Application of Artificial Intelligence (AI) techniques to risk assessment of African Swine Fever (ASF) for Australia

Kevin Liu

African Swine Fever (ASF) has become one of the most dangerous viruses threatening biological security due to its high infection and case fatality rate. Since 2007, ASF has spread into many countries, especially in Europe and Asia. As an important part of prevention, risk assessment can provide preventive measures earlier to prevent the introduction into Australia, which is important to sustain Australia’s livestock industry and global trade reputation. Australia is currently ASF-free, but the disease has been reported in many neighbouring countries, so it is necessary for Australia to raise vigilance against the ASF spread. From the comprehensive literature review, the past models are difficult to directly apply in this research, which is due to the data collection standards being different between the different regions. Therefore, it is necessary to re-establish the model for Australia to perform the ASF risk assessment. In this study, AI techniques using fuzzy rule-based systems and linear regression were used to analyze the Australian ASF risk. Australian ASF risk is classified into three aspects: ASF introduction risk, ASF domestic epidemic risk once ASF has been introduced into Australia, and the seasonal risk in Australia. International passengers (IP) and international import trade (IIT) have been identified as the two main ASF introduction factors based on transmission features from past research. After analyzing the data from 2019 and 2020 through the establishment of a fuzzy risk assessment model, the ASF introduction into Australia based on the collected data belonged to the low risk. To further increase the prevention ability, the second case assumed that the ASF had been introduced into Australia, and another fuzzy risk assessment model was created to assess the epidemic risk in different Australian regions. As a result, although the ASF pandemic risk is relatively low in Australia, it includes a risk of irregular and scattered outbreaks, with Victoria (VIC) and New South Wales (NSW) - Australia Capital Territory (NSW-ACT) being the highest risk for the second case. To establish the third seasonal risk assessment model in Australia, Poland was selected as a template to develop such a risk assessment model, as Australia is an ASF-free country. In this stage of the research, the seasonal risk of ASF could be inferred by analyzing the monthly average rainfall, monthly average temperature, and the number of ASF-wild boar cases using a time series analysis methodology. In this study, linear regression and an automated fuzzy rule generation algorithm (genfis) were used. The results of the two analyses using linear regression and genfis were similar, suggesting that May, June, July, and August are the active months for ASF outbreaks. To sum up, this work provides an understanding of Australia’s ASF risk transmission based on AI techniques using the data collected. To our knowledge, this is the first study to comprehensively analyze the ASF epidemic risk in a country by using fuzzy rule-based modelling. The methodology also provides insights and useful information for the establishment of fuzzy models to perform the ASF risk assessment for other countries.

Application of Artificial Intelligence (AI) techniques to risk assessment of African Swine Fever (ASF) for Australia

Files and links (1)

Abstract

Details

Metrics