Predicting patterns to extract knowledge can be a tough task but it is worth. When you want to accomplish that task you have to take your time analysing all the data you have and you have to adapt it to the algorithms and technologies you are going to use after analysing. So you need to know the type of data that you own. When you have finished making the analysis, you also need to know what you want to find out and, therefore, which methodologies you are going to use to accomplish your objectives. At the end of this chapter you can see a real case making all that process. In particular, a Classification problem is shown as an example when using machine learning methodologies to find out if a hospital patient should be admitted or not in Cardiology department.
TopIntroduction
In this chapter we roughly discuss how to derive conclusions from hospital admission data. We describe a process to identify patterns in the data as well as the description of several concepts needed to carry out that objective. We use different Big Data analytics techniques to achieve our goal. Big Data analytics allowed us to uncover hidden patterns and unknown correlations to start working with available datasets. Then, we are able to improve the operational efficiency and obtain business benefits, in order to follow a system of work. Similar conclusions were reached according to (Powers, Meyer, Roebuck, & Vaziri, 2005), where they use advanced econometric cost modelling techniques to predict healthcare costs using pharmacy data.
Initially, it is advisable to make a study of correlation indexes from the attributes we are going to use. These indexes will give us a better idea about the most appropriate attributes and will allow us to obtain conclusions with the selected dataset.
Once we have decided what are the answers we want to know and the type of study we want to accomplish, we need to begin studying the type of data and the type of structure that we have in our dataset. This means that an important part of the available time to develop the study was devoted to prepare our data for the algorithms we would use. For that reason, section 2 details the type of data we can find and how it usually appears.
After preparing the dataset environment, we needed to use the appropriate Machine Learning techniques depending on the type of data we had and on which conclusions we wanted to obtain. In the present day, other studies are using machine learning techniques to predict behaviours in health systems and they select their appropriate techniques to reach them. As an example of this, some researchers would like to know if patients are going to re-enter during the next twelve months as it is done in (Vaithianathan, Jiang, & Ashton, 2012), where they used multivariate logistic regression. Or, perhaps, they would like to predict hospital admissions depending on patient-specific medical history using several types of classification algorithms, according to Wuyang et al. (2015). Since several years ago, those techniques have gradually been introduced in different studies thanks to their effectiveness when making predictions, as we can observe in (Wuyang et al., 2015) too.