Application of classification methods in business intelligence
Tutor / director / evaluatorVegas Lozano, Esteban
Document typeMaster thesis
Rights accessRestricted access - author's decision
In past few years, many start-ups have started working. The few that survive make very important decisions by exploiting available data. This has amplified the importance of Business Intelligence and their tools, which empowers them to designate many advanced methods to cope with their problems. Churn is one of such problems. Many companies have worked on it and many papers have been published in the literature. Ulabox is a seven-year-old start-up at the time of writing this work and it is facing many challenges one of which is churn. In this work, we are going to apply some techniques of supervised and unsupervised machine learning to predict churn. Normally datasets available have straightforward patterns, natural patterns as in iris, or generated from a natural or logical selection process as in titanic dataset. Machine learning techniques yield good results with high precisions using these datasets; however, our dataset is generated from a fuzzy pattern of whimsical human mind, which makes it hard to make predictions with results as good as the typical datasets.