Table of ContentsSupervised LearningUnsupervised LearningReinforcement LearningTogether LearningMachine learning is the process of automatically extracting patterns from historical data to predict future events and can be used to create predictive models (Kelleher et al., 2015). In Perry (2013), machine learning is defined as automated algorithms for extracting structures from historical data, generalizing results, and predicting future data. As Shouval et al. (2013) further explain that machine learning starts without a predefined model, but the model is created by learning patterns in the data used. Kelleher et al. (2015), Perry (2013) and Shouval et al. (2013) classify machine learning into two main types called supervised and unsupervised learning, with a third type called reinforced learning explained in Sutton and Barto (2015). Several machine learning algorithms can be combined and their results used in a complementary manner in what is called ensemble learning (Polikar, 2010; Brown, 2010). These different types of learning were evaluated in order to select a learning method for use in this research and are described below. Say no to plagiarism. Get a tailor-made essay on “Why violent video games should not be banned”?Get the original essaySupervised learningSupervised learning algorithms build models by learning relationships between descriptive features (input) and target features (output ) based on historical datasets (Kelleher et al., 2015). The algorithm is trained by providing it with known inputs and their corresponding responses, and from the learned relationship it can predict responses for unknown inputs (Shouval et al., 2013). In Shouval et al. (2013), supervised learning can be grouped into regression and classification. Regression algorithms – linear regression is continuous in nature predictions such as time measurements and logical regression predictions are discrete in nature like male/female (Shouval et al., 2013). Regression-like types of machine learning algorithms are therefore not suitable for a model producing an index. Classification algorithms – decision trees are classification algorithms that assemble a decision tree from a labeled data set, with the root at the top branching downward to form branches and ending at the leaves (Suknovic et al., 2011). The root and each of the branches are discrete functions with attribute-value pairs that require a decision to be made to move to the next depth or tree level (Barros, et al., 2015). Like Barros et al. (2015) explain that each branch has only one incoming input and can have two or more output branches, while the root has only output branches and no input. Leaves also called terminals are at the end of the tree and each represents an output class. Suknović et al. (2011) explain that from the labeled dataset of known attribute and corresponding values, the algorithm can learn the classification and be able to predict a class given a future unknown attribute. The advantages of decision trees are ease of use and interpretation, as stated by Shalev-Shwartz and Ben-David (2014). Decision tree algorithms are proposed for use in this study.CARTBayes NaveK-NNUnsupervised learningUnlikesupervised learning, unsupervised algorithms are only given descriptive features without corresponding target features. The algorithm learns by identifying relationships in the inputs and groups them into clusters, associations or detects abnormal behavior. Association algorithms – in machine learning, association algorithms are rule-based algorithms that learn by discovering interesting relationships between data points in a dataset (Rudin et al., 2013). According to Al-Maolegi and Arkok (2014), the extraction of association rules in databases is widely carried out in sales transactions in order to create patterns between items. The basic concepts in association are to detect frequent elements in a dataset and generate association rules based on how the elements occur together (Al-Maolegi & Arkok, 2014). Apriori is the leading association algorithm in unsupervised machine learning (A-Maolegi & Arkok, 2014), with Eclat and FP-Growth also used (Heaton, 2017). The strength of association learning and algorithms lies in the interest in the association between elements and the frequency of occurrence. This will not be useful in the initial modeling of the RPPI. Anomaly Detection Algorithms - Blomquist and Möller (2015) defined anomalies as data patterns that do not conform to what has been specified as normal behavior. Anomaly detection algorithms are used to detect anomalies or data points that do not follow the norm in a data set. These algorithms are derived from statistical methods for pruning and cleaning datasets of outliers (Goldstein & Uchida, 2016). The authors detail how these algorithms are used today to detect fraud in the financial sector and as intrusion detection tools in information security. These algorithms do not correspond to RPPI modeling and are not taken into account for this study. Clustering algorithms – clustering is the partitioning of data consisting of similar objects into groups called clusters. Objects in a group or cluster share similarities and are different from objects in other clusters (Al-Haddad & Aldabbagh, 2015). In Al-Haddad and Aldabbagh (2015), explain that clustering creates structures from unlabeled data by separating data based on their properties. Three types of clustering are presented in Al-Haddad and Aldabbagh (2015), namely exclusive, overlapping and hierarchical clustering. Exclusively, the clustering of an object can only belong to one group and not to another group. In overlapping clustering, an object can belong to multiple groups. With hierarchical clustering, objects are grouped hierarchically and can belong to multiple clusters. The clustering algorithms taken into account for the study are k-means. The K-means algorithm is a clustering algorithm that divides data into a predetermined number of clusters by the user by scanning the dataset to find commonalities (Goswami, 2015). The K-means algorithm is easy to understand and use (Goswami, 2015), and can be used in any research area (Morissette & Chartier, 2013). Al-Haddad and Aldabbagh (2015) and Goswami (2015) explain the process as follows: initially, random cluster centers called centroids are selected and all objects in the dataset are assigned to a centroid closest to them to form a cluster. When all objects have been grouped, new centroids are calculated based on the previous clusters and the »..