A review of the use of machine learning techniques by social media enterprises
Fazle Rabbi
Australian Computer Society, Australia
Abstract
In this review, the current status of research on machine learning techniques by social enterprises was explored. First the popularity of social media, the type of big data generated by them, how they are potentially and actually are used were outlined. The general concept of machine learning, components, methods and its application were discussed. On this background, specific methods and stages used in machine analytics of social media were reviewed with explanatory diagrams reproduced from the works of various authors. Bayesian Network and Support Vector Machine are most commonly used methods due to their advantages over other methods. The methods are used mostly for opinion mining, sentiment analysis and trend analysis. Social network analysis and business management are the most frequent applications, although it has been applied in many other fields also. Many new methods and modifications of current methods are being developed almost continuously to solve the problems in current methods.
There are also limitations in current machine learning techniques. Samples drawn from big data may not always represent of the population. The gigantic size of data, with low value intensity, spread over several sources and their dynamic nature can cause certain biases. Large computing capacities and sophisticated methods of sampling, extraction and analysis are required to handle big data. Accuracy and objectivity of social media data are not always assured and the problem multiplies with use of diverse data sources. Access and ethical issues may arise. Heterogeneity, noise accumulation, spurious correlations and incidental endogeneity are other problems.