Data Visualization and Preprocessing

ML methods, being purely data driven, relies on the availability of high quality dataset. However, in reality, the datasets may have inconsistencies, errors, and may even be incomplete. Further, the choice of an appropriate ML algorithm for a given dataset will depend highly on the nature, size, distribution of the dataset. In this chapter, we discuss the different approaches to visualize data such histogramsHistogram, scatter plotsScatter plot, heat mapsHeat map, and tree mapsTree maps. Further, several measures that quantify the data including central and higher-order measuresHigher-order measures are discussed. Next, we discuss several commonly used outlierOutliers detectionOutlier detection algorithms that enable “data cleaning”. Finally, we discuss data-imputation algorithms such as SMOTESMOTE and ADASYNADASYN for imputing data in imbalanced datasets.