A survey on machine learning and fuzzy inference based prediction with small datasets

Abstract

Different machine learning (ML) algorithms are popular for solving nonlinear knowledge-intensive problems, like predicting natural disasters or disease risk for populations, classifying cancer patients into high or low-risk groups, etc. However, training those algorithms requires sufficient or ever big digital data that is not available in real-world applications. The authors of various articles suggest how to solve the problem of small data when applying ML algorithms in a particular subject area. Nevertheless, the main research question arises what are the main trends in solving the problem of small data in ML and fuzzy inference based prediction? To answer the defined question, this paper presents a survey based on the articles extracted from the Web of Science (WoS) and Scopus databases. The results show that this topic has become more relevant, and popular algorithms, like SVR, clustering, Naïve Bayesian algorithms, decision trees, etc., are adopted to work with small data.