Data mining on real data of an airline

The theme of this thesis is data mining which is becoming an integral part of businesses operations thanks to the possibility of storing large amounts of data. The aim of the thesis is to perform data mining on an airline’s data and its use to improve the satisfaction of its passengers. In the first part, there is described theoretical background, which is important for understanding the practical part. The most represented is description of knowledge discovery in databases phases according to CRISP-DM methodology and subsequently description of used analytical methods. The second part of the thesis is practical, and it follows described CRISP-DM methodology. This part contains data mining in the form of association rules, clustering, decision trees, neural networks, and logistic regression. I have concluded that the airline should use knowledge of confident rules with high support and focus on areas contained in rules. It should also use decision trees which are best of all classification methods for this problem. For the classification tasks I created two models. One covers the whole dataset and the second one covers only attributes that can be known before boarding the plane. First model has an accuracy of 95,5 % and the second one has an accuracy of 82,3 %. Decision trees have the highest accuracy probably due to the bigger number of categorical variables in the dataset.