Multiple Human-Behaviour Indicators for Predicting Lung Cancer Mortality with Support Vector Machine

Sci Rep. 2018 Nov 9;8(1):16596. doi: 10.1038/s41598-018-34945-z.

Abstract

Lung cancer is still one of the most common causes of death around the world, while there is overwhelming evidence that the environment and lifestyle factors are predominant causes of most sporadic cancers. However, when applying human-behaviour indicators to the prediction of cancer mortality (CM), we are often caught in a dilemma with inadequate sample size. Thus, this study extracted 30 human-behaviour indicators of seven categories (air pollution, tobacco smoking & alcohol consumption, socioeconomic status, food structure, working culture, medical level, and demographic structure) from Organization for Economic Cooperation and Development Database and World Health Organization Mortality Database for 13 countries (1998-2013), and employed Support Vector Machine (SVM) to examine the weights of 30 indicators across the 13 countries and the power for predicting lung CM for the years between 2014-2016. The weights of different human-behaviour indicators indicate that every country has its own lung cancer killers, that is, the human-behaviour indicators are country specific; Moreover, SVM has an excellent power in predicting their lung CM. The average accuracy in prediction offered by SVM can be as high as 96.08% for the 13 countries tested between 2014 and 2016.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Aged, 80 and over
  • Alcohol Drinking / adverse effects*
  • Databases, Factual
  • Environmental Exposure / adverse effects*
  • Feeding Behavior*
  • Female
  • Humans
  • Lung Neoplasms / etiology
  • Lung Neoplasms / mortality*
  • Lung Neoplasms / psychology
  • Male
  • Neural Networks, Computer
  • Risk Factors
  • Smoking / adverse effects*
  • Socioeconomic Factors
  • Support Vector Machine*