Predictive analytics issues with real data
In this section, you will learn about the main issues with real data-based predictive analytics solutions. Mainly, we will discuss the following three issues.
Partial and scarce training data
One of the main requirements of predictive analytics models to work well in practice is the availability of large-scale historical data. In sectors such as healthcare, banking, security, and manufacturing, it is not easy to find such datasets. The main reasons behind that are privacy concerns, regulations, and trade secrets. As we know, without sufficient training datasets, ML algorithms simply cannot work well in practice. Thus, predictive analytics methods based on real data work well only in certain fields where data is available. Therefore, augmenting real data with synthetic data can complement small-sized and incomplete real datasets. Thus, it solves one of the main issues of real datasets in these fields, as we will see in the next section...
