. In order to reduce the learning cost and over - fitting, feature selection is required to select effective search terms.

3. **Interpretability**:
- In the infectious disease prediction problem, a reliable predictor not only needs to accurately predict the outbreak, but also needs to make predictions based on meaningful evidence. Therefore, it is very important to select search terms that are semantically related to the target infectious disease so that experts can understand the model.

4. **Universality**:
- Most of the existing studies only focus on one or two major infectious diseases (such as influenza and dengue fever), and do not explore whether their methods are applicable to other infectious diseases. Therefore, it is necessary to develop a method that can be widely applied to multiple infectious diseases.

To solve these problems, the author proposes a feature selection method based on seasonal adjustment. This method improves the stability and accuracy of prediction by decomposing the time series into three components (seasonal component, trend component, and irregular component), and performing feature selection and modeling on each component respectively. In addition, this method also specially designs a feature selection mechanism to ensure that the selected search terms are semantically related to the target infectious disease.

### Method Overview

1. **Seasonal Adjustment**:
- Decompose the time series of the infection rate and the time series of the candidate search terms into the seasonal component \(\hat{S}\), the trend component \(\hat{T}\), and the irregular component \(\hat{I}\).

2. **Feature Scoring**:
- Score each search term of the trend component \(\hat{T}\) and the irregular component \(\hat{I}\) respectively, and rank them according to the score.

3. **Feature Subset Selection**:
- Use the wrapper approach to select the optimal feature subset from the ranked search term list.

4. **Final Prediction Model**:
- Train models for predicting the trend component and the irregular component respectively, and combine the seasonal component for the final prediction.

Through this method, the author has successfully improved the accuracy and stability of prediction, and can select search terms that are semantically related to the target infectious disease in the prediction model.